1 2018-06-01 Richard Biener <rguenther@suse.de>
4 * tree-ssa-structalias.c (get_function_part_constraint):
6 (find_func_aliases_for_call): Properly handle indirect
8 (find_func_clobbers): Likewise.
9 (ipa_pta_execute): Likewise.
10 (create_variable_info_for): For functions that are ifunc_resolver
11 resolve to a varinfo that contains the result of the resolver
13 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
16 2018-05-31 Michael Collison <michael.collison@arm.com>
18 * config/aarch64/aarch64.md:
19 (*fix_to_zero_extenddfdi2): New pattern.
20 * gcc.target/aarch64/fix_extend1.c: New testcase.
22 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
26 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
27 and BUILT_IN_STRNCMP_EQ.
28 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
30 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
31 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
32 (gimple_fold_builtin): Likewise.
33 * tree-ssa-strlen.c (compute_string_length): New function.
34 (determine_min_obsize): New function.
35 (handle_builtin_string_cmp): New function to handle calls to
36 string compare functions.
37 (strlen_optimize_stmt): Add handling to builtin string compare
39 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
40 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
41 * tree.c (build_common_builtin_nodes): Add new defines of
42 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
44 2018-05-31 Jakub Jelinek <jakub@redhat.com>
47 * bb-reorder.c (pass_partition_blocks::gate): Return false for
48 functions with naked attribute.
50 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
52 * config/i386/sse.md (avx_vec_concat<mode>):
53 Substitute concat_tg_mode mode attribute with xtg_mode.
54 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
55 (concat_tg_mode): Remove mode attribute.
57 2018-05-31 Martin Sebor <msebor@redhat.com>
60 * calls.c (alloc_max_size): Correct a logic error/typo.
61 Treat excessive arguments as infinite. Warn for invalid arguments.
62 * doc/invoke.texi (-Walloc-size-larger-than): Update.
64 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
67 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
70 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
71 Cesar Philippidis <cesar@codesourcery.com>
74 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
75 when emitting error on private/firstprivate reductions.
76 * omp-low.c (lower_omp_target): Avoid reference-type processing
77 on pointers for firstprivate clause.
79 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
81 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
84 * config/aarch64/aarch64-simd.md
85 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
86 (aarch64_ld1_x3_<mode>): Likewise
87 (aarch64_st1x2<VALLDIF:mode>): Likewise
88 (aarch64_st1_x2_<mode>): Likewise
89 (aarch64_st1x3<VALLDIF:mode>): Likewise
90 (aarch64_st1_x3_<mode>): Likewise
91 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
92 (vld1_s8_x3): Likewise.
93 (vld1_u16_x3): Likewise.
94 (vld1_s16_x3): Likewise.
95 (vld1_u32_x3): Likewise.
96 (vld1_s32_x3): Likewise.
97 (vld1_u64_x3): Likewise.
98 (vld1_s64_x3): Likewise.
99 (vld1_f16_x3): Likewise.
100 (vld1_f32_x3): Likewise.
101 (vld1_f64_x3): Likewise.
102 (vld1_p8_x3): Likewise.
103 (vld1_p16_x3): Likewise.
104 (vld1_p64_x3): Likewise.
105 (vld1q_u8_x3): Likewise.
106 (vld1q_s8_x3): Likewise.
107 (vld1q_u16_x3): Likewise.
108 (vld1q_s16_x3): Likewise.
109 (vld1q_u32_x3): Likewise.
110 (vld1q_s32_x3): Likewise.
111 (vld1q_u64_x3): Likewise.
112 (vld1q_s64_x3): Likewise.
113 (vld1q_f16_x3): Likewise.
114 (vld1q_f32_x3): Likewise.
115 (vld1q_f64_x3): Likewise.
116 (vld1q_p8_x3): Likewise.
117 (vld1q_p16_x3): Likewise.
118 (vld1q_p64_x3): Likewise.
119 (vst1_s64_x2): Likewise.
120 (vst1_u64_x2): Likewise.
121 (vst1_f64_x2): Likewise.
122 (vst1_s8_x2): Likewise.
123 (vst1_p8_x2): Likewise.
124 (vst1_s16_x2): Likewise.
125 (vst1_p16_x2): Likewise.
126 (vst1_s32_x2): Likewise.
127 (vst1_u8_x2): Likewise.
128 (vst1_u16_x2): Likewise.
129 (vst1_u32_x2): Likewise.
130 (vst1_f16_x2): Likewise.
131 (vst1_f32_x2): Likewise.
132 (vst1_p64_x2): Likewise.
133 (vst1q_s8_x2): Likewise.
134 (vst1q_p8_x2): Likewise.
135 (vst1q_s16_x2): Likewise.
136 (vst1q_p16_x2): Likewise.
137 (vst1q_s32_x2): Likewise.
138 (vst1q_s64_x2): Likewise.
139 (vst1q_u8_x2): Likewise.
140 (vst1q_u16_x2): Likewise.
141 (vst1q_u32_x2): Likewise.
142 (vst1q_u64_x2): Likewise.
143 (vst1q_f16_x2): Likewise.
144 (vst1q_f32_x2): Likewise.
145 (vst1q_f64_x2): Likewise.
146 (vst1q_p64_x2): Likewise.
147 (vst1_s64_x3): Likewise.
148 (vst1_u64_x3): Likewise.
149 (vst1_f64_x3): Likewise.
150 (vst1_s8_x3): Likewise.
151 (vst1_p8_x3): Likewise.
152 (vst1_s16_x3): Likewise.
153 (vst1_p16_x3): Likewise.
154 (vst1_s32_x3): Likewise.
155 (vst1_u8_x3): Likewise.
156 (vst1_u16_x3): Likewise.
157 (vst1_u32_x3): Likewise.
158 (vst1_f16_x3): Likewise.
159 (vst1_f32_x3): Likewise.
160 (vst1_p64_x3): Likewise.
161 (vst1q_s8_x3): Likewise.
162 (vst1q_p8_x3): Likewise.
163 (vst1q_s16_x3): Likewise.
164 (vst1q_p16_x3): Likewise.
165 (vst1q_s32_x3): Likewise.
166 (vst1q_s64_x3): Likewise.
167 (vst1q_u8_x3): Likewise.
168 (vst1q_u16_x3): Likewise.
169 (vst1q_u32_x3): Likewise.
170 (vst1q_u64_x3): Likewise.
171 (vst1q_f16_x3): Likewise.
172 (vst1q_f32_x3): Likewise.
173 (vst1q_f64_x3): Likewise.
174 (vst1q_p64_x3): Likewise.
176 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
178 * config/msp430/msp430.c (msp430_output_labelref): Prepend
179 user_label_prefix to name.
181 * tree-core.h: Update comment about the format of NAME string
182 passed to handler in attribute_spec.
184 * config/msp430/msp430.md: Remove erroneous subreg expression from
185 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
186 zero_extend{q,h}isi2.
188 2018-05-30 Borislav Petkov <bp@suse.de>
190 * doc/extend.texi: Document some architecture specific
191 constraints and sort entries.
193 2018-05-30 Martin Sebor <msebor@redhat.com>
196 * builtins.c (expand_builtin_stpcpy_1): New function.
197 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
198 only if the former succeeds.
200 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
202 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
205 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
207 * doc/invoke.texi (-flinker-output): Document
209 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
211 * passes.c (ipa_write_summaries): Only modify statements if body
213 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
214 incrementally linking.
215 (ipa_passes): Likewise.
216 * lto-cgraph.c (lto_output_node): When incrementally linking do not
217 pass down resolution info.
218 * common.opt (flag_incremental_link): Update info.
219 * gcc.c (plugin specs): Turn flinker-output=* to
220 -plugin-opt=-linker-output-known
221 * toplev.c (compile_file): Also cut compilation when doing incremental
223 * flag-types. (enum lto_partition_model): Add
224 LTO_LINKER_OUTPUT_NOLTOREL.
225 (invoke.texi): Add -flinker-output docs.
226 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
227 link same way as WPA; do not stream in dead initializers.
229 * dwarf2out.c (dwarf2out_die_ref_for_decl,
230 darf2out_register_external_decl): Support incremental link.
232 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
234 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
236 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
238 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
239 it down to simple_object_copy_lto_debug_sections.
240 (run_gcc): Determine incremental LTO link time and configure
241 lto1 into non-wpa mode, disable renaming of debug sections.
243 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
245 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
246 descriptions of various incorrectly documented functions.
248 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
251 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
252 address check not strict.
254 2018-05-30 Richard Biener <rguenther@suse.de>
256 PR tree-optimization/85964
257 * tracer.c (better_p): Drop initialized count check, we only
258 call the function with initialized counts now.
259 (find_best_successor): Do find a best edge if one
260 has uninitialized count.
261 (find_best_predecessor): Likewise. Do BB frequency check only
262 if count is initialized.
264 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
266 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
267 (aarch64_ldrstr_offset_compare): New.
268 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
269 load/store orderings.
270 (aarch64_gen_adjusted_ldpstp): Likewise.
272 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
274 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
275 Check for subset of GENERAL_REGS and FP_REGS.
276 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
279 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
281 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
282 and wi::to_poly_offset. Add the current offset and then check
283 whether the sum fits, rather than using an unchecked addition of
284 a checked term. Check for a shwi rather than a uhwi.
285 * expr.c (get_bit_range): Use tree_to_poly_uint64.
286 (store_constructor): Use poly_int_tree_p.
287 (expand_expr_real_1): Likewise.
288 * function.c (assign_temp): Likewise.
289 * fold-const.c (const_binop): Use poly_int_tree_p and
291 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
293 * ipa-icf-gimple.c (func_checker::compare_operand): Use
294 to_poly_offset for MEM offsets.
295 * ipa-icf.c (sem_variable::equals): Likewise.
296 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
297 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
298 wi::to_poly_offset for BIT_FIELD_REF offsets.
299 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
301 * var-tracking.c (emit_note_insn_var_location): Use
304 2018-05-29 Jim Wilson <jimw@sifive.com>
306 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
308 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
311 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
312 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
314 (sse4_1_round<mode>2): Use nonimmediate_operand
315 for operand 1 predicate.
317 2018-05-29 Martin Sebor <msebor@redhat.com>
318 Richard Biener <rguenther@suse.de>
321 * calls.c (get_size_range): Call determine_value_range instead
323 * tree-vrp.h (determine_value_range): Declared new function.
324 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
326 2018-05-29 Richard Biener <rguenther@suse.de>
328 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
329 sure to use non-pattern stmts for get_earlier_stmt arguments.
330 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
331 called on pattern stmts.
332 (get_later_stmt): Likewise.
334 2018-05-29 Martin Liska <mliska@suse.cz>
336 PR gcov-profile/85759
337 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
340 2018-05-29 Jakub Jelinek <jakub@redhat.com>
342 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
344 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
348 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
349 VEC_PACK_FLOAT_EXPR): New tree codes.
350 * tree-pretty-print.c (op_code_prio): Handle
351 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
352 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
353 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
354 * tree-inline.c (estimate_operator_cost): Likewise.
355 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
356 * fold-const.c (const_binop): Likewise.
357 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
358 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
359 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
360 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
361 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
362 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
363 * expr.c (expand_expr_real_2): Likewise.
364 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
365 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
366 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
368 * optabs.c (expand_widen_pattern_expr): For
369 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
370 sign from result type rather than operand's type.
371 (expand_binop_directly): For vec_packu_float_optab and
372 vec_packs_float_optab allow result type to be different from operand's
374 * optabs-tree.c (optab_for_tree_code): Handle
375 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
376 VEC_PACK_FLOAT_EXPR. Formatting fixes.
377 * tree-vect-generic.c (expand_vector_operations_1): Handle
378 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
380 * tree-vect-stmts.c (supportable_widening_operation): Handle
382 (supportable_narrowing_operation): Handle FLOAT_EXPR.
383 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
384 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
385 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
386 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
388 (vec_pack<floatprefix>_float_<mode>): New expander.
389 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
391 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
392 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
393 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
394 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
395 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
397 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
398 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
399 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
400 VEC_PACK_FLOAT_EXPR): Document.
402 2018-05-29 Richard Biener <rguenther@suse.de>
404 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
406 (stmt_vec_info_vec): Make pointer.
407 (init_stmt_vec_info_vec): Remove.
408 (free_stmt_vec_info_vec): Likewise.
409 (set_stmt_vec_info_vec): New function.
410 (free_stmt_vec_infos): Likewise.
411 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
412 (set_vinfo_for_stmt): Likewise.
413 (get_earlier_stmt): Likewise.
414 (get_later_stmt): Likewise.
415 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
416 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
417 (vec_info::~vec_info): Free stmt_vec_infos.
418 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
419 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
420 (pass_slp_vectorize::execute): Likewise.
421 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
422 (free_stmt_vec_info_vec): Likewise.
423 (set_stmt_vec_info_vec): New function.
424 (free_stmt_vec_infos): Likewise.
425 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
426 the global stmt_vec_info_vec.
427 * tree-parloops.c (gather_scalar_reductions): Use
428 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
431 2018-05-29 Richard Biener <rguenther@suse.de>
433 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
435 2018-05-29 Martin Liska <mliska@suse.cz>
436 David Malcolm <dmalcolm@redhat.com>
438 * vec.c (test_reverse): New.
439 (vec_c_tests): Add new test.
440 * vec.h (vl_ptr>::reverse): New function.
442 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
444 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
446 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
449 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
451 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
453 2018-05-28 Richard Biener <rguenther@suse.de>
455 PR tree-optimization/85933
456 * tree-vect-data-refs.c (vect_record_base_alignments): Only
457 look at stmts marked as vectorizable.
459 2018-05-28 Richard Biener <rguenther@suse.de>
461 PR tree-optimization/85934
462 * tree-vect-generic.c (expand_vector_operations_1): Hoist
463 vector boolean check before scalar optimization.
465 2018-05-28 Jakub Jelinek <jakub@redhat.com>
467 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
470 2018-05-28 Mark Wielaard <mark@klomp.org>
472 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
473 if it is an expression containing a minus sign.
475 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
477 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
479 2018-05-27 Paul Koning <ni1d@arrl.net>
481 * config/pdp11/pdp11.md (truncsihi2): Remove.
483 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
484 Chung-Ju Wu <jasonwucj@gmail.com>
486 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
488 (unaligned_store_dw): Ditto.
489 * config/nds32/nds32-memory-manipulation.c
490 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
491 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
492 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
493 (emit_setmem_word_loop): Rename to ...
494 (emit_setmem_doubleword_loop): ... this.
495 (nds32_gen_dup_4_byte_to_word_value): New function.
496 (nds32_gen_dup_8_byte_to_double_word_value): New function.
497 (nds32_expand_setmem_loop): Refine implementation.
498 (nds32_expand_setmem_loop_v3m): Ditto.
499 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
502 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
504 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
506 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
508 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
509 (nds32_init_machine_status): Initialize machine->attr_naked_p and
510 machine->attr_no_prologue_p.
511 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
512 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
513 (nds32_expand_epilogue): Consider attr_naked_p.
514 (nds32_expand_epilogue_v3pop): Likewise.
515 (nds32_can_use_return_insn): Likewise.
516 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
517 attr_no_prologue_p fields.
518 * config/nds32/nds32.opt (mret-in-naked-func): New option.
520 2018-05-27 Jakub Jelinek <jakub@redhat.com>
523 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
526 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
528 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
530 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
532 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
534 (*<floatsuffix>floatv2div2sf2): Rename to ...
535 (*float<floatunssuffix>v2div2sf2): ... this.
536 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
537 (float<floatunssuffix>v2div2sf2_mask): ... this.
538 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
539 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
540 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
542 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
544 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
546 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
548 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
550 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
552 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
553 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
554 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
555 gen_ufix_truncv8dfv8si2.
556 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
557 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
558 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
559 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
560 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
561 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
562 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
563 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
565 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
569 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
571 2018-05-25 Jim Wilson <jimw@sifive.com>
573 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
574 * config/riscv/riscv.c (struct machine_function): Add
575 interrupt_handler_p and attribute_checked_p fields.
576 (riscv_attribute_table): Add interrupt.
577 (riscv_interrupt_type_p): New.
578 (riscv_save_reg_p): Save extra regs for interrupt handler.
579 (riscv_use_save_libcall): Return false for interrupt handler.
580 (riscv_first_stack_step): Add forward declaration.
581 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
582 for interrupt handler with large frame. Use it for saved reg list.
583 (riscv_expand_prologue): Move flag_stack_usage_info support to
584 eliminate duplication.
585 (riscv_expand_epilogue): Generate mret for interrupt handler.
586 (riscv_epilogue_uses): New.
587 (riscv_can_use_return_insn): Return false for interrupt handler.
588 (riscv_function_ok_for_sibcall): Likewise.
589 (riscv_set_current_function): Add interrupt handler support.
590 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
591 * config/riscv/riscv.md (UNSPECV_MRET): New.
593 (riscv_frflags, riscv_fsflags): Use tab after opcode.
595 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
597 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
599 PR tree-optimization/85712
600 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
601 this candidate has already been replaced in-situ by a copy.
603 2018-05-25 Jason Merrill <jason@redhat.com>
605 PR c++/80485 - inline function non-zero address.
606 * symtab.c (nonzero_address): Check DECL_COMDAT.
608 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
611 * config/alpha/alpha.md (ashlsi3): New insn pattern.
612 (*ashlsi_se): Rename from *ashldi_se. Define as sign
613 extension of SImode operation. Use const123_operand predicate.
615 (*saddl_se_1): Ditto.
617 (*ssubl_se_1): Ditto.
618 * config/alpha/predicates.md (const123_operand): New predicate.
619 * config/alpha/constraints.md (P): Use IN_RANGE.
621 2018-05-25 Richard Biener <rguenther@suse.de>
623 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
625 (ref_maybe_used_by_stmt_p): Likewise.
626 (stmt_may_clobber_ref_p): Likewise.
627 (stmt_may_clobber_ref_p_1): Likewise.
628 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
630 (ref_maybe_used_by_stmt_p): Likewise.
631 (stmt_may_clobber_ref_p): Likewise.
632 (stmt_may_clobber_ref_p_1): Likewise.
633 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
634 the alias oracle to disambiguate DRs with stmts DR analysis
636 (vect_analyze_data_refs): Do not give up on not analyzable
637 DRs for BB vectorization. Remove code truncating the dataref
640 2018-05-25 Jakub Jelinek <jakub@redhat.com>
643 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
644 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
645 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
647 2018-05-25 Richard Biener <rguenther@suse.de>
649 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
650 function, combining stmt data ref gathering and fatal analysis
652 (vect_analyze_data_refs): Remove now redudnant code and simplify.
653 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
654 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
655 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
656 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
658 2018-05-25 Bin Cheng <bin.cheng@arm.com>
660 PR tree-optimization/85720
661 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
662 SCC if all partitions are builtins.
663 (version_loop_by_alias_check): New parameter. Generate cancelable
664 runtime alias check if all partitions are builtins.
665 (distribute_loop): Update call to above function.
667 2018-05-25 Bin Cheng <bin.cheng@arm.com>
669 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
670 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
671 (parm_default_def_partition_arg): Ditto.
672 (set_parm_default_def_partition): Ditto.
673 (get_parm_default_def_partitions): Ditto and make it static.
674 (get_undefined_value_partitions): Ditto and make it static.
675 (remove_ssa_form): Refactor call to init_var_map here.
676 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
677 computation for loop region.
678 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
679 (register_default_def): Delete.
680 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
681 (parm_default_def_partition_arg): Ditto.
682 (set_parm_default_def_partition): Ditto.
683 (get_parm_default_def_partitions): Ditto and make it static.
684 (get_undefined_value_partitions): Ditto and make it static.
685 (coalesce_with_default, coalesce_with_default): Update comment.
686 (create_coalesce_list_for_region): New func factored out from
687 create_outofssa_var_map.
688 (populate_coalesce_list_for_outofssa): New func factored out from
689 create_outofssa_var_map and coalesce_ssa_name.
690 (create_outofssa_var_map): Delete.
691 (coalesce_ssa_name): Refactor to support live range computation.
692 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
693 (get_parm_default_def_partitions): Delete.
694 (get_undefined_value_partitions): Ditto.
695 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
696 computation for loop region.
697 (new_tree_live_info, loe_visit_block): Ditto.
698 (live_worklist, set_var_live_on_entry): Ditto.
699 (calculate_live_on_exit, verify_live_on_entry): Ditto.
700 * tree-ssa-live.h (struct _var_map): New fields.
701 (init_var_map): Change decl.
702 (region_contains_p): New.
704 2018-05-25 Bin Cheng <bin.cheng@arm.com>
706 * tree-ssa-live.h (live_merge_and_clear): Delete.
708 2018-05-25 Richard Biener <rguenther@suse.de>
711 * tree-dump.c (dequeue_and_dump): Remove access to removed
712 operand 2 of a SWITCH_EXPR.
714 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
716 * doc/sourcebuild.texi (vect_double_cond_arith): Include
717 multiplication and division.
718 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
719 (cond_udiv@var{m}, cond_umod@var{m}): Document.
720 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
721 (cond_udiv_optab, cond_umod_optab): New optabs.
722 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
723 (IFN_COND_RDIV): New internal functions.
724 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
725 TRUNC_MOD_EXPR and RDIV_EXPR.
726 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
727 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
729 (SVE_INT_BINARY): Include mult.
730 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
731 (optab, sve_int_op): Handle mult.
732 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
734 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
735 for SVE_INT_BINARY_SD.
737 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
739 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
740 (optab, sve_int_op): Handle div and udiv.
741 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
742 for SVE_INT_BINARY_SD.
743 (*<optab><mode>3): New insn for the same.
745 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
747 * tree-vect-patterns.c: Include predict.h.
748 (vect_recog_divmod_pattern): Restrict check for division support
749 to when optimizing for size.
751 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
753 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
754 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
755 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
756 (gimple_match_op::set_op): Likewise.
757 (gimple_resimplify4): Declare.
758 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
759 (expr::gen_transform): Likewise.
760 (decision_tree::gen): Generate a simplification routine for 4 operands.
761 * gimple-match-head.c (gimple_simplify): Add an overload for
762 4 operands. In the top-level function, handle up to 4 call
763 arguments and call gimple_resimplify4.
764 (gimple_resimplify4): New function.
765 (build_call_internal): Pass a fourth operand.
766 (maybe_push_to_seq): Likewise.
767 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
768 Fold VEC_COND_EXPRs of an operation and a default value into
769 an IFN_COND_* function if possible.
770 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
772 (SVE_COND_FP_BINARY): Include them.
773 (optab, sve_fp_op): Handle them.
774 (SVE_INT_BINARY_REV): New code iterator.
775 (SVE_COND_FP_BINARY_REV): New int iterator.
776 (commutative): New int attribute.
777 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
779 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
781 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
782 (*cond_<optab><mode>): New patterns for reversed operands.
784 2018-05-25 Richard Biener <rguenther@suse.de>
786 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
787 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
788 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
789 (STMT_VINFO_GROUPED_ACCESS): Adjust.
790 * tree-vect-data-refs.c (everywhere): Adjust users.
791 * tree-vect-loop.c (everywhere): Likewise.
792 * tree-vect-slp.c (everywhere): Likewise.
793 * tree-vect-stmts.c (everywhere): Likewise.
794 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
796 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
798 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
800 (gcc_cv_as_section_exclude): ... this.
801 Try Solaris as #exclude syntax.
802 * configure: Regenerate.
803 * config.in: Regenerate.
804 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
806 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
807 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
809 * varasm.c (default_elf_asm_named_section): Don't check if
810 HAVE_GAS_SECTION_EXCLUDE is defined.
812 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
814 * doc/md.texi: Update the documentation of the cond_* optabs
815 to mention the new final operand. Fix GET_MODE_NUNITS call.
816 Describe the scalar case too.
817 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
818 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
820 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
821 (get_conditional_internal_fn): Update comment.
822 * tree-vect-loop.c (vectorizable_reduction): Pass the original
823 accumulator value as a final argument to conditional functions.
824 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
825 a define_expand and add an "else" operand. Assert for now that
826 the else operand is equal to operand 2. Use SVE_INT_BINARY and
827 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
828 (*cond_<optab><mode>): New patterns.
829 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
830 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
831 (UNSPEC_COND_EOR): Delete.
832 (optab): Remove associated mappings.
833 (SVE_INT_BINARY): New code iterator.
834 (sve_int_op): Remove int attribute and add "minus" to the code
836 (SVE_COND_INT_OP): Delete.
837 (SVE_COND_FP_OP): Rename to...
838 (SVE_COND_FP_BINARY): ...this.
840 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
842 * optabs.c (can_reuse_operands_p): New function.
843 (maybe_legitimize_operands): Try to reuse the results for
846 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
848 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
849 Add {q} suffix to insn mnemonic.
851 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
853 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
854 (msp430_warn_func_return): New.
856 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
858 * fold-const.c (tree_nonzero_bits): New function.
859 * fold-const.h (tree_nonzero_bits): Likewise.
860 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
861 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
863 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
867 * varasm.c (assemble_alias): Check ifunc_resolver only on
870 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
873 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
874 when memory input operand is handled.
876 2018-05-24 Luis Machado <luis.machado@linaro.org>
878 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
880 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
882 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
884 * match.pd: Delay FMA folds until after vectorization.
886 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
889 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
890 address check not strict.
892 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
894 * gimple-match.h (gimple_match_op): New class.
895 (mprts_hook): Replace parameters with a gimple_match_op *.
896 (maybe_build_generic_op): Likewise.
897 (gimple_simplified_result_is_gimple_val): Replace parameters with
898 a const gimple_match_op *.
899 (gimple_simplify): Replace code_helper * and tree * parameters with
900 a gimple_match_op * parameter.
901 (gimple_resimplify1): Replace code_helper *, tree and tree *
902 parameters with a gimple_match_op * parameter.
903 (gimple_resimplify2): Likewise.
904 (gimple_resimplify3): Likewise.
905 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
906 parameters with a gimple_match_op * parameter.
907 * gimple-match-head.c (gimple_simplify): Change prototypes of
908 auto-generated functions to take a gimple_match_op * instead of
909 separate code_helper * and tree * parameters. Make the same
910 change in the top-level overload and update calls to the
911 gimple_resimplify routines. Update calls to the auto-generated
912 functions and to maybe_push_res_to_seq in the publicly-facing
913 operation-specific gimple_simplify overloads.
914 (gimple_match_op::MAX_NUM_OPS): Define.
915 (gimple_resimplify1): Replace rcode and ops with a single res_op
916 parameter. Update call to gimple_simplify.
917 (gimple_resimplify2): Likewise.
918 (gimple_resimplify3): Likewise.
919 (mprts_hook): Replace parameters with a gimple_match_op *.
920 (maybe_build_generic_op): Likewise.
921 (build_call_internal): Replace type, nargs and ops with
923 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
924 with a single gimple_match_op *. Update calls to mprts_hook,
925 build_call_internal and gimple_simplified_result_is_gimple_val.
926 Factor out code that is common to the tree_code and combined_fn cases.
927 * genmatch.c (expr::gen_transform): Replace tem_code and
928 tem_ops with a gimple_match_op called tem_op. Update calls
929 to the gimple_resimplify functions and maybe_push_res_to_seq.
930 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
931 res_ops. Update call to the gimple_resimplify functions.
932 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
933 (decision_tree::gen): Make the functions take a gimple_match_op *
934 called res_op instead of separate res_code and res_ops parameters.
935 Update call accordingly.
936 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
937 and ops with a single res_op parameter. Update calls to
938 maybe_build_generic_op and maybe_push_res_to_seq.
939 (fold_stmt_1): Update calls to gimple_simplify and
940 replace_stmt_with_simplification.
941 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
942 and gimple_simplified_result_is_gimple_val.
943 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
945 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
946 with a gimple_match_op *.
947 (vn_nary_build_or_lookup): Likewise. Update call to
948 vn_nary_build_or_lookup_1.
949 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
950 gimple_match_op *. Update calls to the gimple_resimplify routines
951 and to gimple_simplified_result_is_gimple_val.
952 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
953 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
954 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
955 (visit_nary_op): Likewise.
956 (visit_reference_op_load): Likewise.
958 2018-05-23 Luis Machado <luis.machado@linaro.org>
960 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
961 modifier for printing the step amount.
963 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
966 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
969 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
971 * doc/sourcebuild.texi (Endianness): New subsubsection.
973 2018-05-23 Luis Machado <luis.machado@linaro.org>
975 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
976 <prefetch_dynamic_strides>: New const bool field.
977 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
978 prefetch_dynamic_strides.
979 (exynosm1_prefetch_tune): Likewise.
980 (thunderxt88_prefetch_tune): Likewise.
981 (thunderx_prefetch_tune): Likewise.
982 (thunderx2t99_prefetch_tune): Likewise.
983 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
985 (aarch64_override_options_internal): Update to set
986 PARAM_PREFETCH_DYNAMIC_STRIDES.
987 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
988 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
989 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
990 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
991 prefetch-dynamic-strides setting.
993 2018-05-23 Luis Machado <luis.machado@linaro.org>
995 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
996 <minimum_stride>: New const int field.
997 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
998 minimum_stride field defaulting to -1.
999 (exynosm1_prefetch_tune): Likewise.
1000 (thunderxt88_prefetch_tune): Likewise.
1001 (thunderx_prefetch_tune): Likewise.
1002 (thunderx2t99_prefetch_tune): Likewise.
1003 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1004 <default_opt_level>: Set to 3.
1005 (aarch64_override_options_internal): Update to set
1006 PARAM_PREFETCH_MINIMUM_STRIDE.
1007 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1008 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1009 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1010 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1011 stride is constant and is below the minimum stride threshold.
1013 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1015 * config/arm/arm-cpus.in (mode26): Delete.
1016 (armv4): Delete mode26 reference.
1017 * config/arm/arm.c (arm_configure_build_target): Delete use of
1020 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1022 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1024 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1025 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1027 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1028 (floatunsdidf2): Ditto.
1030 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1032 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1033 (fixuns_trunc<mode>si2_avx512f): Ditto.
1034 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1035 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1036 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1038 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1040 PR rtl-optimization/79985
1041 * df-scan.c (df_insn_refs_collect): Remove special case for
1042 global registers and asm statements.
1044 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1046 * extend.texi (Global Register Variables): Rewrite the bullet list.
1047 Note that the register is available for allocation. Note that access
1048 via inline asm must use constraints. Add note about async-signal
1049 handlers. Remove paragraph about automagic register selection.
1051 2018-05-23 Richard Biener <rguenther@suse.de>
1053 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1054 of fixed offset from memset VN.
1056 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1058 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1060 (alloc_cand_and_find_basis): Initialize first_interp field.
1061 (slsr_process_mul): Modify first_interp field.
1062 (slsr_process_add): Likewise.
1063 (slsr_process_cast): Modify first_interp field for each new
1065 (slsr_process_copy): Likewise.
1066 (dump_candidate): Dump first_interp field.
1067 (replace_mult_candidate): Process all interpretations, not just
1069 (replace_rhs_if_not_dup): Likewise.
1070 (replace_one_candidate): Likewise.
1072 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1074 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1076 (aarch64_needs_frame_chain): New function.
1077 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1079 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1082 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1083 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1084 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1085 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1086 as true for strict-align.
1087 (aarch64_can_inline_p): Perform checks even when callee has no
1088 attributes to check for strict alignment.
1089 * doc/extend.texi (AArch64 Function Attributes): Document
1091 * doc/invoke.texi: (AArch64 Options): Likewise.
1093 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1095 PR tree-optimization/85853
1096 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1097 the handling of the root of the node to...
1098 (vect_slp_analyze_node_operations_1): ...this new function,
1099 and run the whole thing with the child nodes' def types
1100 set according to their SLP node's def type.
1102 2018-05-23 Richard Biener <rguenther@suse.de>
1105 * tree-data-ref.c (create_runtime_alias_checks): Defer
1106 and ignore overflow warnings.
1108 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1110 PR tree-optimization/85822
1111 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1114 2018-05-23 Richard Biener <rguenther@suse.de>
1116 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1117 memset constants via native_interpret_expr.
1119 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1122 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1124 (cgraph_node::create_alias): Likewise.
1125 (cgraph_node::get_availability): Check ifunc_resolver instead
1126 of looking up ifunc attribute.
1127 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1128 * varasm.c (do_assemble_alias): Likewise.
1129 (assemble_alias): Likewise.
1130 (default_binds_local_p_3): Likewise.
1131 * cgraph.h (cgraph_node): Add ifunc_resolver.
1132 (cgraph_node::only_called_directly_or_aliased_p): Return false
1134 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1136 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1137 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1138 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1139 instead of looking up ifunc attribute.
1141 2018-05-22 Luis Machado <luis.machado@linaro.org>
1143 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1145 2018-05-22 Martin Sebor <msebor@redhat.com>
1148 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1149 only when expasion succeeds.
1150 (expand_builtin_strcmp): Same.
1151 (expand_builtin_strncmp): Same.
1153 2018-05-22 Martin Sebor <msebor@redhat.com>
1155 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1157 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1158 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1160 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1161 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1163 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1164 Add check for legitimate_address.
1165 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1166 (aarch64_swap_ldrstr_operands): New.
1167 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1170 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1171 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1173 * config/aarch64/aarch64.md: New patterns to generate stp
1175 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1176 single words and double words.
1177 (load_pair_sw, load_pair_dw): Likewise.
1178 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1180 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1182 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1183 for different mode ldpstp and add peephole for merged zero stores.
1185 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1187 (aarch64_gen_store_pair): Rename calls to match new patterns.
1188 (aarch64_gen_load_pair): Rename calls to match new patterns.
1189 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1190 (load_pair<DREG:mode><DREG2:mode>): ... This.
1191 (store_pair<mode>): Rename to...
1192 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1193 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1195 (V_INT_EQUIV): Handle SImode.
1196 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1199 2018-05-22 Martin Sebor <msebor@redhat.com>
1202 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1203 or ajust the presumed bound on an operation to avoid unnecessary
1206 2018-05-22 Martin Sebor <msebor@redhat.com>
1208 PR tree-optimization/85826
1209 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1210 assuming that a DECL necesarily has a constant size.
1212 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1215 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1217 2018-05-22 Richard Biener <rguenther@suse.de>
1219 PR tree-optimization/85834
1220 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1221 non-constant and non-zero memset arguments.
1223 2018-05-22 Martin Liska <mliska@suse.cz>
1226 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1228 2018-05-22 Richard Biener <rguenther@suse.de>
1230 PR tree-optimization/85863
1231 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1232 comparisons when vectype is specified.
1233 (vectorizable_condition): Do not specify vectype for
1234 vect_is_simple_cond when SLP vectorizing.
1236 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1239 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1240 define __ibm128 as long double.
1241 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1242 as a distinct type when IEEE 128-bit support is enabled.
1243 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1244 128-bit types to use the correct functions.
1245 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1246 convert between 128-bit floating point types that have different
1247 modes but the same representation, instead of using gen_lowpart to
1249 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1251 (IFKF_reg): New attributes to give the register constraints for
1253 (extend<mode>tf2_internal): New insns to mark an explicit
1254 conversion between 128-bit floating point types that have a
1255 different mode but share the same representation.
1257 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1259 PR tree-optimization/85814
1260 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1261 a null return from get_strinfo when unsharing the next
1262 strinfo in the chain.
1264 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1267 * varasm.c (weak_finish): Clean up weak_decls.
1269 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1271 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1272 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1273 UNSPEC_UADALP values.
1274 * config/aarch64/iterators.md (ABAL): New int iterator.
1277 (sur): Add mappings for the above.
1278 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1280 (aarch64_<sur>abal<mode>_4): Likewise.
1281 (aarch64_<sur>adalp<mode>_3): Likewise.
1282 (<sur>sadv16qi): New define_expand.
1284 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1286 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1287 (*movdf_internal): Ditto.
1288 (*rcpsf2_sse): Ditto.
1289 (*rsqrtsf2_sse): Ditto.
1290 (*sqrt<mode>2_sse): Ditto.
1292 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1294 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1296 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1297 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1298 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1299 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1301 * config/aarch64/arm_neon.h: Likewise.
1302 * config/aarch64/iterators.md (VQ_I): New.
1304 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1306 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1308 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1311 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1313 * config/nds32/constraints.md (S): New constraint.
1314 * config/nds32/nds32.md (call_internal): Use constraint S.
1315 (call_value_internal): Likewise.
1316 (sibcall_internal): Likewise.
1317 (sibcall_value_internal): Likewise.
1319 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1320 Chung-Ju Wu <jasonwucj@gmail.com>
1322 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1325 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1326 Chung-Ju Wu <jasonwucj@gmail.com>
1328 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1329 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1330 (nds32_rtx_costs_impl): Simplify.
1331 (nds32_address_cost_impl): Simplify.
1332 (nds32_init_rtx_costs): New function.
1333 (nds32_rtx_costs_speed_prefer): Likewise.
1334 (nds32_rtx_costs_size_prefer): Likewise.
1335 (nds32_address_cost_speed_prefer): Likewise.
1336 (nds32_address_cost_speed_fwprop): Likewise.
1337 (nds32_address_cost_size_prefer): Likewise.
1338 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1339 * config/nds32/nds32.c (nds32_option_override): Use
1340 nds32_init_rtx_costs function.
1342 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1345 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1346 (TARGET_PIPELINE_N8): Likewise.
1347 (TARGET_PIPELINE_N10): Likewise.
1348 (TARGET_PIPELINE_N13): Likewise.
1349 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1351 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1353 * config/nds32/nds32-fpu.md: Update copyright year.
1355 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1359 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1361 * config/nds32/nds32.c
1362 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1363 * config/nds32/nds32.opt (minline-asm-r15): New option.
1365 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1367 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1369 * config/nds32/nds32.md (abssi2): New pattern.
1371 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1373 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1374 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1375 Merge insn pattern from sse_cvtsi2ss<round_name> and
1376 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1377 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1378 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1379 using SWI48 mode iterator.
1380 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1381 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1382 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1383 pattern from sse_cvttss2si<round_saeonly_name>
1384 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1385 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1386 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1387 using SWI48 mode iterator.
1388 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1389 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1390 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1391 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1392 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1393 using SWI48 mode iterator.
1394 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1395 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1396 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1397 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1398 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1399 SWI48 mode iterator.
1400 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1401 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1402 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1403 pattern from sse_cvttsd2si<round_saeonly_name>
1404 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1406 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1408 * config/nds32/nds32-md-auxiliary.c
1409 (nds32_valid_smw_lwm_base_p): Refine.
1410 (nds32_output_smw_single_word): Refine.
1411 (nds32_output_smw_double_word): New.
1412 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1414 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1416 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1417 (nds32_output_stack_pop): Refine.
1418 (nds32_expand_unaligned_load): Refine.
1419 (nds32_expand_unaligned_store): Refine.
1421 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1422 Chung-Ju Wu <jasonwucj@gmail.com>
1424 * config/nds32/constants.md: Add TP_REGNUM constant.
1425 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1426 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1428 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1429 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1430 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1431 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1432 and PIC code generation.
1433 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1435 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1437 * config/nds32/nds32.md: Support TLS and PIC.
1438 * config/nds32/nds32.c: Support TLS and PIC.
1439 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1440 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1443 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1445 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1446 mode with E_ prefix.
1448 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1449 Chung-Ju Wu <jasonwucj@gmail.com>
1451 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1452 * config/nds32/nds32-md-auxiliary.c
1453 (symbolic_reference_mentioned_p): New.
1454 (nds32_legitimize_ict_address): New.
1455 (nds32_expand_ict_move): New.
1456 (nds32_indirect_call_referenced_p): New.
1457 (nds32_symbol_binds_local_p): Delete.
1458 (nds32_long_call_p): Modify.
1459 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1460 * config/nds32/nds32-protos.h
1461 (symbolic_reference_mentioned_p): Declare.
1462 (nds32_legitimize_ict_address): Declare.
1463 (nds32_expand_ict_move): Declare.
1464 (nds32_indirect_call_referenced_p): Declare.
1465 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1466 (nds32_relax_group): Use nds32_ict_const_p as condition.
1467 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1468 (nds32_asm_file_start): Output ict_model directive in asm code.
1469 (nds32_legitimate_address_p): Consider indirect call.
1470 (nds32_print_operand): Consider indirect call.
1471 (nds32_print_operand_address): Consider indirect call.
1472 (nds32_insert_attributes): Handle "indirect_call" attribute.
1473 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1474 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1475 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1476 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1477 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1478 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1479 (TARGET_ICT_MODEL_SMALL): Define.
1480 (TARGET_ICT_MODEL_LARGE): Define.
1481 * config/nds32/nds32.md (movsi): Consider ict model.
1482 (call, call_value): Consider ict model.
1483 (sibcall, sibcall_value): Consider ict model.
1484 * config/nds32/nds32.opt (mict-model): New option.
1485 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1488 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1489 Monk Chiang <sh.chiang04@gmail.com>
1490 Jim Wilson <jimw@sifive.com>
1492 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1493 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1494 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1495 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1496 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1497 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1498 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1499 compute save_libcall_adjustment properly.
1500 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1501 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1502 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1503 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1504 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1505 (ABI_SPEC): Handle mabi=ilp32e.
1506 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1507 (RVE): Add RVE mask.
1508 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1509 <-march>: Add rv32e as an example.
1511 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1514 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1515 (intra_create_variable_infos): Handle C++ constructors.
1517 2018-05-18 Martin Liska <mliska@suse.cz>
1519 * passes.def: Remove a redundant pass.
1521 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1524 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1526 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1528 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1530 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1532 (armv2, armv2a, armv3, armv3m): Delete architectures.
1533 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1534 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1535 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1537 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1538 (*mulsidi3adddi): Likewise.
1539 (mulsidi3): Likewise.
1540 (*mulsidi3_nov6): Likewise.
1541 (umulsidi3): Likewise.
1542 (umulsidi3_nov6): Likewise.
1543 (umaddsidi4): Likewise.
1544 (*umulsidi3adddi): Likewise.
1545 (smulsi3_highpart): Likewise.
1546 (*smulsi3_highpart_nov6): Likewise.
1547 (umulsi3_highpart): Likewise.
1548 (*umulsi3_highpart_nov6): Likewise.
1549 * config/arm/arm.h (arm_arch3m): Delete.
1550 * config/arm/arm.c (arm_arch3m): Delete.
1551 (arm_option_override_internal): Update armv3-related comment.
1552 (arm_configure_build_target): Delete use of isa_bit_mode32.
1553 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1554 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1555 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1557 (mulusa3): Likewise.
1558 * config/arm/arm-protos.h (arm_arch3m): Delete.
1559 * config/arm/arm-tables.opt: Regenerate.
1560 * config/arm/arm-tune.md: Likewise.
1561 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1562 deleted architectures.
1564 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1566 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1567 (armv5t, armv5te): New features.
1568 (ARMv5, ARMv5e): Delete fgroups.
1569 (ARMv5t, ARMv5te): Adjust for above changes.
1571 (armv5, armv5e): Delete arches.
1572 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1574 (*call_reg_arm): Likewise.
1575 (*call_value_reg_armv5): Likewise.
1576 (*call_value_reg_arm): Likewise.
1577 (*call_symbol): Likewise.
1578 (*call_value_symbol): Likewise.
1579 (*sibcall_insn): Likewise.
1580 (*sibcall_value_insn): Likewise.
1582 (prefetch): Likewise.
1583 (define_split and define_peephole2 dependent on arm_arch5):
1585 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1587 (TARGET_ARM_QBIT): Likewise.
1588 (TARGET_DSP_MULTIPLY): Likewise.
1589 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1590 (arm_arch5, arm_arch5e): Delete.
1591 (arm_arch5t, arm_arch5te): Declare.
1592 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1593 (arm_arch5t): Declare.
1594 (arm_option_reconfigure_globals): Update for the above.
1595 (arm_options_perform_arch_sanity_checks): Update comment, replace
1596 use of arm_arch5 with arm_arch5t.
1597 (use_return_insn): Likewise.
1598 (arm_emit_call_insn): Likewise.
1599 (output_return_instruction): Likewise.
1600 (arm_final_prescan_insn): Likewise.
1601 (arm_coproc_builtin_available): Likewise.
1602 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1603 arm_arch5e with arm_arch5t and arm_arch5te.
1604 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1605 (arm_arch5t, arm_arch5te): Declare.
1606 * config/arm/arm-tables.opt: Regenerate.
1607 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1608 * config/arm/t-multilib: Likewise.
1609 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1610 instead of arm_arch5.
1611 (*call_reg_thumb1): Likewise.
1612 (*call_value_reg_thumb1_v5): Likewise.
1613 (*call_value_reg_thumb1): Likewise.
1614 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1616 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1618 2018-05-18 Martin Liska <mliska@suse.cz>
1620 PR gcov-profile/84846
1621 * doc/gcov.texi: Document -t option of gcov tool.
1623 2018-05-18 Martin Liska <mliska@suse.cz>
1625 PR gcov-profile/84846
1626 * gcov.c (print_usage): Add new -t option.
1627 (process_args): Handle the option.
1628 (generate_results): Use stdout as output when requested by
1631 2018-05-18 Martin Liska <mliska@suse.cz>
1633 PR gcov-profile/84846
1634 * coverage.c (coverage_init): Write PWD to .gcno file.
1635 * doc/gcov.texi: Document how working directory is printed.
1636 * gcov-dump.c (dump_gcov_file): Print PWD.
1637 * gcov.c (output_intermediate_file): Likewise.
1638 (read_graph_file): Read PWD string.
1639 (output_lines): Print PWD.
1641 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1644 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1645 for retval and return false if all args to phi are zero.
1647 2018-05-18 Richard Biener <rguenther@suse.de>
1649 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1651 (evrp_dom_walker::before_dom_children): Call it.
1653 2018-05-18 Richard Biener <rguenther@suse.de>
1655 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1656 results when processing array refs with variable index.
1658 2018-05-18 Toon Moene <toon@moene.org>
1660 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1661 directly after that of -floop-interchange. Indicate that both
1662 options are enabled by default when specifying -O3.
1664 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1666 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1667 iterator. Delete separate integer-mode vec_set<mode> expander.
1668 (aarch64_simd_vec_setv2di): Delete.
1669 (vec_setv2di): Delete.
1670 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1671 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1672 the "w, r" alternative.
1674 2018-05-18 Martin Liska <mliska@suse.cz>
1676 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1677 * tree-pass.h (make_pass_lower_switch_O0): New function.
1678 * tree-switch-conversion.c (node_has_low_bound): Remove.
1679 (node_has_high_bound): Likewise.
1680 (node_is_bounded): Likewise.
1681 (class pass_lower_switch): Make it a template type and create
1683 (pass_lower_switch::execute): Add template argument.
1684 (make_pass_lower_switch): New function.
1685 (make_pass_lower_switch_O0): New function.
1686 (do_jump_if_equal): Remove.
1687 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1688 all the hard work to tree optimization passes.
1690 2018-05-18 Martin Liska <mliska@suse.cz>
1692 * dbgcnt.c (limit_low): Renamed from limit.
1693 (limit_high): New variable.
1694 (dbg_cnt_is_enabled): Check for upper limit.
1695 (dbg_cnt): Adjust dumping.
1696 (dbg_cnt_set_limit_by_index): Add new argument for high
1698 (dbg_cnt_set_limit_by_name): Likewise.
1699 (dbg_cnt_process_single_pair): Parse new format.
1700 (dbg_cnt_process_opt): Use strtok.
1701 (dbg_cnt_list_all_counters): Remove 'value' and add
1703 * doc/invoke.texi: Document changes.
1705 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1707 * doc/sourcebuild.texi (scalar_all_fma): Document.
1708 * tree.def (FMA_EXPR): Delete.
1709 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1710 * internal-fn.c (ternary_direct): New macro.
1711 (expand_ternary_optab_fn): Likewise.
1712 (direct_ternary_optab_supported_p): Likewise.
1713 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1714 * builtins.c (fold_builtin_fma): Delete.
1715 (fold_builtin_3): Don't call it.
1716 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1717 * expr.c (expand_expr_real_2): Likewise.
1718 * fold-const.c (operand_equal_p): Likewise.
1719 (fold_ternary_loc): Likewise.
1720 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1721 * gimple.c (DEFTREECODE): Likewise.
1722 * gimplify.c (gimplify_expr): Likewise.
1723 * optabs-tree.c (optab_for_tree_code): Likewise.
1724 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1725 * tree-eh.c (operation_could_trap_p): Likewise.
1726 (stmt_could_throw_1_p): Likewise.
1727 * tree-inline.c (estimate_operator_cost): Likewise.
1728 * tree-pretty-print.c (dump_generic_node): Likewise.
1729 (op_code_prio): Likewise.
1730 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1731 * tree-ssa-operands.c (get_expr_operands): Likewise.
1732 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1733 * fold-const-call.h (fold_fma): Delete.
1734 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1735 CFN_FNMA and CFN_FNMS.
1737 * genmatch.c (combined_fn): New enum.
1738 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1739 (commutative_op): New function.
1740 (commutate): Use it. Handle more than 2 operands.
1741 (dt_operand::gen_gimple_expr): Use commutative_op.
1742 (parser::parse_expr): Allow :c to be used with non-binary
1743 operators if the commutative operand is known.
1744 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1745 CFN_FMS, CFN_FNMA and CFN_FNMS.
1746 (backprop::process_assign_use): Remove FMA_EXPR handling.
1747 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1748 (gen_hsa_fma): New function.
1749 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1750 IFN_FNMA and IFN_FNMS.
1751 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1752 * gimple-fold.h (follow_all_ssa_edges): Declare.
1753 * gimple-fold.c (follow_all_ssa_edges): New function.
1754 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1755 gimple_build interface and use follow_all_ssa_edges to fold the result.
1756 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1757 instead of checking for optabs directly.
1758 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1759 rather than FMA_EXPRs.
1760 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1761 call to IFN_FMA instead of an FMA_EXPR.
1763 2018-05-17 Jim Wilson <jimw@sifive.com>
1765 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1766 sign extended promoted subreg, and the range does not have the sign bit
1767 set, then do a sign extend.
1769 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1770 test, check for sign extended subreg and/or constant operands, and
1771 do a sign extend in that case.
1773 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1775 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1776 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1778 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1779 Change logics_shift_reg to logics_shift_imm.
1780 (thunderx2t99_fp_loadpair_basic): Delete.
1781 (thunderx2t99_fp_storepair_basic): Delete.
1782 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1783 (thunderx2t99_asimd_polynomial): Delete.
1784 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1785 and neon_fp_mul_d_scalar_q.
1786 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1787 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1788 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1789 (thunderx2t99_asimd_lut): Add missing tbl types.
1790 (thunderx2t99_asimd_ext): Delete.
1791 (thunderx2t99_asimd_load1_1_mult): Delete.
1792 (thunderx2t99_asimd_load1_2_mult): Delete.
1793 (thunderx2t99_asimd_load1_ldp): New.
1794 (thunderx2t99_asimd_load1): New.
1795 (thunderx2t99_asimd_load2): Add missing *load2* types.
1796 (thunderx2t99_asimd_load3): New.
1797 (thunderx2t99_asimd_load4): New.
1798 (thunderx2t99_asimd_store1_1_mult): Delete.
1799 (thunderx2t99_asimd_store1_2_mult): Delete.
1800 (thunderx2t99_asimd_store2_mult): Delete.
1801 (thunderx2t99_asimd_store2_onelane): Delete.
1802 (thunderx2t99_asimd_store_stp): New.
1803 (thunderx2t99_asimd_store1): New.
1804 (thunderx2t99_asimd_store2): New.
1805 (thunderx2t99_asimd_store3): New.
1806 (thunderx2t99_asimd_store4): New.
1808 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1810 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1811 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1813 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1814 Segher Boessenkool <segher@kernel.crashing.org>
1817 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1820 2018-05-17 Richard Biener <rguenther@suse.de>
1822 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1823 for pruning loop and prune defs feeding only already visited PHIs.
1825 2018-05-17 Richard Biener <rguenther@suse.de>
1827 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1829 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1830 Richard Biener <rguenther@suse.de>
1832 PR tree-optimization/85793
1833 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1834 for VMAT_ELEMENTWISE.
1836 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1838 * internal-fn.h (lookup_internal_fn): Declare
1839 * internal-fn.c (lookup_internal_fn): New function.
1840 * gimple.c (gimple_build_call_from_tree): Handle calls to
1842 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1843 internal function names.
1844 * tree-pretty-print.c (dump_generic_node): Likewise.
1845 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1847 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1849 * gimple-fold.h (gimple_build): Make the function forms take
1850 combined_fn rather than built_in_function.
1851 (gimple_simplify): Likewise.
1852 * gimple-match-head.c (gimple_simplify): Likewise.
1853 * gimple-fold.c (gimple_build): Likewise.
1854 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1855 rather than gimple_build_call_internal.
1856 (get_initial_defs_for_reduction): Likewise.
1857 (vect_create_epilog_for_reduction): Likewise.
1858 (vectorizable_live_operation): Likewise.
1860 2018-05-17 Martin Liska <mliska@suse.cz>
1862 * gimple-ssa-sprintf.c (format_directive): Do not use
1863 space in between 'G_' and '('.
1865 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1868 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1869 even if the mask is not all ones.
1872 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1874 (ix86_gimple_fold_builtin): Likewise.
1877 * config/i386/i386.c: Include tree-vector-builder.h.
1878 (ix86_vector_shift_count): New function.
1879 (ix86_fold_builtin): Fold shift builtins by scalar count.
1880 (ix86_gimple_fold_builtin): Likewise.
1882 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1883 _mm512_setzero): New intrinsics.
1885 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1886 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1888 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1889 code generation for cases where splatting a value is not useful.
1890 * simplify-rtx.c (simplify_ternary_operation): Simplify
1891 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1892 mode to a vec_concat.
1894 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1896 * config.gcc: Support "goldmont-plus".
1897 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1899 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1900 PROCESSOR_GOLDMONT_PLUS.
1901 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1902 (processor_target_table): Add "goldmont-plus".
1903 (PTA_GOLDMONT_PLUS): Define.
1904 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1905 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1906 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1907 (fold_builtin_cpu): Add "goldmont-plus".
1908 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1909 (ix86_option_override_internal): Add "goldmont-plus".
1910 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1911 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1912 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1913 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1915 2018-05-17 Richard Biener <rguenther@suse.de>
1917 PR tree-optimization/85757
1918 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1919 remove defs that only feed that PHI from further processing.
1921 2018-05-16 Jim Wilson <jimw@sifive.com>
1923 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1925 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1927 2018-05-16 Mark Wielaard <mark@klomp.org>
1929 * dwarf2out.c (count_index_strings): New function.
1930 (output_indirect_strings): Call count_index_strings and generate
1931 header for dwarf_version >= 5.
1933 2018-05-16 Mark Wielaard <mark@klomp.org>
1935 * dwarf2out.c (dwarf_FORM): New function.
1936 (set_indirect_string): Use dwarf_FORM.
1937 (reset_indirect_string): Likewise.
1938 (size_of_die): Likewise.
1939 (value_format): Likewise.
1940 (output_die): Likewise.
1941 (add_skeleton_AT_string): Likewise.
1942 (output_macinfo_op): Likewise.
1943 (index_string): Likewise.
1944 (output_index_string_offset): Likewise.
1945 (output_index_string): Likewise.
1946 (count_index_strings): Likewise.
1948 2018-05-16 Carl Love <cel@us.ibm.com>
1950 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1951 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1953 2018-05-16 Martin Jambor <mjambor@suse.cz>
1955 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1956 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1958 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1960 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1961 (fnma<mode>4): Likewise.
1962 (fms<mode>4): Likewise.
1963 (fnms<mode>4): Likewise.
1964 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1965 (aarch64_fnma<mode>4): Likewise.
1966 (aarch64_fms<mode>4): Likewise.
1967 (aarch64_fnms<mode>4): Likewise.
1968 (aarch64_fnmadd<mode>4): Likewise.
1970 2018-05-16 Jason Merrill <jason@redhat.com>
1972 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1974 2018-05-16 Richard Biener <rguenther@suse.de>
1976 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1977 (dump_stmt_cost): Declare.
1978 (add_stmt_cost): Dump cost we add.
1979 (add_stmt_costs): New function.
1980 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1982 (vect_analyze_stmt): Adjust prototype.
1983 (vectorizable_condition): Likewise.
1984 (vectorizable_live_operation): Likewise.
1985 (vectorizable_reduction): Likewise.
1986 (vectorizable_induction): Likewise.
1987 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1988 cost vector to pass to vectorizable_ and record afterwards.
1989 (vect_model_reduction_cost): Take cost vector argument and adjust.
1990 (vect_model_induction_cost): Likewise.
1991 (vectorizable_reduction): Likewise.
1992 (vectorizable_induction): Likewise.
1993 (vectorizable_live_operation): Likewise.
1994 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1995 SLP_TREE_NUMBER_OF_VEC_STMTS.
1996 (vect_analyze_slp_cost_1): Remove.
1997 (vect_analyze_slp_cost): Likewise.
1998 (vect_slp_analyze_node_operations): Take visited args and
1999 a target cost vector. Avoid processing already visited stmt sets.
2000 (vect_slp_analyze_operations): Use a local cost vector to gather
2001 costs and register those of non-discarded instances.
2002 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2003 (vect_schedule_slp_instance): Remove copying of
2004 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2006 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2007 adding cost. Record cost entry location.
2008 (vect_prologue_cost_for_slp_op): Function to compute cost of
2009 a constant or invariant generated for SLP vect in the prologue,
2010 split out from vect_analyze_slp_cost_1.
2011 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2012 (vect_model_promotion_demotion_cost): Likewise.
2013 (vect_model_store_cost): Likewise, make static.
2014 (vect_model_load_cost): Likewise.
2015 (vectorizable_bswap): Add cost vector arg and adjust.
2016 (vectorizable_call): Likewise.
2017 (vectorizable_simd_clone_call): Likewise.
2018 (vectorizable_conversion): Likewise.
2019 (vectorizable_assignment): Likewise.
2020 (vectorizable_shift): Likewise.
2021 (vectorizable_operation): Likewise.
2022 (vectorizable_store): Likewise.
2023 (vectorizable_load): Likewise.
2024 (vectorizable_condition): Likewise.
2025 (vectorizable_comparison): Likewise.
2026 (can_vectorize_live_stmts): Likewise.
2027 (vect_analyze_stmt): Likewise.
2028 (vect_transform_stmt): Adjust calls to vectorizable_*.
2029 * tree-vectorizer.c: Include gimple-pretty-print.h.
2030 (dump_stmt_cost): New function.
2032 2018-05-16 Richard Biener <rguenther@suse.de>
2034 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2035 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2036 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2037 (check_name): New callback.
2038 (dse_classify_store): Track cycles via a visited bitmap of PHI
2039 defs and simplify handling of in-loop and across loop dead stores
2040 and properly fail for loop-variant refs. Handle byte-tracking with
2041 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2044 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2046 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2047 (vect_get_mask_type_for_stmt): Likewise.
2048 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2050 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2051 to determine the statement's vector type and the vector type that
2052 should be used for calculating nunits. Deal with cases in which
2053 the type has to be deferred.
2054 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2055 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2056 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2057 (vect_determine_vf_for_stmt): New functions, split out from...
2058 (vect_determine_vectorization_factor): ...here.
2059 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2060 (vect_get_mask_type_for_stmt): New functions, split out from
2061 vect_determine_vectorization_factor.
2063 2018-05-16 Richard Biener <rguenther@suse.de>
2065 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2066 verify the [VEC_]COND_EXPR embedded comparison.
2068 2018-05-15 Martin Sebor <msebor@redhat.com>
2070 PR tree-optimization/85753
2071 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2072 RECORD_TYPE in addition to ARRAY_TYPE.
2074 2018-05-15 Martin Sebor <msebor@redhat.com>
2077 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2079 2018-05-15 Richard Biener <rguenther@suse.de>
2081 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2082 add by_clobber_p one. Change algorithm to collect all defs
2083 representing uses we need to walk and try reducing them to
2084 a single one before failing.
2085 (dse_dom_walker::dse_optimize_stmt): Adjust.
2087 2018-05-13 Mark Wielaard <mark@klomp.org>
2089 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2090 (size_of_loc_descr): Likewise.
2091 (output_loc_operands): Likewise.
2092 (output_loc_operands_raw): Likewise.
2093 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2094 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2095 (hash_loc_operands): Likewise.
2096 (compare_loc_operands): Likewise.
2098 2018-05-14 Mark Wielaard <mark@klomp.org>
2100 * dwarf2out.c (count_index_addrs): New function.
2101 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2103 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2105 PR tree-optimization/83648
2106 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2107 return value as malloc candidate.
2109 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2112 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2113 as true in call to suggest_attribute.
2115 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2117 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2118 -mreadonly-in-sdata.
2120 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2122 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2124 (aarch64_crypto_aesd_fused): Likewise.
2126 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2128 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2129 (movsi_aarch64): Likewise.
2130 (load_pairsi): Likewise.
2131 (load_pairdi): Likewise.
2132 (store_pairsi): Likewise.
2133 (store_pairdi): Likewise.
2134 (load_pairsf): Likewise.
2135 (load_pairdf): Likewise.
2136 (store_pairsf): Likewise.
2137 (store_pairdf): Likewise.
2138 (zero_extend): Likewise.
2139 (trunc): Swap alternatives.
2140 (fcvt_target): Add '?' to prefer w over r.
2142 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2145 * config/i386/i386.md: Disallow non-commutative arithmetics in
2146 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2147 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2148 in the peephole2 before it.
2150 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2152 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2153 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2154 (ix86_handle_option): Handle -mcldemote.
2155 * config.gcc: New header.
2156 * config/i386/cldemoteintrin.h: New file.
2157 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2158 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2160 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2161 OPTION_MASK_ISA_CLDEMOTE.
2162 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2163 (ix86_valid_target_attribute_inner_p): Ditto.
2164 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2165 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2166 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2167 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2168 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2170 * config/i386/i386.opt: Add -mcldemote.
2171 * config/i386/x86intrin.h: New header.
2172 * doc/invoke.texi: Add -mcldemote.
2174 2018-05-14 Richard Biener <rguenther@suse.de>
2176 * doc/match-and-simplify.texi: Adjust :s documentation.
2178 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2180 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2181 intended memcpy size.
2182 (REORDER_45): Likewise.
2184 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2186 * sort.cc: New file.
2187 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2188 * vec.c (qsort_chk): Use gcc_qsort.
2189 * Makefile.in (OBJS-libcommon): Add sort.o.
2190 (build/sort.o): New target. Use it...
2191 (BUILD_RTL): ... here, and...
2192 (build/gencfn-macros): ... here, and...
2193 (build/genmatch): ... here.
2195 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2196 Chung-Ju Wu <jasonwucj@gmail.com>
2198 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2199 * config/nds32/nds32-graywolf.md: New file.
2200 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2201 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2203 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2204 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2205 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2206 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2207 * config/nds32/pipelines.md: Include n15 settings.
2209 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2210 Chung-Ju Wu <jasonwucj@gmail.com>
2212 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2213 * config/nds32/nds32-n13.md: New file.
2214 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2215 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2217 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2218 * config/nds32/nds32.md (pipeline_model): Add n13.
2219 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2220 * config/nds32/pipelines.md: Include n13 settings.
2222 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2223 Chung-Ju Wu <jasonwucj@gmail.com>
2225 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2226 * config/nds32/nds32-n10.md: New file.
2227 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2228 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2230 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2231 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2232 * config/nds32/nds32.md (pipeline_model): Add n10.
2233 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2234 * config/nds32/pipelines.md: Include n10 settings.
2236 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2237 Kito Cheng <kito.cheng@gmail.com>
2238 Chung-Ju Wu <jasonwucj@gmail.com>
2240 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2241 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2242 Add enum values for DSP extension instructions.
2243 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2245 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2246 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2248 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2249 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2250 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2251 * config/nds32/nds32-intrinsic.md: Likewise.
2252 * config/nds32/nds32_intrinsic.h: Likewise.
2253 * config/nds32/nds32-md-auxiliary.c: Likewise.
2254 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2255 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2256 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2257 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2258 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2259 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2260 TYPE_DMAC in switch statement.
2261 * config/nds32/nds32.c: New checking and implementation for DSP
2262 extension instructions.
2263 * config/nds32/nds32.h: Likewise.
2264 * config/nds32/nds32.md: Likewise.
2265 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2266 * config/nds32/predicates.md: Implement new predicates for DSP
2269 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2271 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2272 Reformat alternatives and attributes so it is easier to identify
2273 which constraints/attributes go with which instruction.
2274 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2275 (mov<mode>_softfloat32, FMOVE64): Likewise.
2276 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2277 (mov<mode>_softfloat64, FMOVE64): Likewise.
2279 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2281 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2283 (Basic PowerPC Built-in Functions): The new name of the
2284 subsection previously known as "PowerPC Built-in Functions".
2285 (Basic PowerPC Built-in Functions Available on all Configurations):
2287 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2288 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2289 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2290 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2292 2018-05-11 Martin Jambor <mjambor@suse.cz>
2295 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2298 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2301 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2303 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2305 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2306 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2307 (ix86_handle_option): Handle -mwaitpkg.
2308 * config.gcc: New header.
2309 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2310 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2311 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2313 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2314 OPTION_MASK_ISA_WAITPKG.
2315 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2316 (ix86_option_override_internal): Add PTA_WAITPKG.
2317 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2318 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2319 IX86_BUILTIN_TPAUSE.
2320 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2321 __builtin_ia32_umwait and __builtin_ia32_tpause.
2322 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2323 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2324 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2325 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2326 UNSPECV_TPAUSE): New.
2327 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2328 * config/i386/i386.opt: Add -mwaitpkg.
2329 * config/i386/waitpkgintrin.h: New file.
2330 * config/i386/x86intrin.h: New header.
2331 * doc/invoke.texi: Add -mwaitpkg.
2333 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2336 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2338 (cortex-m0): Use armv6s-m isa.
2339 (cortex-m0plus): Likewise.
2340 (cortex-m1): Likewise.
2341 (cortex-m0.small-multiply): Likewise.
2342 (cortex-m0plus.small-multiply): Likewise.
2343 (cortex-m1.small-multiply): Likewise.
2345 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2346 Jakub Jelinek <jakub@redhat.com>
2348 PR tree-optimization/85692
2349 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2350 source permute as well.
2352 2018-05-11 Martin Liska <mliska@suse.cz>
2355 * doc/extend.texi: Document LLVM style format for no_sanitize
2358 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2360 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2361 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2362 (mode_supports_vsx_dform_quad): Likewise.
2363 (mode_supports_vmx_dform): Move these functions to be next to the
2364 other mode_supports functions.
2365 (mode_supports_dq_form): Likewise.
2366 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2367 mode_supports_dq_form.
2368 (reg_offset_addressing_ok_p): Likewise.
2369 (offsettable_ok_by_alignment): Likewise.
2370 (rs6000_legitimate_offset_address_p): Likewise.
2371 (legitimate_lo_sum_address_p): Likewise.
2372 (rs6000_legitimize_address): Likewise.
2373 (rs6000_legitimize_reload_address): Likewise.
2374 (rs6000_secondary_reload_inner): Likewise.
2375 (rs6000_preferred_reload_class): Likewise.
2376 (rs6000_output_move_128bit): Likewise.
2378 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2380 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2381 Generate SImode target register for null target.
2382 <case IX86_BUILTIN_XGETBV>: Ditto.
2383 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2384 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2386 2018-05-10 Carl Love <cel@us.ibm.com>
2388 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2389 dcbtt and dcbtstt if operands[2] is 0.
2391 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2394 * config/i386/sse.md (usadv64qi): New expander.
2396 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2398 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2399 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2400 -maltivec=be support.
2401 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2402 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2403 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2404 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2405 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2406 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2407 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2408 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2409 altivec_vsumsws): Adjust.
2410 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2411 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2412 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2414 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2415 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2416 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2417 (altivec_lve<VI_char>x): Delete expand.
2418 (*altivec_lve<VI_char>x_internal): Rename to...
2419 (altivec_lve<VI_char>x): ... this.
2420 (altivec_lvxl_<mode>): Delete expand.
2421 (*altivec_lvxl_<mode>_internal): Rename to ...
2422 (altivec_lvxl_<mode>): ... this.
2423 (altivec_stvxl_<mode>): Delete expand.
2424 (*altivec_stvxl_<mode>_internal): Rename to ...
2425 (altivec_stvxl_<mode>): ... this.
2426 (altivec_stve<VI_char>x): Delete expand.
2427 (*altivec_stve<VI_char>x_internal): Rename to ...
2428 (altivec_stve<VI_char>x): ... this.
2429 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2430 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2431 reduc_plus_scal_<mode>): Adjust.
2432 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2434 (rs6000_cpu_cpp_builtins): Adjust.
2435 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2436 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2437 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2438 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2439 -maltivec=be support.
2440 (rs6000_split_vec_extract_var): Adjust.
2441 (rs6000_split_v4si_init): Adjust.
2442 (swap_selector_for_mode): Delete.
2443 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2444 altivec_expand_stvex_be): Delete.
2445 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2446 -maltivec=be support.
2447 (rs6000_gimple_fold_builtin): Ditto.
2448 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2450 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2451 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2452 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2453 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2454 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2455 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2456 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2457 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2458 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2459 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2460 anonymous split): Adjust.
2461 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2462 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2464 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2466 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2467 when --with-gxx-include-dir is also specified.
2468 * configure: Regenerate.
2470 2018-05-09 Jim Wilson <jimw@sifive.com>
2473 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2474 * config/riscv/t-withmultilib: New.
2475 * config/riscv/withmultilib.h: New.
2476 * doc/install.texi: Document RISC-V --with-multilib-list support.
2478 2018-05-09 Richard Biener <rguenther@suse.de>
2480 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2482 (vect_bb_vectorization_profitable_p): Adjust. Compute
2483 actual scalar cost using the cost vector and the add_stmt_cost
2486 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2488 PR rtl-optimization/85645
2489 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2490 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2492 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2494 PR rtl-optimization/85645
2495 * shrink-wrap.c (spread_components): Return a boolean saying if
2496 anything was changed.
2497 (try_shrink_wrapping_separate): Iterate spread_components until
2498 nothing changes anymore.
2500 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2502 PR rtl-optimization/85645
2503 * regrename.c (build_def_use): Also kill the chains that include the
2504 destination of a REG_CFA_REGISTER note.
2506 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2508 PR rtl-optimization/85645
2509 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2510 insn that has a REG_CFA_REGISTER note.
2512 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2514 * cfgexpand.c (expand_clobber): New function.
2515 (expand_gimple_stmt_1): Use it.
2516 * tree-vect-stmts.c (vect_clobber_variable): New function,
2518 (vectorizable_simd_clone_call): ...here.
2519 (vectorizable_store): Emit a clobber either side of an
2520 IFN_STORE_LANES sequence.
2521 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2523 2018-05-09 Tom de Vries <tom@codesourcery.com>
2526 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2527 (define_insn "trap_if_false"): Add exit after trap.
2529 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2531 PR rtl-optimization/85638
2532 * bb-reorder.c: Include common/common-target.h.
2533 (create_forwarder_block): New function extracted from...
2534 (fix_up_crossing_landing_pad): ...here. Rename into...
2535 (dw2_fix_up_crossing_landing_pad): ...this.
2536 (sjlj_fix_up_crossing_landing_pad): New function.
2537 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2538 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2539 from both partitions and exit the loop after one iteration.
2541 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2544 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2546 (Basic PowerPC Built-in Functions): The new name of the
2547 subsection previously known as "PowerPC Built-in Functions".
2548 (Basic PowerPC Built-in Functions Available on all Configurations):
2550 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2552 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2554 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2556 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2559 2018-05-08 Jim Wilson <jimw@sifive.com>
2561 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2562 (LD_EMUL_SUFFIX): New.
2563 (LINK_SPEC): Use it.
2565 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2567 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2569 (Basic PowerPC Built-in Functions): The new name of the
2570 subsection previously known as "PowerPC Built-in Functions".
2571 (Basic PowerPC Built-in Functions Available on all Configurations):
2573 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2575 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2577 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2579 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2582 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2585 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2586 after cmpelim optimization.
2588 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2590 * config.gcc: Support "goldmont".
2591 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2592 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2594 * config/i386/i386.c (m_GOLDMONT): Define.
2595 (processor_target_table): Add "goldmont".
2596 (PTA_GOLDMONT): Define.
2597 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2598 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2599 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2600 (fold_builtin_cpu): Add "goldmont".
2601 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2602 (ix86_option_override_internal): Add "goldmont".
2603 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2604 (processor_type): Add PROCESSOR_GOLDMONT.
2605 * config/i386/i386.md: Add CPU "glm".
2606 * config/i386/glm.md: New file.
2607 * config/i386/x86-tune.def: Add m_GOLDMONT.
2608 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2610 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2613 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2615 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2616 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2617 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2620 * config/i386/i386.c (ix86_fold_builtin): Handle
2621 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2624 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2625 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2627 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2630 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2631 (check_arch): Likewise.
2632 (check_fpu): Return the result rather than printing it.
2633 (end arch): Fix operator precedence.
2634 (end cpu): Likewise.
2635 (END): Print the result from check_fpu.
2637 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2638 Alan Hayward <alan.hayward@arm.com>
2639 David Sherwood <david.sherwood@arm.com>
2641 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2642 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2643 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2644 (*fcmuo<mode>_and): New patterns.
2646 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2648 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2649 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2650 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2651 (cmp_op, sve_imm_con): New code attributes.
2652 (SVE_COND_INT_CMP, imm_con): Delete.
2653 (cmp_op): Remove above unspecs from int attribute.
2654 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2656 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2657 comparison-specific unspecs.
2658 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2659 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2660 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2661 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2662 (*vec_fcm<cmp_op><mode>): Rename to...
2663 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2664 (*vec_fcmuo<mode>): Rename to...
2665 (*fcmuo<mode>): ...this and adjust likewise.
2666 (*pred_fcm<cmp_op><mode>): New pattern.
2667 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2668 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2670 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2672 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2673 (aarch64_emit_sve_predicated_cond): New function.
2674 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2675 (aarch64_emit_unspec_cond_or): Replace with...
2676 (aarch64_emit_sve_or_conds): ...this new function. Use
2677 aarch64_emit_sve_ptrue_op for the individual comparisons and
2678 aarch64_emit_binop to OR them together.
2679 (aarch64_emit_inverted_unspec_cond): Replace with...
2680 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2681 aarch64_emit_sve_ptrue_op for the comparison and
2682 aarch64_emit_unop to invert the result.
2683 (aarch64_expand_sve_vec_cmp_float): Update after the above
2684 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2686 2018-05-07 Nathan Sidwell <nathan@acm.org>
2688 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2689 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2690 (Backwards Compatibility): Likewise.
2692 2018-05-07 Luis Machado <luis.machado@linaro.org>
2696 2018-05-07 Luis Machado <luis.machado@linaro.org>
2698 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2699 <prefetch_dynamic_strides>: New const bool field.
2700 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2701 prefetch_dynamic_strides.
2702 (exynosm1_prefetch_tune): Likewise.
2703 (thunderxt88_prefetch_tune): Likewise.
2704 (thunderx_prefetch_tune): Likewise.
2705 (thunderx2t99_prefetch_tune): Likewise.
2706 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2707 (aarch64_override_options_internal): Update to set
2708 PARAM_PREFETCH_DYNAMIC_STRIDES.
2709 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2710 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2711 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2712 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2713 prefetch-dynamic-strides setting.
2715 2018-05-07 Luis Machado <luis.machado@linaro.org>
2717 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2718 <minimum_stride>: New const int field.
2719 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2720 minimum_stride field.
2721 (exynosm1_prefetch_tune): Likewise.
2722 (thunderxt88_prefetch_tune): Likewise.
2723 (thunderx_prefetch_tune): Likewise.
2724 (thunderx2t99_prefetch_tune): Likewise.
2725 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2726 (aarch64_override_options_internal): Update to set
2727 PARAM_PREFETCH_MINIMUM_STRIDE.
2728 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2729 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2730 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2731 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2732 stride is constant and is below the minimum stride threshold.
2734 2018-05-07 Luis Machado <luis.machado@linaro.org>
2736 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2739 2018-05-07 Luis Machado <luis.machado@linaro.org>
2741 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2742 <prefetch_dynamic_strides>: New const bool field.
2743 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2744 prefetch_dynamic_strides.
2745 (exynosm1_prefetch_tune): Likewise.
2746 (thunderxt88_prefetch_tune): Likewise.
2747 (thunderx_prefetch_tune): Likewise.
2748 (thunderx2t99_prefetch_tune): Likewise.
2749 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2750 (aarch64_override_options_internal): Update to set
2751 PARAM_PREFETCH_DYNAMIC_STRIDES.
2752 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2753 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2754 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2755 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2756 prefetch-dynamic-strides setting.
2758 2018-05-07 Luis Machado <luis.machado@linaro.org>
2760 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2761 <minimum_stride>: New const int field.
2762 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2763 minimum_stride field.
2764 (exynosm1_prefetch_tune): Likewise.
2765 (thunderxt88_prefetch_tune): Likewise.
2766 (thunderx_prefetch_tune): Likewise.
2767 (thunderx2t99_prefetch_tune): Likewise.
2768 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2769 (aarch64_override_options_internal): Update to set
2770 PARAM_PREFETCH_MINIMUM_STRIDE.
2771 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2772 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2773 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2774 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2775 stride is constant and is below the minimum stride threshold.
2777 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2780 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2781 the type is addressable. Don't force op into register if it has
2784 2018-05-05 Roland McGrath <mcgrathr@google.com>
2787 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2788 any section for which we don't know a specific type it should have,
2789 regardless of name. Previously this was done only for the exact
2790 names ".init_array", ".fini_array", and ".preinit_array".
2791 (default_elf_asm_named_section): Add comment about
2792 relationship with default_section_type_flags and SECTION_NOTYPE.
2793 (get_section): Don't consider it a type conflict if one side has
2794 SECTION_NOTYPE and the other doesn't, as long as neither has the
2795 SECTION_BSS et al used in the default_section_type_flags logic.
2797 2018-05-05 Tom de Vries <tom@codesourcery.com>
2800 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2801 (workaround_barsyncs): New function.
2802 (nvptx_reorg): Use workaround_barsyncs.
2803 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2804 (define_expand "nvptx_membar_cta"): New define_expand.
2805 (define_insn "*nvptx_membar_cta"): New insn.
2807 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2809 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2810 To improve optimization opportunities.
2811 * builtin-types.def: The new needed builtin types for the above.
2813 2018-05-04 Richard Biener <rguenther@suse.de>
2815 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2816 * gimple-ssa-store-merging.c
2817 (imm_store_chain_info::output_merged_store): Remove redundant create,
2818 release split_store vector contents on failure.
2819 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2820 scalar stmt vector on cache hit.
2822 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2824 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2826 * config.gcc (powerpc-xilinx-eabi*): Remove.
2827 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2829 (fusion_addis_mem_combo_load): Ditto.
2830 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2832 (rs6000_cpu_cpp_builtins): Ditto.
2833 * config/rs6000/rs6000-linux.c
2834 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2835 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2836 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2838 (rs6000_setup_reg_addr_masks): Ditto.
2839 (rs6000_init_hard_regno_mode_ok): Ditto.
2840 (rs6000_option_override_internal): Ditto.
2841 (legitimate_lo_sum_address_p): Ditto.
2842 (rs6000_legitimize_address): Ditto.
2843 (rs6000_legitimize_reload_address): Ditto.
2844 (rs6000_legitimate_address_p): Ditto.
2845 (abi_v4_pass_in_fpr): Ditto.
2846 (setup_incoming_varargs): Ditto.
2847 (rs6000_gimplify_va_arg): Ditto.
2848 (rs6000_split_multireg_move): Ditto.
2849 (rs6000_savres_strategy): Ditto.
2850 (rs6000_emit_prologue_components): Ditto.
2851 (rs6000_emit_epilogue_components): Ditto.
2852 (rs6000_emit_prologue): Ditto.
2853 (rs6000_emit_epilogue): Ditto.
2854 (rs6000_elf_file_end): Ditto.
2855 (rs6000_function_value): Ditto.
2856 (rs6000_libcall_value): Ditto.
2857 * config/rs6000/rs6000.h: Ditto.
2858 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2859 (TARGET_MINMAX): ... this. New.
2860 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2861 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2862 (*movsi_internal1_single): Delete.
2863 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2864 mfpu=, mxilinx-fpu): Delete.
2865 * config/rs6000/singlefp.h: Delete.
2866 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2867 * config/rs6000/t-rs6000: Ditto.
2868 * config/rs6000/t-xilinx: Delete.
2869 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2870 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2871 (VStype_simple): Delete.
2872 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2873 * config/rs6000/xfpu.h: Delete.
2874 * config/rs6000/xfpu.md: Delete.
2875 * config/rs6000/xilinx.h: Delete.
2876 * config/rs6000/xilinx.opt: Delete.
2877 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2878 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2880 2018-05-04 Tom de Vries <tom@codesourcery.com>
2883 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2886 2018-05-04 Richard Biener <rguenther@suse.de>
2889 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2890 (expand_complex_div_wide): Likewise.
2891 (expand_complex_operations_1): Likewise.
2892 (expand_complex_libcall): Preserve EH info of the original stmt.
2893 (tree_lower_complex): Handle removed blocks.
2894 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2895 on complex multiplication and division libcall builtins.
2897 2018-05-04 Richard Biener <rguenther@suse.de>
2900 * fold-const.c (negate_expr_p): Restrict negation of operand
2901 zero of a division to when we know that can happen without
2903 (fold_negate_expr_1): Likewise.
2905 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2908 * real.h (real_nextafter): Declare.
2909 * real.c (real_nextafter): New function.
2910 * fold-const-call.c (fold_const_nextafter): New function.
2911 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2912 CASE_CFN_NEXTTOWARD.
2913 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2914 even when arg1_mode is different from arg0_mode.
2916 2018-05-03 Nathan Sidwell <nathan@acm.org>
2918 * doc/extend.texi (Deprecated Features): Remove
2920 (Backwards Compatibility): Likewise.
2921 * doc/invoke.texi (C++ Language Options): Likewise.
2922 (C++ Dialect Options): Likewise.
2924 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2927 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2928 _mm512_mask_mullox_epi64): New intrinsics.
2930 2018-05-03 Tom de Vries <tom@codesourcery.com>
2933 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2934 dump files): Add offload-tree.
2936 2018-05-03 Richard Biener <rguenther@suse.de>
2938 PR tree-optimization/85615
2939 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2940 to loops not nested in BBs loop father to avoid creating multi-entry
2943 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2945 PR tree-optimization/70291
2946 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2947 arguments. Change return type to tree. Emit libcall as a new
2948 statement rather than replacing existing one when inplace_p is true.
2949 (expand_complex_multiplication_components): New function.
2950 (expand_complex_multiplication): Expand floating-point complex
2951 multiplication using the above.
2952 (expand_complex_division): Rename inner_type parameter to type.
2953 Update expand_complex_libcall call-site.
2954 (expand_complex_operations_1): Update expand_complex_multiplication
2955 and expand_complex_division call-sites.
2957 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2960 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2961 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2962 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2963 the highest significant bit of the shift count mask is clear. In
2964 check whether and[sq]i3 is needed verify that all significant bits
2965 of the shift count other than the highest are set.
2967 2018-05-02 Tom de Vries <tom@codesourcery.com>
2970 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2971 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2972 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2973 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2974 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2975 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2976 __builtin_goacc_parlevel_size.
2978 2018-05-02 Richard Biener <rguenther@suse.de>
2980 PR tree-optimization/85597
2981 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2982 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2985 2018-05-02 Tom de Vries <tom@codesourcery.com>
2988 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2989 dump files): Add ltrans-tree.
2991 2018-05-02 Tom de Vries <tom@codesourcery.com>
2994 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2995 dump files): Add wpa-ipa.
2997 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2999 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3000 powerpc*-*-linux*paired* target.
3001 * config/rs6000/750cl.h: Delete.
3002 * config/rs6000/paired.h: Delete.
3003 * config/rs6000/paired.md: Delete.
3004 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3006 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3007 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3008 comment. Remove paired float support.
3009 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3010 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3012 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3013 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3015 * config/rs6000/rs6000.c: Remove paired float support.
3016 (paired_expand_vector_init, paired_expand_vector_move,
3017 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3018 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3019 paired_expand_builtin, paired_expand_predicate_builtin,
3020 paired_init_builtins): Delete.
3021 * config/rs6000/rs6000.h: Remove paired float support.
3022 * config/rs6000/rs6000.md: Remove paired float support.
3023 (move_from_CR_ov_bit): Delete.
3024 * config/rs6000/rs6000.opt (mpaired): Delete.
3025 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3026 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3028 2018-05-02 Richard Biener <rguenther@suse.de>
3031 * gimplify.c (gimplify_save_expr): When in SSA form allow
3032 SAVE_EXPRs to compute to SSA vars.
3034 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3037 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3038 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3039 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3040 clobber operands[2], instead use a new pseudo. Formatting fixes.
3042 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3044 PR tree-optimization/85586
3045 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3046 exit early for statements in the same group if the accesses are
3049 2018-05-02 Tom de Vries <tom@codesourcery.com>
3052 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3055 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3057 PR tree-optimization/85143
3058 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3060 2018-05-01 Tom de Vries <tom@codesourcery.com>
3063 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3064 not found" error message.
3066 2018-05-01 Tom de Vries <tom@codesourcery.com>
3069 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3070 * vec.c (test_ordered_remove_if): New function.
3071 (vec_c_tests): Call test_ordered_remove_if.
3072 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3073 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3074 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3075 VEC_ORDERED_REMOVE_IF.
3077 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3079 PR tree-optimization/82665
3080 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3081 pointer subtraction where arguments come from a memchr call.
3083 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3085 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3086 --push-state --as-needed and --pop-state instead of --as-needed and
3087 --no-as-needed if ld supports it.
3088 * configure: Regenerated.
3091 * doc/install.texi2html: Replace _002d with - and _002a with * in
3092 generated html files using sed.
3094 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3097 * gcc-rich-location.c (blank_line_before_p): New function.
3098 (use_new_line): New function.
3099 (gcc_rich_location::add_fixit_insert_formatted): New function.
3100 * gcc-rich-location.h
3101 (gcc_rich_location::add_fixit_insert_formatted): New function.
3103 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3105 * selftest.c (assert_streq): Rename "expected" and "actual" to
3106 "val1" and "val2". Extend NULL-handling to cover both inputs
3107 symmetrically, while still requiring both to be non-NULL for a pass.
3108 * selftest.h (assert_streq): Rename "expected" and "actual" to
3110 (ASSERT_EQ): Likewise.
3111 (ASSERT_EQ_AT): Likewise.
3112 (ASSERT_KNOWN_EQ): Likewise.
3113 (ASSERT_KNOWN_EQ_AT): Likewise.
3114 (ASSERT_NE): Likewise.
3115 (ASSERT_MAYBE_NE): Likewise.
3116 (ASSERT_MAYBE_NE_AT): Likewise.
3117 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3118 the assertion to pass.
3119 (ASSERT_STREQ_AT): Likewise.
3121 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3123 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3124 interaction with -pie.
3126 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3128 * selftest.h: Fix alphabetization of per-source-file selftest
3131 2018-04-30 Jason Merrill <jason@redhat.com>
3133 PR c++/61982 - dead stores to destroyed objects.
3134 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3137 2018-04-30 Jason Merrill <jason@redhat.com>
3139 * tree.c (build_clobber): New.
3140 * tree.h: Declare it.
3141 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3143 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3145 * diagnostic-show-locus.c (layout::layout): Update for
3146 location_get_source_line returning a char_span.
3147 (struct char_span): Move to input.h.
3148 (struct correction): Update for fields in char_span becoming
3150 (struct source_line): Update for location_get_source_line
3151 returning a char_span.
3152 (layout::print_line): Likewise.
3153 * edit-context.c (edited_file::print_content): Likewise.
3154 (edited_file::print_diff_hunk): Likewise.
3155 (edited_file::print_run_of_changed_lines): Likewise.
3156 (edited_file::get_num_lines): Likewise.
3157 (edited_line::edited_line): Likewise.
3158 * final.c (asm_show_source): Likewise.
3159 * input.c (location_get_source_line): Convert return type
3160 from const char * to char_span, losing the final "line_len"
3162 (dump_location_info): Update for the above.
3163 (get_substring_ranges_for_loc): Likewise. Use a char_span
3164 when handling the literal within the line.
3165 (test_reading_source_line): Update for location_get_source_line
3166 returning a char_span.
3167 * input.h (class char_span): Move here from
3168 diagnostic-show-locus.c, converting from a struct to a class.
3169 Make data members private.
3170 (char_span::operator bool): New.
3171 (char_span::length): New.
3172 (char_span::get_buffer): New.
3173 (char_span::operator[]): New.
3174 (char_span::subspan): Make const.
3175 (char_span::xstrdup): New.
3176 (location_get_source_line): Convert return type from const char *
3177 to char_span, losing the final "line_size" param.
3179 2018-04-30 Jan Hubicka <jh@suse.cz>
3181 * lto-wrapper.c (ltrans_priorities): New static var.
3182 (cmp_priority): New.
3183 (run_gcc): Read priorities and if doing parallel build order
3184 the Makefile by them.
3186 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3188 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3190 2018-04-30 Richard Biener <rguenther@suse.de>
3192 * tree-cfg.c (verify_address): Remove base argument, add
3193 flag whether to check TREE_ADDRESSABLE and do that.
3194 (verify_expr): Remove.
3195 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3196 (verify_gimple_assign_single): Likewise.
3197 (verify_gimple_switch): Likewise.
3198 (verify_expr_location_1): Dereference tp once. Add (disabled)
3199 piece from verify_expr.
3200 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3202 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3204 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3206 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3208 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3209 (small_data_pattern): Likewise.
3210 (arc_rewrite_small_data): Likewise.
3211 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3212 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3213 (get_symbol_alignment): New function.
3214 (legitimate_small_data_address_p): Likewise.
3215 (legitimate_scaled_address): Update, call
3216 legitimate_small_data_address_p.
3217 (output_sdata): New static variable.
3218 (arc_print_operand): Update how we handle small data operands.
3219 (arc_print_operand_address): Likewise.
3220 (arc_legitimate_address_p): Update, use
3221 legitimate_small_data_address_p.
3222 (arc_rewrite_small_data_p): Remove.
3223 (arc_rewrite_small_data_1): Likewise.
3224 (arc_rewrite_small_data): Likewise.
3225 (small_data_pattern): Likewise.
3226 (compact_sda_memory_operand): Update to use
3227 legitimate_small_data_address_p and get_symbol_alignment.
3228 (prepare_move_operands): Don't rewite sdata pattern.
3229 (prepare_extend_operands): Remove.
3230 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3232 (zero_extendqisi2): Likewise.
3233 (zero_extendhisi2): Likewise.
3234 (extendqihi2): Likewise.
3235 (extendqisi2): Likewise.
3236 (extendhisi2): Likewise.
3240 * config/arc/constraints.md (Usd): Change it to memory constraint.
3242 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3244 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3245 as source of std instructions.
3246 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3247 allow 6-bit constants as source for store instructions.
3248 (movdi_insn): Update instruction pattern to allow 6-bit constants
3249 as source for store instructions.
3251 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3253 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3255 2018-04-30 Nathan Sidwell <nathan@acm.org>
3256 Sandra Loosemore <sandra@codesourcery.com>
3258 * dumpfile.c (dump_open): Allow '-' for stdout.
3259 * doc/invoke.texi (Developer Options): Document dump filename
3260 determination early. Document stdin/stdout selection.
3262 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3264 Microblaze Target: PIC data text relative
3266 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3267 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3269 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3270 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3271 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3272 New addressing mode for data-text relative position indepenedent code.
3273 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3274 'ADDRESS_SYMBOLIC_TXT_REL'.
3275 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3276 (microblaze_legitimate_pic_operand): Exclude function calls from
3277 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3278 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3280 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3281 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3282 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3283 for 'address + offset'.
3284 (microblaze_expand_prologue): Add new function prologue call for
3286 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3287 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3288 table in case of TARGET_PIC_DATA_TEXT_REL.
3289 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3290 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3291 Add new macros 'UNSPEC_TEXT',
3292 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3293 + exclude function calls from 'UNSPEC_PLT' in case of data text
3295 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3296 new target hook for generating address diff vector tables in case of
3298 * doc/tm.texi : Regenerate.
3299 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3300 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3301 of addr diff vector generation.
3302 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3303 target hook definition.
3304 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3305 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3306 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3307 option for data text relative.
3309 2018-04-30 Richard Biener <rguenther@suse.de>
3311 * tree-chrec.h (evolution_function_is_constant_p): Remove
3313 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3315 2018-04-30 Richard Biener <rguenther@suse.de>
3318 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3320 2018-04-30 Richard Biener <rguenther@suse.de>
3322 PR tree-optimization/28364
3323 PR tree-optimization/85275
3324 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3325 copying first exit test.
3327 2018-04-28 Mark Wielaard <mark@klomp.org>
3329 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3331 (dwarf_AT): Handle DW_AT_addr_base.
3332 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3334 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3337 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3338 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3339 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3340 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3342 2018-04-28 Richard Biener <rguenther@suse.de>
3344 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3345 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3346 to reflect use. Only add interesting stmts.
3348 2018-04-27 Martin Jambor <mjambor@suse.cz>
3351 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3352 the jump function allows for passing through aggregate values.
3354 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3356 * input.h (in_system_header_at): Convert from macro to inline
3358 (from_macro_expansion_at): Likewise.
3359 (from_macro_definition_at): Likewise.
3361 2018-04-27 Jeff Law <law@redhat.com>
3363 * config.gcc: Mark tile* targets as deprecated/obsolete.
3365 2018-04-27 Richard Biener <rguenther@suse.de>
3367 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3370 2018-04-27 Richard Biener <rguenther@suse.de>
3372 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3374 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3376 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3377 with Yd constraint. Set "preferred_for_speed" attribute from
3378 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3380 (*movdi_internal): Ditto.
3381 (movti_interunit splitters): Remove
3382 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3383 (movdi_interunit splitters): Ditto.
3384 * config/i386/constraints.md (Ye): Remove.
3385 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3387 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3390 * config/aarch64/constraints.md (Usg): Limit to 31.
3393 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3395 PR tree-optimization/85529
3396 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3397 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3398 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3399 zero extension or masking of the MSB bit.
3400 (optimize_range_tests): Add FIRST_BB argument, pass it through
3401 to optimize_range_tests_var_bound.
3402 (maybe_optimize_range_tests, reassociate_bb): Adjust
3403 optimize_range_tests callers.
3405 2018-04-26 Richard Biener <rguenther@suse.de>
3406 Jakub Jelinek <jakub@redhat.com>
3408 * cgraph.h (symbol_table): Just declare debug method here.
3409 * symtab.c (symbol_table::debug): Define.
3411 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3413 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3415 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3417 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3418 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3419 (*movdi_internal): Substitute Yi and Yj constraint with x
3420 and Ym and Yn constraint with y constraint. Update "isa"
3421 attribute and set "preferred_for_speed" attribute from
3422 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3423 (*movsi_internal): Ditto.
3424 (*movdf_internal): Ditto.
3425 (*movsf_internal): Ditto.
3426 (*zero_extendsidi2): Ditto.
3427 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3428 (sse2_loadld): Ditto.
3429 (*vec_extract<ssevecmodelower>_0): Ditto.
3430 (*vec_extractv4si_0_zext_sse4): Ditto.
3431 (vec_concatv2di): Ditto.
3432 (*vec_dup<mode>): Ditto.
3433 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3434 * config/i386/constraints.md (Yi): Remove.
3439 2018-04-26 Nathan Sidwell <nathan@acm.org>
3441 * dumpfile.c (dump_open): New.
3442 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3443 (dump_finish): Detect stdio/stderr by value not name.
3445 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3447 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3449 2018-04-26 Tom de Vries <tom@codesourcery.com>
3452 * config/nvptx/nvptx.c (verify_neutering_jumps)
3453 (verify_neutering_labels): New function
3454 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3456 2018-04-26 Tom de Vries <tom@codesourcery.com>
3459 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3460 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3463 2018-04-26 Richard Biener <rguenther@suse.de>
3464 Tom de Vries <tom@codesourcery.com>
3467 * lto-streamer-out.c (output_function): Fixup loops if required to match
3468 discovery done in the reader.
3470 2018-04-26 Richard Biener <rguenther@suse.de>
3472 PR tree-optimization/85116
3473 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3474 have a loop exit from the single latch predecessor. Remove
3475 case of header with just condition.
3476 (ch_base::copy_headers): Exclude infinite loops from any
3478 (pass_ch::execute): Record exits.
3480 2018-04-26 Richard Biener <rguenther@suse.de>
3482 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3483 prologue cost vector and pass it to vect_get_load_cost.
3484 (vect_get_peeling_costs_all_drs): Likewise.
3485 (vect_peeling_hash_get_lowest_cost): Likewise.
3486 (vect_enhance_data_refs_alignment): Likewise.
3488 2018-04-26 Richard Biener <rguenther@suse.de>
3491 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3492 checking of integer<->pointer conversions.
3493 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3494 sign-/zero-extending pointer types.
3495 (expand_omp_for_static_chunk): Likewise.
3497 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3498 Jean Lee <xiaoyur347@gmail.com>
3500 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3501 (TARGET_ASAN_SHADOW_OFFSET): Define.
3502 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3503 true for -fsanitize=address.
3505 2018-04-25 Mark Wielaard <mark@klomp.org>
3507 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3510 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3512 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3513 than "alu", remove explicit "memory" and "imm_disp" attributes.
3514 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3517 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3518 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3519 gen_lowpart_no_emit.
3521 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3524 * config/i386/i386.c (ix86_expand_builtin): Change memory
3525 operand to XI, extend p0 to Pmode.
3526 * config/i386/i386.md: Change unspec volatile and operand
3527 1 mode to XI, change operand 0 mode to P.
3529 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3531 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3532 GET_MODE_MASK before any checking.
3533 (nds32_can_use_bset_p): Likewise.
3534 (nds32_can_use_btgl_p): Likewise.
3536 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3538 * config/nds32/nds32-doubleword.md: New define_split pattern for
3539 illegal register number.
3541 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3543 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3545 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3547 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3549 2018-04-25 Richard Biener <rguenther@suse.de>
3551 * lto-streamer.h (LTO_major_version): Bump to 8.
3553 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3555 * BASE-VER: Set to 9.0.0.
3557 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3559 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3560 in __abskf2 and __powikf2.
3562 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3565 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3566 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3567 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3568 Use the above on operand 2. Reindent.
3569 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3571 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3574 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3576 * config/i386/i386.opt (mcet): Removed.
3577 * doc/install.texi: Remove -mcet documentation.
3578 * doc/invoke.texi: Likewise.
3580 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3583 * doc/install.texi: Remove -mcet from bootstrap-cet.
3585 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3588 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3589 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3593 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3594 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3595 containing a CONST_VECTOR.
3597 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3599 * doc/install.texi: Update newlib dependency for nvptx.
3601 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3604 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3605 instead of INTVAL when shifting x left.
3607 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3609 PR tree-optimization/85478
3610 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3611 vect_grouped_store_supported for single element vectors.
3613 2018-04-24 Richard Biener <rguenther@suse.de>
3616 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3617 load cost increase to the case of non-constant step.
3619 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3622 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3623 destination if any_malformed_asm.
3625 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3628 * expr.c (store_field): In the bitfield case, if the value comes from
3629 a function call and is returned in registers by means of a PARALLEL,
3630 do not change the mode of the temporary unless BLKmode and VOIDmode.
3632 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3634 PR rtl-optimization/85423
3635 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3636 dependencies to debug insns when the previous insn is non-debug.
3638 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3640 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3641 enums into a single definition.
3642 (fls): Fix predicates and printing.
3645 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3647 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3648 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3649 and short u6 immediate.
3650 (check_if_valid_sleep_operand): Remove.
3651 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3653 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3655 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3656 flag_always_save_lp condition.
3657 * config/nds32/nds32.opt (malways-save-lp): New option.
3659 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3661 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3662 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3663 * config/nds32/nds32.h
3664 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3665 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3667 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3669 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3670 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3672 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3673 Chung-Ju Wu <jasonwucj@gmail.com>
3675 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3677 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3678 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3680 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3682 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3684 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3686 * config/nds32/nds32-protos.h (nds32_data_alignment,
3687 nds32_local_alignment): Declare.
3688 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3689 nds32_local_alignment): New functions.
3690 (TARGET_CONSTANT_ALIGNMENT): Define.
3691 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3693 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3695 * config/nds32/nds32.c
3696 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3697 (TARGET_MODES_TIEABLE_P): Likewise.
3699 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3701 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3704 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3705 Chung-Ju Wu <jasonwucj@gmail.com>
3707 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3708 for unaligned access.
3709 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3711 * config/nds32/nds32-intrinsic.md: Likewise.
3712 * config/nds32/nds32_intrinsic.h: Likewise.
3713 * config/nds32/nds32.h (nds32_builtins): Likewise.
3714 * config/nds32/nds32.opt (munaligned-access): New option.
3715 * config/nds32/nds32.c (nds32_asm_file_start): Display
3716 flag_unaligned_access status.
3718 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3720 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3721 -mno-relax is present.
3722 * config/riscv/linux.h (LINK_SPEC): Ditto.
3724 2018-04-20 Martin Sebor <msebor@redhat.com>
3727 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3729 (gimple_fold_builtin_stxcpy_chk): Same.
3730 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3732 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3735 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3736 __powikf2 when long double is IEEE 128-bit.
3738 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3740 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3741 step to make sure stack always aligned.
3743 2018-04-20 Carl Love <cel@us.ibm.com>
3746 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3747 size check for arg0.
3749 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3750 Tom de Vries <tom@codesourcery.com>
3753 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3754 Emit insns for calls too.
3755 (nvptx_find_par): Always look for worker-level predecessor insn.
3756 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3758 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3759 (nvptx_process_pars): Propagate frames for calls.
3761 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3764 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3766 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3767 (ix86_handle_option): Don't handle OPT_mibt.
3768 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3770 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3772 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3773 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3774 (ix86_target_macros): Define __CET__ with flag_cf_protection
3775 for -fcf-protection.
3776 * config/i386/i386.c (isa2_opts): Remove -mibt.
3777 * config/i386/i386.h (TARGET_IBT): Removed.
3778 (TARGET_IBT_P): Likewise.
3779 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3780 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3781 * config/i386/i386.opt (mcet): Update help message.
3784 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3785 -mcet as an alias for -mshstk.
3787 2018-04-20 Richard Biener <rguenther@suse.de>
3790 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3791 complexity by forcing a single use of the multiply operand.
3793 2018-04-20 Martin Jambor <mjambor@suse.cz>
3796 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3797 recursion dependency to only apply to non-clones.
3799 2018-04-20 Martin Jambor <mjambor@suse.cz>
3802 * ipa-cp.c (create_specialized_node): Check that clones of
3803 self-recursive edges exist during IPA-CP.
3805 2018-04-19 Toon Moene <toon@moene.org>
3807 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3810 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3812 PR tree-optimization/85467
3813 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3814 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3815 VECTOR_CST element to type.
3817 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3820 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3821 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3822 (builtin_longjmp): Likewise.
3823 (save_stack_nonlocal): New pattern.
3824 (restore_stack_nonlocal): Likewise.
3826 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3829 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3830 Replace ASM_OUTPUT_LABEL with fprintf.
3832 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3835 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3836 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3837 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3838 define __IBT__ and __SHSTK__ for -fcf-protection.
3839 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3841 (ix86_trampoline_init): Likewise.
3842 (x86_output_mi_thunk): Likewise.
3843 (ix86_notrack_prefixed_insn_p): Likewise.
3844 (ix86_option_override_internal): Don't disallow -fcf-protection.
3845 * config/i386/i386.md (rdssp<mode>): Also enable for
3847 (incssp<mode>): Likewise.
3848 (nop_endbr): Likewise.
3849 * config/i386/i386.opt (mcet): Change help message to built-in
3853 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3854 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3855 enable CET built-in functions.
3857 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3859 * common/config/i386/i386-common.c
3860 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3861 OPTION_MASK_ISA_MOVDIRI_UNSET,
3862 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3863 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3864 * config.gcc (movdirintrin.h): New header.
3865 * config/i386/cpuid.h (bit_MOVDIRI,
3866 bit_MOVDIR64B): New bits.
3867 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3869 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3870 (VOID, PVOID, PCVOID)): New function types.
3871 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3872 __builtin_ia32_directstoreu_u64,
3873 __builtin_ia32_movdir64b): New builtins.
3874 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3875 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3877 (ix86_valid_target_attribute_inner_p): Ditto.
3878 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3879 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3880 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3881 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3882 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3883 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3884 (movdiri<mode>, movdir64b_<mode>): New.
3885 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3886 * config/i386/immintrin.h: Include movdirintrin.h.
3887 * config/i386/movdirintrin.h: New file.
3888 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3890 2018-04-19 Richard Biener <rguenther@suse.de>
3893 * cfg.c (clear_bb_flags): When loop state says we have
3894 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3896 2018-04-19 Richard Biener <rguenther@suse.de>
3898 PR tree-optimization/84737
3899 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3900 copying restrict info.
3901 (vect_setup_realignment): Use it.
3902 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3903 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3904 the first DR to all generated stores.
3905 (vectorizable_load): Likewise for loads.
3907 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3909 PR tree-optimization/85446
3910 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3911 the integral and pointer types to have the same precision.
3913 * doc/install.texi: Document --disable-cet being the default and
3916 2018-04-18 Martin Liska <mliska@suse.cz>
3918 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3921 2018-04-18 Martin Liska <mliska@suse.cz>
3924 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3927 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3928 arguments if they are comparable.
3930 2018-04-18 Martin Liska <mliska@suse.cz>
3933 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3936 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3939 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3942 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3943 ENDBR after calling __morestack.
3945 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3948 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3949 by using gcc_base_ver to generate a gcc_driver_version, and use
3950 it when generating GCC_DRIVER_NAME.
3951 * configure: Regenerate.
3953 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3956 * config.gcc: Obsolete powerpc*-*-*spe*.
3958 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3961 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3962 (stabstr_D): Change type of unum from unsigned int to
3963 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3966 2018-04-17 Jim Wilson <jimw@sifive.com>
3969 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3970 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3971 Set arg_pointer_offset after using pretend_args_size.
3973 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3975 PR rtl-optimization/85431
3976 * dse.c (record_store): Ignore zero width stores.
3979 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3980 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3981 __builtin_stack_restore rather than after it.
3982 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3983 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3984 argument instead of virtual_dynamic_stack_rtx.
3986 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3988 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3990 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3991 Add note to error message to explain internal mapping of overloaded
3992 built-in function name to non-overloaded built-in function name.
3993 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3996 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3999 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4000 where the inputs overlap with the output.
4002 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4004 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4005 (=v, v) alternative and explicit "memory" attribute.
4006 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4007 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4009 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4010 "sselog1" type instead of "sselog".
4011 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4012 "sselog". Remove explicit "memory" attribute.
4013 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4014 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4016 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4017 "sselog1" type instead of "sselog". Remove explicit "memory"
4019 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4020 use "sselog1" type instead of "sselog". Remove explicit "memory"
4022 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4023 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4025 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4026 "sselog1" type instead of "sselog". Remove explicit "memory"
4028 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4029 use "sselog1" type instead of "sselog". Remove explicit "memory"
4033 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4036 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4039 2018-04-17 Martin Jambor <mjambor@suse.cz>
4042 * ipa-cp.c (create_specialized_node): Call
4043 expand_all_artificial_thunks if necessary.
4045 2018-04-17 Martin Liska <mliska@suse.cz>
4048 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4049 in message, remote space in between '_G' and '('.
4051 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4054 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4055 avx512f_vmcmp<mode>3<round_saeonly_name>,
4056 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4057 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4058 avx512f_rndscale<mode><round_saeonly_name>,
4059 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4060 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4061 Use %<iptr>2 instead of %2 for -masm=intel.
4062 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4063 avx512f_vcvttss2usi<round_saeonly_name>,
4064 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4066 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4067 avx512f_vcvttsd2usi<round_saeonly_name>,
4068 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4069 Use %q1 instead of %1 for -masm=intel.
4070 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4071 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4072 of %3 for -masm=intel.
4073 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4075 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4077 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4079 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4080 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4082 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4084 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4085 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4086 %g1 and one with %0 and %1.
4087 (avx512er_vmrcp28<mode><round_saeonly_name>,
4088 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4090 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4091 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4092 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4093 of %0 and %{%4%} for -masm=intel.
4094 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4095 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4096 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4097 order of %0 and %{%5%}%{z%} for -masm=intel.
4099 2018-04-17 Jan Hubicka <jh@suse.cz>
4102 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4104 2018-04-17 Martin Liska <mliska@suse.cz>
4107 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4108 for target_clone error message. Make default implementation
4109 clone to be a local declaration.
4110 (separate_attrs): Add new argument and check for an empty
4112 (expand_target_clones): Handle it.
4113 (ipa_target_clone): Make redirection just for target_clones
4116 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4117 Tom de Vries <tom@codesourcery.com>
4120 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4121 tiled basic blocks without omp continue statements.
4123 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4126 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4127 vec_extract expression as having side effects to make sure it gets
4130 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4133 * config/i386/i386.c (get_builtin_code_for_version): Check
4136 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4139 * gcc/config.gcc: Support "skylake".
4140 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4142 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4143 (processor_target_table): Add "skylake".
4144 (ix86_option_override_internal): Add "skylake".
4145 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4146 PROCESSOR_CANNONLAKE.
4147 (get_builtin_code_for_version): Fix priority for
4148 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4149 PROCESSOR_SKYLAKE-AVX512.
4150 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4151 (processor_type): Add PROCESSOR_SKYLAKE.
4153 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4154 Jason Merrill <jason@redhat.com>
4157 * convert.c (convert_to_integer_1): Use direct recursion for
4158 enumeral types and types with a precision less than the number
4159 of bits in their mode.
4161 2018-04-16 Julia Koval <julia.koval@intel.com>
4164 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4165 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4167 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4170 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4171 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4172 and -mno-direct-move.
4174 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4177 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4178 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4179 Ensure negative shifts result in {0}.
4181 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4183 PR rtl-optimization/79916
4184 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4185 regs (if any) to define how to gnerate SD moves when LRA is in
4188 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4190 PR rtl-optimization/85393
4191 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4192 * except.c (expand_dw2_landing_pad_for_region): Make static.
4193 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4194 a label and unconditional jump to old_bb, rather than
4195 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4198 PR rtl-optimization/85376
4199 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4200 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4201 instead of a specific value.
4203 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4204 Bin Cheng <bin.cheng@arm.com>
4206 PR tree-optimization/82965
4207 PR tree-optimization/83991
4208 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4209 by_profile_only parameter.
4210 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4211 information if the loop was predicted to iterate too many times.
4212 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4214 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4217 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4220 2018-04-13 Martin Liska <mliska@suse.cz>
4221 Jakub Jelinek <jakub@redhat.com>
4224 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4225 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4226 * builtins.c (expand_builtin_memory_copy_args): Use
4227 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4228 handle dest_addr == pc_rtx.
4230 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4233 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4234 asked to not generate direct moves.
4235 (fix_trunc<mode>si2_stfiwx): Similar.
4236 (fix_trunc<mode>si2_internal): Similar.
4238 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4241 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4242 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4243 lookup if dest in some wider mode is known to be const0_rtx and
4244 if so, record permanent equivalence for it to be ZERO_EXTEND of
4245 the narrower mode destination.
4247 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4249 * lto-streamer-out.c (output_function): Revert 259346.
4250 * omp-expand.c (expand_oacc_for): Likewise.
4252 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4254 PR rtl-optimization/85354
4255 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4256 * sel-sched.c (sel_global_init): ... here.
4258 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4261 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4262 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4263 mode for PE-COFF targets.
4264 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4265 (i386_pe_asm_lto_end): Likewise.
4266 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4267 (TARGET_ASM_LTO_END): Likewise.
4268 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4269 (i386_pe_asm_lto_start): New function.
4270 (i386_pe_asm_lto_end): Likewise.
4272 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4273 Richard Biener <rguenther@suse.de>
4276 * lto-streamer-out.c (output_function): Fix CFG loop state before
4278 * omp-expand.c (expand_oacc_for): Handle calls to internal
4279 functions like regular functions.
4281 2018-04-12 Richard Biener <rguenther@suse.de>
4284 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4285 for the early LTO debug to properly generate references to it
4286 during DIE emission. Do not re-use that for the skeleton for
4288 (dwarf2out_early_finish): Likewise.
4290 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4293 * config/i386/sse.md
4294 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4295 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4296 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4297 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4298 and output is a reg, avoid creating invalid lowpart subreg, but
4299 instead split into a 512-bit move. Don't split if not AVX512VL,
4300 input is xmm16+ reg and output is a mem.
4301 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4302 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4303 xmm16+ reg and output is a mem.
4305 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4307 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4308 also for flag_dwarf2_cfi_asm.
4310 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4312 PR rtl-optimization/85342
4313 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4314 a bool scalar var inside of the loop instead. Don't try to update
4315 recog_data.operand after failed apply_change_group.
4317 2018-04-12 Tom de Vries <tom@codesourcery.com>
4320 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4321 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4322 array with flexible array member as array without given dimension.
4323 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4324 argument for undefined param to true.
4326 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4329 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4330 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4331 from PowerPC section.
4332 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4333 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4334 help text that is too long.
4335 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4336 help text that is too long.
4337 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4338 help text that is too long.
4340 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4342 * config/alpha/alpha.md (stack_probe_internal): Rename
4343 from "probe_stack". Update all callers.
4345 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4347 PR rtl-optimization/84566
4348 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4349 sched_macro_fuse_insns.
4351 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4354 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4355 (compute_block_dependences): ... from here.
4357 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4359 PR tree-optimization/85331
4360 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4361 from int to HOST_WIDE_INT.
4363 2018-04-11 Martin Jambor <mjambor@suse.cz>
4366 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4367 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4368 not the same as the source val.
4369 (cgraph_edge_brings_value_p): New parameter.
4370 (gather_edges_for_value): Pass destination value to
4371 cgraph_edge_brings_value_p.
4372 (perhaps_add_new_callers): Likewise.
4373 (get_info_about_necessary_edges): Likewise and exclude values brought
4374 only by self-recursive edges.
4375 (create_specialized_node): Redirect only clones of self-calling edges.
4376 (+self_recursive_pass_through_p): New function.
4377 (find_more_scalar_values_for_callers_subset): Use it.
4378 (find_aggregate_values_for_callers_subset): Likewise.
4379 (known_aggs_to_agg_replacement_list): Removed.
4380 (decide_whether_version_node): Re-calculate known constants for all
4381 remaining context clones.
4383 2018-04-11 Richard Biener <rguenther@suse.de>
4386 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4387 from early DWARF output.
4388 (dwarf2out_early_finish): Output line info unconditionally into
4389 early DWARF and add reference to it.
4391 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4394 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4395 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4396 other than V2DFmode using iptr mode attribute.
4397 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4399 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4401 PR rtl-optimization/84659
4402 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4404 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4407 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4409 (output_loc_list): Pass address of a dummy size variable even in the
4410 locview handling loop.
4411 (index_location_lists): Add comment on why skip_loc_list_entry can't
4414 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4417 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4420 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4423 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4424 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4425 and -mstring-compare-inline-limit.
4427 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4430 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4431 for stack clash protection in a register whenever we need it to be in
4434 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4436 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4437 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4439 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4442 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4444 (mlong-double-): Ditto.
4445 * config/rs6000/sysv4.opt (msdata=): Ditto.
4446 (mtls-size=): Ditto.
4448 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4450 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4451 erroneous entries for
4452 "vector int vec_ldl (int, long int *)", and
4453 "vector unsigned int vec_ldl (int, unsigned long int *)".
4454 Add comments and entries for
4455 "vector bool char vec_ldl (int, bool char *)",
4456 "vector bool short vec_ldl (int, bool short *)",
4457 "vector bool int vec_ldl (int, bool int *)",
4458 "vector bool long long vec_ldl (int, bool long long *)",
4459 "vector pixel vec_ldl (int, pixel *)",
4460 "vector long long vec_ldl (int, long long *)",
4461 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4462 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4463 type tree bool_long_long_type_node and correct definition of
4464 bool_V2DI_type_node to make reference to this new type tree.
4465 (rs6000_mangle_type): Replace erroneous reference to
4466 bool_long_type_node with bool_long_long_type_node.
4467 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4468 comments to emphasize sign distinctions for char and int types and
4469 replace RS6000_BTI_bool_long constant with
4470 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4471 use of RS6000_BTI_pixel.
4472 (bool_long_type_node): Remove this macro definition.
4473 (bool_long_long_type_node): New macro definition
4475 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4477 PR rtl-optimization/85300
4478 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4479 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4480 simplify_unary_operation fails.
4482 2018-04-10 Martin Liska <mliska@suse.cz>
4484 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4485 cgraph_edge and ipa_ref.
4487 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4491 * config/i386/sse.md
4492 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4493 computation of the VEC_MERGE selector from mask.
4494 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4495 Fix decoding of the VEC_MERGE selector into mask.
4497 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4499 PR tree-optimization/85286
4500 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4502 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4504 * final.c (final_1): Set insn_last_address as well as
4505 insn_current_address.
4507 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4510 * explow.c (emit_stack_probe): Call validize_mem on memory location
4511 before passing it to gen_probe_stack. Create address operand and
4512 legitimize it for the probe_stack_address case.
4514 2018-04-09 Jan Hubicka <jh@suse.cz>
4517 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4518 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4519 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4521 (free_lang_data): Rebuild type inheritance graph.
4523 2018-04-09 Martin Sebor <msebor@redhat.com>
4525 * invoke.texi (-finline-small-functions): Mention other optimization
4527 (-findirect-inlining, -fpartial-inlining): Same.
4528 (-finline-functions-called-once): Same.
4529 (-freorder-blocks-and-partition): Same.
4531 2018-04-09 Jan Hubicka <jh@suse.cz>
4534 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4535 jumps; choose last target that matches the criteria (i.e.
4536 no partition changes for non-crossing jumps).
4537 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4538 support for redirecting crossing jumps to non-crossing.
4540 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4542 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4543 also for naked functions.
4545 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4547 * config/arc/arc.md (add_shift): New pattern.
4548 (add_shift2): Likewise.
4549 (sub_shift): Likewise.
4550 (sub_shift_cmp0_noout): Likewise.
4551 (compare_si_ashiftsi): Likewise.
4552 (xbfu_cmp0_noout): New combine pattern.
4553 (xbfu_cmp0"): Likewise.
4554 (movsi_set_cc_insn): Place the predicable variant first.
4555 (commutative_binary_cmp0_noout): Remove clobber.
4556 (commutative_binary_cmp0): New pattern.
4557 (noncommutative_binary_cmp0): Likewise.
4558 (noncommutative_binary_cmp0_noout): Likewise.
4559 (noncommutative_binary_comparison_result_used): Removed.
4560 (rsub_cmp0): New pattern.
4561 (rsub_cmp0_noout): Likewise.
4562 (extzvsi): Changed, keep only meaningful variants.
4563 (SQH, SEZ): New iterators.
4564 (SQH_postfix): New mode attribute.
4565 (SEZ_prefix): New code attribute.
4566 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4567 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4568 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4570 (noncommutative_operator): Check the availability of barrel
4573 2018-04-09 Richard Biener <rguenther@suse.de>
4575 PR tree-optimization/85284
4576 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4577 Only use the niter constraining form of simple_iv when the exit
4580 2018-04-09 Tom de Vries <tom@codesourcery.com>
4583 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4584 (define_expand "*memory_barrier"): New define_expand.
4585 (define_insn "memory_barrier"): New insn.
4587 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4589 PR rtl-optimization/80463
4590 PR rtl-optimization/83972
4591 PR rtl-optimization/83480
4593 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4594 correct producer for the insn.
4595 (tidy_control_flow): Fixup seqnos in case of debug insns.
4597 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4599 PR rtl-optimization/83913
4601 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4602 different sched-times when merging exprs.
4604 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4606 PR rtl-optimization/83962
4608 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4609 tidy_fallthru_edge and tidy_control_flow.
4611 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4613 PR rtl-optimization/83530
4615 * sel-sched.c (force_next_insn): New global variable.
4616 (remove_insn_for_debug): When force_next_insn is true, also leave only
4617 next insn in the ready list.
4618 (sel_sched_region): When the region wasn't scheduled, make another pass
4619 over it with force_next_insn set to 1.
4621 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4623 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4625 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4626 for interrupt control.
4627 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4628 functions for interrupt control.
4629 * config/nds32/nds32-intrinsic.md: Likewise.
4630 * config/nds32/nds32_intrinsic.h: Likewise.
4631 * config/nds32/nds32.h (nds32_builtins): Likewise.
4633 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32.c (nds32_init_machine_status,
4636 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4637 strict_aligned_p field.
4638 (nds32_expand_to_rtl_hook): New function.
4639 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4640 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4642 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4643 Chung-Ju Wu <jasonwucj@gmail.com>
4645 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4646 * config/nds32/nds32-n7.md: New file.
4647 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4648 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4650 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4651 * config/nds32/nds32.md (pipeline_model): Add n7.
4652 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4653 * config/nds32/pipelines.md: Include n7 settings.
4655 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4656 Chung-Ju Wu <jasonwucj@gmail.com>
4658 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4659 * config/nds32/nds32-e8.md: New file.
4660 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4661 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4663 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4664 * config/nds32/nds32.md (pipeline_model): Add e8.
4665 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4666 * config/nds32/pipelines.md: Include e8 settings.
4668 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4669 Chung-Ju Wu <jasonwucj@gmail.com>
4671 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4672 * config/nds32/nds32-n8.md: New file.
4673 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4674 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4676 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4677 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4678 * config/nds32/nds32.md (pipeline_model): Add n8.
4679 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4680 * config/nds32/pipelines.md: Include n8 settings.
4682 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4683 Chung-Ju Wu <jasonwucj@gmail.com>
4685 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4686 * config/nds32/nds32-n9-2r1w.md: New file.
4687 * config/nds32/nds32-n9-3r2w.md: New file.
4688 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4689 nds32_register_ports): New or modify for cpu n9.
4690 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4692 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4693 * config/nds32/nds32-utils.c: New file.
4694 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4695 TARGET_MUL_SLOW): Define.
4696 * config/nds32/nds32.md (pipeline_model): New attribute.
4697 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4698 New options that support cpu n9.
4699 * config/nds32/pipelines.md: Include n9 settings.
4700 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4702 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4704 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4705 information if necessary.
4706 (output_cond_branch_compare_zero): Likewise.
4707 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4708 (nds32_target_alignment): Refine for alignment.
4709 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4710 (FUNCTION_BOUNDARY): Modify.
4711 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4713 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4715 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4717 * config/nds32/constants.md (unspec_volatile_element): Add values for
4718 TLB operation and data prefetch.
4719 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4720 functions for TLB operation and data prefetch.
4721 * config/nds32/nds32-intrinsic.md: Likewise.
4722 * config/nds32/nds32_intrinsic.h: Likewise.
4723 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4724 (nds32_print_operand): Likewise.
4725 * config/nds32/nds32.h (nds32_builtins): Likewise.
4727 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4728 Andrew Pinski <pinsika@gcc.gnu.org>
4731 * match.pd: Use constant_boolean_node of correct type instead of
4732 boolean_true_node or boolean_false_node for simplifying
4733 pointer comparisons to zero.
4735 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4737 PR tree-optimization/80021
4738 * tree.c (verify_type_variant): Make error call in verify_variant_match
4739 translatable and remove final full stop.
4741 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4743 * config/nds32/constants.md (unspec_volatile_element): Add
4744 UNSPEC_VOLATILE_EH_RETURN.
4745 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4746 nds32_output_stack_pop): Support dwarf exception handling process.
4747 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4748 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4749 exception handling process.
4750 (nds32_compute_stack_frame): Likewise.
4751 (nds32_return_addr_rtx): Likewise.
4752 (nds32_initial_elimination_offset): Likewise.
4753 (nds32_expand_prologue): Likewise.
4754 (nds32_expand_epilogue): Likewise.
4755 (nds32_dynamic_chain_address): New function.
4756 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4758 (DYNAMIC_CHAIN_ADDRESS): Define.
4759 (EH_RETURN_DATA_REGNO): Define.
4760 (EH_RETURN_STACKADJ_RTX): Define.
4761 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4762 patterns for dwarf exception handling.
4764 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4766 * config/nds32/nds32.h: Clean up obsolete macros.
4768 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4770 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4771 Add enum values for particular instructions.
4772 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4773 particular intrinsic functions.
4774 * config/nds32/nds32-intrinsic.md: Likewise.
4775 * config/nds32/nds32_intrinsic.h: Likewise.
4776 * config/nds32/nds32.h (nds32_builtins): Likewise.
4777 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4778 (btst, ave): New patterns for particular instructions.
4780 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4782 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4783 Add enum values for atomic load/store and memory sync.
4784 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4786 * config/nds32/nds32-intrinsic.md: Likewise.
4787 * config/nds32/nds32_intrinsic.h: Likewise.
4788 * config/nds32/nds32.h (nds32_builtins): Likewise.
4790 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4792 PR tree-optimization/85257
4793 * fold-const.c (native_encode_vector): If not all elts could fit
4794 and off is -1, return 0 rather than offset.
4795 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4796 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4797 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4798 adjust buffer in native_interpret_expr call.
4800 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4802 * config/nds32/constants.md (unspec_volatile_element): Add cache
4803 control enum values.
4804 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4805 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4806 * config/nds32/nds32.c (nds32_cctl_names): New.
4807 (nds32_print_operand): Handle cache control register names.
4808 * config/nds32/nds32.h (nds32_builtins): New enum values.
4809 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4811 * config/nds32/nds32.md (type): Add mmu.
4812 * config/nds32/pipelines.md (simple_insn): Add mmu.
4814 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32.md (type): Remove call.
4817 * config/nds32/pipelines.md (simple_insn): Likewise.
4819 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4821 * config/nds32/constants.md (unspec_volatile_element): Add
4822 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4823 UNSPEC_VOLATILE_FMFCFG.
4824 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4825 description for fmfcfg and fmfcsr.
4826 (bdesc_1arg): Add fmtcsr.
4827 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4828 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4829 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4830 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4831 unspec_fmfcfg): New patterns.
4832 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4833 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4834 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4835 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4836 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4837 __nds32__fmfcfg): Define.
4839 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4841 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4842 intrinsic register names.
4843 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4844 intrinsic register enum values and macros.
4846 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4848 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4849 for load/store addressing form.
4850 (nds32_print_operand_address): Likewise.
4852 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4855 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4856 based on LABEL_REF. Remove useless assertion.
4857 (pic_address_needs_scratch): Fix formatting.
4858 (sparc_legitimize_pic_address): Minor tweaks.
4859 (sparc_delegitimize_address): Adjust assertion accordingly.
4860 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4861 into symbolic_operand.
4862 (movsi_high_pic_label_ref): Likewise.
4863 (movsi_lo_sum_pic_label_ref): Likewise.
4864 (movdi_pic_label_ref): Likewise.
4865 (movdi_high_pic_label_ref): Likewise.
4866 (movdi_lo_sum_pic_label_ref): Likewise.
4868 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4870 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4871 custom LIB_SPEC setup.
4873 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4874 Kito Cheng <kito.cheng@gmail.com>
4876 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4877 * config/riscv/freebsd.h: New.
4879 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4881 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4882 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4885 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4886 Kito Cheng <kito.cheng@gmail.com>
4888 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4889 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4890 * config/nds32/nds32-protos.h (nds32_output_call,
4891 nds32_output_return): Declare.
4892 * config/nds32/nds32.md: Refine all the call and return patterns.
4894 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4897 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4898 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4900 PR rtl-optimization/84872
4901 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4902 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4905 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4907 * expr.c (copy_blkmode_to_reg): Revert 254862.
4908 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4910 2018-04-06 Richard Biener <rguenther@suse.de>
4913 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4914 after seeing a component reference with an adjacent field. Treat
4915 refs to arrays at struct end of external decls similar to
4916 refs to unconstrained commons.
4918 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4921 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4922 look through SAVE_EXPRs with non-side-effects argument. Adjust
4924 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4927 2018-04-06 Richard Biener <rguenther@suse.de>
4930 * alias.c (find_base_term): New wrapper around find_base_term
4931 unwinding CSELIB_VAL_PTR changes.
4932 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4935 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4937 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4939 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4940 constant definitions.
4941 ("nop"): lr 0,0 -> nopr r0
4942 ("nop_lr0", "nop_lr1"): New insn definitions.
4944 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4946 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4947 NDS32_V3PUSH_AVAILABLE_P macro.
4949 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4950 Chung-Ju Wu <jasonwucj@gmail.com>
4952 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4953 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4954 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4955 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4956 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4957 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4958 * config/nds32/constraints.md: New constraints and checking for hard
4959 float configuration.
4960 * config/nds32/iterators.md: New mode iterator and attribute for hard
4961 float configuration.
4962 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4964 * config/nds32/nds32-fpu.md: New file.
4965 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4966 deal with hard float code generation.
4967 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4969 (abi_type, float_reg_number): New enum type.
4970 * config/nds32/nds32-predicates.c: New predicates for hard float.
4971 * config/nds32/nds32-protos.h: Declare functions for hard float.
4972 * config/nds32/nds32.c: Implementation for hard float configuration.
4973 * config/nds32/nds32.h: Definitions for hard float configuration.
4974 * config/nds32/nds32.md: Include hard float machine description and
4975 modify patterns for hard float configuration.
4976 * config/nds32/nds32.opt: New options for hard float configuration.
4977 * config/nds32/predicates.md: New predicates for hard float
4980 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4982 * common/config/nds32/nds32-common.c
4983 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4985 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4988 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4989 CONSTRUCTOR_ELT (ctor, ...)->value.
4991 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4994 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4996 2018-04-05 Tom de Vries <tom@codesourcery.com>
4999 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5002 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5003 Kito Cheng <kito.cheng@gmail.com>
5005 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5006 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5007 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5008 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5010 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5011 Kito Cheng <kito.cheng@gmail.com>
5013 * config/nds32/constraints.md (Ufe): New memory constraint.
5014 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5015 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5016 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5018 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5019 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5021 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5023 * config/nds32/nds32.md: Use optimize_size in the condition for
5024 alu-shift instructions.
5026 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5028 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5030 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5032 * config/nds32/nds32.md (negsi2): Refine pattern.
5034 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5035 Chung-Ju Wu <jasonwucj@gmail.com>
5037 * config/nds32/iterators.md (shift_rotate): New code iterator.
5038 (shift): New code attribute.
5039 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5040 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5041 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5042 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5043 bit-wise operations.
5044 (andsi3, *andsi3): Ditto.
5045 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5046 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5047 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5048 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5049 nds32_ior_operand, nds32_xor_operand): New predicates.
5051 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5053 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5054 (addsi3, subsi3): ... this.
5056 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5058 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5060 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5062 * config/nds32/nds32.md: Adjust indention.
5064 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5066 * config/nds32/nds32.md (feature): New attribute.
5068 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5070 * config/nds32/nds32.md (subtype): New attribute.
5072 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5075 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5076 expansion to perform a bitwise AND of the argument followed by a
5077 boolean negation of the result.
5079 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5081 PR rtl-optimization/84878
5082 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5083 the basic block. Assert the use reference is not artificial and that
5084 it has an associated insn.
5086 2018-04-04 Michael Matz <matz@suse.de>
5088 * builtins.c (compute_objsize): Pass correct operand
5089 to array_at_struct_end_p.
5091 2018-04-04 Richard Biener <rguenther@suse.de>
5094 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5095 from contexts for DINFO_LEVEL_TERSE and below.
5097 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5099 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5100 resiter_operand condition.
5101 * config/nds32/nds32.md (*move<mode>): Ditto.
5103 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5104 Monk Chiang <sh.chiang04@gmail.com>
5106 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5108 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5110 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5112 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5113 Kito Cheng <kito.cheng@gmail.com>
5115 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5116 nds32_cond_code_str, output_cond_branch,
5117 output_cond_branch_compare_zero, nds32_expand_cbranch,
5118 nds32_expand_cstore, nds32_expand_movcc,
5119 nds32_output_cbranchsi4_equality_zero,
5120 nds32_output_cbranchsi4_equality_reg,
5121 nds32_output_cbranchsi4_equality_reg_or_const_int,
5122 nds32_output_cbranchsi4_greater_less_zero: New functions.
5123 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5124 nds32_expand_cstore, nds32_expand_movcc,
5125 nds32_output_cbranchsi4_equality_zero,
5126 nds32_output_cbranchsi4_equality_reg,
5127 nds32_output_cbranchsi4_equality_reg_or_const_int,
5128 nds32_output_cbranchsi4_greater_less_zero): Declare.
5129 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5130 nds32_rimm11s_operand): New predicates.
5131 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5132 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5135 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5137 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5138 * config/nds32/nds32.md: Ditto.
5139 * config/nds32/pipelines.md: Ditto.
5141 2018-04-04 Richard Biener <rguenther@suse.de>
5143 PR tree-optimization/85168
5144 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5145 propagating abnormals.
5147 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5149 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5151 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5152 Kito Cheng <kito.cheng@gmail.com>
5154 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5155 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5156 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5157 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5158 * config/nds32/nds32.md (sibcall_internal): New.
5159 (sibcall_register): Remove.
5160 (sibcall_immediate): Remove.
5161 (sibcall_value_internal): New.
5162 (sibcall_value_register): Remove.
5163 (sibcall_value_immediate): Remove.
5164 * config/nds32/predicates.md (nds32_general_register_operand): New.
5165 (nds32_call_address_operand): New.
5167 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5169 PR rtl-optimization/85167
5170 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5171 bb_defs if *split_p, instead preinitialize it to NULL.
5173 PR tree-optimization/85156
5174 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5175 evaluating the argument multiple times.
5177 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5179 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5181 (_mm_cvtpd_ps): Likewise.
5182 (_mm_cvttpd_epi32): Likewise.
5183 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5184 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5185 vector, pixel, and bool following altivec.h include.
5187 2018-04-03 Martin Sebor <msebor@redhat.com>
5189 * doc/extend.texi (Common Function Attributes): Clarify.
5190 (const attribute): Likewise.
5191 (pure attribute): Likewise.
5193 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5196 * config/i386/i386.c (ix86_expand_vector_set): Use
5197 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5199 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5201 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5202 instructions when changing rounding bits to preserve precision bits
5203 in the x87 control word.
5205 2018-04-03 Martin Liska <mliska@suse.cz>
5207 PR tree-optimization/82491
5208 * rtl.h (strip_offset_and_add): Replace += suboffset with
5209 poly_uint64 () + suboffset.
5211 2018-03-29 Martin Liska <mliska@suse.cz>
5212 Martin Jambor <mjambor@suse.cz>
5215 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5216 param_type is not an integral or pointer type.
5218 2018-04-03 Richard Biener <rguenther@suse.de>
5220 * sese.h (recompute_all_dominators): Remove.
5222 2018-04-02 Martin Sebor <msebor@redhat.com>
5224 * doc/invoke.texi (-Wrestrict): Fix typos.
5226 2018-04-02 Jim Wilson <jimw@sifive.com>
5228 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5229 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5230 (<optab>di3, <optab>si3_extend): Likewise.
5231 (<optab>si3_mask, <optab>si3_mask_1): New.
5232 (<optab>di3_mask, <optab>di3_mask_1): New.
5233 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5234 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5235 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5237 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5239 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5242 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5244 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5245 (nds32_canonicalize_comparison): New function.
5247 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5248 Kito Cheng <kito.cheng@gmail.com>
5249 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5251 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5252 * config/nds32/constants.md (unspec_volatile_element): Add
5253 UNSPEC_VOLATILE_RELAX_GROUP.
5254 * config/nds32/nds32-relax-opt.c: New file.
5255 * config/nds32/nds32-predicates.c
5256 (nds32_symbol_load_store_p): New function.
5257 * config/nds32/nds32-protos.h
5258 (nds32_symbol_load_store_p): Declare function.
5259 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5260 * config/nds32/nds32.c
5261 (nds32_register_pass): New function to register pass.
5262 (nds32_register_passes): New function to register passes.
5263 * config/nds32/nds32.md (relax_group): New pattern.
5264 * config/nds32/nds32.opt (mrelax-hint): New option.
5265 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5267 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5269 * config/nds32/t-nds32: Modify files dependency.
5271 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5273 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5274 (PROFILE_HOOK): Define its implementation.
5276 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5278 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5279 type and 32-bit size.
5281 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5284 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5285 (V_128_256): New mode iterator.
5286 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5287 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5288 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5290 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5291 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5293 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5296 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5297 NaN inputs correctly.
5299 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5302 * config/rs6000/vsx.md (??r): New mode attribute.
5303 (*vsx_mov<mode>_64bit): Use it.
5304 (*vsx_mov<mode>_32bit): Likewise.
5306 2018-03-30 Martin Sebor <msebor@redhat.com>
5308 PR tree-optimization/84818
5309 * builtins.c (check_access): Use warning_n.
5311 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5314 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5316 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5319 2018-03-30 Julia Koval <julia.koval@intel.com>
5322 * x86-tune.def (movx, partial_reg_dependency): Enable for
5325 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5328 * lra-constraints.c (process_alt_operands): Move setting
5329 this_alternative_matches below.
5331 2018-03-29 Martin Liska <mliska@suse.cz>
5334 * doc/invoke.texi: Document how LTO works with debug info.
5335 Describe auto-load support of binutils. Mention 'x86-64'
5336 as valid option value of -march option.
5338 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5340 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5343 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5344 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5345 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5348 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5351 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5352 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5353 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5354 for RS6000_BTM_POWERPC64.
5355 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5356 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5357 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5362 2018-03-28 Carl Love <cel@us.ibm.com>
5365 2017-09-27 Carl Love <cel@us.ibm.com>
5367 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5368 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5369 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5372 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5374 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5375 instead of __vector bool.
5376 (_mm_max_pu8): Likewise.
5377 (_mm_min_pi16): Likewise.
5379 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5382 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5383 (DIVWEUO): Likewise.
5385 (DIVDEUO): Likewise.
5386 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5387 DIVWEUO and DIVDEUO.
5388 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5389 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5390 (div_extend): Likewise.
5391 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5393 (__builtin_divweuo): Likewise.
5394 (__builtin_divdeo): Likewise.
5395 (__builtin_divdeuo): Likewise.
5397 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5400 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5401 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5403 PR tree-optimization/82004
5404 * gimple-match-head.c (optimize_pow_to_exp): New function.
5405 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5406 Don't fold to exp if optimize_pow_to_exp is false.
5408 2018-03-28 Martin Liska <mliska@suse.cz>
5411 * calls.c (initialize_argument_information): Fix trailing space.
5412 * common.opt: Fix typo and provide better explanation for
5413 -fsanitize-coverage option.
5414 * config/i386/i386.opt: Fix typo.
5416 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5417 Martin Liska <mliska@suse.cz>
5420 * gimplify.c (asan_poison_variable): Don't do the check for
5421 gimplify_omp_ctxp here.
5422 (gimplify_decl_expr): Do it here.
5423 (gimplify_target_expr): Likewise.
5425 2018-03-28 Martin Liska <mliska@suse.cz>
5428 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5429 chkp_type_bounds_count if MPX is not enabled.
5431 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5433 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5435 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5438 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5439 function to create the function decl for complex long double
5440 multiply and divide for -mabi=ieeelongdouble.
5441 (init_float128_ieee): Call it.
5443 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5446 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5447 -fcf-protection=branch -mibt.
5448 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5450 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5453 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5455 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5458 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5459 extern array declarations.
5461 2018-03-27 Richard Biener <rguenther@suse.de>
5464 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5465 explicit single_use checks.
5467 2018-03-27 Richard Biener <rguenther@suse.de>
5469 PR tree-optimization/85082
5470 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5473 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5475 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5476 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5477 Turn on fasynchronous-unwind-tables and funwind-tables.
5479 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5482 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5483 (*bmi_blsr_<mode>_ccz): Ditto.
5485 2018-03-26 Tom de Vries <tom@codesourcery.com>
5487 PR tree-optimization/85063
5488 * omp-general.c (offloading_function_p): New function. Factor out
5490 * omp-offload.c (pass_omp_target_link::gate): ... here.
5491 * omp-general.h (offloading_function_p): Declare.
5492 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5493 with attribute omp declare target for offloading functions.
5495 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5497 PR tree-optimization/84005
5498 * tree-data-ref.h (get_base_for_alignment): Declare.
5499 * tree-data-ref.c (get_base_for_alignment_1): New function.
5500 (get_base_for_alignment): Likewise.
5501 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5502 get_base_for_alignment to find a suitable base object, instead
5503 of always using drb->base_address.
5505 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5508 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5509 known size by default.
5511 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5514 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5515 and non BLKmode operands.
5517 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5520 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5521 Clean up attributes.
5523 2018-03-23 Richard Biener <rguenther@suse.de>
5526 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5527 we are going to emit early debug for LTO.
5529 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5532 * function.c (match_asm_constraints_1): Don't optimize if input
5533 doesn't satisfy general_operand predicate for output's mode.
5536 * alias.c (write_dependence_p): Don't require for x_canonicalized
5537 non-VOIDmode if x has VOIDmode.
5540 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5541 just don't try to optimize it rather than assert it never happens.
5543 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5545 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5546 macro expansions for definition of ST_INTERNAL_<mode> and
5547 LD_INTERNAL_<mode> builtins.
5548 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5550 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5552 (altivec_expand_st_builtin): Likewise.
5553 (altivec_expand_builtin): Remove calls to deleted functions.
5554 (rs6000_address_for_altivec): Delete this function.
5555 * config/rs6000/vector.md: Remove expands for
5556 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5558 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5561 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5562 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5563 re-computing once computed.
5564 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5565 (arm_init_machine_status): Initialize
5566 machine->static_chain_stack_bytes.
5568 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5571 * doc/extend.texi: Add four new prototypes for vec_ld.
5572 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5573 definitions for more logical presentation.
5574 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5575 entries for V1TI variants of __builtin_altivec_ld builtin.
5576 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5577 handling of V1TI variant of LVX icode pattern.
5578 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5579 (rs6000_gimple_fold_builtin): Likewise.
5580 (altivec_init_builtins): Add code to define
5581 __builtin_altivec_lvx_v1ti function.
5583 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5586 * function.c (match_asm_constraints_1): Don't do the optimization
5587 if input isn't a REG, SUBREG, MEM or constant.
5589 2018-03-22 Tom de Vries <tom@codesourcery.com>
5591 PR tree-optimization/84956
5592 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5593 bb_has_abnormal_pred.
5595 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5598 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5599 DECL_INITIAL (decl) to decl at the end.
5600 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5603 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5605 * doc/extend.texi (__builtin_tgmath): Document when complex
5606 integer types are treated as _Complex _Float64.
5608 2018-03-21 Tom de Vries <tom@codesourcery.com>
5610 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5612 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5614 PR tree-optimization/84960
5615 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5616 if it is ENTRY block, move them into single succ of ENTRY in that case.
5618 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5620 PR tree-optimization/84811
5621 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5622 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5623 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5624 (known_subrange_p): Update accordingly. Cast each value involved
5625 in the size comparison, rather than casting the result of the
5628 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5630 PR tree-optimization/84982
5631 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5632 by flipping the least significant bit rather than all bits from
5633 bitpos to bitpos + bitsize - 1.
5635 2018-03-21 Nathan Sidwell <nathan@acm.org>
5637 * doc/extend.texi (Deprecated Features): Remove mention of
5638 long-deleted deprecations.
5640 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5643 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5644 * configure: Regenerate.
5646 2018-03-21 Tom de Vries <tom@codesourcery.com>
5648 PR tree-optimization/83126
5649 * tree-parloops.c (num_phis): New function.
5650 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5652 2018-03-21 Nathan Sidwell <nathan@acm.org>
5654 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5655 mention anon-struct/union members and trailing attributes.
5657 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5659 PR tree-optimization/84969
5660 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5661 builtin memset partitions if they set different rhs values.
5663 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5665 PR rtl-optimization/84989
5666 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5667 VEC_DUPLICATE with scalar result mode.
5669 2018-03-21 Martin Liska <mliska@suse.cz>
5672 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5673 not intended return statement.
5675 2018-03-21 Martin Liska <mliska@suse.cz>
5678 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5679 (chkp_find_bound_slots_1): Limit number of iterations.
5681 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5684 * Minor grammar fixes for x86 options.
5686 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5689 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5690 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5692 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5695 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5696 (altivec_lvx_<mode>_1op): Likewise.
5697 (altivec_stvx_<mode>_2op): Likewise.
5698 (altivec_stvx_<mode>_1op): Likewise.
5699 (altivec_lvx_<VM2:mode>): New define_expand.
5700 (altivec_stvx_<VM2:mode>): Likewise.
5701 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5702 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5703 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5704 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5705 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5706 (rs6000_gen_lvx): Likewise.
5707 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5708 (altivec_expand_stv_builtin): Likewise.
5709 (altivec_expand_builtin): Likewise.
5710 * config/rs6000/vector.md: Likewise.
5712 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5715 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5718 2018-03-20 Richard Biener <rguenther@suse.de>
5721 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5722 sign-conversions as zero, fall back to standard scalar_stmt
5725 2018-03-20 Martin Liska <mliska@suse.cz>
5728 * predict.c (rebuild_frequencies): Handle case when we have
5729 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5731 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5734 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5735 flag_section_anchors.
5736 * varasm.c (use_blocks_for_decl_p): Remove hack for
5737 dw2_force_const_mem.
5740 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5742 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5743 be created, use lowpart_subreg of operands[0] rather than operands[0]
5745 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5746 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5747 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5748 and n constraint instead of aarch64_shift_imm_di and Usd.
5749 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5750 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5752 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5755 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5756 to favor GPR over NEON registers.
5757 (<shift>di3_neon): Likewise.
5759 2018-03-20 Tom de Vries <tom@codesourcery.com>
5762 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5763 (nvptx_process_pars): Emit bar.sync asap and alap.
5765 2018-03-20 Tom de Vries <tom@codesourcery.com>
5768 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5769 seen_label if seen_label is already set.
5771 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5774 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5775 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5776 Use 1U instead of 1. Formatting fixes.
5779 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5780 instead of TREE_TYPE (s1) for the return value.
5782 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5784 PR tree-optimization/84946
5785 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5786 bitsize + bitsize in poly_uint64 rather than poly_int64.
5789 * dwarf2asm.c: Include fold-const.c.
5790 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5791 of decl rather than decl itself.
5793 PR rtl-optimization/84643
5794 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5796 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5799 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5800 calling assemble_variable.
5802 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5805 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5806 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5808 2018-03-19 Jim Wilson <jimw@sifive.com>
5811 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5812 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5813 (riscv_first_stack_step): Likewise.
5814 (riscv_option_override): Use STACK_BOUNDARY instead of
5816 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5818 (BIGGEST_ALIGNMENT): Set to 128.
5819 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5820 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5823 2018-03-19 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/84933
5826 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5827 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5829 2018-03-19 Richard Biener <rguenther@suse.de>
5831 PR tree-optimization/84859
5832 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5833 (cond_if_else_store_replacement): Perform sinking operation on
5834 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5835 Generalize what a BB with a single eligible store is.
5837 2018-03-19 Richard Biener <rguenther@suse.de>
5839 PR tree-optimization/84929
5840 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5841 chrec_is_positive against non-chrec arg.
5843 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5846 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5848 2018-03-18 Martin Liska <mliska@suse.cz>
5850 PR rtl-optimization/84635
5851 * regrename.c (build_def_use): Use matches_mode only when
5854 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5856 PR tree-optimization/84913
5857 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5858 vectorize chains of COND_EXPRs.
5860 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5862 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5864 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5866 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5868 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5870 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5872 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5873 Kito Cheng <kito.cheng@gmail.com>
5875 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5876 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5877 (nds32_adjust_reg_alloc_order): New function.
5878 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5880 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5882 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5883 nds32_print_operand, nds32_print_operand_address): Use
5884 HOST_WIDE_INT_PRINT_DEC instead.
5886 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5888 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5890 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5893 * config/i386/i386.c (initial_ix86_tune_features,
5894 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5896 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5897 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5898 rather than 1u << ix86_tune. Formatting fix.
5899 (ix86_option_override_internal): Change ix86_arch_mask from
5900 unsigned int to unsigned HOST_WIDE_INT, initialize to
5901 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5902 (ix86_function_specific_restore): Likewise.
5904 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5907 * postreload.c (reload_combine_recognize_pattern): Perform
5908 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5909 truncate_int_for_mode the result for the destination's mode.
5912 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5913 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5915 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5918 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5921 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5922 regno_allocno_class_array and sorted_pseudos.
5923 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5924 insns where regno is used.
5926 2018-03-16 Martin Liska <mliska@suse.cz>
5929 * multiple_target.c (create_dispatcher_calls): Redirect
5930 reference in the symbol table.
5932 2018-03-16 Martin Liska <mliska@suse.cz>
5935 * multiple_target.c (create_dispatcher_calls): Redirect also
5938 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5942 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5943 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5944 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5946 2018-03-16 Julia Koval <julia.koval@intel.com>
5948 * doc/invoke.texi (Skylake Server): Add CLWB.
5949 Cannonlake): Remove CLWB.
5951 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5953 PR tree-optimization/84841
5954 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5956 (FLOAT_ONE_CONST_TYPE): Define.
5957 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5958 (sort_by_operand_rank): Put entries with higher constant_type last
5959 rather than first to match comments.
5961 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5963 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5966 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5969 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5970 check and use address of target always.
5972 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5975 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5976 (indirect_thunk_bnd_needed): Likewise.
5977 (indirect_thunks_used): Likewise.
5978 (indirect_thunks_bnd_used): Likewise.
5979 (indirect_return_needed): New.
5980 (indirect_return_bnd_needed): Likewise.
5981 (output_indirect_thunk_function): Add a bool argument for
5983 (output_indirect_thunk_function): Don't generate alias for
5984 function return thunk.
5985 (ix86_code_end): Call output_indirect_thunk_function to generate
5986 function return thunks.
5987 (ix86_output_function_return): Set indirect_return_bnd_needed
5988 and indirect_return_needed instead of indirect_thunk_bnd_needed
5989 and indirect_thunk_needed.
5991 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5993 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5994 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5995 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5997 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5998 Paul Hua <paul.hua.gm@gmail.com>
6001 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6003 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6005 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6006 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6009 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6012 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6013 instead of GET_MODE_SIZE when comparing Units.
6015 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6018 * varasm.c (hash_section): Return an unchangeble hash value
6019 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6020 Return !aarch64_can_use_per_function_literal_pools_p ().
6022 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6025 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6026 rather than address of cmode as last argument to prepare_cmp_insn.
6028 2018-03-15 Julia Koval <julia.koval@intel.com>
6030 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6031 F_AVX512VNNI, F_AVX512BITALG): New.
6033 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6036 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6037 insn for floating-point loads and stores.
6039 2018-03-14 Carl Love <cel@us.ibm.com>
6041 * config/rs6000/rs6000-c.c: Add macro definitions for
6042 ALTIVEC_BUILTIN_VEC_PERMXOR.
6043 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6044 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6045 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6046 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6048 * config/doc/extend.texi: Add prototypes for vec_permxor.
6050 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6053 * diagnostic-show-locus.c (class layout_point): Convert m_line
6054 from int to linenum_type.
6055 (line_span::comparator): Use linenum "compare" function when
6056 comparing line numbers.
6057 (test_line_span): New function.
6058 (layout_range::contains_point): Convert param "row" from int to
6060 (layout_range::intersects_line_p): Likewise.
6061 (layout::will_show_line_p): Likewise.
6062 (layout::print_source_line): Likewise.
6063 (layout::should_print_annotation_line_p): Likewise.
6064 (layout::print_annotation_line): Likewise.
6065 (layout::print_leading_fixits): Likewise.
6066 (layout::annotation_line_showed_range_p): Likewise.
6067 (struct line_corrections): Likewise for field m_row.
6068 (line_corrections::line_corrections): Likewise for param "row".
6069 (layout::print_trailing_fixits): Likewise.
6070 (layout::get_state_at_point): Likewise.
6071 (layout::get_x_bound_for_row): Likewise.
6072 (layout::print_line): Likewise.
6073 (diagnostic_show_locus): Likewise for locals "last_line" and
6075 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6076 * input.c (selftest::test_linenum_comparisons): New function.
6077 (selftest::input_c_tests): Call it.
6078 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6079 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6080 * selftest.h (ASSERT_GT): New macro.
6081 (ASSERT_GT_AT): New macro.
6082 (ASSERT_LT): New macro.
6083 (ASSERT_LT_AT): New macro.
6085 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6087 PR rtl-optimization/84780
6088 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6090 2018-03-14 Martin Liska <mliska@suse.cz>
6092 * tree.c (record_node_allocation_statistics): Use
6093 get_stats_node_kind.
6094 (get_stats_node_kind): New function extracted from
6095 record_node_allocation_statistics.
6096 (free_node): Use get_stats_node_kind.
6098 2018-03-14 Richard Biener <rguenther@suse.de>
6100 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6101 that the value-set of ANTIC_IN doesn't grow.
6104 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6106 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6107 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6109 2018-03-14 Julia Koval <julia.koval@intel.com>
6111 * config.gcc (icelake-client, icelake-server): New.
6113 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6114 (initial_ix86_arch_features): Ditto.
6115 (PTA_SKYLAKE): Add SGX.
6116 (PTA_ICELAKE): Remove.
6117 (PTA_ICELAKE_CLIENT): New.
6118 (PTA_ICELAKE_SERVER): New.
6119 (ix86_option_override_internal): Split up icelake on icelake client and
6121 (get_builtin_code_for_version): Ditto.
6122 (fold_builtin_cpu): Ditto.
6123 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6124 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6125 * config/i386/i386.h (processor_type): Ditto.
6126 * doc/invoke.texi: Ditto.
6128 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6131 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6132 INTEGER_CST offset, add it together with bitpos / 8 and
6133 sign extend based on POINTER_SIZE.
6137 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6140 * config/i386/constraints.md (Yc): New register constraint.
6141 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6142 Use Yc constraint for alternative 2 of operand 0. Remove
6143 preferred_for_speed attribute.
6145 2018-03-14 Richard Biener <rguenther@suse.de>
6147 PR tree-optimization/84830
6148 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6149 with the old one to avoid oscillations.
6151 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6154 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6156 (assign_by_spills): Return a flag of reload assignment failure.
6157 Do not process the reload assignment failures. Do not spill other
6158 reload pseudos if they has the same reg class. Update n if
6160 (lra_assign): Add a return arg. Set up from the result of
6161 assign_by_spills call.
6162 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6163 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6164 usage_insns if it is not NULL.
6165 (spill_hard_reg_in_range): New function.
6166 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6167 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6168 function prototypes.
6169 (lra_assign): Change prototype.
6170 * lra.c (lra): Add code to deal with fails by splitting hard reg
6173 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6175 * config/riscv/riscv.opt (mrelax): New option.
6176 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6177 "norelax" when riscv_mrelax is disabled.
6178 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6180 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6183 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6184 reassociation for int modes.
6186 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6188 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6189 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6191 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6192 * config/aarch64/aarch64-sve.md
6193 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6194 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6195 (*extend<mode><Vwide>2): Rename to...
6196 (aarch64_sve_extend<mode><Vwide>2): ...this.
6197 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6198 renaming the old pattern to...
6199 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6201 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6202 define_expand, renaming the old pattern to...
6203 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6204 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6205 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6206 account when deciding which SVE instruction the optab should use.
6207 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6209 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6211 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6212 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6213 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6214 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6215 (tlsdesc_small_<mode>): Turn a define_expand and use
6216 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6217 (tlsdesc_small_advsimd_<mode>): ...this.
6218 (tlsdesc_small_sve_<mode>): New pattern.
6220 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6222 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6223 (UNSPEC_UMUL_HIGHPART): New constants.
6224 (MUL_HIGHPART): New int iteraor.
6225 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6226 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6228 (*<su>mul<mode>3_highpart): New define_insn.
6230 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6233 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6236 2018-03-13 Martin Liska <mliska@suse.cz>
6239 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6241 (sem_item_optimizer::~sem_item_optimizer): Release it.
6242 (sem_item_optimizer::merge_classes): Register variable aliases.
6243 (sem_item_optimizer::fixup_pt_set): New function.
6244 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6245 * ipa-icf.h: Declare new variables and functions.
6247 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6250 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6251 integer_pow2p@2 and test integer_pow2p in condition.
6252 (A < 0 ? C : 0): Similarly for @1.
6255 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6256 characters starting at p contain '\0' character, don't look beyond
6260 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6261 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6264 * reg-stack.c (change_stack): Change update_end var from int to
6265 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6266 also call set_block_for_insn on the newly added insns and rescan.
6269 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6270 on the last operand.
6273 * tree.c (stabilize_reference_1): Return save_expr (e) for
6274 STATEMENT_LIST even if it doesn't have side-effects.
6276 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6278 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6280 2018-03-12 Renlin Li <renlin.li@arm.com>
6282 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6283 aarch64_output_scalar_simd_mov_immediate.
6285 2018-03-12 Martin Sebor <msebor@redhat.com>
6287 PR tree-optimization/83456
6288 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6289 for perfectly overlapping calls to memcpy.
6290 (gimple_fold_builtin_memory_chk): Same.
6291 (gimple_fold_builtin_strcpy): Handle no-warning.
6292 (gimple_fold_builtin_stxcpy_chk): Same.
6293 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6295 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6297 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6298 parameter. Use it for SFmode.
6299 (rs6000_function_arg_advance_1): Adjust.
6300 (rs6000_function_arg): Adjust.
6301 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6303 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6305 PR rtl-optimization/84169
6306 PR rtl-optimization/84780
6307 * combine.c (can_combine_p): Check for a 2-insn combination whether
6308 the destination register is used between the two insns, too.
6310 2018-03-12 Richard Biener <rguenther@suse.de>
6312 PR tree-optimization/84803
6313 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6314 for refs DR analysis didn't process.
6316 2018-03-12 Richard Biener <rguenther@suse.de>
6318 PR tree-optimization/84777
6319 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6320 force-vectorize loops ignore whether we are optimizing for size.
6322 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6324 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6325 (TARGET_MD_ASM_ADJUST): Define.
6327 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6328 Kito Cheng <kito.cheng@gmail.com>
6329 Chung-Ju Wu <jasonwucj@gmail.com>
6331 * config/nds32/nds32.c (nds32_compute_stack_frame,
6332 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6333 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6334 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6335 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6336 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6337 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6338 * config/nds32/nds32.md (prologue, epilogue): Use macro
6339 NDS32_V3PUSH_AVAILABLE_P to do checking.
6341 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6344 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6345 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6346 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6347 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6348 addition of most attributes on !orig_type_die or the attribute not
6349 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6351 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6352 Chung-Ju Wu <jasonwucj@gmail.com>
6354 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6356 * config/nds32/nds32.opt (mvh): New option.
6358 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6359 Chung-Ju Wu <jasonwucj@gmail.com>
6361 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6363 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6364 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6367 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6368 Chung-Ju Wu <jasonwucj@gmail.com>
6370 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6372 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6373 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6375 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6376 Kito Cheng <kito.cheng@gmail.com>
6377 Chung-Ju Wu <jasonwucj@gmail.com>
6379 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6380 UNSPEC_FFMISM and UNSPEC_FLMISM.
6381 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6382 for ffb, ffmism and flmism.
6383 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6384 (unspec_ffmism): Ditto.
6385 (unspec_flmism): Ditto.
6386 (nds32_expand_builtin_impl): Check if string extension is available.
6387 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6388 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6390 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6393 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6396 * lra-assigns.c (assign_by_spills): Return a flag of reload
6397 assignment failure. Do not process the reload assignment
6398 failures. Do not spill other reload pseudos if they has the same
6400 (lra_assign): Add a return arg. Set up from the result of
6401 assign_by_spills call.
6402 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6403 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6404 usage_insns if it is not NULL.
6405 (spill_hard_reg_in_range): New function.
6406 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6407 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6408 function prototypes.
6409 (lra_assign): Change prototype.
6410 * lra.c (lra): Add code to deal with fails by splitting hard reg
6413 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6416 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6418 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6421 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6422 (dw_val_node): Add val_symbolic_view.
6423 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6424 (symview_upper_bound): New.
6425 (new_line_info_table): Initialize symviews_since_reset.
6426 (dwarf2out_source_line): Count symviews_since_reset and set
6427 symview_upper_bound.
6428 (dw_val_equal_p): Handle symview.
6429 (add_AT_symview): New.
6430 (print_dw_val): Handle symview.
6431 (attr_checksum, attr_checksum_ordered): Likewise.
6432 (same_dw_val_p, size_of_die): Likewise.
6433 (value_format, output_die): Likewise.
6434 (add_high_low_attributes): Use add_AT_symview for entry_view.
6435 (dwarf2out_finish): Reset symview_upper_bound, clear
6438 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6441 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6442 Add strict argument and use it.
6443 (rs6000_split_multireg_move): Update for new strict argument.
6444 (mem_operand_gpr): Disallow all non-offsettable addresses.
6445 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6447 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6450 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6451 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6452 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6455 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6456 decl, use remap_type if we want to use the type.
6458 2018-03-09 Martin Sebor <msebor@redhat.com>
6460 PR tree-optimization/84526
6461 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6463 (builtin_access::generic_overlap): Be prepared to handle non-array
6466 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6468 PR rtl-optimization/84682
6469 * lra-constraints.c (process_address_1): Check is_address flag
6470 for address constraints.
6471 (process_alt_operands): Likewise.
6472 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6473 preprocess_constraints.
6474 * recog.h (preprocess_constraints): Add oploc parameter.
6476 * recog.c (preprocess_constraints): Test address_operand for
6477 CT_ADDRESS constraints.
6479 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6482 * lra-assigns.c (assign_by_spills): Return a flag of reload
6483 assignment failure. Do not process the reload assignment
6484 failures. Do not spill other reload pseudos if they has the same
6486 (lra_assign): Add a return arg. Set up from the result of
6487 assign_by_spills call.
6488 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6489 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6490 usage_insns if it is not NULL.
6491 (spill_hard_reg_in_range): New function.
6492 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6493 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6494 function prototypes.
6495 (lra_assign): Change prototype.
6496 * lra.c (lra): Add code to deal with fails by splitting hard reg
6499 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6502 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6503 Accept complain bool parameter. Only emit errors if it is true.
6504 (arm_parse_cpu_option_name): Likewise.
6505 (arm_target_thumb_only): Adjust callers of the above.
6506 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6507 prototype to take a default true bool parameter.
6508 (arm_parse_arch_option_name): Likewise.
6510 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6511 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6515 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6516 * configure: Regenerate.
6517 * configure.ac ("linker --version-script option"): New.
6518 ("linker soname option"): New.
6520 2018-03-09 Richard Biener <rguenther@suse.de>
6522 PR tree-optimization/84775
6523 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6524 immediate uses of predicate stmts and mark them modified.
6527 PR tree-optimization/84178
6528 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6530 (version_loop_for_if_conversion): Delay update_ssa call.
6531 (tree_if_conversion): Delay update_ssa until after predicate
6534 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6537 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6538 when the function accesses prior frames.
6540 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6543 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6544 gen_llsym, otherwise call maybe_gen_llsym.
6547 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6548 has ',' character inside of it.
6550 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6553 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6554 as clobbering CC_REGNUM.
6556 2018-03-08 Richard Biener <rguenther@suse.de>
6559 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6560 (follow_copies_to_constant): Do not follow SSA names registered
6563 2018-03-08 Richard Biener <rguenther@suse.de>
6565 PR tree-optimization/84178
6566 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6568 (version_loop_for_if_conversion): Delay update_ssa call.
6569 (tree_if_conversion): Delay update_ssa until after predicate
6572 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6574 PR tree-optimization/84178
6575 * tree-if-conv.c (release_bb_predicate): Remove the
6576 the assertion that the stmts have NULL use_ops.
6577 Discard the statements, asserting that they haven't
6578 yet been added to a BB.
6580 2018-03-08 Richard Biener <rguenther@suse.de>
6582 PR tree-optimization/84746
6583 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6584 (phi_translate): Pass in destination ANTIC_OUT set.
6585 (phi_translate_1): Likewise. For a simplified result lookup
6586 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6587 (phi_translate_set): Adjust.
6588 (do_pre_regular_insertion): Likewise.
6589 (do_pre_partial_partial_insertion): Likewise.
6591 2018-03-08 Martin Liska <mliska@suse.cz>
6593 PR gcov-profile/84735
6594 * doc/gcov.texi: Document usage of profile files.
6595 * gcov-io.h: Document changes in the format.
6597 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6601 * dwarf2out.c (struct dw_line_info_table): Update comments for
6603 (FORCE_RESET_NEXT_VIEW): New.
6604 (FORCE_RESETTING_VIEW_P): New.
6605 (RESETTING_VIEW_P): Check for -1 too.
6606 (ZERO_VIEW_P): Likewise.
6607 (new_line_info_table): Force-reset next view.
6608 (dwarf2out_begin_function): Likewise.
6609 (dwarf2out_source_line): Simplify zero_view_p initialization.
6610 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6611 view directly. Omit view when omitting .loc at line 0.
6613 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6615 PR tree-optimization/84740
6616 * tree-switch-conversion.c (process_switch): Call build_constructors
6617 only if info.phi_count is non-zero.
6619 PR tree-optimization/84739
6620 * tree-tailcall.c (find_tail_calls): Check call arguments against
6621 DECL_ARGUMENTS (current_function_decl) rather than
6622 DECL_ARGUMENTS (func) when checking for tail recursion.
6624 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6626 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6627 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6628 Volker Reichelt's entry and add entries for people that perform
6629 GCC fuzzy testing and report numerous bugs.
6631 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6634 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6635 readonly data in sdata, if that is disabled.
6636 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6637 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6638 -mreadonly-in-sdata option.
6640 2018-03-07 Martin Sebor <msebor@redhat.com>
6642 PR tree-optimization/84468
6643 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6644 basic block when looking for nul assignment.
6646 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6649 * except.h (output_function_exception_table): Adjust prototype.
6650 * except.c (output_function_exception_table): Remove FNNAME parameter
6651 and add SECTION parameter. Ouput one part of the table at a time.
6652 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6653 the first part of the exception table and emit unwind directives.
6654 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6655 (i386_pe_seh_cold_init): Likewise.
6656 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6657 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6658 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6659 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6660 * config/i386/winnt.c: Include except.h.
6661 (struct seh_frame_state): Add reg_offset, after_prologue and
6662 in_cold_section fields.
6663 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6664 (i386_pe_seh_cold_init): New function.
6665 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6666 to seh->in_cold_section.
6667 (seh_emit_push): Record the offset of the push.
6668 (seh_emit_save): Record the offet of the save.
6669 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6670 Test seh->after_prologue to disregard the epilogue.
6671 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6672 (i386_pe_end_cold_function): New function.
6674 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6677 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6678 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6681 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6683 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6684 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6687 * multiple_target.c: Include tree-inline.h and intl.h.
6688 (expand_target_clones): Diagnose and fail if node->definition and
6689 !tree_versionable_function_p (node->decl).
6691 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6693 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6695 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6696 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6697 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6699 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6702 * combine.c (try_combine): Use reg_or_subregno instead of handling
6703 just paradoxical SUBREGs and REGs.
6705 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6707 * config/arc/arc.c (arc_finalize_pic): Remove function.
6708 (arc_must_save_register): We use single base PIC register, remove
6709 checks to save/restore the PIC register.
6710 (arc_expand_prologue): Likewise.
6711 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6713 (arc_verify_short): Likewise.
6714 (arc_attr_type): Likewise.
6715 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6716 (walk_stores): Likewise.
6717 (arc_address_cost): Make it static.
6718 (arc_verify_short): Likewise.
6719 (branch_dest): Likewise.
6720 (arc_attr_type): Likewise.
6721 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6722 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6723 (arc_final_prescan_insn): Remove inserting the nops due to
6724 hardware hazards. It is done in reorg step.
6725 (insn_length_variant_t): Remove.
6726 (insn_length_parameters_t): Likewise.
6727 (arc_insn_length_parameters): Likewise.
6728 (arc_get_insn_variants): Likewise.
6729 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6731 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6734 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6737 PR tree-optimization/84687
6738 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6740 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6742 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6744 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6745 Rename to ppc_speculation_barrier.
6746 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6747 __builtin_ppc_speculation_barrier.
6749 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6752 * combine.c (combine_simplify_rtx): Don't try to simplify if
6753 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6756 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6758 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6759 to 32 bytes when compiling for POWER9.
6761 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6764 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6765 regparm >= 3 with no arg reg available also for calls with
6766 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6769 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6771 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6773 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6776 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6778 2018-03-05 Richard Biener <rguenther@suse.de>
6780 PR tree-optimization/84486
6781 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6782 When inserting a __builtin_assume_aligned call set the LHS
6783 SSA name alignment info accordingly.
6785 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6787 PR tree-optimization/84114
6788 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6789 Avoid reassociation of FLOAT_MODE addition.
6791 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6793 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6794 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6795 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6796 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6797 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6798 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6799 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6801 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6802 __builtin_ia32_wbinvd): New builtins.
6803 (SPECIAL_ARGS2): New.
6804 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6805 (SPECIAL_ARGS2): New.
6806 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6807 (ix86_valid_target_attribute_inner_p): Ditto.
6808 (ix86_init_mmx_sse_builtins): Add special_args2.
6809 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6810 TARGET_WBNOINVD_P): New.
6811 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6812 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6813 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6814 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6815 * config/i386/pconfigintrin.h: New file.
6816 * config/i386/wbnoinvdintrin.h: Ditto.
6817 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6818 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6820 2018-03-05 Richard Biener <rguenther@suse.de>
6822 PR tree-optimization/84670
6823 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6825 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6826 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6827 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6828 doesn't grow if all successors (recursively) were visited at least
6831 2018-03-05 Richard Biener <rguenther@suse.de>
6833 PR tree-optimization/84650
6834 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6835 if executed in the loop pipeline.
6837 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6839 * doc/configfiles.texi (Configuration Files): Move info about
6840 conditionalizing $target-protos.h to...
6841 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6842 differs from $target-protos.h.
6844 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6845 Chung-Ju Wu <jasonwucj@gmail.com>
6847 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6848 * config/nds32/nds32-multiple.md (setmemsi): Define.
6849 * config/nds32/nds32-memory-manipulation.c
6850 (nds32_gen_dup_4_byte_to_word_value): New.
6851 (emit_setmem_word_loop): New.
6852 (emit_setmem_byte_loop): New.
6853 (nds32_expand_setmem_loop): New.
6854 (nds32_expand_setmem_loop_v3m): New.
6855 (nds32_expand_setmem_unroll): New.
6856 (nds32_expand_setmem): New.
6858 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6859 Chung-Ju Wu <jasonwucj@gmail.com>
6861 * config/nds32/nds32-memory-manipulation.c
6862 (nds32_emit_load_store): New.
6863 (nds32_emit_post_inc_load_store): New.
6864 (nds32_emit_mem_move): New.
6865 (nds32_emit_mem_move_block): New.
6866 (nds32_expand_movmemsi_loop_unknown_size): New.
6867 (nds32_expand_movmemsi_loop_known_size): New.
6868 (nds32_expand_movmemsi_loop): New.
6869 (nds32_expand_movmemsi_unroll): New.
6870 (nds32_expand_movmemqi): Rename ...
6871 (nds32_expand_movmemsi): ... to this.
6872 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6873 (movmemsi): ... to this.
6874 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6875 (nds32_expand_movmemsi): ... to this.
6877 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6878 Monk Chiang <sh.chiang04@gmail.com>
6879 Chung-Ju Wu <jasonwucj@gmail.com>
6881 * config/nds32/nds32-protos.h
6882 (nds32_expand_load_multiple): New arguments.
6883 (nds32_expand_store_multiple): Ditto.
6884 (nds32_valid_multiple_load_store): Rename ...
6885 (nds32_valid_multiple_load_store_p): ... to this.
6886 * config/nds32/nds32-memory-manipulation.c
6887 (nds32_expand_load_multiple): Refine implementation.
6888 (nds32_expand_store_multiple): Ditto.
6889 * config/nds32/nds32-multiple.md
6890 (load_multiple): Update nds32_expand_load_multiple interface.
6891 (store_multiple): Update nds32_expand_store_multiple interface.
6892 * config/nds32/nds32-predicates.c
6893 (nds32_valid_multiple_load_store): Rename ...
6894 (nds32_valid_multiple_load_store_p): ... to this and refine
6896 * config/nds32/predicates.md
6897 (nds32_load_multiple_and_update_address_operation): New predicate.
6898 (nds32_store_multiple_and_update_address_operation): New predicate.
6900 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6901 Chung-Ju Wu <jasonwucj@gmail.com>
6903 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6904 (combo): New attribute.
6905 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6907 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6909 * config/nds32/nds32.opt: Change -mcmodel= default value.
6911 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6912 Monk Chiang <sh.chiang04@gmail.com>
6913 Chung-Ju Wu <jasonwucj@gmail.com>
6915 * config/nds32/constants.md (unspec_element): New enum.
6916 * config/nds32/constraints.md (Umw): New constraint.
6917 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6918 * config/nds32/nds32-intrinsic.md: Likewise.
6919 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6920 (nds32_valid_smw_lwm_base_p): New.
6921 (nds32_output_smw_single_word): New.
6922 (nds32_output_lmw_single_word): New.
6923 (nds32_expand_unaligned_load): New.
6924 (nds32_expand_unaligned_store): New.
6925 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6926 (nds32_output_smw_single_word): Declare.
6927 (nds32_output_lmw_single_word): Declare.
6928 (nds32_expand_unaligned_load): Declare.
6929 (nds32_expand_unaligned_store): Declare.
6930 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6931 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6932 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6933 NDS32_BUILTIN_UASTORE_DW.
6934 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6937 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6938 Kito Cheng <kito.cheng@gmail.com>
6939 Chung-Ju Wu <jasonwucj@gmail.com>
6941 * config/nds32/nds32-intrinsic.c
6942 (nds32_expand_builtin_null_ftype_reg): Delete.
6943 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6944 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6945 (nds32_read_argument): New.
6946 (nds32_legitimize_target): Ditto.
6947 (nds32_legitimize_argument): Ditto.
6948 (nds32_check_constant_argument): Ditto.
6949 (nds32_expand_unop_builtin): Ditto.
6950 (nds32_expand_unopimm_builtin): Ditto.
6951 (nds32_expand_binop_builtin): Ditto.
6952 (nds32_builtin_decl_impl): Ditto.
6953 (builtin_description): Ditto.
6954 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6955 (nds32_init_builtins_impl): Ditto.
6956 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6957 (nds32_builtin_decl): New.
6958 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6959 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6961 2018-03-02 Jeff Law <law@redhat.com>
6963 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6964 (redundant_insn, fill_simple_delay_slots): Likewise.
6965 (fill_slots_from_thread): Likewise.
6966 * resource.c (mark_referenced_resources): Likewise.
6967 (mark_set_resources, find_dead_or_set_registers): Likewise.
6969 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6971 * substring-locations.h (format_warning_va): Formatting fix for
6973 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6975 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6976 * substring-locations.c: Include intl.h.
6977 (format_warning_va): Turned into small wrapper around
6978 format_warning_n_va, renamed to ...
6979 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6980 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6982 (format_warning_at_substring_n): New function.
6983 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6984 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6985 format_warning_at_substring with just a shorter name instead of
6986 const function pointer.
6987 (fmtwarn_n): New function.
6988 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6989 appropriate, get rid of all the fmtstr temporaries, move conditionals
6990 with G_() wrapped string literals directly into fmtwarn arguments,
6991 cast dir.len to (int), formatting fixes.
6993 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6995 * doc/invoke.texi: Remove "Cilk Plus" references.
6997 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6998 Richard Biener <rguenther@suse.de>
7001 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7002 for error or warning attributes if CALL_FROM_THUNK_P is set.
7005 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7008 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7009 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7012 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7013 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7015 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7016 instead of -1U in last predictors element's probability member.
7018 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7021 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7022 arguments if they are comparable.
7024 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7026 PR tree-optimization/84634
7027 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7028 masks and masked_loop_p with a single loop_masks, making sure it's
7029 null for bb vectorization.
7031 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7033 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7034 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7035 loop->force_vectorize to check whether there is no alias.
7037 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7040 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7042 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7044 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7045 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7046 instead of a loop around prev_real_insn.
7047 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7051 * config/i386/i386.c (ix86_print_operand): Use conditional
7052 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7055 2018-03-02 Richard Biener <rguenther@suse.de>
7057 PR tree-optimization/84427
7058 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7059 (bitmap_set_subtract_values): Rewrite to handle multiple
7062 (prune_clobbered_mems): Likewise.
7063 (phi_translate): Take edge instead of pred/phiblock.
7064 (phi_translate_1): Likewise.
7065 (phi_translate_set): Likewise. Insert all translated
7066 exprs for a value into the set, keeping possibly multiple
7067 expressions per value.
7068 (compute_antic_aux): Adjust for phi_translate changes.
7069 When intersecting union the expressions and prune those
7070 not in the final value set, keeping possibly multiple
7071 expressions per value. Do not use value-insertion
7072 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7073 all expressions. Add verification that the value-sets
7074 only shrink during iteration.
7075 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7076 (do_pre_regular_insertion): Likewise.
7077 (do_pre_partial_partial_insertion): Likewise.
7079 2018-03-02 Richard Biener <rguenther@suse.de>
7082 * config/darwin.c (saved_debug_info_level): New static global.
7083 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7084 (darwin_asm_lto_end): Restore debug info generation settings.
7086 2018-03-01 Martin Liska <mliska@suse.cz>
7089 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7092 2018-03-01 Richard Biener <rguenther@suse.de>
7095 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7096 type annotation with LTO when debug was disabled at compile-time.
7098 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7100 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7102 (mips_final_postscan_insn): Likewise.
7104 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7106 PR rtl-optimization/84528
7107 * alias.c (init_alias_target): Add commentary.
7108 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7109 a unique base value if the frame pointer is not eliminated
7110 to the stack pointer.
7112 2018-03-01 Tom de Vries <tom@codesourcery.com>
7114 PR rtl-optimization/83327
7115 * lra-int.h (hard_regs_spilled_into): Declare.
7116 * lra.c (hard_regs_spilled_into): Define.
7117 (init_reg_info): Init hard_regs_spilled_into.
7118 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7119 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7120 (process_bb_lives): Handle hard_regs_spilled_into.
7121 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7122 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7124 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7126 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7127 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7128 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7129 * config/rs6000/aix72.h: New file.
7131 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7133 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7134 instead of warning_at with conditional singular and plural messages
7138 * stor-layout.c (update_alignment_for_field): For
7139 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7140 && !DECL_PACKED (field), do the alignment update, just use
7141 only desired_align instead of MAX (type_align, desired_align)
7143 (place_field): Don't do known_align < desired_align handling
7144 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7145 is non-NULL, instead do it after rli->prev_field handling and
7146 only if not within a bitfield word. For DECL_PACKED (field)
7147 use type_align of BITS_PER_UNIT.
7149 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7151 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7152 superfluous parentheses and trailing spaces.
7154 2018-02-28 Richard Biener <rguenther@suse.de>
7156 PR tree-optimization/84584
7157 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7158 SCoPs with fake exit edge.
7160 2018-02-28 Martin Liska <mliska@suse.cz>
7163 * timevar.c (timer::print): Fix format to properly print 100%
7166 2018-02-28 Richard Biener <rguenther@suse.de>
7169 * genmatch.c (capture_info::walk_match): Do not mark
7170 captured expressions without operands as expr_p given
7171 they act more like predicates and should be subject to
7172 "lost tail" side-effect preserving.
7174 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7176 PR rtl-optimization/81611
7177 * auto-inc-dec.c (attempt_change): Move dead note from
7178 mem_insn if it's the next use of regno
7179 (find_address): Take address use of reg holding
7180 non-incremented value. Add parm to limit search to the named
7182 (merge_in_block): Attempt to use a mem insn that is the next
7183 use of the original regno.
7185 2018-02-27 Martin Sebor <msebor@redhat.com>
7188 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7189 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7191 2018-02-27 Martin Sebor <msebor@redhat.com>
7193 PR translation/84207
7194 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7195 n argument to unsigned HOST_WIDE_INT.
7196 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7197 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7198 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7199 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7201 2018-02-27 Richard Biener <rguenther@suse.de>
7203 PR tree-optimization/84512
7204 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7205 Do not use the estimate returned from record_stmt_cost for
7206 the scalar iteration cost but sum properly using add_stmt_cost.
7208 2018-02-27 Richard Biener <rguenther@suse.de>
7210 PR tree-optimization/84466
7211 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7212 Adjust last change to less strictly validate use operands.
7214 2018-02-27 Martin Liska <mliska@suse.cz>
7216 PR gcov-profile/84548
7217 * gcov.c (process_file): Allow partial overlap and consider it
7218 also as group functions.
7219 (output_lines): Properly calculate range of lines for a group.
7221 2018-02-27 Martin Liska <mliska@suse.cz>
7223 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7224 'ggc' suffixes. Change first column width.
7225 (timer::print): Fix formatting of the column.
7227 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7229 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7230 preserve inline entry blocks for the sake of debug inline
7231 entry point markers alone.
7232 (remove_unused_locals): Suggest in comments a better place to
7233 force the preservation of inline entry blocks that are
7234 otherwise unused, but do not preserve them.
7236 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7238 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7240 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7243 * config/i386/constraints.md (Bs): Replace
7244 ix86_indirect_branch_register with
7245 TARGET_INDIRECT_BRANCH_REGISTER.
7247 * config/i386/i386.md (indirect_jump): Likewise.
7248 (tablejump): Likewise.
7249 (*sibcall_memory): Likewise.
7250 (*sibcall_value_memory): Likewise.
7251 Peepholes of indirect call and jump via memory: Likewise.
7252 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7253 (*sibcall_value_GOT_32): Likewise.
7254 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7255 (GOT_memory_operand): Likewise.
7256 (call_insn_operand): Likewise.
7257 (sibcall_insn_operand): Likewise.
7258 (GOT32_symbol_operand): Likewise.
7259 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7261 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7263 PR rtl-optimization/83496
7264 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7265 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7266 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7267 redundant insn, if any.
7268 (relax_delay_slots): Likewise.
7269 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7271 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7273 PR tree-optimization/83965
7274 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7275 that grouped statements are part of a reduction chain. Return
7276 true if the statement is not marked as a reduction itself but
7278 (vect_recog_dot_prod_pattern): Don't check whether the statement
7279 is part of a group here.
7280 (vect_recog_sad_pattern): Likewise.
7281 (vect_recog_widen_sum_pattern): Likewise.
7283 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7286 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7288 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7291 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7293 (ix86_output_indirect_function_return): New prototype.
7294 (ix86_split_simple_return_pop_internal): Likewise.
7295 * config/i386/i386.c (indirect_return_via_cx): New.
7296 (indirect_return_via_cx_bnd): Likewise.
7297 (indirect_thunk_name): Handle return va CX_REG.
7298 (output_indirect_thunk_function): Create alias for
7299 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7300 (ix86_output_indirect_jmp): Remove the bool argument.
7301 (ix86_output_indirect_function_return): New function.
7302 (ix86_split_simple_return_pop_internal): Likewise.
7303 * config/i386/i386.md (*indirect_jump): Don't pass false
7304 to ix86_output_indirect_jmp.
7305 (*tablejump_1): Likewise.
7306 (simple_return_pop_internal): Change it to define_insn_and_split.
7307 Call ix86_split_simple_return_pop_internal to split it for
7309 (simple_return_indirect_internal): Call
7310 ix86_output_indirect_function_return instead of
7311 ix86_output_indirect_jmp.
7313 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7316 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7317 memset and value initialization afterwards.
7319 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7321 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7323 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7326 * common/config/aarch64/aarch64-common.c
7327 (aarch_option_optimization_table[]): Switch
7328 off fomit-frame-pointer
7330 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7331 Chung-Ju Wu <jasonwucj@gmail.com>
7333 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7335 (store_multiple): Ditto.
7337 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7339 * config.gcc: Add --with-cpu support for nds32 target.
7340 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7341 * config/nds32/nds32.opt: Add -mcpu= option.
7343 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7345 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7346 isel=yes): Warn for these deprecated options.
7348 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7350 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7351 ISA_2_5_MASKS_EMBEDDED.
7353 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7355 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7356 p->max as pointers rather than using iterative_hash_expr.
7358 2018-02-23 Carl Love <cel@us.ibm.com>
7360 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7361 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7363 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7364 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7365 P8V_BUILTIN_VEC_VUNSIGNED2.
7367 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7370 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7371 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7373 * lra-constraints.c (curr_insn_transform): Set up
7374 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7376 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7377 * lra-eliminations.c (spill_pseudos): Ditto.
7378 (process_insn_for_elimination): Ditto.
7379 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7380 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7381 LRA_NON_CLOBBERED_ALT.
7383 2018-02-22 Martin Sebor <msebor@redhat.com>
7385 PR tree-optimization/84480
7386 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7387 to maybe_diag_stxncpy_trunc. Call it.
7388 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7389 from gimple_fold_builtin_strcpy. Print inlining stack.
7390 (handle_builtin_stxncpy): Print inlining stack.
7391 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7393 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7396 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7397 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7398 and -fcheck-pointer-bounds are used together.
7399 (indirect_thunk_prefix): New enum.
7400 (indirect_thunk_need_prefix): New function.
7401 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7402 "_nt" instead of "_bnd" for NOTRACK prefix.
7403 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7404 (output_indirect_thunk_function): Likewise.
7406 (ix86_code_end): Update output_indirect_thunk_function calls.
7407 (ix86_output_indirect_branch_via_reg): Replace
7408 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7409 (ix86_output_indirect_branch_via_push): Likewise.
7410 (ix86_output_function_return): Likewise.
7411 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7412 incompatible with -fcf-protection=branch and
7413 -fcheck-pointer-bounds.
7415 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7418 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7419 Change gcc_assert call to output_operand_lossage.
7421 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7423 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7425 2018-02-22 DJ Delorie <dj@redhat.com>
7426 Sebastian Perta <sebastian.perta@renesas.com>
7427 Oleg Endo <olegendo@gcc.gnu.org>
7429 * config/rx/rx.c (rx_rtx_costs): New function.
7430 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7432 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7434 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7436 2018-02-22 Martin Liska <mliska@suse.cz>
7439 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7440 Add "native" as a possible value.
7442 2018-02-22 Martin Liska <mliska@suse.cz>
7445 * config/i386/i386.c (ix86_option_override_internal):
7446 Add "native" as a possible value for -march and -mtune.
7448 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7451 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7452 to all type variants.
7454 PR tree-optimization/84503
7455 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7456 width as info->bitpos + info->bitsize - start.
7457 (merged_store_group::merge_overlapping): Simplify width computation.
7458 (check_no_overlap): New function.
7459 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7460 start + width and last_order of the group, fail if check_no_overlap
7462 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7463 to group if check_no_overlap fails.
7465 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7467 * config/rs6000/altivec.md: Delete contraint arguments to
7468 define_expand, define_split, and define_peephole2, and in
7469 define_insn_and_split if always unused.
7470 * config/rs6000/darwin.md: Ditto.
7471 * config/rs6000/dfp.md: Ditto.
7472 * config/rs6000/rs6000.md: Ditto.
7473 * config/rs6000/sync.md: Ditto.
7474 * config/rs6000/vector.md: Ditto.
7475 * config/rs6000/vsx.md: Ditto.
7477 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7479 * config/rs6000/altivec.md: Write output control strings as braced
7480 blocks instead of double-quoted strings.
7481 * config/rs6000/darwin.md: Ditto.
7482 * config/rs6000/rs6000.md: Ditto.
7483 * config/rs6000/vector.md: Ditto.
7484 * config/rs6000/vsx.md: Ditto.
7486 2018-02-21 Jason Merrill <jason@redhat.com>
7488 PR c++/84314 - ICE with templates and fastcall attribute.
7489 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7491 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7493 * ipa-cp.c (determine_versionability): Fix comment typos.
7495 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7498 * ipa-cp.c (determine_versionability): Do not version functions caling
7501 2018-02-21 Martin Liska <mliska@suse.cz>
7504 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7505 Add "native" as a possible value.
7506 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7507 the macro when native cpu detection is available.
7509 2018-02-21 Martin Liska <mliska@suse.cz>
7512 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7513 Add "native" as a possible value.
7514 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7515 when native cpu detection is available.
7517 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7518 Martin Sebor <msebor@redhat.com>
7520 PR tree-optimization/84478
7521 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7523 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7524 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7525 support which is conservatively correct, for 2 only stay conservative
7526 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7527 argument to the 2 argument get_range_strlen, adjust 6 arg
7528 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7530 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7531 (gimple_fold_builtin_strlen): Pass true as last argument to
7534 2018-02-20 Martin Sebor <msebor@redhat.com>
7537 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7538 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7539 (builtin_memref::builtin_memref): Factor out parts into
7540 set_base_and_offset and call it.
7542 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7545 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7546 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7547 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7548 search at the associated MODE_INT.
7550 2018-02-20 Jeff Law <law@redhat.com>
7553 PR tree-optimization/81592
7555 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7556 for range data rather than using global data.
7557 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7558 range data rather than using global data.
7559 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7560 pass it to children as needed.
7561 (struct directive::fmtresult): Similarly.
7562 (struct directive::set_width): Similarly.
7563 (struct directive::set_precision): Similarly.
7564 (format_integer, format_directive, parse_directive): Similarly.
7565 (format_none): Accept unnamed vr_values parameter.
7566 (format_percent, format_floating, format_character): Similarly.
7567 (format_string, format_plain): Similarly.
7568 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7569 the EVRP range analyzer for range data rather than using global data.
7570 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7571 gimple-ssa-evrp-analyze.h
7572 (class sprintf_dom_walker): Add after_dom_children member function.
7573 Add evrp_range_analyzer member.
7574 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7575 range analyzer as needed.
7576 (sprintf_dom_walker::after_dom_children): New member function.
7577 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7579 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7580 (evrp_range_analyzer::pop_to_marker): Likewise.
7582 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7584 PR tree-optimization/84419
7585 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7586 with the required type if its current type is compatible but
7589 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7592 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7593 after vectorization.
7595 2018-02-20 Martin Liska <mliska@suse.cz>
7598 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7599 possible values if we don't have a hint.
7601 2018-02-20 Martin Liska <mliska@suse.cz>
7605 * final.c (shorten_branches): Build align_tab array with one
7607 * opts.c (finish_options): Add alignment option limit check.
7608 (MAX_CODE_ALIGN): Likewise.
7609 (MAX_CODE_ALIGN_VALUE): Likewise.
7610 * doc/invoke.texi: Document maximum allowed option value for
7611 all -falign-* options.
7613 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7616 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7617 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7618 * var-tracking.c (emit_note_insn_var_location): Remove all references
7619 to NOTE_INSN_CALL_ARG_LOCATION.
7620 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7621 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7622 Use copy_rtx_if_shared.
7623 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7624 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7625 (dwarf2out_var_location): Remove handling of
7626 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7628 * final.c (final_scan_insn): Remove all references to
7629 NOTE_INSN_CALL_ARG_LOCATION.
7630 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7631 before dumping final insns.
7632 * except.c (emit_note_eh_region_end): Remove all references to
7633 NOTE_INSN_CALL_ARG_LOCATION.
7634 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7635 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7636 * config/arc/arc.c (hwloop_optimize): Likewise.
7637 * config/arm/arm.c (create_fix_barrier): Likewise.
7638 * config/s390/s390.c (s390_chunkify_start): Likewise.
7639 * config/sh/sh.c (find_barrier): Likewise.
7640 * config/i386/i386.c (rest_of_insert_endbranch,
7641 ix86_seh_fixup_eh_fallthru): Likewise.
7642 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7643 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7644 * config/frv/frv.c (frv_function_prologue): Likewise.
7645 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7647 (note_outside_basic_block_p): Remove all references to
7648 NOTE_INSN_CALL_ARG_LOCATION.
7649 * gengtype.c (adjust_field_rtx_def): Likewise.
7650 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7652 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7653 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7656 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7659 PR tree-optimization/84452
7660 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7661 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7664 2018-02-19 Martin Liska <mliska@suse.cz>
7667 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7669 2018-02-19 Martin Liska <mliska@suse.cz>
7670 Richard Sandiford <richard.sandiford@linaro.org>
7672 PR tree-optimization/82491
7673 * gimple-fold.c (get_base_constructor): Make earlier bail out
7676 2018-02-19 Carl Love <cel@us.ibm.com>
7678 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7679 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7681 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7682 P8V_BUILTIN_VEC_NEG.
7684 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7686 * config/rl78/rl78.md (movdf): New define expand.
7688 2018-02-19 Martin Liska <mliska@suse.cz>
7691 * doc/invoke.texi: Fix typo.
7692 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7694 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7696 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7697 handle rs6000_single_float and rs6000_double_float specially for
7700 2018-02-16 Jeff Law <law@redhat.com>
7702 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7705 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7708 * ipa-inline.c (inline_small_functions): Fix a typo.
7710 2018-02-16 Nathan Sidwell <nathan@acm.org>
7712 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7714 2018-02-16 Carl Love <cel@us.ibm.com>
7716 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7717 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7718 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7719 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7720 expansion to P8V_BUILTIN_VEC_FLOAT2.
7722 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7724 PR rtl-optimization/70023
7725 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7726 src_regno into account.
7728 2018-02-16 Carl Love <cel@us.ibm.com>
7730 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7731 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7732 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7733 * config/rs6000/rs6000.c: Remove case statements for
7734 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7735 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7736 and P9V_BUILTIN_VEC_VINSERT4B.
7737 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7738 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7739 * config/rs6000/vsx.md:
7740 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7743 2018-02-16 Carl Love <cel@us.ibm.com>
7745 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7747 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7749 * config/rs6000/rs6000-c.c: Add the definitions for
7750 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7751 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7752 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7753 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7754 definition for insert4b and define insn *insert3b_internal.
7755 * doc/extend.texi: Add documentation for vec_extract4b.
7757 2018-02-16 Nathan Sidwell <nathan@acm.org>
7759 * doc/extend.texi (Backwards Compatibility): Mention friend
7760 injection. Note for-scope is deprecated.
7761 * doc/invoke.texi (-ffriend-injection): Deprecate.
7763 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7765 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7766 that moved to I2, also allow destinations that are a paradoxical
7767 subreg (instead of a normal reg).
7769 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7772 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7775 2018-02-16 Richard Biener <rguenther@suse.de>
7777 PR tree-optimization/84037
7778 PR tree-optimization/84016
7780 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7781 Adjust vec_construct for the fact we need additional higher latency
7782 128bit inserts for AVX256 and AVX512 vector builds.
7783 (ix86_add_stmt_cost): Scale vector construction cost for
7786 2018-02-16 Richard Biener <rguenther@suse.de>
7788 PR tree-optimization/84417
7789 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7790 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7791 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7793 2018-02-16 Martin Liska <mliska@suse.cz>
7796 * internal-fn.def (ASAN_CHECK): Set proper flags.
7797 (ASAN_MARK): Likewise.
7799 2018-02-16 Julia Koval <julia.koval@intel.com>
7801 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7802 from PTA_CANNONLAKE.
7804 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7807 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7808 Use ++iter rather than iter++ for std::list iterators.
7809 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7810 defer deleting them until all nodes in the forest are processed. Do
7811 free even leaf nodes. Change to_process into auto_vec.
7814 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7815 * vec.h (vec_default_construct): Use memset instead of placement new
7816 if BROKEN_VALUE_INITIALIZATION is defined.
7817 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7818 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7821 PR rtl-optimization/83723
7822 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7823 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7824 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7826 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7828 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7830 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7832 PR rtl-optimization/81443
7833 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7834 from inner REGs to paradoxical SUBREGs.
7836 2018-02-16 Richard Biener <rguenther@suse.de>
7838 PR tree-optimization/84399
7839 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7840 For operands we can analyze at their definition make sure we can
7841 analyze them at each use as well.
7843 2018-02-16 Richard Biener <rguenther@suse.de>
7845 PR tree-optimization/84190
7846 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7847 volatile accesses if the decl isn't volatile.
7849 2018-02-15 Jason Merrill <jason@redhat.com>
7851 PR c++/84314 - ICE with templates and fastcall attribute.
7852 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7853 TYPE_CANONICAL on an existing type.
7855 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7857 PR tree-optimization/84383
7858 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7859 dstoff nor call operand_equal_p if dstbase is NULL.
7861 PR tree-optimization/84334
7862 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7863 also a CONSTANT_CLASS_P, punt.
7865 2018-02-14 Jim Wilson <jimw@sifive.com>
7867 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7868 first SMALL_OPERAND check. New local min_second_step. Move assert
7869 to where locals are set. Add TARGET_RVC support.
7870 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7872 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7874 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7876 2018-02-14 Martin Sebor <msebor@redhat.com>
7878 PR tree-optimization/83698
7879 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7880 arrays constrain the offset range to their bounds.
7881 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7882 (builtin_access::overlap): Avoid setting the size of overlap if it's
7884 (maybe_diag_overlap): Also consider arrays when deciding what values
7885 of offsets to include in diagnostics.
7887 2018-02-14 Martin Sebor <msebor@redhat.com>
7890 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7891 that correspond to the kind of a declaration.
7893 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7896 * config/pa/pa.md: Load address of PIC label using the linkage table
7897 if the label is nonlocal.
7899 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7901 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7902 warning message if user requests -maltivec=be.
7903 * doc/invoke.texi: Document deprecation of -maltivec=be.
7905 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7908 * config/rs6000/rs6000-c.c: Update definitions for
7909 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7910 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7912 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7915 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7916 add _get_ssp intrinsics. Remove argument from
7917 __builtin_ia32_rdssp[d|q].
7918 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7919 * config/i386/i386-builtin.def: Remove argument from
7920 __builtin_ia32_rdssp[d|q].
7921 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7922 ix86_expand_special_args_builtin for _rdssp[d|q].
7923 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7924 Clear register before usage.
7925 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7926 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7928 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7930 PR tree-optimization/84357
7931 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7932 operand 1 of an ARRAY_REF too.
7934 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7937 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7938 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7940 (set_of_reg): New struct.
7941 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7942 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7943 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7945 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7946 Split into bitclr, bitset, bitinvert patterns if appropriate.
7947 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7948 use rx_fuse_in_memory_bitop.
7949 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7950 to named insn, correct maximum insn length.
7952 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7955 * machmode.def: Define a complex mode for PARTIAL_INT.
7956 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7958 * doc/rtl.texi: Document CSPImode.
7959 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7961 (msp430_hard_regno_nregs_with_padding): Likewise.
7963 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7966 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7968 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7970 PR rtl-optimization/84169
7971 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7972 we generated a parallel as new i3 and we split that to new i2 and i3
7973 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7974 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7975 those to i2, not i1. Partially rewrite this scan code.
7977 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7980 * stor-layout.c (place_field): For variable length fields, adjust
7981 offset_align afterwards not just based on the field's alignment,
7982 but also on the size.
7985 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7986 of exps and logs in the use_exp2 case.
7988 2018-02-13 Jeff Law <law@redhat.com>
7990 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7993 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7996 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8000 * common.opt (gas-loc-support, gas-locview-support): New.
8001 (ginline-points, ginternal-reset-location-views): New.
8002 * doc/invoke.texi: Document them. Use @itemx where intended.
8003 (gvariable-location-views): Adjust.
8004 * target.def (reset_location_view): New.
8005 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8006 (TARGET_RESET_LOCATION_VIEW): New.
8007 * doc/tm.texi: Rebuilt.
8008 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8009 (dwarf2out_default_as_locview_support): New.
8010 (output_asm_line_debug_info): Use option variables.
8011 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8012 (output_loc_list): Likewise.
8013 (add_high_low_attributes): Check option variables.
8014 Don't output entry view attribute in strict mode.
8015 (gen_inlined_subroutine_die): Check option variables.
8016 (dwarf2out_inline_entry): Likewise.
8017 (init_sections_and_labels): Likewise.
8018 (dwarf2out_early_finish): Likewise.
8019 (maybe_reset_location_view): New, from...
8020 (dwarf2out_var_location): ... here. Call it.
8021 * debug.h (dwarf2out_default_as_loc_support): Declare.
8022 (dwarf2out_default_as_locview_support): Declare.
8023 * hooks.c (hook_int_rtx_insn_0): New.
8024 * hooks.h (hook_int_rtx_insn_0): Declare.
8025 * toplev.c (process_options): Take -gas-loc-support and
8026 -gas-locview-support from dwarf2out. Enable
8027 -gvariable-location-views by default only with locview
8028 assembler support. Enable -ginternal-reset-location-views by
8029 default only if the target defines the corresponding hook.
8030 Enable -ginline-points by default if location views are
8031 enabled; force it disabled if statement frontiers are
8033 * tree-inline.c (expand_call_inline): Check option variables.
8034 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8036 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8038 PR tree-optimization/84321
8039 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8040 handling. Also check whether the anti-range contains any values
8041 that satisfy the mask; switch to a VR_RANGE if not.
8043 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8046 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8048 2018-02-13 Martin Jambor <mjambor@suse.cz>
8051 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8052 of call statements, also set location of a load to a temporary.
8054 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8056 * config/rl78/rl78.c (add_vector_labels): New function.
8057 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8058 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8059 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8060 which checks that no arguments are passed.
8061 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8062 * doc/extend.texi: Documentation for the new attribute.
8064 2018-02-13 Andreas Schwab <schwab@suse.de>
8066 * config/riscv/linux.h (CPP_SPEC): Define.
8068 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8071 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8072 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8073 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8074 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8075 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8076 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8077 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8078 temporarily for AES and PCLMUL builtins.
8080 PR tree-optimization/84339
8081 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8082 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8086 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8087 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8088 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8090 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8092 * omp-simd-clone.h: New file.
8093 * omp-simd-clone.c: Include omp-simd-clone.h.
8094 (expand_simd_clones): No longer static.
8095 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8096 cgraph.h and omp-simd-clone.h.
8097 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8098 (vect_recog_widen_shift_pattern): Formatting fix.
8099 (vect_pattern_recog_1): Don't check optab for calls.
8102 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8103 operands[2] into a REG before using gen_lowpart on it.
8105 2018-02-12 Jeff Law <law@redhat.com>
8108 * config/sh/sh.c (find_barrier): Consider a sibling call
8111 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8112 successfully back substituting a reg.
8114 2018-02-12 Richard Biener <rguenther@suse.de>
8116 PR tree-optimization/84037
8117 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8118 parameter, move visited init to caller.
8119 (vect_slp_analyze_operations): Separate cost from validity
8120 check, initialize visited once for all instances.
8121 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8123 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8125 (vect_model_promotion_demotion_cost): Likewise.
8126 (vectorizable_bswap): Guard cost modeling with !slp_node
8127 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8129 (vectorizable_call): Likewise.
8130 (vectorizable_conversion): Likewise.
8131 (vectorizable_assignment): Likewise.
8132 (vectorizable_shift): Likewise.
8133 (vectorizable_operation): Likewise.
8134 (vectorizable_store): Likewise.
8135 (vectorizable_load): Likewise.
8136 (vectorizable_condition): Likewise.
8137 (vectorizable_comparison): Likewise.
8139 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8142 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8143 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8144 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8146 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8149 * params.def (inline-min-speedup): Increase from 8 to 15.
8150 (max-inline-insns-auto): Decrease from 40 to 30.
8151 * ipa-split.c (consider_split): Add some buffer for function to
8152 be considered inlining candidate.
8153 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8156 2018-02-12 Richard Biener <rguenther@suse.de>
8158 PR tree-optimization/84037
8159 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8160 matched stmts if we cannot swap the non-matched ones.
8162 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8164 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8165 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8166 _mm_maskz_scalef_round_ss): New intrinsics.
8167 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8168 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8169 __builtin_ia32_scalefss_round): Remove.
8170 (__builtin_ia32_scalefsd_mask_round,
8171 __builtin_ia32_scalefss_mask_round): New intrinsics.
8172 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8173 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8174 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8175 "<round_constraint>")): Changed to ...
8176 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8177 "<round_scalar_constraint>")): ... this.
8178 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8179 %0, %1, %2<round_op3>}"): Changed to ...
8180 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8181 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8182 %2<round_scalar_mask_op3>}"): ... this.
8183 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8185 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8187 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8188 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8189 (_mm_maskz_sqrt_round_ss): New intrinsics.
8190 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8191 (__builtin_ia32_sqrtsd_mask_round)
8192 (__builtin_ia32_sqrtss_mask_round): New builtins.
8193 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8194 (__builtin_ia32_sqrtss_round): Remove.
8195 (__builtin_ia32_sqrtsd_mask_round)
8196 (__builtin_ia32_sqrtss_mask_round): New builtins.
8197 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8198 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8199 ((match_operand:VF_128 1 "vector_operand"
8200 "xBm,<round_constraint>")): Changed to ...
8201 ((match_operand:VF_128 1 "vector_operand"
8202 "xBm,<round_scalar_constraint>")): ... this.
8203 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8204 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8205 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8206 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8207 %<iptr>1<round_scalar_mask_op3>}): ... this.
8208 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8209 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8211 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8214 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8215 Cast vec_cmpeq result to correct type.
8216 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8217 Cast vec_cmpgt result to correct type.
8219 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8221 * final.c (final_scan_insn_1): Renamed from...
8222 (final_scan_insn): ... this. New wrapper, to recover
8223 seen from the outermost call in recursive ones.
8224 * config/sparc/sparc.c (output_return): Drop seen from call.
8225 (output_sibcall): Likewise.
8226 * config/visium/visium.c (output_branch): Likewise.
8228 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8230 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8233 2018-02-10 Alan Modra <amodra@gmail.com>
8236 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8237 Specify LR as an input.
8239 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8242 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8243 remove_member_access_dummy_vars): New functions.
8244 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8245 lower_omp_1, execute_lower_omp): Use them.
8247 PR rtl-optimization/84308
8248 * shrink-wrap.c (spread_components): Release todo vector.
8250 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8252 PR rtl-optimization/57193
8253 * ira-color.c (struct allocno_color_data): Add member
8254 conflict_allocno_hard_prefs.
8255 (update_conflict_allocno_hard_prefs): New.
8256 (bucket_allocno_compare_func): Add a preference based on
8257 conflict_allocno_hard_prefs.
8258 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8259 (color_allocnos): Remove a dead code. Initiate
8260 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8262 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8265 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8266 constraint from =wa to wa. Avoid a subreg on the output operand,
8267 instead use a pseudo and subreg it in a move.
8268 (p9_xxbrd_<mode>): Changed to ...
8269 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8270 (p9_xxbrd_v2df): New expander.
8271 (p9_xxbrw_<mode>): Changed to ...
8272 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8273 (p9_xxbrw_v4sf): New expander.
8275 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8277 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8279 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8282 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8283 multiply in 32-bit mode.
8284 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8285 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8288 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8290 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8291 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8292 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8293 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8295 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8298 * dwarf2out.c (is_trivial_indirect_ref): New function.
8299 (dwarf2out_late_global_decl): Do not generate a location
8300 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8301 and that are not defined in the current unit.
8303 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8305 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8306 instead of a libcall for UNORDERED.
8308 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8311 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8312 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8314 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8317 * config/s390/s390.c (s390_set_current_function): Invoke
8318 s390_indirect_branch_settings also if fndecl didn't change.
8320 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8322 * config/rs6000/rs6000.md (blockage): Set length to zero.
8324 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8326 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8328 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8331 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8332 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8336 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8337 PARALLEL incoming that failed vt_get_decl_and_offset check.
8340 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8341 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8343 (get_variable_section): For decls in named .bss* sections pass true as
8344 second argument to bss_initializer_p.
8346 2018-02-09 Marek Polacek <polacek@redhat.com>
8347 Jakub Jelinek <jakub@redhat.com>
8350 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8351 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8352 Sync some changes from cxx_fold_indirect_ref.
8354 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8356 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8358 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8359 (BLOCK_INLINE_ENTRY_LABEL): New.
8360 (dwarf2out_var_location): Disregard inline entry markers.
8361 (inline_entry_data): New struct.
8362 (inline_entry_data_hasher): New hashtable type.
8363 (inline_entry_data_hasher::hash): New.
8364 (inline_entry_data_hasher::equal): New.
8365 (inline_entry_data_table): New variable.
8366 (add_high_low_attributes): Add DW_AT_entry_pc and
8367 DW_AT_GNU_entry_view attributes if a pending entry is found
8368 in inline_entry_data_table. Add old entry_pc attribute only
8369 if debug nonbinding markers are disabled.
8370 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8371 markers are enabled.
8372 (block_within_block_p, dwarf2out_inline_entry): New.
8373 (dwarf2out_finish): Check that no entries remained in
8374 inline_entry_data_table.
8375 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8376 (final_scan_insn, notice_source_line): Likewise.
8377 (rest_of_clean_state): Skip inline entry markers.
8378 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8380 * gimple.c (gimple_build_debug_inline_entry): New.
8381 * gimple.h (enum gimple_debug_subcode): Add
8382 GIMPLE_DEBUG_INLINE_ENTRY.
8383 (gimple_build_debug_inline_entry): Declare.
8384 (gimple_debug_inline_entry_p): New.
8385 (gimple_debug_nonbind_marker_p): Adjust.
8386 * insn-notes.def (INLINE_ENTRY): New.
8387 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8388 inline entry marker notes.
8389 (print_insn): Likewise.
8390 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8391 (INSN_DEBUG_MARKER_KIND): Likewise.
8392 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8393 * tree-inline.c (expand_call_inline): Build and insert
8394 debug_inline_entry stmt.
8395 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8396 inline entry blocks early, if nonbind markers are enabled.
8397 (dump_scope_block): Dump fragment info.
8398 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8399 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8400 (gimple_build_debug_inline_entry): New.
8401 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8402 Enable/disable inline entry points too.
8403 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8404 (DEBUG_INSN): Describe inline entry markers.
8406 * common.opt (gvariable-location-views): New.
8407 (gvariable-location-views=incompat5): New.
8408 * config.in: Rebuilt.
8409 * configure: Rebuilt.
8410 * configure.ac: Test assembler for view support.
8411 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8412 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8413 * dwarf2out.c (var_loc_view): New typedef.
8414 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8415 (dwarf2out_locviews_in_attribute): New.
8416 (dwarf2out_locviews_in_loclist): New.
8417 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8418 (enum dw_line_info_opcode): Add LI_adv_address.
8419 (struct dw_line_info_table): Add view.
8420 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8421 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8422 (zero_view_p): New variable.
8423 (ZERO_VIEW_P): New macro.
8424 (output_asm_line_debug_info): New.
8425 (struct var_loc_node): Add view.
8426 (add_AT_view_list, AT_loc_list): New.
8427 (add_var_loc_to_decl): Add view param. Test it against last.
8428 (new_loc_list): Add view params. Record them.
8429 (AT_loc_list_ptr): Handle loc and view lists.
8430 (view_list_to_loc_list_val_node): New.
8431 (print_dw_val): Handle dw_val_class_view_list.
8432 (size_of_die): Likewise.
8433 (value_format): Likewise.
8434 (loc_list_has_views): New.
8435 (gen_llsym): Set vl_symbol too.
8436 (maybe_gen_llsym, skip_loc_list_entry): New.
8437 (dwarf2out_maybe_output_loclist_view_pair): New.
8438 (output_loc_list): Output view list or entries too.
8439 (output_view_list_offset): New.
8440 (output_die): Handle dw_val_class_view_list.
8441 (output_dwarf_version): New.
8442 (output_compilation_unit_header): Use it.
8443 (output_skeleton_debug_sections): Likewise.
8444 (output_rnglists, output_line_info): Likewise.
8445 (output_pubnames, output_aranges): Update version comments.
8446 (output_one_line_info_table): Output view numbers in asm comments.
8447 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8448 Call maybe_gen_llsym.
8449 (loc_list_from_tree_1): Adjust.
8450 (add_AT_location_description): Create view list attribute if
8451 needed, check it's absent otherwise.
8452 (convert_cfa_to_fb_loc_list): Adjust.
8453 (maybe_emit_file): Call output_asm_line_debug_info for test.
8454 (dwarf2out_var_location): Reset views as needed. Precompute
8455 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8456 attribute. Set view.
8457 (new_line_info_table): Reset next view.
8458 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8459 (dwarf2out_source_line): Likewise. Output view resets and labels to
8460 the assembler, or select appropriate line info opcodes.
8461 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8462 (optimize_string_length): Catch it. Adjust.
8463 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8464 dw_val_class_view_list, and remove it if no longer needed.
8465 (hash_loc_list): Hash view numbers.
8466 (loc_list_hasher::equal): Compare them.
8467 (optimize_location_lists): Check whether a view list symbol is
8468 needed, and whether the locview attribute is present, and
8469 whether they match. Remove the locview attribute if no longer
8471 (index_location_lists): Call skip_loc_list_entry for test.
8472 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8473 Use output_dwarf_version.
8474 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8475 (struct dw_val_node): Add val_view_list.
8476 * final.c (SEEN_NEXT_VIEW): New.
8477 (set_next_view_needed): New.
8478 (clear_next_view_needed): New.
8479 (maybe_output_next_view): New.
8480 (final_start_function): Rename to...
8481 (final_start_function_1): ... this. Take pointer to FIRST,
8482 add SEEN parameter. Emit param bindings in the initial view.
8483 (final_start_function): Reintroduce SEEN-less interface.
8484 (final): Rename to...
8485 (final_1): ... this. Take SEEN parameter. Output final pending
8486 next view at the end.
8487 (final): Reintroduce seen-less interface.
8488 (final_scan_insn): Output pending next view before switching
8489 sections or ending a block. Mark the next view as needed when
8490 outputting variable locations. Notify debug backend of section
8491 changes, and of location view changes.
8492 (rest_of_handle_final): Adjust.
8493 * toplev.c (process_options): Autodetect value for debug variable
8494 location views option. Warn on incompat5 without -gdwarf-5.
8495 * doc/invoke.texi (gvariable-location-views): New.
8496 (gvariable-location-views=incompat5): New.
8497 (gno-variable-location-views): New.
8499 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8501 PR tree-optimization/84136
8502 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8503 that the result of find_edge is non-NULL.
8505 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8508 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8509 storing integer register in SImode. Fix cost of 256 and 512
8510 byte aligned SSE register store.
8512 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8514 * config/i386/i386.c (ix86_multiplication_cost): Fix
8515 multiplication cost for TARGET_AVX512DQ.
8517 2018-02-08 Marek Polacek <polacek@redhat.com>
8519 PR tree-optimization/84238
8520 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8523 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8525 PR tree-optimization/84265
8526 * tree-vect-stmts.c (vectorizable_store): Don't treat
8527 VMAT_CONTIGUOUS accesses as grouped.
8528 (vectorizable_load): Likewise.
8530 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8532 PR tree-optimization/81635
8533 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8534 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8535 (test_round_for_mask): New functions.
8536 (wide_int_cc_tests): Call test_round_for_mask.
8537 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8538 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8539 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8540 range returned by get_range_info.
8542 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8545 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8546 * symtab.c: Include builtins.h
8547 (symtab_node::output_to_lto_symbol_table_p): Move here
8548 from lto-streamer-out.c:output_symbol_p.
8549 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8550 (output_symbol_p): Move all logic to symtab.c
8551 (produce_symtab): Update.
8553 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8555 * config/s390/s390-opts.h (enum indirect_branch): Define.
8556 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8557 (s390_indirect_branch_via_thunk)
8558 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8559 (enum s390_indirect_branch_type): Define.
8560 * config/s390/s390.c (struct s390_frame_layout, struct
8561 machine_function): Remove.
8562 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8563 (indirect_branch_table_label_no, indirect_branch_table_name):
8565 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8566 (enum s390_indirect_branch_option): Define.
8567 (s390_return_addr_from_memory): New function.
8568 (s390_handle_string_attribute): New function.
8569 (s390_attribute_table): Add new attribute handler.
8570 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8571 (s390_indirect_branch_via_thunk): New function.
8572 (s390_indirect_branch_via_inline_thunk): New function.
8573 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8574 sibling call optimization for non z10 compiles.
8575 (s390_emit_call): Force indirect branch target to be a single
8576 register. Add r1 clobber for non-z10 compiles.
8577 (s390_emit_epilogue): Emit return jump via return_use expander.
8578 (s390_reorg): Handle JUMP_INSNs as execute targets.
8579 (s390_option_override_internal): Perform validity checks for the
8580 new command line options.
8581 (s390_indirect_branch_attrvalue): New function.
8582 (s390_indirect_branch_settings): New function.
8583 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8584 (s390_output_indirect_thunk_function): New function.
8585 (s390_code_end): Implement target hook.
8586 (s390_case_values_threshold): Implement target hook.
8587 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8589 * config/s390/s390.h (struct s390_frame_layout)
8590 (struct machine_function): Move here from s390.c.
8591 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8592 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8593 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8594 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8595 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8596 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8597 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8598 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8599 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8600 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8601 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8602 (mnemonic attribute): Add values which aren't recognized
8604 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8605 pattern for branch conversion. Fix mnemonic attribute.
8606 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8607 indirect branch via thunk if requested.
8608 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8609 ("*indirect_jump"): Disable for branch conversion using out of
8611 ("indirect_jump_via_thunk<mode>_z10")
8612 ("indirect_jump_via_thunk<mode>")
8613 ("indirect_jump_via_inlinethunk<mode>_z10")
8614 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8615 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8616 ("casesi_jump_via_inlinethunk<mode>_z10")
8617 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8618 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8619 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8620 ("*indirect2_jump"): Disable for branch conversion.
8621 ("casesi_jump"): Turn into expander and expand patterns for branch
8623 ("return_use"): New expander.
8624 ("*return"): Emit return via thunk and rename it to ...
8625 ("*return<mode>"): ... this one.
8626 * config/s390/s390.opt: Add new options and and enum for the
8629 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8631 * lra-constraints.c (match_reload): Unconditionally use
8632 gen_lowpart_SUBREG, rather than selecting between that
8633 and equivalent gen_rtx_SUBREG code.
8635 2018-02-08 Richard Biener <rguenther@suse.de>
8637 PR tree-optimization/84233
8638 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8639 changed flag instead of boguously re-using phi_inserted.
8641 2018-02-08 Martin Jambor <mjambor@suse.cz>
8643 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8644 static local variables.
8646 2018-02-08 Richard Biener <rguenther@suse.de>
8648 PR tree-optimization/84278
8649 * tree-vect-stmts.c (vectorizable_store): When looking for
8650 smaller vector types to perform grouped strided loads/stores
8651 make sure the mode is supported by the target.
8652 (vectorizable_load): Likewise.
8654 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8656 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8657 Increase LDP/STP opportunities by adding adjacent callee-saves.
8659 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8661 PR rtl-optimization/84068
8662 PR rtl-optimization/83459
8663 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8665 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8667 PR tree-optimization/84224
8668 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8669 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8672 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8675 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8676 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8677 count, remove one USE.
8679 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8681 * doc/install.texi (Configuration): Document the
8682 --with-long-double-format={ibm,ieee} PowerPC configuration
8686 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8687 Convert from define_expand to be define_insn_and_split. Rework
8688 float/double/_Float128 conversions to QI/HI/SImode to work with
8689 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8690 conversions to QI/HImode types did a store and then a load to
8691 truncate the value. For conversions to VSX registers, don't split
8692 the insn, instead emit the code directly. Use the code iterator
8693 any_fix to combine signed and unsigned conversions.
8694 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8695 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8696 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8697 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8698 (fix_<mode>di2_hw): Likewise.
8699 (fixuns_<mode>di2_hw): Likewise.
8700 (fix_<mode>si2_hw): Likewise.
8701 (fixuns_<mode>si2_hw): Likewise.
8702 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8703 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8704 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8705 fix<uns>_trunc<SFDF:mode>si2_p8.
8706 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8708 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8709 (fix<uns>_<mode>_mem): Likewise.
8710 (fctiw<u>z_<mode>_mem): Likewise.
8711 (fix<uns>_<mode>_mem): Likewise.
8712 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8713 the register allocator from doing a direct move to the GPRs to do
8714 a store, and instead use the ISA 3.0 store byte/half-word from
8715 vector register instruction. For IEEE 128-bit floating point,
8716 also optimize stores of 32-bit ints.
8717 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8719 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8721 * genextract.c (push_pathstr_operand): New function to support
8723 (walk_rtx): Call push_pathstr_operand.
8724 (print_path): Support [a-zA-Z].
8726 2018-02-07 Richard Biener <rguenther@suse.de>
8728 PR tree-optimization/84037
8729 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8730 (cse_and_gimplify_to_preheader): Declare.
8731 (vect_get_place_in_interleaving_chain): Likewise.
8732 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8734 (_loop_vec_info::~_loop_vec_info): Delete it.
8735 (cse_and_gimplify_to_preheader): New function.
8736 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8737 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8738 (vectorizable_load): Likewise. For grouped stores always base
8739 the IV on the first element.
8740 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8741 condition before gimplifying.
8743 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8745 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8746 *DIV_EXPR and *MOD_EXPR.
8748 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8751 * config/i386/i386.c (ix86_option_override_internal): Mask out
8752 the CF_SET bit when checking -fcf-protection.
8754 2018-02-07 Tom de Vries <tom@codesourcery.com>
8757 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8760 2018-02-07 Richard Biener <rguenther@suse.de>
8762 PR tree-optimization/84204
8763 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8766 PR tree-optimization/84205
8767 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8768 special-case isl_ast_op_zdiv_r.
8770 PR tree-optimization/84223
8771 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8772 Only add conditions from within the region.
8773 (gather_bbs::after_dom_children): Adjust.
8775 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8778 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8779 * config/avr/avr.md: Only post-reload split REG-REG moves if
8780 either register is GENERAL_REG_P.
8782 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8784 PR tree-optimization/84235
8785 * tree-ssa-scopedtables.c
8786 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8787 if the subtraction is performed in floating point type where NaNs are
8788 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8789 build 1. Formatting fix.
8791 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8794 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8795 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8796 and skip it regardless of bb boundaries. Use CALL_P macro,
8797 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8798 unnecessarily, formatting fix.
8800 2018-02-06 Michael Collison <michael.collison@arm.com>
8802 * config/arm/thumb2.md:
8803 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8804 (*thumb_mov_notscc): Ditto.
8806 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8809 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8810 unsigned_fix, not "s".
8812 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8814 * configure.ac (gcc_fn_eh_frame_ro): New function.
8815 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8816 correct .eh_frame permissions.
8817 * configure: Regenerate.
8819 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8821 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8824 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8826 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8827 Display warning message for -mno-speculate-indirect-jumps.
8829 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8831 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8833 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8835 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8837 PR tree-optimization/84225
8838 * tree-eh.c (find_trapping_overflow): Only call
8839 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8841 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8844 * config/i386/i386.c: Reimplement the check of possible options
8845 -mibt/-mshstk conbination. Change error messages.
8846 * doc/invoke.texi: Fix a typo: remove extra '='.
8848 2018-02-06 Marek Polacek <polacek@redhat.com>
8850 PR tree-optimization/84228
8851 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8853 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8856 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8857 emitted arch directives.
8858 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8859 __ARM_FEATURE_COPROC before changing architectures.
8861 2018-02-06 Richard Biener <rguenther@suse.de>
8863 * config/i386/i386.c (print_reg): Fix typo.
8864 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8866 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8868 * configure: Regenerate.
8870 2018-02-05 Martin Sebor <msebor@redhat.com>
8872 PR tree-optimization/83369
8873 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8876 2018-02-05 Martin Liska <mliska@suse.cz>
8878 * doc/invoke.texi: Cherry-pick upstream r323995.
8880 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8882 * ira.c (ira_init_register_move_cost): Adjust comment.
8884 2018-02-05 Martin Liska <mliska@suse.cz>
8886 PR gcov-profile/84137
8887 * doc/gcov.texi: Fix typo in documentation.
8889 2018-02-05 Martin Liska <mliska@suse.cz>
8891 PR gcov-profile/83879
8892 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8893 using dlopen functionality.
8895 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8897 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8898 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8899 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8900 _mm_maskz_range_round_ss): New intrinsics.
8901 (__builtin_ia32_rangesd128_round)
8902 (__builtin_ia32_rangess128_round): Remove.
8903 (__builtin_ia32_rangesd128_mask_round,
8904 __builtin_ia32_rangess128_mask_round): New builtins.
8905 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8906 __builtin_ia32_rangess128_round): Remove.
8907 (__builtin_ia32_rangesd128_mask_round,
8908 __builtin_ia32_rangess128_mask_round): New builtins.
8909 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8910 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8911 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8912 "<round_saeonly_constraint>")): Changed to ...
8913 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8914 "<round_saeonly_scalar_constraint>")): ... this.
8915 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8916 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8917 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8918 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8919 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8921 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8923 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8925 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8926 Remove all values except native, 8540 and 8548.
8928 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8930 * config/i386/i386.c (ix86_output_function_return): Pass
8931 INVALID_REGNUM, instead of -1, as invalid register number to
8932 indirect_thunk_name and output_indirect_thunk.
8934 2018-02-02 Julia Koval <julia.koval@intel.com>
8936 * config.gcc: Add -march=icelake.
8937 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8938 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8939 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8940 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8941 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8942 (processor_target_table): Add icelake.
8943 (ix86_option_override_internal): Handle new PTAs.
8944 (get_builtin_code_for_version): Handle icelake.
8945 (M_INTEL_COREI7_ICELAKE): New.
8946 (fold_builtin_cpu): Handle icelake.
8947 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8948 * doc/invoke.texi: Add -march=icelake.
8950 2018-02-02 Julia Koval <julia.koval@intel.com>
8952 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8953 to wide_int_bitmask.
8954 * wide-int-bitmask.h: New.
8956 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8959 * config/i386/i386.md: Replace Pmode with word_mode in
8960 builtin_setjmp_setup and builtin_longjmp to support x32.
8962 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8966 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8968 (rs6000_supported_cpu_names): New static variable.
8969 (linux_cpu_translation_table): Likewise.
8970 (elf_platform) <cpu>: Define new static variable and use it.
8971 Translate kernel AT_PLATFORM name to canonical name if needed.
8972 Error if platform name is unknown.
8974 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8977 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8979 2018-02-01 Jeff Law <law@redhat.com>
8982 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8983 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8984 the scratch if RELEASE_VIA_POP is false.
8985 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8986 If we have to save a temporary register, decrement SIZE appropriately.
8987 Pass new arguments to release_scratch_register_on_entry.
8988 (ix86_adjust_stack_and_probe): Likewise.
8989 (ix86_emit_probe_stack_range): Pass new arguments to
8990 release_scratch_register_on_entry.
8992 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8994 PR rtl-optimization/84157
8995 * combine.c (change_zero_ext): Use REG_P predicate in
8996 front of HARD_REGISTER_P predicate.
8998 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9000 * config/avr/avr.c (avr_option_override): Move disabling of
9001 -fdelete-null-pointer-checks to...
9002 * common/config/avr/avr-common.c (avr_option_optimization_table):
9005 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9007 PR tree-optimization/81635
9008 * tree-data-ref.c (split_constant_offset_1): For types that
9009 wrap on overflow, try to use range info to prove that wrapping
9012 2018-02-01 Renlin Li <renlin.li@arm.com>
9015 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9017 (aarch64_register_move_cost): Likewise.
9018 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9020 (REG_CLASS_NAMES): Likewise.
9021 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9022 TAILCALL_ADDR_REGS. Remove IP registers.
9023 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9025 2018-02-01 Richard Biener <rguenther@suse.de>
9027 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9028 for specifying RPO order and allow NULL for that.
9029 * domwalk.c (dom_walker::dom_walker): Likewise.
9030 (dom_walker::walk): Handle NULL RPO order.
9031 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9033 (rewrite_update_dom_walker): Likewise.
9034 (mark_def_dom_walker): Likewise.
9036 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9038 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9039 (aarch64_maybe_expand_sve_subreg_move): Declare.
9040 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9041 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9043 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9044 that are semantically a reverse operation.
9045 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9046 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9047 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9049 (aarch64_can_change_mode_class): For big-endian, forbid changes
9050 between two SVE modes if they have different element sizes.
9052 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9054 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9055 the TImode handling for big-endian targets.
9057 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9059 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9060 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9062 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9063 Remove BSWAP handing for big-endian targets and use the form of
9064 LD1RQ appropariate for the mode.
9066 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9068 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9069 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9072 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9075 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9076 check for operands that need to go through aarch64_sve_reload_be.
9078 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9080 PR tree-optimization/81661
9081 PR tree-optimization/84117
9082 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9083 * tree-eh.c: Include gimplify.h.
9084 (find_trapping_overflow, replace_trapping_overflow,
9085 rewrite_to_non_trapping_overflow): New functions.
9086 * tree-vect-loop.c: Include tree-eh.h.
9087 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9088 * tree-data-ref.c: Include tree-eh.h.
9089 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9091 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9093 PR rtl-optimization/84123
9094 * combine.c (change_zero_ext): Check if hard register satisfies
9095 can_change_dest_mode before calling gen_lowpart_SUBREG.
9097 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9100 * ira.c (ira_init_register_move_cost): Remove assert.
9102 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9104 PR rtl-optimization/84071
9105 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9106 * doc/tm.texi: Regenerate.
9108 2018-01-31 Richard Biener <rguenther@suse.de>
9110 PR tree-optimization/84132
9111 * tree-data-ref.c (analyze_miv_subscript): Properly
9112 check whether evolution_function_is_affine_multivariate_p
9113 before calling gcd_of_steps_may_divide_p.
9115 2018-01-31 Julia Koval <julia.koval@intel.com>
9118 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9119 * config/i386/i386.md (rdpid_rex64) New.
9120 (rdpid): Make 32bit only.
9122 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9125 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9126 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9128 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9131 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9133 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9135 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9137 PR rtl-optimization/84071
9138 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9139 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9141 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9143 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9144 (arc_attribute_table): Add 'aux' attribute.
9145 (arc_in_small_data_p): Consider aux like variables.
9146 (arc_is_aux_reg_p): New function.
9147 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9148 (arc_get_aux_arg): New function.
9149 (prepare_move_operands): Handle aux-register access.
9150 (arc_handle_aux_attribute): New function.
9151 * doc/extend.texi (ARC Variable attributes): Add subsection.
9153 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9155 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9156 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9157 (arc_attribute_table): Add 'uncached' attribute.
9158 (arc_print_operand): Print '.di' flag for uncached memory
9160 (arc_in_small_data_p): Do not consider for small data the uncached
9162 (arc_is_uncached_mem_p): New function.
9163 * config/arc/predicates.md (compact_store_memory_operand): Check
9164 for uncached memory accesses.
9165 (nonvol_nonimm_operand): Likewise.
9166 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9168 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9171 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9172 falign-loops=): Add Optimization flag.
9174 2018-01-30 Jeff Law <law@redhat.com>
9177 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9178 INT_REGISTERS_SAVED. Check it prior to calling
9179 get_scratch_register_on_entry.
9180 (ix86_adjust_stack_and_probe): Similarly.
9181 (ix86_emit_probe_stack_range): Similarly.
9182 (ix86_expand_prologue): Corresponding changes.
9184 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9187 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9188 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9190 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9193 * lra-constraints.c (curr_insn_transform): Process AND in the
9196 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9198 PR rtl-optimization/83986
9199 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9200 dependence against last_pending_memory_flush in addition to
9203 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9205 PR tree-optimization/81611
9206 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9209 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9212 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9215 2018-01-30 Richard Biener <rguenther@suse.de>
9216 Jakub Jelinek <jakub@redhat.com>
9218 PR tree-optimization/84111
9219 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9220 inner loops added during recursion, as they don't have up-to-date
9223 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9226 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9227 (can_inline_edge_by_limits_p): ... here.
9228 (can_early_inline_edge_p, check_callers,
9229 update_caller_keys, update_callee_keys, recursive_inlining,
9230 add_new_edges_to_heap, speculation_useful_p,
9231 inline_small_functions,
9232 inline_small_functions, flatten_function,
9233 inline_to_all_callers_1): Update.
9235 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9237 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9240 2018-01-30 Richard Biener <rguenther@suse.de>
9242 PR tree-optimization/83008
9243 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9244 invariant and constant vector uses in stmts when they need
9247 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9250 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9251 * configure: Regenerate.
9253 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9255 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9257 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9258 Use gen_rtx_REG rather than gen_lowpart.
9260 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9262 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9263 rather than 0 when creating partial subregs.
9265 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9267 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9270 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9273 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9274 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9275 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9276 flags. This restores the settings used before the 2017-07-24.
9277 Turning off pre increment/decrement/modify allows IVOPTS to
9278 optimize DF/SF loops where the index is an int.
9280 2018-01-29 Richard Biener <rguenther@suse.de>
9281 Kelvin Nilsen <kelvin@gcc.gnu.org>
9284 * tree-vect-stmts.c (vectorizable_call): Don't call
9285 targetm.vectorize_builtin_md_vectorized_function if callee is
9288 2018-01-22 Carl Love <cel@us.ibm.com>
9290 * doc/extend.tex: Fix typo in second arg in
9291 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9293 2018-01-29 Richard Biener <rguenther@suse.de>
9295 PR tree-optimization/84086
9296 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9297 (flush_ssaname_freelist): When SSA names were released reset
9298 the SCEV hash table.
9300 2018-01-29 Richard Biener <rguenther@suse.de>
9302 PR tree-optimization/84057
9303 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9304 removed paths when removing edges.
9306 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9308 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9309 -mfunction-return=@var{choice}.
9311 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9314 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9316 (layout::print_source_line): Likewise.
9317 (test_get_line_width_without_trailing_whitespace): Add test cases.
9319 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9322 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9325 2018-01-26 Jim Wilson <jimw@sifive.com>
9327 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9329 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9332 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9334 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9335 and CMP + SUB-immediate -> SUBS.
9337 2018-01-26 Martin Sebor <msebor@redhat.com>
9339 PR tree-optimization/83896
9340 * tree-ssa-strlen.c (get_string_len): Rename...
9341 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9342 Avoid assuming length is constant.
9343 (handle_char_store): Use HOST_WIDE_INT for string length.
9345 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9348 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9349 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9351 2018-01-26 Richard Biener <rguenther@suse.de>
9353 PR rtl-optimization/84003
9354 * dse.c (record_store): Only record redundant stores when
9355 the earlier store aliases at least all accesses the later one does.
9357 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9359 PR rtl-optimization/83985
9360 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9361 REG_CFA_RESTORE insns.
9362 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9363 REG_CFA_RESTORE insns here.
9366 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9367 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9369 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9371 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9372 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9373 (arc_init): Likewise.
9374 (arc_override_options): Likewise.
9375 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9377 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9379 * config/arc/arc.h (TARGET_DBNZ): Define.
9380 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9381 properly set the tune attribute.
9382 (dbnz): Use TARGET_DBNZ guard.
9383 * config/arc/arc.opt (mtune): Add core3 option.
9385 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9387 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9388 recognize new pic like addresses.
9389 (arc_delegitimize_address): Clean up.
9391 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9393 * config/arc/arc-arches.def: Option mrf16 valid for all
9395 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9396 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9397 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9398 * config/arc/arc-tables.opt: Regenerate.
9399 * config/arc/arc.c (arc_conditional_register_usage): Handle
9400 reduced register file case.
9401 (arc_file_start): Set must have build attributes.
9402 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9404 * config/arc/arc.opt (mrf16): Add new option.
9405 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9406 * config/arc/genmultilib.awk: Handle new mrf16 option.
9407 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9408 * config/arc/t-multilib: Regenerate.
9409 * doc/invoke.texi (ARC Options): Document mrf16 option.
9411 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9413 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9414 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9415 (arc_attribute_table): Add 'secure_call' attribute.
9416 (arc_print_operand): Print secure call operand.
9417 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9419 (arc_is_secure_call_p): New function. * config/arc/arc.md
9420 (call_i): Add support for sjli instruction.
9421 (call_value_i): Likewise.
9422 * config/arc/constraints.md (Csc): New constraint.
9424 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9425 John Eric Martin <John.Martin@emmicro-us.com>
9427 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9428 * config/arc/arc.c (_arc_jli_section): New struct.
9429 (arc_jli_section): New type.
9430 (rc_jli_sections): New static variable.
9431 (arc_handle_jli_attribute): New function.
9432 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9433 (arc_file_end): New function.
9434 (TARGET_ASM_FILE_END): Define.
9435 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9436 (arc_add_jli_section): New function.
9437 (jli_call_scan): Likewise.
9438 (arc_reorg): Call jli_call_scan.
9439 (arc_output_addsi): Remove 'S' from printing asm operand.
9440 (arc_is_jli_call_p): New function.
9441 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9443 (movhi_insn): Likewise.
9444 (movsi_insn): Likewise.
9445 (movsi_set_cc_insn): Likewise.
9446 (loadqi_update): Likewise.
9447 (load_zeroextendqisi_update): Likewise.
9448 (load_signextendqisi_update): Likewise.
9449 (loadhi_update): Likewise.
9450 (load_zeroextendhisi_update): Likewise.
9451 (load_signextendhisi_update): Likewise.
9452 (loadsi_update): Likewise.
9453 (loadsf_update): Likewise.
9454 (movsicc_insn): Likewise.
9455 (bset_insn): Likewise.
9456 (bxor_insn): Likewise.
9457 (bclr_insn): Likewise.
9458 (bmsk_insn): Likewise.
9459 (bicsi3_insn): Likewise.
9460 (cmpsi_cc_c_insn): Likewise.
9461 (movsi_ne): Likewise.
9462 (movsi_cond_exec): Likewise.
9463 (clrsbsi2): Likewise.
9474 (call_i): Remove 'S' asm letter, add jli instruction.
9475 (call_value_i): Likewise.
9476 * config/arc/arc.op (mjli-always): New option.
9477 * config/arc/constraints.md (Cji): New constraint.
9478 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9480 (subsf3_fpx): Likewise.
9481 (mulsf3_fpx): Likewise.
9482 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9484 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9486 * doc/invoke.texi (ARC): Document mjli-always option.
9488 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9490 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9491 avoid addition with 0 and use incw and decw where possible.
9493 2018-01-26 Richard Biener <rguenther@suse.de>
9495 PR tree-optimization/81082
9496 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9497 association if it requires casting to unsigned.
9498 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9499 from fold_plusminus_mult_expr to catch important cases late when
9500 range info is available.
9502 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9504 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9505 * configure.ac (hidden_linkonce): New test.
9506 * configure: Regenerate.
9507 * config.in: Regenerate.
9509 2018-01-26 Julia Koval <julia.koval@intel.com>
9511 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9512 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9513 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9514 _mm_mask_bitshuffle_epi64_mask): Fix type.
9515 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9516 USI_FTYPE_V4DI_V4DI_USI): Remove.
9517 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9518 __builtin_ia32_vpshufbitqmb256_mask,
9519 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9520 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9521 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9523 2018-01-26 Alan Modra <amodra@gmail.com>
9526 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9527 UNSPEC_VBPERMQ. Sort other unspecs.
9529 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9531 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9533 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9536 * predict.c (drop_profile): Do not push/pop cfun; update also
9538 (handle_missing_profiles): Fix logic looking for zero profiles.
9540 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9543 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9544 on functions with #pragma omp declare simd or functions with simd
9546 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9547 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9548 Remove trailing \n from warning_at calls.
9550 2018-01-25 Tom de Vries <tom@codesourcery.com>
9553 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9554 for neutered workers.
9556 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9559 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9560 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9562 2018-01-24 Jeff Law <law@redhat.com>
9565 * i386.c (get_probe_interval): Move to earlier point.
9566 (ix86_compute_frame_layout): If -fstack-clash-protection and
9567 the frame is larger than the probe interval, then use pushes
9568 to save registers rather than reg->mem moves.
9569 (ix86_expand_prologue): Remove conditional for int_registers_saved
9572 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9575 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9576 min/max for never referenced object.
9578 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9581 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9583 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9584 attributes from DECL_ATTRIBUTES (decl) without affecting
9585 DECL_ATTRIBUTES (current_function_decl).
9586 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9587 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9589 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9591 PR tree-optimization/83979
9592 * fold-const.c (fold_comparison): Use constant_boolean_node
9593 instead of boolean_{true,false}_node.
9595 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9597 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9600 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9602 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9603 Simplify the clause that sets the length attribute.
9604 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9605 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9606 clause that sets the length attribute.
9607 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9609 2018-01-24 Tom de Vries <tom@codesourcery.com>
9612 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9613 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9614 Add strict parameter.
9615 (prevent_branch_around_nothing): Insert dummy insn between branch to
9616 label and label with no ptx insn inbetween.
9617 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9619 2018-01-24 Tom de Vries <tom@codesourcery.com>
9622 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9623 for neutered threads in warp.
9624 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9626 2018-01-24 Richard Biener <rguenther@suse.de>
9628 PR tree-optimization/83176
9629 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9632 2018-01-24 Richard Biener <rguenther@suse.de>
9634 PR tree-optimization/82819
9635 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9636 code generating pluses that are no-ops in the target precision.
9638 2018-01-24 Richard Biener <rguenther@suse.de>
9641 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9643 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9645 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9646 to merge probabilities.
9647 * predict.c (probably_never_executed): Also mark as cold functions
9648 with global 0 profile and guessed local profile.
9649 * profile-count.c (profile_probability::combine_with_count): New
9651 * profile-count.h (profile_probability::operator*,
9652 profile_probability::operator*=, profile_probability::operator/,
9653 profile_probability::operator/=): Reduce precision to adjusted
9654 and set value to guessed on contradictory divisions.
9655 (profile_probability::combine_with_freq): Remove.
9656 (profile_probability::combine_wiht_count): Declare.
9657 (profile_count::force_nonzero):: Set to adjusted.
9658 (profile_count::probability_in):: Set quality to adjusted.
9659 * tree-ssa-tail-merge.c (replace_block_by): Use
9662 2018-01-23 Andrew Waterman <andrew@sifive.com>
9663 Jim Wilson <jimw@sifive.com>
9665 * config/riscv/riscv.c (riscv_stack_boundary): New.
9666 (riscv_option_override): Set riscv_stack_boundary. Handle
9667 riscv_preferred_stack_boundary_arg.
9668 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9669 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9670 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9671 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9672 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9673 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9675 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9678 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9679 of struct ix86_frame.
9680 (ix86_expand_epilogue): Likewise. Add a local variable for
9681 the reg_save_offset field in struct ix86_frame.
9683 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9685 PR tree-optimization/82604
9686 * tree-loop-distribution.c (enum partition_kind): New enum item
9687 PKIND_PARTIAL_MEMSET.
9688 (partition_builtin_p): Support above new enum item.
9689 (generate_code_for_partition): Ditto.
9690 (compute_access_range): Differentiate cases that equality can be
9691 proven at all loops, the innermost loops or no loops.
9692 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9693 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9694 (finalize_partitions, distribute_loop): Don't fuse partition of
9695 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9696 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9699 2018-01-23 Martin Liska <mliska@suse.cz>
9701 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9702 order to ignore the predictor.
9703 (PRED_POLYMORPHIC_CALL): Likewise.
9704 (PRED_RECURSIVE_CALL): Likewise.
9706 2018-01-23 Martin Liska <mliska@suse.cz>
9708 * tree-profile.c (tree_profiling): Print function header to
9709 aware reader which function we are working on.
9710 * value-prof.c (gimple_find_values_to_profile): Do not print
9711 not interesting value histograms.
9713 2018-01-23 Martin Liska <mliska@suse.cz>
9715 * profile-count.h (enum profile_quality): Add
9716 profile_uninitialized as the first value. Do not number values
9717 as they are zero based.
9718 (profile_count::verify): Update sanity check.
9719 (profile_probability::verify): Likewise.
9721 2018-01-23 Nathan Sidwell <nathan@acm.org>
9723 * doc/invoke.texi (ffor-scope): Deprecate.
9725 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9727 PR tree-optimization/83510
9728 * domwalk.c (set_all_edges_as_executable): New function.
9729 (dom_walker::dom_walker): Convert bool param
9730 "skip_unreachable_blocks" to enum reachability. Move setup of
9731 edge flags to set_all_edges_as_executable and only do it when
9732 reachability is REACHABLE_BLOCKS.
9733 * domwalk.h (enum dom_walker::reachability): New enum.
9734 (dom_walker::dom_walker): Convert bool param
9735 "skip_unreachable_blocks" to enum reachability.
9736 (set_all_edges_as_executable): New decl.
9737 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9738 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9740 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9741 but converting true to REACHABLE_BLOCKS.
9742 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9744 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9745 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9746 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9748 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9749 if check_all_array_refs will be called.
9751 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9753 * tree.c (selftest::test_location_wrappers): Add more test
9756 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9758 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9759 (selftest::test_bit_in_range): Likewise.
9761 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9764 * doc/sourcebuild.texi (vect_float): Say that the selector
9765 only describes the situation when -funsafe-math-optimizations is on.
9766 (vect_float_strict): Document.
9768 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9770 PR tree-optimization/83965
9771 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9772 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9773 instead of checking only for a reduction.
9774 (vect_recog_widen_sum_pattern): Likewise.
9776 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9778 * predict.c (probably_never_executed): Only use precise profile info.
9779 (compute_function_frequency): Skip after inlining hack since we now
9780 have quality checking.
9782 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9784 * profile-count.h (profile_probability::very_unlikely,
9785 profile_probability::unlikely, profile_probability::even): Set
9786 precision to guessed.
9788 2018-01-23 Richard Biener <rguenther@suse.de>
9790 PR tree-optimization/83963
9791 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9792 Properly terminate dominator walk when crossing the exit edge not
9793 when visiting its source block.
9795 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9798 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9799 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9801 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9803 PR tree-optimization/83957
9804 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9805 semicolon after for body surrounded by braces.
9807 PR tree-optimization/83081
9808 * profile-count.h (profile_probability::split): New method.
9809 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9810 Use profile_probability::split.
9811 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9812 when splitting a single conditional jump into 2.
9814 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9816 PR tree-optimization/69452
9817 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9820 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9822 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9823 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9824 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9826 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9828 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9829 * config/rl78/rl78.md (movdi): New define_expand.
9830 * config/rl78/rl78.c (rl78_split_movdi): New function.
9832 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9835 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9837 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9838 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9839 128-bit to produce an UNSPEC move to get the double word with the
9840 signbit and then a shift directly to do signbit.
9841 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9842 implementation with a new version that just does either a direct
9843 move or a regular move. Move memory interface to separate insns.
9844 Move insns so they are next to the expander.
9845 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9846 with signbit move. Split big and little endian case.
9847 (signbit<mode>2_dm_mem_le): Likewise.
9848 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9849 (signbit<mode>2_dm2): Likewise.
9851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9853 * config/rl78/rl78.md (anddi3): New define_expand.
9855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9857 * config/rl78/rl78.md (umindi3): New define_expand.
9859 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9861 * config/rl78/rl78.md (smindi3): New define_expand.
9863 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9865 * config/rl78/rl78.md (smaxdi3): New define_expand.
9867 2018-01-22 Carl Love <cel@us.ibm.com>
9869 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9870 LVX_V1TI): Add macro expansion.
9871 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9872 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9873 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9874 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9875 Change check to determine if the instruction is a byte reversing
9876 entry. Fix typo in comment.
9877 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9878 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9879 Add def_builtin calls for new builtins.
9880 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9881 Add define_insn expansion.
9883 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9885 * config/rl78/rl78.md (umaxdi3): New define_expand.
9887 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9889 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9890 for non-QImode registers.
9892 2018-01-22 Richard Biener <rguenther@suse.de>
9894 PR tree-optimization/83963
9895 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9896 including the loop exit block.
9897 (scop_detection::merge_sese): Likewise.
9898 (scop_detection::add_scop): Do it here instead.
9900 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9902 * doc/sourcebuild.texi (arm_softfloat): Document.
9904 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9907 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9908 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9909 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9911 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9912 David Edelsohn <dje.gcc@gmail.com>
9915 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9916 Change "crset eq" to "crset 2".
9917 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9918 (*call_indirect_aix<mode>_nospec): Likewise.
9919 (*call_value_indirect_aix<mode>_nospec): Likewise.
9920 (*call_indirect_elfv2<mode>_nospec): Likewise.
9921 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9922 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9923 change assembly output from . to $.
9924 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9925 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9926 (*tablejump<mode>_internal1_nospec): Likewise.
9928 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9931 * config/sh/sh_optimize_sett_clrt.cc:
9932 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9934 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9936 PR tree-optimization/83940
9937 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9938 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9939 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9940 use it to pass back the definition type.
9941 (vect_check_store_rhs): Likewise rhs_dt_out.
9942 (vect_build_gather_load_calls): Add a mask_dt argument and use
9943 it instead of a call to vect_is_simple_use.
9944 (vectorizable_store): Update calls to vect_check_load_store_mask
9945 and vect_check_store_rhs. Use the dt returned by the latter instead
9946 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9947 instead of calls to vect_is_simple_use. Pass the scalar rather
9948 than the vector operand to vect_is_simple_use when handling
9949 second and subsequent copies of an rhs value.
9950 (vectorizable_load): Update calls to vect_check_load_store_mask
9951 and vect_build_gather_load_calls. Use the cached mask_dt and
9952 gs_info.offset_dt instead of calls to vect_is_simple_use.
9954 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9957 * tree-emutls.c: Include gimplify.h.
9958 (lower_emutls_2): New function.
9959 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9960 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9961 it before further processing.
9964 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9965 UINTVAL (trueop1) instead of INTVAL (op1).
9967 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9971 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9972 INCOMING_FRAME_SP_OFFSET if not defined.
9973 (scan_trace): Add ENTRY argument. If true and
9974 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9975 emit a note to adjust the CFA offset.
9976 (create_cfi_notes): Adjust scan_trace callers.
9977 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9978 INCOMING_FRAME_SP_OFFSET in the CIE.
9979 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9980 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9982 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9983 * doc/tm.texi: Regenerated.
9985 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9987 PR rtl-optimization/83147
9988 * lra-constraints.c (remove_inheritance_pseudos): Use
9989 lra_substitute_pseudo_within_insn.
9991 2018-01-19 Tom de Vries <tom@codesourcery.com>
9992 Cesar Philippidis <cesar@codesourcery.com>
9995 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9997 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10000 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10001 spaces for function labels.
10003 2018-01-19 Martin Liska <mliska@suse.cz>
10005 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10006 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10007 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10008 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10009 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10010 (PRED_CONST_RETURN): Change from 69 to 65.
10011 (PRED_NULL_RETURN): Change from 91 to 71.
10012 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10013 (PRED_LOOP_GUARD): Change from 66 to 73.
10015 2018-01-19 Martin Liska <mliska@suse.cz>
10017 * predict.c (predict_insn_def): Add new assert.
10018 (struct branch_predictor): Change type to signed integer.
10019 (test_prediction_value_range): Amend test to cover
10020 PROB_UNINITIALIZED.
10021 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10022 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10023 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10024 (PRED_LOOP_IV_COMPARE): Likewise.
10025 * predict.h (PROB_UNINITIALIZED): Define new constant.
10027 2018-01-19 Martin Liska <mliska@suse.cz>
10029 * predict.c (dump_prediction): Add new format for
10030 analyze_brprob.py script which is enabled with -details
10032 * profile-count.h (precise_p): New function.
10034 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10036 PR tree-optimization/83922
10037 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10038 there are no statements that need masking.
10039 (vect_active_double_reduction_p): New function.
10040 (vect_analyze_loop_operations): Use it when handling phis that
10041 are not in the loop header.
10043 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10045 PR tree-optimization/83914
10046 * tree-vect-loop.c (vectorizable_induction): Don't convert
10047 init_expr or apply the peeling adjustment for inductions
10048 that are nested within the vectorized loop.
10050 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10052 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10055 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10059 * function.h (gimplify_parameters): Add gimple_seq * argument.
10060 * function.c: Include gimple.h and options.h.
10061 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10062 for the added local temporaries if needed.
10063 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10064 if there are any parameter cleanups, wrap whole body into a
10065 try/finally with the cleanups.
10067 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10070 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10071 Use GET_MODE_CLASS for scalar floating point.
10073 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10077 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10078 Fix call of call_cgraph_insertion_hooks.
10080 2018-01-18 Martin Sebor <msebor@redhat.com>
10082 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10084 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10087 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10090 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10093 * common.opt: (-ffile-prefix-map): New option.
10094 * opts.c (common_handle_option): Defer it.
10095 * opts-global.c (handle_common_deferred_options): Handle it.
10096 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10097 * file-prefix-map.h: New file.
10098 (remap_debug_filename, add_debug_prefix_map): ...here.
10099 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10100 * final.c (debug_prefix_map, add_debug_prefix_map
10101 remap_debug_filename): Move to...
10102 * file-prefix-map.c: New file.
10103 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10104 generalize, get rid of alloca(), use strrchr() instead of strchr().
10105 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10106 Implement in terms of add_prefix_map().
10107 (remap_macro_filename, remap_debug_filename): Implement in term of
10109 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10110 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10111 * dbxout.c: Include file-prefix-map.h.
10112 * varasm.c: Likewise.
10113 * vmsdbgout.c: Likewise.
10114 * xcoffout.c: Likewise.
10115 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10116 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10117 * doc/invoke.texi (-ffile-prefix-map): Document.
10118 (-fdebug-prefix-map): Update description.
10120 2018-01-18 Martin Liska <mliska@suse.cz>
10122 * config/i386/i386.c (indirect_thunk_name): Document that also
10124 (output_indirect_thunk): Document why both instructions
10125 (pause and lfence) are generated.
10127 2018-01-18 Richard Biener <rguenther@suse.de>
10129 PR tree-optimization/83887
10130 * graphite-scop-detection.c
10131 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10132 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10133 (scop_detection::merge_sese): Re-implement with a flood-fill
10134 algorithm that properly finds a SESE region if it exists.
10136 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10139 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10140 pointer_diff optimizations use view_convert instead of convert.
10142 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10144 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10145 Generate different code for -mno-speculate-indirect-jumps.
10146 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10147 (*call_indirect_aix<mode>): Disable for
10148 -mno-speculate-indirect-jumps.
10149 (*call_indirect_aix<mode>_nospec): New define_insn.
10150 (*call_value_indirect_aix<mode>): Disable for
10151 -mno-speculate-indirect-jumps.
10152 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10153 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10154 -mno-speculate-indirect-jumps.
10155 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10157 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10159 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10160 long double type, set the flags for noting the default long double
10161 type, even if we don't pass or return a long double type.
10163 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10166 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10169 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10171 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10172 support for merge[hl].
10173 (fold_mergehl_helper): New helper function.
10174 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10175 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10176 (altivec_vmrglw_direct): Add xxmrglw insn.
10178 2018-01-17 Andrew Waterman <andrew@sifive.com>
10180 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10181 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10183 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10186 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10187 call the lto_location_cache before reading the
10188 DECL_SOURCE_LOCATION of the types.
10190 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10191 Richard Sandiford <richard.sandiford@linaro.org>
10193 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10194 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10195 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10196 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10197 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10199 * config/aarch64/constraints.md (aarch64_movti_operand):
10201 * config/aarch64/predicates.md (Uti): Add new constraint.
10203 2018-01-17 Carl Love <cel@us.ibm.com>
10205 * config/rs6000/vsx.md (define_expand xl_len_r,
10206 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10207 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10209 (define_expand, define_insn): Move the shift left from the
10210 define_insn to the define_expand for lxvl and stxvl instructions.
10211 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10212 and XL_LEN_R definitions to PURE.
10214 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10216 * config/i386/i386.c (indirect_thunk_name): Declare regno
10217 as unsigned int. Compare regno with INVALID_REGNUM.
10218 (output_indirect_thunk): Ditto.
10219 (output_indirect_thunk_function): Ditto.
10220 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10221 in the call to output_indirect_thunk_function.
10223 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10225 PR middle-end/83884
10226 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10227 rather than the size of inner_type to determine the stack slot size
10228 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10230 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10233 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10236 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10238 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10239 endian Linux systems to optionally enable multilibs for selecting
10240 the long double type if the user configured an explicit type.
10241 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10242 have no long double multilibs if not defined.
10243 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10244 warn if the user used -mabi={ieee,ibm}longdouble and we built
10245 multilibs for long double.
10246 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10247 appropriate multilib option.
10248 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10250 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10251 for building long double multilibs.
10252 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10254 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10256 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10259 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10261 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10264 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10267 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10270 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10272 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10273 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10275 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10277 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10278 different rtl trees depending on TARGET_64BIT.
10279 (rs6000_gen_lvx): Likewise.
10281 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10283 * config/visium/visium.md (nop): Tweak comment.
10284 (hazard_nop): Likewise.
10286 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10288 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10289 -mspeculate-indirect-jumps.
10290 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10291 for -mno-speculate-indirect-jumps.
10292 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10293 (*call_value_indirect_elfv2<mode>): Disable for
10294 -mno-speculate-indirect-jumps.
10295 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10296 (indirect_jump): Emit different RTL for
10297 -mno-speculate-indirect-jumps.
10298 (*indirect_jump<mode>): Disable for
10299 -mno-speculate-indirect-jumps.
10300 (*indirect_jump<mode>_nospec): New define_insn.
10301 (tablejump): Emit different RTL for
10302 -mno-speculate-indirect-jumps.
10303 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10304 (tablejumpsi_nospec): New define_expand.
10305 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10306 (tablejumpdi_nospec): New define_expand.
10307 (*tablejump<mode>_internal1): Disable for
10308 -mno-speculate-indirect-jumps.
10309 (*tablejump<mode>_internal1_nospec): New define_insn.
10310 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10313 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10315 * caller-save.c (insert_save): Drop unnecessary parameter. All
10318 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10319 Richard Biener <rguenth@suse.de>
10322 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10323 return early, inline manually is_gimple_sizepos. Make sure if we
10324 call gimplify_expr we don't end up with a gimple constant.
10325 * tree.c (variably_modified_type_p): Don't return true for
10326 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10327 * gimplify.h (is_gimple_sizepos): Remove.
10329 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10331 PR tree-optimization/83857
10332 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10333 vectorizable_live_operation for pure SLP statements.
10334 (vectorizable_live_operation): Handle PHIs.
10336 2018-01-16 Richard Biener <rguenther@suse.de>
10338 PR tree-optimization/83867
10339 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10340 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10342 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10345 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10346 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10347 If off is not INTEGER_CST, issue a may not be aligned warning
10348 rather than isn't aligned. Use isn%'t rather than isn't.
10349 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10351 <case MULT_EXPR>: Improve the case when bottom and one of the
10352 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10353 operand, in that case check if the other operand is multiple of
10354 bottom divided by the INTEGER_CST operand.
10356 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10359 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10360 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10361 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10362 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10363 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10364 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10365 (pa_function_arg_size): New function.
10367 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10369 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10370 in a separate statement.
10372 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10374 PR tree-optimization/83847
10375 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10376 group gathers and scatters.
10378 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10380 PR rtl-optimization/86620
10381 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10383 PR rtl-optimization/83213
10384 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10385 to last if both are JUMP_INSNs.
10387 PR tree-optimization/83843
10388 * gimple-ssa-store-merging.c
10389 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10390 store_immediate_info for bswap/nop orig_stores.
10392 2018-01-15 Andrew Waterman <andrew@sifive.com>
10394 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10396 <UDIV>: Increase cost if !TARGET_DIV.
10398 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10400 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10401 (define_attr "cr_logical_3op"): New.
10402 (cceq_ior_compare): Adjust.
10403 (cceq_ior_compare_complement): Adjust.
10404 (*cceq_rev_compare): Adjust.
10405 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10406 (is_cracked_insn): Adjust.
10407 (insn_must_be_first_in_group): Adjust.
10408 * config/rs6000/40x.md: Adjust.
10409 * config/rs6000/440.md: Adjust.
10410 * config/rs6000/476.md: Adjust.
10411 * config/rs6000/601.md: Adjust.
10412 * config/rs6000/603.md: Adjust.
10413 * config/rs6000/6xx.md: Adjust.
10414 * config/rs6000/7450.md: Adjust.
10415 * config/rs6000/7xx.md: Adjust.
10416 * config/rs6000/8540.md: Adjust.
10417 * config/rs6000/cell.md: Adjust.
10418 * config/rs6000/e300c2c3.md: Adjust.
10419 * config/rs6000/e500mc.md: Adjust.
10420 * config/rs6000/e500mc64.md: Adjust.
10421 * config/rs6000/e5500.md: Adjust.
10422 * config/rs6000/e6500.md: Adjust.
10423 * config/rs6000/mpc.md: Adjust.
10424 * config/rs6000/power4.md: Adjust.
10425 * config/rs6000/power5.md: Adjust.
10426 * config/rs6000/power6.md: Adjust.
10427 * config/rs6000/power7.md: Adjust.
10428 * config/rs6000/power8.md: Adjust.
10429 * config/rs6000/power9.md: Adjust.
10430 * config/rs6000/rs64.md: Adjust.
10431 * config/rs6000/titan.md: Adjust.
10433 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10435 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10436 ix86_indirect_branch_register logic.
10438 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10440 * config/i386/constraints.md (Bs): Update
10441 ix86_indirect_branch_register check. Don't check
10442 ix86_indirect_branch_register with GOT_memory_operand.
10444 * config/i386/predicates.md (GOT_memory_operand): Don't check
10445 ix86_indirect_branch_register here.
10446 (GOT32_symbol_operand): Likewise.
10448 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10450 * config/i386/predicates.md (constant_call_address_operand):
10451 Rewrite ix86_indirect_branch_register logic.
10452 (sibcall_insn_operand): Likewise.
10454 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10456 * config/i386/constraints.md (Bs): Replace
10457 ix86_indirect_branch_thunk_register with
10458 ix86_indirect_branch_register.
10460 * config/i386/i386.md (indirect_jump): Likewise.
10461 (tablejump): Likewise.
10462 (*sibcall_memory): Likewise.
10463 (*sibcall_value_memory): Likewise.
10464 Peepholes of indirect call and jump via memory: Likewise.
10465 * config/i386/i386.opt: Likewise.
10466 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10467 (GOT_memory_operand): Likewise.
10468 (call_insn_operand): Likewise.
10469 (sibcall_insn_operand): Likewise.
10470 (GOT32_symbol_operand): Likewise.
10472 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10474 PR middle-end/83837
10475 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10476 type rather than type addr's type points to.
10477 (expand_omp_atomic_mutex): Likewise.
10478 (expand_omp_atomic): Likewise.
10480 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10483 * config/i386/i386.c (output_indirect_thunk_function): Use
10484 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10485 for __x86_return_thunk.
10487 2018-01-15 Richard Biener <rguenther@suse.de>
10489 PR middle-end/83850
10490 * expmed.c (extract_bit_field_1): Fix typo.
10492 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10495 * config/arm/iterators.md (VF): New mode iterator.
10496 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10497 Remove integer-related logic from pattern.
10498 (neon_vabd<mode>_3): Likewise.
10500 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10502 PR middle-end/82694
10503 * common.opt (fstrict-overflow): No longer an alias.
10504 (fwrapv-pointer): New option.
10505 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10506 also for pointer types based on flag_wrapv_pointer.
10507 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10508 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10509 opts->x_flag_wrapv got set.
10510 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10511 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10512 POINTER_TYPE_OVERFLOW_UNDEFINED.
10513 * match.pd: Likewise in address comparison pattern.
10514 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10516 2018-01-15 Richard Biener <rguenther@suse.de>
10519 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10520 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10521 Reset type names to their identifier if their TYPE_DECL doesn't
10522 have linkage (and thus is used for ODR and devirt).
10523 (save_debug_info_for_decl): Remove.
10524 (save_debug_info_for_type): Likewise.
10525 (add_tree_to_fld_list): Adjust.
10526 * tree-pretty-print.c (dump_generic_node): Make dumping of
10527 type names more robust.
10529 2018-01-15 Richard Biener <rguenther@suse.de>
10531 * BASE-VER: Bump to 8.0.1.
10533 2018-01-14 Martin Sebor <msebor@redhat.com>
10536 * builtins.c (check_access): Avoid warning when the no-warning bit
10539 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10541 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10542 * ira-color (allocno_hard_regs_compare): Likewise.
10544 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10547 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10548 Use .pushsection/.popsection.
10550 2018-01-14 Martin Sebor <msebor@redhat.com>
10553 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10555 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10557 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10558 entry from extra_headers.
10559 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10560 extra_headers, make the list bitwise identical to the i?86-*-* one.
10562 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10564 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10565 -mcmodel=large with -mindirect-branch=thunk,
10566 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10567 -mfunction-return=thunk-extern.
10568 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10569 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10570 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10572 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10574 * config/i386/i386.c (print_reg): Print the name of the full
10575 integer register without '%'.
10576 (ix86_print_operand): Handle 'V'.
10577 * doc/extend.texi: Document 'V' modifier.
10579 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10581 * config/i386/constraints.md (Bs): Disallow memory operand for
10582 -mindirect-branch-register.
10584 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10585 (GOT_memory_operand): Likewise.
10586 (call_insn_operand): Likewise.
10587 (sibcall_insn_operand): Likewise.
10588 (GOT32_symbol_operand): Likewise.
10589 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10590 for -mindirect-branch-register.
10591 (tablejump): Likewise.
10592 (*sibcall_memory): Likewise.
10593 (*sibcall_value_memory): Likewise.
10594 Disallow peepholes of indirect call and jump via memory for
10595 -mindirect-branch-register.
10596 (*call_pop): Replace m with Bw.
10597 (*call_value_pop): Likewise.
10598 (*sibcall_pop_memory): Replace m with Bs.
10599 * config/i386/i386.opt (mindirect-branch-register): New option.
10600 * doc/invoke.texi: Document -mindirect-branch-register option.
10602 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10604 * config/i386/i386-protos.h (ix86_output_function_return): New.
10605 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10606 set function_return_type.
10607 (indirect_thunk_name): Add ret_p to indicate thunk for function
10609 (output_indirect_thunk_function): Pass false to
10610 indirect_thunk_name.
10611 (ix86_output_indirect_branch_via_reg): Likewise.
10612 (ix86_output_indirect_branch_via_push): Likewise.
10613 (output_indirect_thunk_function): Create alias for function
10614 return thunk if regno < 0.
10615 (ix86_output_function_return): New function.
10616 (ix86_handle_fndecl_attribute): Handle function_return.
10617 (ix86_attribute_table): Add function_return.
10618 * config/i386/i386.h (machine_function): Add
10619 function_return_type.
10620 * config/i386/i386.md (simple_return_internal): Use
10621 ix86_output_function_return.
10622 (simple_return_internal_long): Likewise.
10623 * config/i386/i386.opt (mfunction-return=): New option.
10624 (indirect_branch): Mention -mfunction-return=.
10625 * doc/extend.texi: Document function_return function attribute.
10626 * doc/invoke.texi: Document -mfunction-return= option.
10628 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10630 * config/i386/i386-opts.h (indirect_branch): New.
10631 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10632 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10633 with local indirect jump when converting indirect call and jump.
10634 (ix86_set_indirect_branch_type): New.
10635 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10636 (indirectlabelno): New.
10637 (indirect_thunk_needed): Likewise.
10638 (indirect_thunk_bnd_needed): Likewise.
10639 (indirect_thunks_used): Likewise.
10640 (indirect_thunks_bnd_used): Likewise.
10641 (INDIRECT_LABEL): Likewise.
10642 (indirect_thunk_name): Likewise.
10643 (output_indirect_thunk): Likewise.
10644 (output_indirect_thunk_function): Likewise.
10645 (ix86_output_indirect_branch_via_reg): Likewise.
10646 (ix86_output_indirect_branch_via_push): Likewise.
10647 (ix86_output_indirect_branch): Likewise.
10648 (ix86_output_indirect_jmp): Likewise.
10649 (ix86_code_end): Call output_indirect_thunk_function if needed.
10650 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10652 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10653 (ix86_attribute_table): Add indirect_branch.
10654 * config/i386/i386.h (machine_function): Add indirect_branch_type
10655 and has_local_indirect_jump.
10656 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10658 (tablejump): Likewise.
10659 (*indirect_jump): Use ix86_output_indirect_jmp.
10660 (*tablejump_1): Likewise.
10661 (simple_return_indirect_internal): Likewise.
10662 * config/i386/i386.opt (mindirect-branch=): New option.
10663 (indirect_branch): New.
10666 (thunk-inline): Likewise.
10667 (thunk-extern): Likewise.
10668 * doc/extend.texi: Document indirect_branch function attribute.
10669 * doc/invoke.texi: Document -mindirect-branch= option.
10671 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10674 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10676 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10678 * ipa-inline.c (want_inline_small_function_p): Return false if
10679 inlining has already failed with CIF_FINAL_ERROR.
10680 (update_caller_keys): Call want_inline_small_function_p before
10682 (update_callee_keys): Likewise.
10684 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10686 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10688 (rs6000_quadword_masked_address_p): Likewise.
10689 (quad_aligned_load_p): Likewise.
10690 (quad_aligned_store_p): Likewise.
10691 (const_load_sequence_p): Add comment to describe the outer-most loop.
10692 (mimic_memory_attributes_and_flags): New function.
10693 (rs6000_gen_stvx): Likewise.
10694 (replace_swapped_aligned_store): Likewise.
10695 (rs6000_gen_lvx): Likewise.
10696 (replace_swapped_aligned_load): Likewise.
10697 (replace_swapped_load_constant): Capitalize argument name in
10698 comment describing this function.
10699 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10700 and stores that access quad-word aligned addresses and replace
10701 with stvx or lvx instructions when appropriate.
10702 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10703 New function prototype.
10704 (rs6000_quadword_masked_address_p): Likewise.
10705 (rs6000_gen_lvx): Likewise.
10706 (rs6000_gen_stvx): Likewise.
10707 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10708 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10709 when memory address is aligned.
10710 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10711 this split to select lvx instruction when memory address is aligned.
10712 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10713 instruction when memory address is aligned.
10714 (*vsx_le_perm_load_v16qi): Likewise.
10715 (four unnamed splitters): Modify to select the stvx instruction
10716 when memory is aligned.
10718 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10720 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10721 which appears in the queue multiple times.
10723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10724 Alan Hayward <alan.hayward@arm.com>
10725 David Sherwood <david.sherwood@arm.com>
10727 * tree-vectorizer.h (vec_lower_bound): New structure.
10728 (_loop_vec_info): Add check_nonzero and lower_bounds.
10729 (LOOP_VINFO_CHECK_NONZERO): New macro.
10730 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10731 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10732 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10733 fields. Make seg_len the distance travelled, not including the
10735 (dr_direction_indicator): Declare.
10736 (dr_zero_step_indicator): Likewise.
10737 (dr_known_forward_stride_p): Likewise.
10738 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10740 (runtime_alias_check_p): Allow runtime alias checks with
10742 (operator ==): Compare access_size and align.
10743 (prune_runtime_alias_test_list): Rework for new distinction between
10744 the access_size and seg_len.
10745 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10747 (get_segment_min_max): New function.
10748 (create_intersect_range_checks): Use it.
10749 (dr_step_indicator): New function.
10750 (dr_direction_indicator): Likewise.
10751 (dr_zero_step_indicator): Likewise.
10752 (dr_known_forward_stride_p): Likewise.
10753 * tree-loop-distribution.c (data_ref_segment_size): Return
10754 DR_STEP * (niters - 1).
10755 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10757 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10758 (vect_preserves_scalar_order_p): New function, split out from...
10759 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10760 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10761 (vect_vfa_access_size): New function.
10762 (vect_vfa_align): Likewise.
10763 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10764 (dump_lower_bound): New function.
10765 (vect_check_lower_bound): Likewise.
10766 (vect_small_gap_p): Likewise.
10767 (vectorizable_with_step_bound_p): Likewise.
10768 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10769 depencies if the vectorization factor is 1. Convert the checks
10770 for nonzero steps into checks on the bounds of DR_STEP. Try using
10771 a bunds check for variable steps if the minimum required step is
10772 relatively small. Update calls to the dr_with_seg_len
10773 constructor and to vect_compile_time_alias.
10774 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10776 (vect_loop_versioning): Call it.
10777 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10779 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10782 Alan Hayward <alan.hayward@arm.com>
10783 David Sherwood <david.sherwood@arm.com>
10785 * doc/sourcebuild.texi (vect_scatter_store): Document.
10786 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10788 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10790 * genopinit.c (main): Add supports_vec_scatter_store and
10791 supports_vec_scatter_store_cached to target_optabs.
10792 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10793 IFN_MASK_SCATTER_STORE.
10794 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10796 * internal-fn.h (internal_store_fn_p): Declare.
10797 (internal_fn_stored_value_index): Likewise.
10798 * internal-fn.c (scatter_store_direct): New macro.
10799 (expand_scatter_store_optab_fn): New function.
10800 (direct_scatter_store_optab_supported_p): New macro.
10801 (internal_store_fn_p): New function.
10802 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10803 IFN_MASK_SCATTER_STORE.
10804 (internal_fn_mask_index): Likewise.
10805 (internal_fn_stored_value_index): New function.
10806 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10807 for scatter stores.
10808 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10809 * optabs-query.c (supports_vec_scatter_store_p): New function.
10810 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10811 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10812 true for scatter stores.
10813 (vect_gather_scatter_fn_p): Handle scatter stores too.
10814 (vect_check_gather_scatter): Consider using scatter stores if
10815 supports_vec_scatter_store_p.
10816 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10817 scatter stores too.
10818 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10819 internal_fn_stored_value_index.
10820 (check_load_store_masking): Handle scatter stores too.
10821 (vect_get_store_rhs): Make public.
10822 (vectorizable_call): Use internal_store_fn_p.
10823 (vectorizable_store): Handle scatter store internal functions.
10824 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10825 when deciding whether the end of the group has been reached.
10826 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10827 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10828 (mask_scatter_store<mode>): New insns.
10830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10831 Alan Hayward <alan.hayward@arm.com>
10832 David Sherwood <david.sherwood@arm.com>
10834 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10835 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10836 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10838 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10839 Use vect_truncate_gather_scatter_offset if we can't treat the
10840 operation as a normal gather load or scatter store.
10841 (get_group_load_store_type): Take the gather_scatter_info
10842 as argument. Try using a gather load or scatter store for
10843 single-element groups.
10844 (get_load_store_type): Update calls to get_group_load_store_type
10845 and vect_use_strided_gather_scatters_p.
10847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10848 Alan Hayward <alan.hayward@arm.com>
10849 David Sherwood <david.sherwood@arm.com>
10851 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10852 optional tree argument.
10853 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10855 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10856 but continue to use the current value as a fallback.
10857 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10858 to compare the updates.
10859 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10860 (get_load_store_type): Use it when handling a strided access.
10861 (vect_get_strided_load_store_ops): New function.
10862 (vect_get_data_ptr_increment): Likewise.
10863 (vectorizable_load): Handle strided gather loads. Always pass
10864 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10867 Alan Hayward <alan.hayward@arm.com>
10868 David Sherwood <david.sherwood@arm.com>
10870 * doc/md.texi (gather_load@var{m}): Document.
10871 (mask_gather_load@var{m}): Likewise.
10872 * genopinit.c (main): Add supports_vec_gather_load and
10873 supports_vec_gather_load_cached to target_optabs.
10874 * optabs-tree.c (init_tree_optimization_optabs): Use
10875 ggc_cleared_alloc to allocate target_optabs.
10876 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10877 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10879 * internal-fn.h (internal_load_fn_p): Declare.
10880 (internal_gather_scatter_fn_p): Likewise.
10881 (internal_fn_mask_index): Likewise.
10882 (internal_gather_scatter_fn_supported_p): Likewise.
10883 * internal-fn.c (gather_load_direct): New macro.
10884 (expand_gather_load_optab_fn): New function.
10885 (direct_gather_load_optab_supported_p): New macro.
10886 (direct_internal_fn_optab): New function.
10887 (internal_load_fn_p): Likewise.
10888 (internal_gather_scatter_fn_p): Likewise.
10889 (internal_fn_mask_index): Likewise.
10890 (internal_gather_scatter_fn_supported_p): Likewise.
10891 * optabs-query.c (supports_at_least_one_mode_p): New function.
10892 (supports_vec_gather_load_p): Likewise.
10893 * optabs-query.h (supports_vec_gather_load_p): Declare.
10894 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10895 and memory_type field.
10896 (NUM_PATTERNS): Bump to 15.
10897 * tree-vect-data-refs.c: Include internal-fn.h.
10898 (vect_gather_scatter_fn_p): New function.
10899 (vect_describe_gather_scatter_call): Likewise.
10900 (vect_check_gather_scatter): Try using internal functions for
10901 gather loads. Recognize existing calls to a gather load function.
10902 (vect_analyze_data_refs): Consider using gather loads if
10903 supports_vec_gather_load_p.
10904 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10905 (vect_get_gather_scatter_offset_type): Likewise.
10906 (vect_convert_mask_for_vectype): Likewise.
10907 (vect_add_conversion_to_patterm): Likewise.
10908 (vect_try_gather_scatter_pattern): Likewise.
10909 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10910 (vect_vect_recog_func_ptrs): Add it.
10911 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10912 internal_fn_mask_index and internal_gather_scatter_fn_p.
10913 (check_load_store_masking): Take the gather_scatter_info as an
10914 argument and handle gather loads.
10915 (vect_get_gather_scatter_ops): New function.
10916 (vectorizable_call): Check internal_load_fn_p.
10917 (vectorizable_load): Likewise. Handle gather load internal
10919 (vectorizable_store): Update call to check_load_store_masking.
10920 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10921 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10922 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10923 (aarch64_gather_scale_operand_d): New predicates.
10924 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10925 (mask_gather_load<mode>): New insns.
10927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10928 Alan Hayward <alan.hayward@arm.com>
10929 David Sherwood <david.sherwood@arm.com>
10931 * optabs.def (fold_left_plus_optab): New optab.
10932 * doc/md.texi (fold_left_plus_@var{m}): Document.
10933 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10934 * internal-fn.c (fold_left_direct): Define.
10935 (expand_fold_left_optab_fn): Likewise.
10936 (direct_fold_left_optab_supported_p): Likewise.
10937 * fold-const-call.c (fold_const_fold_left): New function.
10938 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10939 * tree-parloops.c (valid_reduction_p): New function.
10940 (gather_scalar_reductions): Use it.
10941 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10942 (vect_finish_replace_stmt): Declare.
10943 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10944 (needs_fold_left_reduction_p): New function, split out from...
10945 (vect_is_simple_reduction): ...here. Accept reductions that
10946 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10947 (vect_force_simple_reduction): Also store the reduction type in
10948 the assignment's STMT_VINFO_REDUC_TYPE.
10949 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10950 (merge_with_identity): New function.
10951 (vect_expand_fold_left): Likewise.
10952 (vectorize_fold_left_reduction): Likewise.
10953 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10954 scalar phi in place for it. Check for target support and reject
10955 cases that would reassociate the operation. Defer the transform
10956 phase to vectorize_fold_left_reduction.
10957 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10958 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10959 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10963 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10964 call to ifc_temp_var.
10966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10967 Alan Hayward <alan.hayward@arm.com>
10968 David Sherwood <david.sherwood@arm.com>
10970 * target.def (legitimize_address_displacement): Take the original
10971 offset as a poly_int.
10972 * targhooks.h (default_legitimize_address_displacement): Update
10974 * targhooks.c (default_legitimize_address_displacement): Likewise.
10975 * doc/tm.texi: Regenerate.
10976 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10977 as an argument, moving assert of ad->disp == ad->disp_term to...
10978 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10979 Try calling targetm.legitimize_address_displacement before expanding
10980 the address rather than afterwards, and adjust for the new interface.
10981 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10982 Match the new hook interface. Handle SVE addresses.
10983 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10984 new hook interface.
10986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10988 * Makefile.in (OBJS): Add early-remat.o.
10989 * target.def (select_early_remat_modes): New hook.
10990 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10991 * doc/tm.texi: Regenerate.
10992 * targhooks.h (default_select_early_remat_modes): Declare.
10993 * targhooks.c (default_select_early_remat_modes): New function.
10994 * timevar.def (TV_EARLY_REMAT): New timevar.
10995 * passes.def (pass_early_remat): New pass.
10996 * tree-pass.h (make_pass_early_remat): Declare.
10997 * early-remat.c: New file.
10998 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11000 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11003 Alan Hayward <alan.hayward@arm.com>
11004 David Sherwood <david.sherwood@arm.com>
11006 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11007 vfm1 with a bound_epilog parameter.
11008 (vect_do_peeling): Update calls accordingly, and move the prologue
11009 call earlier in the function. Treat the base bound_epilog as 0 for
11010 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11011 this base when peeling for gaps.
11012 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11013 with fully-masked loops.
11014 (vect_estimate_min_profitable_iters): Handle the single peeled
11015 iteration in that case.
11017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11018 Alan Hayward <alan.hayward@arm.com>
11019 David Sherwood <david.sherwood@arm.com>
11021 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11022 single-element interleaving even if the size is not a power of 2.
11023 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11024 accesses for single-element interleaving if the group size is
11027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11028 Alan Hayward <alan.hayward@arm.com>
11029 David Sherwood <david.sherwood@arm.com>
11031 * doc/md.texi (fold_extract_last_@var{m}): Document.
11032 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11033 * optabs.def (fold_extract_last_optab): New optab.
11034 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11035 * internal-fn.c (fold_extract_direct): New macro.
11036 (expand_fold_extract_optab_fn): Likewise.
11037 (direct_fold_extract_optab_supported_p): Likewise.
11038 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11039 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11040 EXTRACT_LAST_REDUCTION.
11041 (get_initial_def_for_reduction): Do not create an initial vector
11042 for EXTRACT_LAST_REDUCTION reductions.
11043 (vectorizable_reduction): Leave the scalar phi in place for
11044 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11045 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11046 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11047 transform phase to vectorizable_condition.
11048 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11050 (vect_finish_stmt_generation): ...here.
11051 (vect_finish_replace_stmt): New function.
11052 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11053 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11055 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11058 Alan Hayward <alan.hayward@arm.com>
11059 David Sherwood <david.sherwood@arm.com>
11061 * doc/md.texi (extract_last_@var{m}): Document.
11062 * optabs.def (extract_last_optab): New optab.
11063 * internal-fn.def (EXTRACT_LAST): New internal function.
11064 * internal-fn.c (cond_unary_direct): New macro.
11065 (expand_cond_unary_optab_fn): Likewise.
11066 (direct_cond_unary_optab_supported_p): Likewise.
11067 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11068 loops using EXTRACT_LAST.
11069 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11070 (extract_last_<mode>): ...this optab.
11071 (vec_extract<mode><Vel>): Update accordingly.
11073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11074 Alan Hayward <alan.hayward@arm.com>
11075 David Sherwood <david.sherwood@arm.com>
11077 * target.def (empty_mask_is_expensive): New hook.
11078 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11079 * doc/tm.texi: Regenerate.
11080 * targhooks.h (default_empty_mask_is_expensive): Declare.
11081 * targhooks.c (default_empty_mask_is_expensive): New function.
11082 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11083 if the target says that empty masks are expensive.
11084 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11086 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11089 Alan Hayward <alan.hayward@arm.com>
11090 David Sherwood <david.sherwood@arm.com>
11092 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11093 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11094 (vect_use_loop_mask_for_alignment_p): New function.
11095 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11096 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11097 niters_skip argument. Make sure that the first niters_skip elements
11098 of the first iteration are inactive.
11099 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11100 Update call to vect_set_loop_masks_directly.
11101 (get_misalign_in_elems): New function, split out from...
11102 (vect_gen_prolog_loop_niters): ...here.
11103 (vect_update_init_of_dr): Take a code argument that specifies whether
11104 the adjustment should be added or subtracted.
11105 (vect_update_init_of_drs): Likewise.
11106 (vect_prepare_for_masked_peels): New function.
11107 (vect_do_peeling): Skip prologue peeling if we're using a mask
11108 instead. Update call to vect_update_inits_of_drs.
11109 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11111 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11112 alignment. Do not include the number of peeled iterations in
11113 the minimum threshold in that case.
11114 (vectorizable_induction): Adjust the start value down by
11115 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11116 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11117 Take the number of skipped iterations into account when calculating
11119 * tree-vect-stmts.c (vect_gen_while_not): New function.
11121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11122 Alan Hayward <alan.hayward@arm.com>
11123 David Sherwood <david.sherwood@arm.com>
11125 * doc/sourcebuild.texi (vect_fully_masked): Document.
11126 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11127 default value to 0.
11128 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11130 (vect_analyze_loop_2): ...here. Don't check the vectorization
11131 factor against the number of loop iterations if the loop is
11134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11135 Alan Hayward <alan.hayward@arm.com>
11136 David Sherwood <david.sherwood@arm.com>
11138 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11139 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11140 (dump_groups): Update accordingly.
11141 (iv_use::mem_type): New member variable.
11142 (address_p): New function.
11143 (record_use): Add a mem_type argument and initialize the new
11145 (record_group_use): Add a mem_type argument. Use address_p.
11146 Remove obsolete null checks of base_object. Update call to record_use.
11147 (find_interesting_uses_op): Update call to record_group_use.
11148 (find_interesting_uses_cond): Likewise.
11149 (find_interesting_uses_address): Likewise.
11150 (get_mem_type_for_internal_fn): New function.
11151 (find_address_like_use): Likewise.
11152 (find_interesting_uses_stmt): Try find_address_like_use before
11153 calling find_interesting_uses_op.
11154 (addr_offset_valid_p): Use the iv mem_type field as the type
11155 of the addressed memory.
11156 (add_autoinc_candidates): Likewise.
11157 (get_address_cost): Likewise.
11158 (split_small_address_groups_p): Use address_p.
11159 (split_address_groups): Likewise.
11160 (add_iv_candidate_for_use): Likewise.
11161 (autoinc_possible_for_pair): Likewise.
11162 (rewrite_groups): Likewise.
11163 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11164 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11165 (get_alias_ptr_type_for_ptr_address): New function.
11166 (rewrite_use_address): Rewrite address uses in calls that were
11167 identified by find_address_like_use.
11169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11170 Alan Hayward <alan.hayward@arm.com>
11171 David Sherwood <david.sherwood@arm.com>
11173 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11175 * gimple-expr.h (is_gimple_addressable: Likewise.
11176 * gimple-expr.c (is_gimple_address): Likewise.
11177 * internal-fn.c (expand_call_mem_ref): New function.
11178 (expand_mask_load_optab_fn): Use it.
11179 (expand_mask_store_optab_fn): Likewise.
11181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11186 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11187 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11188 (cond_umax@var{mode}): Document.
11189 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11190 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11191 (cond_umin_optab, cond_umax_optab): New optabs.
11192 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11193 (COND_IOR, COND_XOR): New internal functions.
11194 * internal-fn.h (get_conditional_internal_fn): Declare.
11195 * internal-fn.c (cond_binary_direct): New macro.
11196 (expand_cond_binary_optab_fn): Likewise.
11197 (direct_cond_binary_optab_supported_p): Likewise.
11198 (get_conditional_internal_fn): New function.
11199 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11200 Cope with reduction statements that are vectorized as calls rather
11202 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11203 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11204 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11205 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11206 (UNSPEC_COND_EOR): New unspecs.
11207 (optab): Add mappings for them.
11208 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11209 (sve_int_op, sve_fp_op): New int attributes.
11211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11212 Alan Hayward <alan.hayward@arm.com>
11213 David Sherwood <david.sherwood@arm.com>
11215 * optabs.def (while_ult_optab): New optab.
11216 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11217 * internal-fn.def (WHILE_ULT): New internal function.
11218 * internal-fn.h (direct_internal_fn_supported_p): New override
11219 that takes two types as argument.
11220 * internal-fn.c (while_direct): New macro.
11221 (expand_while_optab_fn): New function.
11222 (convert_optab_supported_p): Likewise.
11223 (direct_while_optab_supported_p): New macro.
11224 * wide-int.h (wi::udiv_ceil): New function.
11225 * tree-vectorizer.h (rgroup_masks): New structure.
11226 (vec_loop_masks): New typedef.
11227 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11228 and fully_masked_p.
11229 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11230 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11231 (vect_max_vf): New function.
11232 (slpeel_make_loop_iterate_ntimes): Delete.
11233 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11234 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11235 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11236 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11237 internal-fn.h, stor-layout.h and optabs-query.h.
11238 (vect_set_loop_mask): New function.
11239 (add_preheader_seq): Likewise.
11240 (add_header_seq): Likewise.
11241 (interleave_supported_p): Likewise.
11242 (vect_maybe_permute_loop_masks): Likewise.
11243 (vect_set_loop_masks_directly): Likewise.
11244 (vect_set_loop_condition_masked): Likewise.
11245 (vect_set_loop_condition_unmasked): New function, split out from
11246 slpeel_make_loop_iterate_ntimes.
11247 (slpeel_make_loop_iterate_ntimes): Rename to..
11248 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11249 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11250 (vect_do_peeling): Update call accordingly.
11251 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11253 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11254 mask_compare_type, can_fully_mask_p and fully_masked_p.
11255 (release_vec_loop_masks): New function.
11256 (_loop_vec_info): Use it to free the loop masks.
11257 (can_produce_all_loop_masks_p): New function.
11258 (vect_get_max_nscalars_per_iter): Likewise.
11259 (vect_verify_full_masking): Likewise.
11260 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11261 retries, and free the mask rgroups before retrying. Check loop-wide
11262 reasons for disallowing fully-masked loops. Make the final decision
11263 about whether use a fully-masked loop or not.
11264 (vect_estimate_min_profitable_iters): Do not assume that peeling
11265 for the number of iterations will be needed for fully-masked loops.
11266 (vectorizable_reduction): Disable fully-masked loops.
11267 (vectorizable_live_operation): Likewise.
11268 (vect_halve_mask_nunits): New function.
11269 (vect_double_mask_nunits): Likewise.
11270 (vect_record_loop_mask): Likewise.
11271 (vect_get_loop_mask): Likewise.
11272 (vect_transform_loop): Handle the case in which the final loop
11273 iteration might handle a partial vector. Call vect_set_loop_condition
11274 instead of slpeel_make_loop_iterate_ntimes.
11275 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11276 (check_load_store_masking): New function.
11277 (prepare_load_store_mask): Likewise.
11278 (vectorizable_store): Handle fully-masked loops.
11279 (vectorizable_load): Likewise.
11280 (supportable_widening_operation): Use vect_halve_mask_nunits for
11282 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11283 (vect_gen_while): New function.
11284 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11285 (aarch64_uqdec<mode>): New insn.
11287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11288 Alan Hayward <alan.hayward@arm.com>
11289 David Sherwood <david.sherwood@arm.com>
11291 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11292 (reduc_xor_scal_optab): New optabs.
11293 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11294 (reduc_xor_scal_@var{m}): Document.
11295 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11296 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11297 internal functions.
11298 * fold-const-call.c (fold_const_call): Handle them.
11299 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11300 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11301 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11302 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11303 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11304 (UNSPEC_XORV): New unspecs.
11305 (optab): Add entries for them.
11306 (BITWISEV): New int iterator.
11307 (bit_reduc_op): New int attributes.
11309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11310 Alan Hayward <alan.hayward@arm.com>
11311 David Sherwood <david.sherwood@arm.com>
11313 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11314 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11315 * optabs.def (vec_shl_insert_optab): New optab.
11316 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11317 (duplicate_and_interleave): Likewise.
11318 * tree-vect-loop.c: Include internal-fn.h.
11319 (neutral_op_for_slp_reduction): New function, split out from
11320 get_initial_defs_for_reduction.
11321 (get_initial_def_for_reduction): Handle option 2 for variable-length
11322 vectors by loading the neutral value into a vector and then shifting
11323 the initial value into element 0.
11324 (get_initial_defs_for_reduction): Replace the code argument with
11325 the neutral value calculated by neutral_op_for_slp_reduction.
11326 Use gimple_build_vector for constant-length vectors.
11327 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11328 but the first group_size elements have a neutral value.
11329 Use duplicate_and_interleave otherwise.
11330 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11331 Update call to get_initial_defs_for_reduction. Handle SLP
11332 reductions for variable-length vectors by creating one vector
11333 result for each scalar result, with the elements associated
11334 with other scalar results stubbed out with the neutral value.
11335 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11336 Require IFN_VEC_SHL_INSERT for double reductions on
11337 variable-length vectors, or SLP reductions that have
11338 a neutral value. Require can_duplicate_and_interleave_p
11339 support for variable-length unchained SLP reductions if there
11340 is no neutral value, such as for MIN/MAX reductions. Also require
11341 the number of vector elements to be a multiple of the number of
11342 SLP statements when doing variable-length unchained SLP reductions.
11343 Update call to vect_create_epilog_for_reduction.
11344 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11345 and remove initial values.
11346 (duplicate_and_interleave): Make public.
11347 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11348 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11351 Alan Hayward <alan.hayward@arm.com>
11352 David Sherwood <david.sherwood@arm.com>
11354 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11355 (can_duplicate_and_interleave_p): New function.
11356 (vect_get_and_check_slp_defs): Take the vector of statements
11357 rather than just the current one. Remove excess parentheses.
11358 Restriction rejectinon of vect_constant_def and vect_external_def
11359 for variable-length vectors to boolean types, or types for which
11360 can_duplicate_and_interleave_p is false.
11361 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11362 (duplicate_and_interleave): New function.
11363 (vect_get_constant_vectors): Use gimple_build_vector for
11364 constant-length vectors and suitable variable-length constant
11365 vectors. Use duplicate_and_interleave for other variable-length
11366 vectors. Don't defer the update when inserting new statements.
11368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11369 Alan Hayward <alan.hayward@arm.com>
11370 David Sherwood <david.sherwood@arm.com>
11372 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11373 min_profitable_iters doesn't go negative.
11375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11376 Alan Hayward <alan.hayward@arm.com>
11377 David Sherwood <david.sherwood@arm.com>
11379 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11380 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11381 * optabs.def (vec_mask_load_lanes_optab): New optab.
11382 (vec_mask_store_lanes_optab): Likewise.
11383 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11384 (MASK_STORE_LANES): Likewise.
11385 * internal-fn.c (mask_load_lanes_direct): New macro.
11386 (mask_store_lanes_direct): Likewise.
11387 (expand_mask_load_optab_fn): Handle masked operations.
11388 (expand_mask_load_lanes_optab_fn): New macro.
11389 (expand_mask_store_optab_fn): Handle masked operations.
11390 (expand_mask_store_lanes_optab_fn): New macro.
11391 (direct_mask_load_lanes_optab_supported_p): Likewise.
11392 (direct_mask_store_lanes_optab_supported_p): Likewise.
11393 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11395 (vect_load_lanes_supported): Likewise.
11396 * tree-vect-data-refs.c (strip_conversion): New function.
11397 (can_group_stmts_p): Likewise.
11398 (vect_analyze_data_ref_accesses): Use it instead of checking
11399 for a pair of assignments.
11400 (vect_store_lanes_supported): Take a masked_p parameter.
11401 (vect_load_lanes_supported): Likewise.
11402 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11403 vect_store_lanes_supported and vect_load_lanes_supported.
11404 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11405 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11406 parameter. Don't allow gaps for masked accesses.
11407 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11408 and vect_load_lanes_supported.
11409 (get_load_store_type): Take a masked_p parameter and update
11410 call to get_group_load_store_type.
11411 (vectorizable_store): Update call to get_load_store_type.
11412 Handle IFN_MASK_STORE_LANES.
11413 (vectorizable_load): Update call to get_load_store_type.
11414 Handle IFN_MASK_LOAD_LANES.
11416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11417 Alan Hayward <alan.hayward@arm.com>
11418 David Sherwood <david.sherwood@arm.com>
11420 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11422 * config/aarch64/aarch64-protos.h
11423 (aarch64_sve_struct_memory_operand_p): Declare.
11424 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11425 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11426 (VPRED, vpred): Handle SVE structure modes.
11427 * config/aarch64/constraints.md (Utx): New constraint.
11428 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11429 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11430 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11431 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11432 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11433 structure modes. Split into pieces after RA.
11434 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11435 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11437 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11438 SVE structure modes.
11439 (aarch64_classify_address): Likewise.
11440 (sizetochar): Move earlier in file.
11441 (aarch64_print_operand): Handle SVE register lists.
11442 (aarch64_array_mode): New function.
11443 (aarch64_sve_struct_memory_operand_p): Likewise.
11444 (TARGET_ARRAY_MODE): Redefine.
11446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11447 Alan Hayward <alan.hayward@arm.com>
11448 David Sherwood <david.sherwood@arm.com>
11450 * target.def (array_mode): New target hook.
11451 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11452 * doc/tm.texi: Regenerate.
11453 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11454 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11455 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11456 targetm.array_mode.
11457 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11461 Alan Hayward <alan.hayward@arm.com>
11462 David Sherwood <david.sherwood@arm.com>
11464 * fold-const.c (fold_binary_loc): Check the argument types
11465 rather than the result type when testing for a vector operation.
11467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11469 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11470 * doc/tm.texi: Regenerate.
11472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11477 (sve): Document new AArch64 extension.
11478 * doc/md.texi (w): Extend the description of the AArch64
11479 constraint to include SVE vectors.
11480 (Upl, Upa): Document new AArch64 predicate constraints.
11481 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11483 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11484 (msve-vector-bits=): New option.
11485 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11486 SVE when these are disabled.
11487 (sve): New extension.
11488 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11489 modes. Adjust their number of units based on aarch64_sve_vg.
11490 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11491 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11492 aarch64_addr_query_type.
11493 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11494 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11495 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11496 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11497 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11498 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11499 (aarch64_simd_imm_zero_p): Delete.
11500 (aarch64_check_zero_based_sve_index_immediate): Declare.
11501 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11502 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11503 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11504 (aarch64_sve_float_mul_immediate_p): Likewise.
11505 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11506 rather than an rtx.
11507 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11508 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11509 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11510 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11511 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11512 (aarch64_regmode_natural_size): Likewise.
11513 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11514 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11516 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11517 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11518 for VG and the SVE predicate registers.
11519 (V_ALIASES): Add a "z"-prefixed alias.
11520 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11521 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11522 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11523 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11524 (REG_CLASS_NAMES): Add entries for them.
11525 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11526 and the predicate registers.
11527 (aarch64_sve_vg): Declare.
11528 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11529 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11530 (REGMODE_NATURAL_SIZE): Define.
11531 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11533 * config/aarch64/aarch64.c: Include cfgrtl.h.
11534 (simd_immediate_info): Add a constructor for series vectors,
11535 and an associated step field.
11536 (aarch64_sve_vg): New variable.
11537 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11538 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11539 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11540 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11541 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11542 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11543 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11544 (aarch64_get_mask_mode): New functions.
11545 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11546 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11547 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11548 predicate modes and predicate registers. Explicitly restrict
11549 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11550 to store a vector mode if it is recognized by
11551 aarch64_classify_vector_mode.
11552 (aarch64_regmode_natural_size): New function.
11553 (aarch64_hard_regno_caller_save_mode): Return the original mode
11555 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11556 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11557 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11558 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11560 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11561 does not overlap dest if the function is frame-related. Handle
11563 (aarch64_split_add_offset): New function.
11564 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11565 them aarch64_add_offset.
11566 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11567 and update call to aarch64_sub_sp.
11568 (aarch64_add_cfa_expression): New function.
11569 (aarch64_expand_prologue): Pass extra temporary registers to the
11570 functions above. Handle the case in which we need to emit new
11571 DW_CFA_expressions for registers that were originally saved
11572 relative to the stack pointer, but now have to be expressed
11573 relative to the frame pointer.
11574 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11576 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11577 IP0 and IP1 values for SVE frames.
11578 (aarch64_expand_vec_series): New function.
11579 (aarch64_expand_sve_widened_duplicate): Likewise.
11580 (aarch64_expand_sve_const_vector): Likewise.
11581 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11582 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11583 into the register, rather than emitting a SET directly.
11584 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11585 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11586 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11587 (offset_9bit_signed_scaled_p): New functions.
11588 (aarch64_replicate_bitmask_imm): New function.
11589 (aarch64_bitmask_imm): Use it.
11590 (aarch64_cannot_force_const_mem): Reject expressions involving
11591 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11592 (aarch64_classify_index): Handle SVE indices, by requiring
11593 a plain register index with a scale that matches the element size.
11594 (aarch64_classify_address): Handle SVE addresses. Assert that
11595 the mode of the address is VOIDmode or an integer mode.
11596 Update call to aarch64_classify_symbol.
11597 (aarch64_classify_symbolic_expression): Update call to
11598 aarch64_classify_symbol.
11599 (aarch64_const_vec_all_in_range_p): New function.
11600 (aarch64_print_vector_float_operand): Likewise.
11601 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11602 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11603 and the FP immediates 1.0 and 0.5.
11604 (aarch64_print_address_internal): Handle SVE addresses.
11605 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11606 (aarch64_regno_regclass): Handle predicate registers.
11607 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11609 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11610 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11611 (aarch64_convert_sve_vector_bits): New function.
11612 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11613 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11614 rather than an rtx.
11615 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11616 Handle SVE vector and predicate modes. Accept VL-based constants
11617 that need only one temporary register, and VL offsets that require
11618 no temporary registers.
11619 (aarch64_conditional_register_usage): Mark the predicate registers
11620 as fixed if SVE isn't available.
11621 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11622 Return true for SVE vector and predicate modes.
11623 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11624 rather than an unsigned int. Handle SVE modes.
11625 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11627 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11629 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11630 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11631 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11632 (aarch64_sve_float_mul_immediate_p): New functions.
11633 (aarch64_sve_valid_immediate): New function.
11634 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11635 Explicitly reject structure modes. Check for INDEX constants.
11636 Handle PTRUE and PFALSE constants.
11637 (aarch64_check_zero_based_sve_index_immediate): New function.
11638 (aarch64_simd_imm_zero_p): Delete.
11639 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11640 vector modes. Accept constants in the range of CNT[BHWD].
11641 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11642 ask for an Advanced SIMD mode.
11643 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11644 (aarch64_simd_vector_alignment): Handle SVE predicates.
11645 (aarch64_vectorize_preferred_vector_alignment): New function.
11646 (aarch64_simd_vector_alignment_reachable): Use it instead of
11648 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11649 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11651 (MAX_VECT_LEN): Delete.
11652 (expand_vec_perm_d): Add a vec_flags field.
11653 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11654 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11655 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11657 (aarch64_evpc_rev): Rename to...
11658 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11659 (aarch64_evpc_rev_global): New function.
11660 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11661 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11663 (aarch64_evpc_sve_tbl): New function.
11664 (aarch64_expand_vec_perm_const_1): Update after rename of
11665 aarch64_evpc_rev. Handle SVE permutes too, trying
11666 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11667 than aarch64_evpc_tbl.
11668 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11669 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11670 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11671 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11672 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11673 (aarch64_expand_sve_vcond): New functions.
11674 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11675 of aarch64_vector_mode_p.
11676 (aarch64_dwarf_poly_indeterminate_value): New function.
11677 (aarch64_compute_pressure_classes): Likewise.
11678 (aarch64_can_change_mode_class): Likewise.
11679 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11680 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11681 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11682 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11683 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11684 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11685 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11686 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11688 (Dn, Dl, Dr): Accept const as well as const_vector.
11689 (Dz): Likewise. Compare against CONST0_RTX.
11690 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11691 of "vector" where appropriate.
11692 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11693 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11694 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11695 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11696 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11697 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11698 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11699 (v_int_equiv): Extend to SVE modes.
11700 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11702 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11703 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11704 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11705 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11706 (SVE_COND_FP_CMP): New int iterators.
11707 (perm_hilo): Handle the new unpack unspecs.
11708 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11710 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11711 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11712 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11713 (aarch64_equality_operator, aarch64_constant_vector_operand)
11714 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11715 (aarch64_sve_nonimmediate_operand): Likewise.
11716 (aarch64_sve_general_operand): Likewise.
11717 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11718 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11719 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11720 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11721 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11722 (aarch64_sve_float_arith_immediate): Likewise.
11723 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11724 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11725 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11726 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11727 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11728 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11729 (aarch64_sve_float_arith_operand): Likewise.
11730 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11731 (aarch64_sve_float_mul_operand): Likewise.
11732 (aarch64_sve_vec_perm_operand): Likewise.
11733 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11734 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11735 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11736 as well as const_vector.
11737 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11738 in file. Use CONST0_RTX and CONSTM1_RTX.
11739 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11740 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11741 Use aarch64_simd_imm_zero.
11742 * config/aarch64/aarch64-sve.md: New file.
11743 * config/aarch64/aarch64.md: Include it.
11744 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11745 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11746 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11747 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11748 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11749 (sve): New attribute.
11750 (enabled): Disable instructions with the sve attribute unless
11752 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11753 aarch64_expand_mov_immediate.
11754 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11755 CNT[BHSD] immediates.
11756 (movti): Split CONST_POLY_INT moves into two halves.
11757 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11758 Split additions that need a temporary here if the destination
11759 is the stack pointer.
11760 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11761 (*add<mode>3_poly_1): New instruction.
11762 (set_clobber_cc): New expander.
11764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11766 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11767 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11768 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11769 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11770 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11771 Change innermode from fixed_mode_size to machine_mode.
11772 (simplify_subreg): Update call accordingly. Handle a constant-sized
11773 subreg of a variable-length CONST_VECTOR.
11775 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11776 Alan Hayward <alan.hayward@arm.com>
11777 David Sherwood <david.sherwood@arm.com>
11779 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11780 (add_offset_to_base): New function, split out from...
11781 (create_mem_ref): ...here. When handling a scale other than 1,
11782 check first whether the address is valid without the offset.
11783 Add it into the base if so, leaving the index and scale as-is.
11785 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11788 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11789 fold_for_warn before checking if arg2 is INTEGER_CST.
11791 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11793 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11794 (store_multiple_operation): Delete.
11795 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11796 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11797 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11798 guarded by TARGET_STRING.
11799 (rs6000_output_load_multiple): Delete.
11800 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11801 OPTION_MASK_STRING / TARGET_STRING handling.
11802 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11803 (const rs6000_opt_masks) <"string">: Change mask to 0.
11804 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11805 (MASK_STRING): Delete.
11806 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11808 (load_multiple): Delete.
11815 (store_multiple): Delete.
11822 (movmemsi_8reg): Delete.
11823 (corresponding unnamed define_insn): Delete.
11824 (movmemsi_6reg): Delete.
11825 (corresponding unnamed define_insn): Delete.
11826 (movmemsi_4reg): Delete.
11827 (corresponding unnamed define_insn): Delete.
11828 (movmemsi_2reg): Delete.
11829 (corresponding unnamed define_insn): Delete.
11830 (movmemsi_1reg): Delete.
11831 (corresponding unnamed define_insn): Delete.
11832 * config/rs6000/rs6000.opt (mno-string): New.
11833 (mstring): Replace by deprecation warning stub.
11834 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11836 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11838 * regrename.c (regrename_do_replace): If replacing the same
11839 reg multiple times, try to reuse last created gen_raw_REG.
11842 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11843 main to workaround a bug in GDB.
11845 2018-01-12 Tom de Vries <tom@codesourcery.com>
11848 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11850 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11852 PR rtl-optimization/80481
11853 * ira-color.c (get_cap_member): New function.
11854 (allocnos_conflict_by_live_ranges_p): Use it.
11855 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11856 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11858 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11861 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11862 (*saddl_se_1): Ditto.
11863 (*ssubsi_1): Ditto.
11864 (*ssubl_se_1): Ditto.
11866 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11868 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11869 rather than wi::to_widest for DR_INITs.
11870 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11871 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11872 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11874 (vect_analyze_group_access_1): Note that here.
11876 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11878 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11879 polynomial type sizes.
11881 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11883 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11884 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11885 (gimple_add_tmp_var): Likewise.
11887 2018-01-12 Martin Liska <mliska@suse.cz>
11889 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11890 (gimple_alloc_sizes): Likewise.
11891 (dump_gimple_statistics): Use PRIu64 in printf format.
11892 * gimple.h: Change uint64_t to int.
11894 2018-01-12 Martin Liska <mliska@suse.cz>
11896 * tree-core.h: Use uint64_t instead of int.
11897 * tree.c (tree_node_counts): Likewise.
11898 (tree_node_sizes): Likewise.
11899 (dump_tree_statistics): Use PRIu64 in printf format.
11901 2018-01-12 Martin Liska <mliska@suse.cz>
11903 * Makefile.in: As qsort_chk is implemented in vec.c, add
11904 vec.o to linkage of gencfn-macros.
11905 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11906 passing the info to record_node_allocation_statistics.
11907 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11909 * ggc-common.c (struct ggc_usage): Add operator== and use
11910 it in operator< and compare function.
11911 * mem-stats.h (struct mem_usage): Likewise.
11912 * vec.c (struct vec_usage): Remove operator< and compare
11913 function. Can be simply inherited.
11915 2018-01-12 Martin Jambor <mjambor@suse.cz>
11918 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11919 * tree-ssa-math-opts.c: Include domwalk.h.
11920 (convert_mult_to_fma_1): New function.
11921 (fma_transformation_info): New type.
11922 (fma_deferring_state): Likewise.
11923 (cancel_fma_deferring): New function.
11924 (result_of_phi): Likewise.
11925 (last_fma_candidate_feeds_initial_phi): Likewise.
11926 (convert_mult_to_fma): Added deferring logic, split actual
11927 transformation to convert_mult_to_fma_1.
11928 (math_opts_dom_walker): New type.
11929 (math_opts_dom_walker::after_dom_children): New method, body moved
11930 here from pass_optimize_widening_mul::execute, added deferring logic
11932 (pass_optimize_widening_mul::execute): Moved most of code to
11933 math_opts_dom_walker::after_dom_children.
11934 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11935 * config/i386/i386.c (ix86_option_override_internal): Added
11936 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11938 2018-01-12 Richard Biener <rguenther@suse.de>
11941 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11942 inline instance vars.
11944 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11947 * config/rx/rx.c (rx_is_restricted_memory_address):
11948 Handle SUBREG case.
11950 2018-01-12 Richard Biener <rguenther@suse.de>
11952 PR tree-optimization/80846
11953 * target.def (split_reduction): New target hook.
11954 * targhooks.c (default_split_reduction): New function.
11955 * targhooks.h (default_split_reduction): Declare.
11956 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11957 target requests first reduce vectors by combining low and high
11959 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11960 (get_vectype_for_scalar_type_and_size): Export.
11961 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11962 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11963 * doc/tm.texi: Regenerate.
11964 * config/i386/i386.c (ix86_split_reduction): Implement
11965 TARGET_VECTORIZE_SPLIT_REDUCTION.
11967 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11970 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11971 in PIC mode except for TARGET_VXWORKS_RTP.
11972 * config/sparc/sparc.c: Include cfgrtl.h.
11973 (TARGET_INIT_PIC_REG): Define.
11974 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11975 (sparc_pic_register_p): New predicate.
11976 (sparc_legitimate_address_p): Use it.
11977 (sparc_legitimize_pic_address): Likewise.
11978 (sparc_delegitimize_address): Likewise.
11979 (sparc_mode_dependent_address_p): Likewise.
11980 (gen_load_pcrel_sym): Remove 4th parameter.
11981 (load_got_register): Adjust call to above. Remove obsolete stuff.
11982 (sparc_expand_prologue): Do not call load_got_register here.
11983 (sparc_flat_expand_prologue): Likewise.
11984 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11985 (sparc_use_pseudo_pic_reg): New function.
11986 (sparc_init_pic_reg): Likewise.
11987 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11988 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11990 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11992 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11993 Add item for branch_cost.
11995 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11997 PR rtl-optimization/83565
11998 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11999 not extend the result to a larger mode for rotate operations.
12000 (num_sign_bit_copies1): Likewise.
12002 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12005 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12007 Use values-Xc.o for -pedantic.
12008 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12010 2018-01-12 Martin Liska <mliska@suse.cz>
12013 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12015 (possible_polymorphic_call_targets): Use it.
12016 (ipa_devirt): Likewise.
12018 2018-01-12 Martin Liska <mliska@suse.cz>
12020 * profile-count.h (enum profile_quality): Use 0 as invalid
12021 enum value of profile_quality.
12023 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12025 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12026 -mext-string options.
12028 2018-01-12 Richard Biener <rguenther@suse.de>
12030 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12031 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12032 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12034 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12036 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12038 * configure.ac (--with-long-double-format): Add support for the
12039 configuration option to change the default long double format on
12041 * config.gcc (powerpc*-linux*-*): Likewise.
12042 * configure: Regenerate.
12043 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12044 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12045 used without modification.
12047 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12049 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12050 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12051 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12052 MISC_BUILTIN_SPEC_BARRIER.
12053 (rs6000_init_builtins): Likewise.
12054 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12056 (speculation_barrier): New define_insn.
12057 * doc/extend.texi: Document __builtin_speculation_barrier.
12059 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12062 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12063 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12064 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12066 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12067 integral modes instead of "ss" and "sd".
12068 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12069 vectors with 32-bit and 64-bit elements.
12070 (vecdupssescalarmodesuffix): New mode attribute.
12071 (vec_dup<mode>): Use it.
12073 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12076 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12077 frame if argument is passed on stack.
12079 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12082 * ree.c (combine_reaching_defs): Optimize also
12083 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12084 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12086 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12088 PR middle-end/83189
12089 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12091 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12093 PR middle-end/83718
12094 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12095 after they are computed.
12097 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12099 PR tree-optimization/83695
12100 * gimple-loop-linterchange.cc
12101 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12102 reset cached scev information after interchange.
12103 (pass_linterchange::execute): Remove call to scev_reset_htab.
12105 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12107 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12108 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12109 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12110 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12111 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12112 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12113 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12114 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12115 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12116 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12117 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12118 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12119 (V_lane_reg): Likewise.
12120 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12122 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12123 (vfmal_lane_low<mode>_intrinsic,
12124 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12125 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12126 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12127 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12128 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12129 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12131 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12133 * config/arm/arm-cpus.in (fp16fml): New feature.
12134 (ALL_SIMD): Add fp16fml.
12135 (armv8.2-a): Add fp16fml as an option.
12136 (armv8.3-a): Likewise.
12137 (armv8.4-a): Add fp16fml as part of fp16.
12138 * config/arm/arm.h (TARGET_FP16FML): Define.
12139 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12141 * config/arm/arm-modes.def (V2HF): Define.
12142 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12143 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12144 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12145 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12146 vfmsl_low, vfmsl_high): New set of builtins.
12147 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12148 (vfml_op): New code attribute.
12149 (VFMLHALVES): New int iterator.
12150 (VFML, VFMLSEL): New mode attributes.
12151 (V_reg): Define mapping for V2HF.
12152 (V_hi, V_lo): New mode attributes.
12153 (VF_constraint): Likewise.
12154 (vfml_half, vfml_half_selector): New int attributes.
12155 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12157 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12158 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12160 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12161 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12162 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12163 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12165 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12166 Document new effective target and option set.
12168 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12170 * config/arm/arm-cpus.in (armv8_4): New feature.
12171 (ARMv8_4a): New fgroup.
12172 (armv8.4-a): New arch.
12173 * config/arm/arm-tables.opt: Regenerate.
12174 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12175 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12176 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12177 Add matching rules for -march=armv8.4-a and extensions.
12178 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12180 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12183 * config/rx/rx.md (BW): New mode attribute.
12184 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12186 2018-01-11 Richard Biener <rguenther@suse.de>
12188 PR tree-optimization/83435
12189 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12190 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12191 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12193 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12194 Alan Hayward <alan.hayward@arm.com>
12195 David Sherwood <david.sherwood@arm.com>
12197 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12199 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12200 (aarch64_print_address_internal): Use it to check for a zero offset.
12202 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12203 Alan Hayward <alan.hayward@arm.com>
12204 David Sherwood <david.sherwood@arm.com>
12206 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12207 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12208 Return a poly_int64 rather than a HOST_WIDE_INT.
12209 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12210 rather than a HOST_WIDE_INT.
12211 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12212 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12213 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12214 final_offset from HOST_WIDE_INT to poly_int64.
12215 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12216 to_constant when getting the number of units in an Advanced SIMD
12218 (aarch64_builtin_vectorized_function): Check for a constant number
12220 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12222 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12223 attribute instead of GET_MODE_NUNITS.
12224 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12225 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12226 GET_MODE_SIZE for fixed-size registers.
12227 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12228 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12229 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12230 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12231 (aarch64_print_operand, aarch64_print_address_internal)
12232 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12233 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12234 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12235 Handle polynomial GET_MODE_SIZE.
12236 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12237 wider than SImode without modification.
12238 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12239 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12240 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12241 passing and returning SVE modes.
12242 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12243 rather than GEN_INT.
12244 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12245 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12246 (aarch64_allocate_and_probe_stack_space): Likewise.
12247 (aarch64_layout_frame): Cope with polynomial offsets.
12248 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12249 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12250 polynomial offsets.
12251 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12252 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12253 poly_int64 rather than a HOST_WIDE_INT.
12254 (aarch64_get_separate_components, aarch64_process_components)
12255 (aarch64_expand_prologue, aarch64_expand_epilogue)
12256 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12257 (aarch64_anchor_offset): New function, split out from...
12258 (aarch64_legitimize_address): ...here.
12259 (aarch64_builtin_vectorization_cost): Handle polynomial
12260 TYPE_VECTOR_SUBPARTS.
12261 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12263 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12264 number of elements from the PARALLEL rather than the mode.
12265 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12266 rather than GET_MODE_BITSIZE.
12267 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12268 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12269 (aarch64_expand_vec_perm_const_1): Handle polynomial
12270 d->perm.length () and d->perm elements.
12271 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12272 Apply to_constant to d->perm elements.
12273 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12274 polynomial CONST_VECTOR_NUNITS.
12275 (aarch64_move_pointer): Take amount as a poly_int64 rather
12277 (aarch64_progress_pointer): Avoid temporary variable.
12278 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12279 the mode attribute instead of GET_MODE.
12281 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12282 Alan Hayward <alan.hayward@arm.com>
12283 David Sherwood <david.sherwood@arm.com>
12285 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12286 x exists before using it.
12287 (aarch64_add_constant_internal): Rename to...
12288 (aarch64_add_offset_1): ...this. Replace regnum with separate
12289 src and dest rtxes. Handle the case in which they're different,
12290 including when the offset is zero. Replace scratchreg with an rtx.
12291 Use 2 additions if there is no spare register into which we can
12292 move a 16-bit constant.
12293 (aarch64_add_constant): Delete.
12294 (aarch64_add_offset): Replace reg with separate src and dest
12295 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12296 Use aarch64_add_offset_1.
12297 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12298 an rtx rather than an int. Take the delta as a poly_int64
12299 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12300 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12301 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12302 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12303 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12304 and aarch64_add_sp.
12305 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12306 aarch64_add_constant.
12308 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12310 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12311 Use scalar_float_mode.
12313 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12315 * config/aarch64/aarch64-simd.md
12316 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12317 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12318 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12319 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12320 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12321 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12322 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12323 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12324 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12325 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12327 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12330 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12331 targ_options->x_arm_arch_string is non NULL.
12333 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12335 * config/aarch64/aarch64.h
12336 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12338 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12341 * expmed.c (emit_store_flag_force): Swap if const op0
12342 and change VOIDmode to mode of op0.
12344 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12346 PR rtl-optimization/83761
12347 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12348 than bytes to mode_for_size.
12350 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12352 PR middle-end/83189
12353 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12354 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12357 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12359 PR middle-end/83575
12360 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12361 when in layout mode.
12362 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12363 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12366 2018-01-10 Michael Collison <michael.collison@arm.com>
12368 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12369 * config/aarch64/aarch64-option-extension.def: Add
12370 AARCH64_OPT_EXTENSION of 'fp16fml'.
12371 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12372 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12373 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12374 * config/aarch64/constraints.md (Ui7): New constraint.
12375 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12376 (VFMLA_SEL_W): Ditto.
12379 (VFMLA16_LOW): New int iterator.
12380 (VFMLA16_HIGH): Ditto.
12381 (UNSPEC_FMLAL): New unspec.
12382 (UNSPEC_FMLSL): Ditto.
12383 (UNSPEC_FMLAL2): Ditto.
12384 (UNSPEC_FMLSL2): Ditto.
12385 (f16mac): New code attribute.
12386 * config/aarch64/aarch64-simd-builtins.def
12387 (aarch64_fmlal_lowv2sf): Ditto.
12388 (aarch64_fmlsl_lowv2sf): Ditto.
12389 (aarch64_fmlalq_lowv4sf): Ditto.
12390 (aarch64_fmlslq_lowv4sf): Ditto.
12391 (aarch64_fmlal_highv2sf): Ditto.
12392 (aarch64_fmlsl_highv2sf): Ditto.
12393 (aarch64_fmlalq_highv4sf): Ditto.
12394 (aarch64_fmlslq_highv4sf): Ditto.
12395 (aarch64_fmlal_lane_lowv2sf): Ditto.
12396 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12397 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12398 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12399 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12400 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12401 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12402 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12403 (aarch64_fmlal_lane_highv2sf): Ditto.
12404 (aarch64_fmlsl_lane_highv2sf): Ditto.
12405 (aarch64_fmlal_laneq_highv2sf): Ditto.
12406 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12407 (aarch64_fmlalq_lane_highv4sf): Ditto.
12408 (aarch64_fmlsl_lane_highv4sf): Ditto.
12409 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12410 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12411 * config/aarch64/aarch64-simd.md:
12412 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12413 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12414 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12415 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12416 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12417 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12418 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12419 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12420 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12421 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12422 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12423 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12424 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12425 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12426 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12427 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12428 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12429 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12430 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12431 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12432 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12433 (vfmlsl_low_u32): Ditto.
12434 (vfmlalq_low_u32): Ditto.
12435 (vfmlslq_low_u32): Ditto.
12436 (vfmlal_high_u32): Ditto.
12437 (vfmlsl_high_u32): Ditto.
12438 (vfmlalq_high_u32): Ditto.
12439 (vfmlslq_high_u32): Ditto.
12440 (vfmlal_lane_low_u32): Ditto.
12441 (vfmlsl_lane_low_u32): Ditto.
12442 (vfmlal_laneq_low_u32): Ditto.
12443 (vfmlsl_laneq_low_u32): Ditto.
12444 (vfmlalq_lane_low_u32): Ditto.
12445 (vfmlslq_lane_low_u32): Ditto.
12446 (vfmlalq_laneq_low_u32): Ditto.
12447 (vfmlslq_laneq_low_u32): Ditto.
12448 (vfmlal_lane_high_u32): Ditto.
12449 (vfmlsl_lane_high_u32): Ditto.
12450 (vfmlal_laneq_high_u32): Ditto.
12451 (vfmlsl_laneq_high_u32): Ditto.
12452 (vfmlalq_lane_high_u32): Ditto.
12453 (vfmlslq_lane_high_u32): Ditto.
12454 (vfmlalq_laneq_high_u32): Ditto.
12455 (vfmlslq_laneq_high_u32): Ditto.
12456 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12457 (AARCH64_FL_FOR_ARCH8_4): New.
12458 (AARCH64_ISA_F16FML): New ISA flag.
12459 (TARGET_F16FML): New feature flag for fp16fml.
12460 (doc/invoke.texi): Document new fp16fml option.
12462 2018-01-10 Michael Collison <michael.collison@arm.com>
12464 * config/aarch64/aarch64-builtins.c:
12465 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12466 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12467 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12468 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12469 (AARCH64_ISA_SHA3): New ISA flag.
12470 (TARGET_SHA3): New feature flag for sha3.
12471 * config/aarch64/iterators.md (sha512_op): New int attribute.
12472 (CRYPTO_SHA512): New int iterator.
12473 (UNSPEC_SHA512H): New unspec.
12474 (UNSPEC_SHA512H2): Ditto.
12475 (UNSPEC_SHA512SU0): Ditto.
12476 (UNSPEC_SHA512SU1): Ditto.
12477 * config/aarch64/aarch64-simd-builtins.def
12478 (aarch64_crypto_sha512hqv2di): New builtin.
12479 (aarch64_crypto_sha512h2qv2di): Ditto.
12480 (aarch64_crypto_sha512su0qv2di): Ditto.
12481 (aarch64_crypto_sha512su1qv2di): Ditto.
12482 (aarch64_eor3qv8hi): Ditto.
12483 (aarch64_rax1qv2di): Ditto.
12484 (aarch64_xarqv2di): Ditto.
12485 (aarch64_bcaxqv8hi): Ditto.
12486 * config/aarch64/aarch64-simd.md:
12487 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12488 (aarch64_crypto_sha512su0qv2di): Ditto.
12489 (aarch64_crypto_sha512su1qv2di): Ditto.
12490 (aarch64_eor3qv8hi): Ditto.
12491 (aarch64_rax1qv2di): Ditto.
12492 (aarch64_xarqv2di): Ditto.
12493 (aarch64_bcaxqv8hi): Ditto.
12494 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12495 (vsha512h2q_u64): Ditto.
12496 (vsha512su0q_u64): Ditto.
12497 (vsha512su1q_u64): Ditto.
12498 (veor3q_u16): Ditto.
12499 (vrax1q_u64): Ditto.
12500 (vxarq_u64): Ditto.
12501 (vbcaxq_u16): Ditto.
12502 * config/arm/types.md (crypto_sha512): New type attribute.
12503 (crypto_sha3): Ditto.
12504 (doc/invoke.texi): Document new sha3 option.
12506 2018-01-10 Michael Collison <michael.collison@arm.com>
12508 * config/aarch64/aarch64-builtins.c:
12509 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12510 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12511 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12512 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12513 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12514 (AARCH64_ISA_SM4): New ISA flag.
12515 (TARGET_SM4): New feature flag for sm4.
12516 * config/aarch64/aarch64-simd-builtins.def
12517 (aarch64_sm3ss1qv4si): Ditto.
12518 (aarch64_sm3tt1aq4si): Ditto.
12519 (aarch64_sm3tt1bq4si): Ditto.
12520 (aarch64_sm3tt2aq4si): Ditto.
12521 (aarch64_sm3tt2bq4si): Ditto.
12522 (aarch64_sm3partw1qv4si): Ditto.
12523 (aarch64_sm3partw2qv4si): Ditto.
12524 (aarch64_sm4eqv4si): Ditto.
12525 (aarch64_sm4ekeyqv4si): Ditto.
12526 * config/aarch64/aarch64-simd.md:
12527 (aarch64_sm3ss1qv4si): Ditto.
12528 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12529 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12530 (aarch64_sm4eqv4si): Ditto.
12531 (aarch64_sm4ekeyqv4si): Ditto.
12532 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12533 (sm3part_op): Ditto.
12534 (CRYPTO_SM3TT): Ditto.
12535 (CRYPTO_SM3PART): Ditto.
12536 (UNSPEC_SM3SS1): New unspec.
12537 (UNSPEC_SM3TT1A): Ditto.
12538 (UNSPEC_SM3TT1B): Ditto.
12539 (UNSPEC_SM3TT2A): Ditto.
12540 (UNSPEC_SM3TT2B): Ditto.
12541 (UNSPEC_SM3PARTW1): Ditto.
12542 (UNSPEC_SM3PARTW2): Ditto.
12543 (UNSPEC_SM4E): Ditto.
12544 (UNSPEC_SM4EKEY): Ditto.
12545 * config/aarch64/constraints.md (Ui2): New constraint.
12546 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12547 * config/arm/types.md (crypto_sm3): New type attribute.
12548 (crypto_sm4): Ditto.
12549 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12550 (vsm3tt1aq_u32): Ditto.
12551 (vsm3tt1bq_u32): Ditto.
12552 (vsm3tt2aq_u32): Ditto.
12553 (vsm3tt2bq_u32): Ditto.
12554 (vsm3partw1q_u32): Ditto.
12555 (vsm3partw2q_u32): Ditto.
12556 (vsm4eq_u32): Ditto.
12557 (vsm4ekeyq_u32): Ditto.
12558 (doc/invoke.texi): Document new sm4 option.
12560 2018-01-10 Michael Collison <michael.collison@arm.com>
12562 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12563 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12564 (AARCH64_FL_FOR_ARCH8_4): New.
12565 (AARCH64_FL_V8_4): New flag.
12566 (doc/invoke.texi): Document new armv8.4-a option.
12568 2018-01-10 Michael Collison <michael.collison@arm.com>
12570 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12571 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12572 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12573 * config/aarch64/aarch64-option-extension.def: Add
12574 AARCH64_OPT_EXTENSION of 'sha2'.
12575 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12576 (crypto): Disable sha2 and aes if crypto disabled.
12577 (crypto): Enable aes and sha2 if enabled.
12578 (simd): Disable sha2 and aes if simd disabled.
12579 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12581 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12582 (TARGET_SHA2): New feature flag for sha2.
12583 (TARGET_AES): New feature flag for aes.
12584 * config/aarch64/aarch64-simd.md:
12585 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12586 conditional on TARGET_AES.
12587 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12588 (aarch64_crypto_sha1hsi): Make pattern conditional
12590 (aarch64_crypto_sha1hv4si): Ditto.
12591 (aarch64_be_crypto_sha1hv4si): Ditto.
12592 (aarch64_crypto_sha1su1v4si): Ditto.
12593 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12594 (aarch64_crypto_sha1su0v4si): Ditto.
12595 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12596 (aarch64_crypto_sha256su0v4si): Ditto.
12597 (aarch64_crypto_sha256su1v4si): Ditto.
12598 (doc/invoke.texi): Document new aes and sha2 options.
12600 2018-01-10 Martin Sebor <msebor@redhat.com>
12602 PR tree-optimization/83781
12603 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12606 2018-01-11 Martin Sebor <msebor@gmail.com>
12607 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12609 PR tree-optimization/83501
12610 PR tree-optimization/81703
12612 * tree-ssa-strlen.c (get_string_cst): Rename...
12613 (get_string_len): ...to this. Handle global constants.
12614 (handle_char_store): Adjust.
12616 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12617 Jim Wilson <jimw@sifive.com>
12619 * config/riscv/riscv-protos.h (riscv_output_return): New.
12620 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12621 (riscv_attribute_table, riscv_output_return),
12622 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12623 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12624 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12626 (riscv_expand_prologue): Add early return for naked function.
12627 (riscv_expand_epilogue): Likewise.
12628 (riscv_function_ok_for_sibcall): Return false for naked function.
12629 (riscv_set_current_function): New.
12630 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12631 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12632 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12633 * doc/extend.texi (RISC-V Function Attributes): New.
12635 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12637 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12638 check for 128-bit long double before checking TCmode.
12639 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12640 128-bit long doubles before checking TFmode or TCmode.
12641 (FLOAT128_IBM_P): Likewise.
12643 2018-01-10 Martin Sebor <msebor@redhat.com>
12645 PR tree-optimization/83671
12646 * builtins.c (c_strlen): Unconditionally return zero for the empty
12648 Use -Warray-bounds for warnings.
12649 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12650 for non-constant array indices with COMPONENT_REF, arrays of
12651 arrays, and pointers to arrays.
12652 (gimple_fold_builtin_strlen): Determine and set length range for
12653 non-constant character arrays.
12655 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12657 PR middle-end/81897
12658 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12661 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12663 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12665 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12668 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12669 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12670 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12671 indexed_or_indirect_operand predicate.
12672 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12673 (*vsx_le_perm_load_v8hi): Likewise.
12674 (*vsx_le_perm_load_v16qi): Likewise.
12675 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12676 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12677 (*vsx_le_perm_store_v8hi): Likewise.
12678 (*vsx_le_perm_store_v16qi): Likewise.
12679 (eight unnamed splitters): Likewise.
12681 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12683 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12684 * config/rs6000/emmintrin.h: Likewise.
12685 * config/rs6000/mmintrin.h: Likewise.
12686 * config/rs6000/xmmintrin.h: Likewise.
12688 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12691 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12693 * tree.c (tree_nop_conversion): Return true for location wrapper
12695 (maybe_wrap_with_location): New function.
12696 (selftest::check_strip_nops): New function.
12697 (selftest::test_location_wrappers): New function.
12698 (selftest::tree_c_tests): Call it.
12699 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12700 (maybe_wrap_with_location): New decl.
12701 (EXPR_LOCATION_WRAPPER_P): New macro.
12702 (location_wrapper_p): New inline function.
12703 (tree_strip_any_location_wrapper): New inline function.
12705 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12708 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12709 stack_realign_offset for the largest alignment of stack slot
12711 (ix86_find_max_used_stack_alignment): New function.
12712 (ix86_finalize_stack_frame_flags): Use it. Set
12713 max_used_stack_alignment if we don't realign stack.
12714 * config/i386/i386.h (machine_function): Add
12715 max_used_stack_alignment.
12717 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12719 * config/arm/arm.opt (-mbranch-cost): New option.
12720 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12723 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12726 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12727 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12729 2018-01-10 Richard Biener <rguenther@suse.de>
12732 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12733 early out so it also covers the case where we have a non-NULL
12736 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12738 PR tree-optimization/83753
12739 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12740 for non-strided grouped accesses if the number of elements is 1.
12742 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12745 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12746 * i386.h (TARGET_USE_GATHER): Define.
12747 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12749 2018-01-10 Martin Liska <mliska@suse.cz>
12752 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12753 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12755 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12756 CLEANUP_NO_PARTITIONING is not set.
12758 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12760 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12761 for vectors, as a partial revert of r254296.
12762 * rtl.h (const_vec_p): Delete.
12763 (const_vec_duplicate_p): Don't test for vector CONSTs.
12764 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12765 * expmed.c (make_tree): Likewise.
12768 * common.md (E, F): Use CONSTANT_P instead of checking for
12770 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12771 checking for CONST_VECTOR.
12773 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12775 PR middle-end/83575
12776 * predict.c (force_edge_cold): Handle in more sane way edges
12777 with no prediction.
12779 2018-01-09 Carl Love <cel@us.ibm.com>
12781 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12783 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12784 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12785 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12786 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12787 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12788 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12789 * config/rs6000/rs6000-protos.h: Add extern defition for
12790 rs6000_generate_float2_double_code.
12791 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12793 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12794 (float2_v2df): Add define_expand.
12796 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12799 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12800 op_mode in the force_to_mode call.
12802 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12804 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12805 instead of checking each element individually.
12806 (aarch64_evpc_uzp): Likewise.
12807 (aarch64_evpc_zip): Likewise.
12808 (aarch64_evpc_ext): Likewise.
12809 (aarch64_evpc_rev): Likewise.
12810 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12811 instead of checking each element individually. Return true without
12813 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12814 whether all selected elements come from the same input, instead of
12815 checking each element individually. Remove calls to gen_rtx_REG,
12816 start_sequence and end_sequence and instead assert that no rtl is
12819 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12821 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12822 order of HIGH and CONST checks.
12824 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12826 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12827 if the destination isn't an SSA_NAME.
12829 2018-01-09 Richard Biener <rguenther@suse.de>
12831 PR tree-optimization/83668
12832 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12834 (canonicalize_loop_form): ... here, renamed from ...
12835 (canonicalize_loop_closed_ssa_form): ... this and amended to
12836 swap successor edges for loop exit blocks to make us use
12837 the RPO order we need for initial schedule generation.
12839 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12841 PR tree-optimization/64811
12842 * match.pd: When optimizing comparisons with Inf, avoid
12843 introducing or losing exceptions from comparisons with NaN.
12845 2018-01-09 Martin Liska <mliska@suse.cz>
12848 * asan.c (shadow_mem_size): Add gcc_assert.
12850 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12852 Don't save registers in main().
12855 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12856 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12857 * config/avr/avr.c (avr_set_current_function): Don't error if
12858 naked, OS_task or OS_main are specified at the same time.
12859 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12861 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12863 * common/config/avr/avr-common.c (avr_option_optimization_table):
12864 Switch on -mmain-is-OS_task for optimizing compilations.
12866 2018-01-09 Richard Biener <rguenther@suse.de>
12868 PR tree-optimization/83572
12869 * graphite.c: Include cfganal.h.
12870 (graphite_transform_loops): Connect infinite loops to exit
12871 and remove fake edges at the end.
12873 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12875 * ipa-inline.c (edge_badness): Revert accidental checkin.
12877 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12880 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12881 symbols; not inline clones.
12883 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12886 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12887 hard registers. Formatting fixes.
12889 PR preprocessor/83722
12890 * gcc.c (try_generate_repro): Pass
12891 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12892 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12895 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12896 Kito Cheng <kito.cheng@gmail.com>
12898 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12899 (riscv_leaf_function_p): Delete.
12900 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12902 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12904 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12906 (do_ifelse): New function.
12907 (do_isel): New function.
12908 (do_sub3): New function.
12909 (do_add3): New function.
12910 (do_load_mask_compare): New function.
12911 (do_overlap_load_compare): New function.
12912 (expand_compare_loop): New function.
12913 (expand_block_compare): Call expand_compare_loop() when appropriate.
12914 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12915 option description.
12916 (-mblock-compare-inline-loop-limit): New option.
12918 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12921 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12922 Reverse order of second and third operands in first alternative.
12923 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12924 of first and second elements in UNSPEC_VPERMR vector.
12925 (altivec_expand_vec_perm_le): Likewise.
12927 2018-01-08 Jeff Law <law@redhat.com>
12929 PR rtl-optimizatin/81308
12930 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12931 (process_switch): If group_case_labels makes a change, then set
12933 (pass_convert_switch::execute): If a switch is converted, then
12934 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12936 PR rtl-optimization/81308
12937 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12940 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12942 PR target/83663 - Revert r255946
12943 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12944 generation for cases where splatting a value is not useful.
12945 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12946 across a vec_duplicate and a paradoxical subreg forming a vector
12947 mode to a vec_concat.
12949 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12951 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12952 -march=armv8.3-a variants.
12953 * config/arm/t-multilib: Likewise.
12954 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12956 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12958 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12960 (cceq_ior_compare_complement): Give it a name so I can use it, and
12961 change boolean_or_operator predicate to boolean_operator so it can
12962 be used to generate a crand.
12963 (eqne): New code iterator.
12964 (bd/bd_neg): New code_attrs.
12965 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12966 a single define_insn.
12967 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12968 decrement (bdnzt/bdnzf/bdzt/bdzf).
12969 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12970 with the new names of the branch decrement patterns, and added the
12971 names of the branch decrement conditional patterns.
12973 2018-01-08 Richard Biener <rguenther@suse.de>
12975 PR tree-optimization/83563
12976 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12979 2018-01-08 Richard Biener <rguenther@suse.de>
12981 PR middle-end/83713
12982 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12984 2018-01-08 Richard Biener <rguenther@suse.de>
12986 PR tree-optimization/83685
12987 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12988 references to abnormals.
12990 2018-01-08 Richard Biener <rguenther@suse.de>
12993 * dwarf2out.c (output_indirect_strings): Handle empty
12994 skeleton_debug_str_hash.
12995 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12997 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12999 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13000 (emit_store_direct): Likewise.
13001 (arc_trampoline_adjust_address): Likewise.
13002 (arc_asm_trampoline_template): New function.
13003 (arc_initialize_trampoline): Use asm_trampoline_template.
13004 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13005 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13006 * config/arc/arc.md (flush_icache): Delete pattern.
13008 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13010 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13011 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13014 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13017 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13018 by not USED_FOR_TARGET.
13019 (make_pass_resolve_sw_modes): Likewise.
13021 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13023 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13026 2018-01-08 Richard Biener <rguenther@suse.de>
13028 PR middle-end/83580
13029 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13031 2018-01-08 Richard Biener <rguenther@suse.de>
13033 PR middle-end/83517
13034 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13036 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13038 PR middle-end/81897
13039 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13040 basic blocks with a small number of successors.
13041 (convert_control_dep_chain_into_preds): Improve handling of
13043 (dump_predicates): Split apart into...
13044 (dump_pred_chain): ...here...
13045 (dump_pred_info): ...and here.
13046 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13047 (can_chain_union_be_invalidated_p): Improve check for invalidation
13049 (uninit_uses_cannot_happen): Avoid unnecessary if
13050 convert_control_dep_chain_into_preds yielded nothing.
13052 2018-01-06 Martin Sebor <msebor@redhat.com>
13054 PR tree-optimization/83640
13055 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13056 subtracting negative offset from size.
13057 (builtin_access::overlap): Adjust offset bounds of the access to fall
13058 within the size of the object if possible.
13060 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13062 PR rtl-optimization/83699
13063 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13064 extract_bit_field_as_subreg to cases in which the extracted
13065 value is also a vector.
13067 * lra-constraints.c (process_alt_operands): Test for the equivalence
13068 substitutions when detecting a possible reload cycle.
13070 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13073 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13074 by default if flag_selective_schedling{,2}. Formatting fixes.
13076 PR rtl-optimization/83682
13077 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13078 if it has non-VECTOR_MODE element mode.
13079 (vec_duplicate_p): Likewise.
13081 PR middle-end/83694
13082 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13083 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13085 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13088 * config/i386/i386-builtin.def
13089 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13090 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13091 Require also OPTION_MASK_ISA_AVX512F in addition to
13092 OPTION_MASK_ISA_GFNI.
13093 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13094 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13095 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13096 to OPTION_MASK_ISA_GFNI.
13097 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13098 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13099 OPTION_MASK_ISA_AVX512BW.
13100 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13101 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13102 addition to OPTION_MASK_ISA_GFNI.
13103 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13104 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13105 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13106 to OPTION_MASK_ISA_GFNI.
13107 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13108 a requirement for all ISAs rather than any of them with a few
13110 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13112 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13113 bitmasks to be enabled with 3 exceptions, instead of requiring any
13114 enabled ISA with lots of exceptions.
13115 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13116 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13117 Change avx512bw in isa attribute to avx512f.
13118 * config/i386/sgxintrin.h: Add license boilerplate.
13119 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13120 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13121 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13122 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13124 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13125 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13126 temporarily sse2 rather than sse if not enabled already.
13129 * config/i386/sse.md (VI248_VLBW): Rename to ...
13130 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13131 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13132 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13133 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13134 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13135 mode iterator instead of VI248_VLBW.
13137 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13139 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13140 (record_modified): Skip clobbers; add debug output.
13141 (param_change_prob): Use sreal frequencies.
13143 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13145 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13146 punt for user-aligned variables.
13148 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13150 * tree-chrec.c (chrec_contains_symbols): Return true for
13153 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13156 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13157 of (x|y) == x for BICS pattern.
13159 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13161 PR tree-optimization/83605
13162 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13163 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13166 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13168 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13169 * config/epiphany/rtems.h: New file.
13171 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13172 Uros Bizjak <ubizjak@gmail.com>
13175 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13176 QIreg_operand instead of register_operand predicate.
13177 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13178 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13179 comments instead of -fmitigate[-_]rop.
13181 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13184 * cgraphunit.c (symbol_table::compile): Switch to text_section
13185 before calling assembly_start debug hook.
13186 * run-rtl-passes.c (run_rtl_passes): Likewise.
13189 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13191 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13192 range_int_cst_p rather than !symbolic_range_p before calling
13193 extract_range_from_multiplicative_op_1.
13195 2018-01-04 Jeff Law <law@redhat.com>
13197 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13198 redundant test in assertion.
13200 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13202 * doc/rtl.texi: Document machine_mode wrapper classes.
13204 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13206 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13207 using tree_to_uhwi.
13209 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13211 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13212 the VEC_PERM_EXPR fold to fail.
13214 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13217 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13218 to switched_sections.
13220 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13223 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13224 test for d.testing.
13226 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13229 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13230 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13232 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13235 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13236 is BLKmode and bitpos not zero or mode change is needed.
13238 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13241 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13244 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13247 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13248 instead of MULT rtx. Update all corresponding splitters.
13249 (*saddl_se): Ditto.
13250 (*ssub<modesuffix>): Ditto.
13251 (*ssubl_se): Ditto.
13252 (*cmp_sadd_di): Update split patterns.
13253 (*cmp_sadd_si): Ditto.
13254 (*cmp_sadd_sidi): Ditto.
13255 (*cmp_ssub_di): Ditto.
13256 (*cmp_ssub_si): Ditto.
13257 (*cmp_ssub_sidi): Ditto.
13258 * config/alpha/predicates.md (const23_operand): New predicate.
13259 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13260 Look for ASHIFT, not MULT inner operand.
13261 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13263 2018-01-04 Martin Liska <mliska@suse.cz>
13265 PR gcov-profile/83669
13266 * gcov.c (output_intermediate_file): Add version to intermediate
13268 * doc/gcov.texi: Document new field 'version' in intermediate
13269 file format. Fix location of '-k' option of gcov command.
13271 2018-01-04 Martin Liska <mliska@suse.cz>
13274 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13276 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13278 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13280 2018-01-03 Martin Sebor <msebor@redhat.com>
13282 PR tree-optimization/83655
13283 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13284 checking calls with invalid arguments.
13286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13288 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13289 (vectorizable_mask_load_store): Delete.
13290 (vectorizable_call): Return false for masked loads and stores.
13291 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13292 instead of gimple_assign_rhs1.
13293 (vectorizable_load): Handle IFN_MASK_LOAD.
13294 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13298 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13300 (vectorizable_mask_load_store): ...here.
13301 (vectorizable_load): ...and here.
13303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13305 * tree-vect-stmts.c (vect_build_all_ones_mask)
13306 (vect_build_zero_merge_argument): New functions, split out from...
13307 (vectorizable_load): ...here.
13309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13311 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13313 (vectorizable_mask_load_store): ...here.
13314 (vectorizable_store): ...and here.
13316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13318 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13320 (vectorizable_mask_load_store): ...here.
13322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13324 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13325 (vect_model_store_cost): Take a vec_load_store_type instead of a
13327 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13328 (vect_model_store_cost): Take a vec_load_store_type instead of a
13330 (vectorizable_mask_load_store): Update accordingly.
13331 (vectorizable_store): Likewise.
13332 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13336 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13337 IFN_MASK_LOAD calls here rather than...
13338 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13341 Alan Hayward <alan.hayward@arm.com>
13342 David Sherwood <david.sherwood@arm.com>
13344 * expmed.c (extract_bit_field_1): For vector extracts,
13345 fall back to extract_bit_field_as_subreg if vec_extract
13348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13349 Alan Hayward <alan.hayward@arm.com>
13350 David Sherwood <david.sherwood@arm.com>
13352 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13353 they are variable or constant sized.
13354 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13355 slots for constant-sized data.
13357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13358 Alan Hayward <alan.hayward@arm.com>
13359 David Sherwood <david.sherwood@arm.com>
13361 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13362 handling COND_EXPRs with boolean comparisons, try to find a better
13363 basis for the mask type than the boolean itself.
13365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13367 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13368 is calculated and how it can be overridden.
13369 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13370 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13372 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13376 Alan Hayward <alan.hayward@arm.com>
13377 David Sherwood <david.sherwood@arm.com>
13379 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13380 Remove the mode argument.
13381 (aarch64_simd_valid_immediate): Remove the mode and inverse
13383 * config/aarch64/iterators.md (bitsize): New iterator.
13384 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13385 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13386 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13387 aarch64_simd_valid_immediate.
13388 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13389 (aarch64_reg_or_bic_imm): Likewise.
13390 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13391 with an insn_type enum and msl with a modifier_type enum.
13392 Replace element_width with a scalar_mode. Change the shift
13393 to unsigned int. Add constructors for scalar_float_mode and
13394 scalar_int_mode elements.
13395 (aarch64_vect_float_const_representable_p): Delete.
13396 (aarch64_can_const_movi_rtx_p)
13397 (aarch64_simd_scalar_immediate_valid_for_move)
13398 (aarch64_simd_make_constant): Update call to
13399 aarch64_simd_valid_immediate.
13400 (aarch64_advsimd_valid_immediate_hs): New function.
13401 (aarch64_advsimd_valid_immediate): Likewise.
13402 (aarch64_simd_valid_immediate): Remove mode and inverse
13403 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13404 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13405 and aarch64_float_const_representable_p on the result.
13406 (aarch64_output_simd_mov_immediate): Remove mode argument.
13407 Update call to aarch64_simd_valid_immediate and use of
13408 simd_immediate_info.
13409 (aarch64_output_scalar_simd_mov_immediate): Update call
13412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13413 Alan Hayward <alan.hayward@arm.com>
13414 David Sherwood <david.sherwood@arm.com>
13416 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13417 (mode_nunits): Likewise CONST_MODE_NUNITS.
13418 * machmode.def (ADJUST_NUNITS): Document.
13419 * genmodes.c (mode_data::need_nunits_adj): New field.
13420 (blank_mode): Update accordingly.
13421 (adj_nunits): New variable.
13422 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13424 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13425 listed in adj_nunits.
13426 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13427 listed in adj_nunits. Don't emit case statements for such modes.
13428 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13429 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13430 nothing if adj_nunits is nonnull.
13431 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13432 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13433 (emit_mode_fbit): Update use of print_maybe_const_decl.
13434 (emit_move_size): Likewise. Treat the array as non-const
13436 (emit_mode_adjustments): Handle adj_nunits.
13438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13440 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13441 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13442 (VECTOR_MODES): Use it.
13443 (make_vector_modes): Take the prefix as an argument.
13445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13446 Alan Hayward <alan.hayward@arm.com>
13447 David Sherwood <david.sherwood@arm.com>
13449 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13450 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13451 for MODE_VECTOR_BOOL.
13452 * machmode.def (VECTOR_BOOL_MODE): Document.
13453 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13454 (make_vector_bool_mode): New function.
13455 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13457 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13458 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13460 * stor-layout.c (int_mode_for_mode): Likewise.
13461 * tree.c (build_vector_type_for_mode): Likewise.
13462 * varasm.c (output_constant_pool_2): Likewise.
13463 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13464 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13465 for MODE_VECTOR_BOOL.
13466 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13467 of mode class checks.
13468 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13469 instead of a list of mode class checks.
13470 (expand_vector_scalar_condition): Likewise.
13471 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13474 Alan Hayward <alan.hayward@arm.com>
13475 David Sherwood <david.sherwood@arm.com>
13477 * machmode.h (mode_size): Change from unsigned short to
13479 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13480 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13481 or if measurement_type is not polynomial.
13482 (fixed_size_mode::includes_p): Check for constant-sized modes.
13483 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13484 return a poly_uint16 rather than an unsigned short.
13485 (emit_mode_size): Change the type of mode_size from unsigned short
13486 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13487 (emit_mode_adjustments): Cope with polynomial vector sizes.
13488 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13490 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13492 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13493 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13494 * caller-save.c (setup_save_areas): Likewise.
13495 (replace_reg_with_saved_mem): Likewise.
13496 * calls.c (emit_library_call_value_1): Likewise.
13497 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13498 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13499 (gen_lowpart_for_combine): Likewise.
13500 * convert.c (convert_to_integer_1): Likewise.
13501 * cse.c (equiv_constant, cse_insn): Likewise.
13502 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13503 (cselib_subst_to_values): Likewise.
13504 * dce.c (word_dce_process_block): Likewise.
13505 * df-problems.c (df_word_lr_mark_ref): Likewise.
13506 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13507 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13508 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13509 (rtl_for_decl_location): Likewise.
13510 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13511 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13512 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13513 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13514 (expand_expr_real_1): Likewise.
13515 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13516 (pad_below): Likewise.
13517 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13518 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13519 * ira.c (get_subreg_tracking_sizes): Likewise.
13520 * ira-build.c (ira_create_allocno_objects): Likewise.
13521 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13522 (ira_sort_regnos_for_alter_reg): Likewise.
13523 * ira-costs.c (record_operand_costs): Likewise.
13524 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13525 (resolve_simple_move): Likewise.
13526 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13527 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13528 (lra_constraints): Likewise.
13529 (CONST_POOL_OK_P): Reject variable-sized modes.
13530 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13531 (add_pseudo_to_slot, lra_spill): Likewise.
13532 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13533 * optabs-query.c (get_best_extraction_insn): Likewise.
13534 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13535 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13536 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13537 * recog.c (offsettable_address_addr_space_p): Likewise.
13538 * regcprop.c (maybe_mode_change): Likewise.
13539 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13540 * regrename.c (build_def_use): Likewise.
13541 * regstat.c (dump_reg_info): Likewise.
13542 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13543 (find_reloads, find_reloads_subreg_address): Likewise.
13544 * reload1.c (eliminate_regs_1): Likewise.
13545 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13546 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13547 (simplify_binary_operation_1, simplify_subreg): Likewise.
13548 * targhooks.c (default_function_arg_padding): Likewise.
13549 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13550 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13551 (verify_gimple_assign_ternary): Likewise.
13552 * tree-inline.c (estimate_move_cost): Likewise.
13553 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13554 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13555 (get_address_cost_ainc): Likewise.
13556 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13557 (vect_supportable_dr_alignment): Likewise.
13558 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13559 (vectorizable_reduction): Likewise.
13560 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13561 (vectorizable_operation, vectorizable_load): Likewise.
13562 * tree.c (build_same_sized_truth_vector_type): Likewise.
13563 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13564 * var-tracking.c (emit_note_insn_var_location): Likewise.
13565 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13566 (ADDR_VEC_ALIGN): Likewise.
13568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13569 Alan Hayward <alan.hayward@arm.com>
13570 David Sherwood <david.sherwood@arm.com>
13572 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13574 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13575 or if measurement_type is polynomial.
13576 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13577 * combine.c (make_extraction): Likewise.
13578 * dse.c (find_shift_sequence): Likewise.
13579 * dwarf2out.c (mem_loc_descriptor): Likewise.
13580 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13581 (extract_bit_field, extract_low_bits): Likewise.
13582 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13583 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13584 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13585 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13586 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13587 * reload.c (find_reloads): Likewise.
13588 * reload1.c (alter_reg): Likewise.
13589 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13590 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13591 * tree-if-conv.c (predicate_mem_writes): Likewise.
13592 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13593 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13594 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13595 * valtrack.c (dead_debug_insert_temp): Likewise.
13596 * varasm.c (mergeable_constant_section): Likewise.
13597 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13600 Alan Hayward <alan.hayward@arm.com>
13601 David Sherwood <david.sherwood@arm.com>
13603 * expr.c (expand_assignment): Cope with polynomial mode sizes
13604 when assigning to a CONCAT.
13606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13607 Alan Hayward <alan.hayward@arm.com>
13608 David Sherwood <david.sherwood@arm.com>
13610 * machmode.h (mode_precision): Change from unsigned short to
13612 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13614 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13615 or if measurement_type is not polynomial.
13616 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13617 in which the mode is already known to be a scalar_int_mode.
13618 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13619 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13621 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13622 for GET_MODE_PRECISION.
13623 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13624 for GET_MODE_PRECISION.
13625 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13627 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13628 (expand_field_assignment, make_extraction): Likewise.
13629 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13630 (get_last_value): Likewise.
13631 * convert.c (convert_to_integer_1): Likewise.
13632 * cse.c (cse_insn): Likewise.
13633 * expr.c (expand_expr_real_1): Likewise.
13634 * lra-constraints.c (simplify_operand_subreg): Likewise.
13635 * optabs-query.c (can_atomic_load_p): Likewise.
13636 * optabs.c (expand_atomic_load): Likewise.
13637 (expand_atomic_store): Likewise.
13638 * ree.c (combine_reaching_defs): Likewise.
13639 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13640 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13641 * tree.h (type_has_mode_precision_p): Likewise.
13642 * ubsan.c (instrument_si_overflow): Likewise.
13644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13645 Alan Hayward <alan.hayward@arm.com>
13646 David Sherwood <david.sherwood@arm.com>
13648 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13649 polynomial numbers of units.
13650 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13651 (valid_vector_subparts_p): New function.
13652 (build_vector_type): Remove temporary shim and take the number
13653 of units as a poly_uint64 rather than an int.
13654 (build_opaque_vector_type): Take the number of units as a
13655 poly_uint64 rather than an int.
13656 * tree.c (build_vector_from_ctor): Handle polynomial
13657 TYPE_VECTOR_SUBPARTS.
13658 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13659 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13660 (build_vector_from_val): If the number of units is variable,
13661 use build_vec_duplicate_cst for constant operands and
13662 VEC_DUPLICATE_EXPR otherwise.
13663 (make_vector_type): Remove temporary is_constant ().
13664 (build_vector_type, build_opaque_vector_type): Take the number of
13665 units as a poly_uint64 rather than an int.
13666 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13668 * cfgexpand.c (expand_debug_expr): Likewise.
13669 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13670 (store_constructor, expand_expr_real_1): Likewise.
13671 (const_scalar_mask_from_tree): Likewise.
13672 * fold-const-call.c (fold_const_reduction): Likewise.
13673 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13674 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13675 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13676 (fold_relational_const): Likewise.
13677 (native_interpret_vector): Likewise. Change the size from an
13678 int to an unsigned int.
13679 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13680 TYPE_VECTOR_SUBPARTS.
13681 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13682 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13683 duplicating a non-constant operand into a variable-length vector.
13684 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13685 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13686 * ipa-icf.c (sem_variable::equals): Likewise.
13687 * match.pd: Likewise.
13688 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13689 * print-tree.c (print_node): Likewise.
13690 * stor-layout.c (layout_type): Likewise.
13691 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13692 * tree-cfg.c (verify_gimple_comparison): Likewise.
13693 (verify_gimple_assign_binary): Likewise.
13694 (verify_gimple_assign_ternary): Likewise.
13695 (verify_gimple_assign_single): Likewise.
13696 * tree-pretty-print.c (dump_generic_node): Likewise.
13697 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13698 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13699 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13700 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13701 (vect_shift_permute_load_chain): Likewise.
13702 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13703 (expand_vector_condition, optimize_vector_constructor): Likewise.
13704 (lower_vec_perm, get_compute_type): Likewise.
13705 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13706 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13707 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13708 (vect_recog_mask_conversion_pattern): Likewise.
13709 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13710 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13711 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13712 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13713 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13714 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13715 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13716 (supportable_widening_operation): Likewise.
13717 (supportable_narrowing_operation): Likewise.
13718 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13720 * varasm.c (output_constant): Likewise.
13722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13723 Alan Hayward <alan.hayward@arm.com>
13724 David Sherwood <david.sherwood@arm.com>
13726 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13727 so that both the length == 3 and length != 3 cases set up their
13728 own permute vectors. Add comments explaining why we know the
13729 number of elements is constant.
13730 (vect_permute_load_chain): Likewise.
13732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13733 Alan Hayward <alan.hayward@arm.com>
13734 David Sherwood <david.sherwood@arm.com>
13736 * machmode.h (mode_nunits): Change from unsigned char to
13738 (ONLY_FIXED_SIZE_MODES): New macro.
13739 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13740 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13741 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13743 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13744 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13745 or if measurement_type is not polynomial.
13746 * genmodes.c (ZERO_COEFFS): New macro.
13747 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13749 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13750 Use ZERO_COEFFS when emitting initializers.
13751 * data-streamer.h (bp_pack_poly_value): New function.
13752 (bp_unpack_poly_value): Likewise.
13753 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13754 for GET_MODE_NUNITS.
13755 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13756 for GET_MODE_NUNITS.
13757 * tree.c (make_vector_type): Remove temporary shim and make
13758 the real function take the number of units as a poly_uint64
13759 rather than an int.
13760 (build_vector_type_for_mode): Handle polynomial nunits.
13761 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13762 * emit-rtl.c (const_vec_series_p_1): Likewise.
13763 (gen_rtx_CONST_VECTOR): Likewise.
13764 * fold-const.c (test_vec_duplicate_folding): Likewise.
13765 * genrecog.c (validate_pattern): Likewise.
13766 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13767 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13768 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13769 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13770 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13771 * rtlanal.c (subreg_get_info): Likewise.
13772 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13773 (vect_grouped_load_supported): Likewise.
13774 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13775 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13776 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13777 (simplify_const_unary_operation, simplify_binary_operation_1)
13778 (simplify_const_binary_operation, simplify_ternary_operation)
13779 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13780 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13781 instead of CONST_VECTOR_NUNITS.
13782 * varasm.c (output_constant_pool_2): Likewise.
13783 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13784 explicit-encoded elements in the XVEC for variable-length vectors.
13786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13788 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13791 Alan Hayward <alan.hayward@arm.com>
13792 David Sherwood <david.sherwood@arm.com>
13794 * coretypes.h (fixed_size_mode): Declare.
13795 (fixed_size_mode_pod): New typedef.
13796 * builtins.h (target_builtins::x_apply_args_mode)
13797 (target_builtins::x_apply_result_mode): Change type to
13798 fixed_size_mode_pod.
13799 * builtins.c (apply_args_size, apply_result_size, result_vector)
13800 (expand_builtin_apply_args_1, expand_builtin_apply)
13801 (expand_builtin_return): Update accordingly.
13803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13805 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13806 * cselib.c (cselib_hash_rtx): Likewise.
13807 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13808 CONST_VECTOR encoding.
13810 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13811 Jeff Law <law@redhat.com>
13814 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13815 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13816 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13817 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13820 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13821 explicitly probe *sp in a noreturn function if there were any callee
13822 register saves or frame pointer is needed.
13824 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13827 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13828 BLKmode for ternary, binary or unary expressions.
13831 * var-tracking.c (delete_vta_debug_insn): New inline function.
13832 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13833 insns from get_insns () to NULL instead of each bb separately.
13834 Use delete_vta_debug_insn. No longer static.
13835 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13836 delete_vta_debug_insns callers.
13837 * rtl.h (delete_vta_debug_insns): Declare.
13838 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13839 instead of variable_tracking_main.
13841 2018-01-03 Martin Sebor <msebor@redhat.com>
13843 PR tree-optimization/83603
13844 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13845 arguments past the endof the argument list in functions declared
13846 without a prototype.
13847 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13848 Avoid checking when arguments are null.
13850 2018-01-03 Martin Sebor <msebor@redhat.com>
13853 * doc/extend.texi (attribute const): Fix a typo.
13854 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13855 issuing -Wsuggest-attribute for void functions.
13857 2018-01-03 Martin Sebor <msebor@redhat.com>
13859 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13860 offset_int::from instead of wide_int::to_shwi.
13861 (maybe_diag_overlap): Remove assertion.
13862 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13863 * gimple-ssa-sprintf.c (format_directive): Same.
13864 (parse_directive): Same.
13865 (sprintf_dom_walker::compute_format_length): Same.
13866 (try_substitute_return_value): Same.
13868 2018-01-03 Jeff Law <law@redhat.com>
13870 PR middle-end/83654
13871 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13872 non-constant residual for zero at runtime and avoid probing in
13873 that case. Reorganize code for trailing problem to mirror handling
13876 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13878 PR tree-optimization/83501
13879 * tree-ssa-strlen.c (get_string_cst): New.
13880 (handle_char_store): Call get_string_cst.
13882 2018-01-03 Martin Liska <mliska@suse.cz>
13884 PR tree-optimization/83593
13885 * tree-ssa-strlen.c: Include tree-cfg.h.
13886 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13887 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13888 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13890 (strlen_dom_walker::before_dom_children): Call
13891 gimple_purge_dead_eh_edges. Dump tranformation with details
13893 (strlen_dom_walker::before_dom_children): Update call by adding
13894 new argument cleanup_eh.
13895 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13897 2018-01-03 Martin Liska <mliska@suse.cz>
13900 * cif-code.def (VARIADIC_THUNK): New enum value.
13901 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13904 2018-01-03 Jan Beulich <jbeulich@suse.com>
13906 * sse.md (mov<mode>_internal): Tighten condition for when to use
13907 vmovdqu<ssescalarsize> for TI and OI modes.
13909 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13911 Update copyright years.
13913 2018-01-03 Martin Liska <mliska@suse.cz>
13916 * ipa-visibility.c (function_and_variable_visibility): Skip
13917 functions with noipa attribure.
13919 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13921 * gcc.c (process_command): Update copyright notice dates.
13922 * gcov-dump.c (print_version): Ditto.
13923 * gcov.c (print_version): Ditto.
13924 * gcov-tool.c (print_version): Ditto.
13925 * gengtype.c (create_file): Ditto.
13926 * doc/cpp.texi: Bump @copying's copyright year.
13927 * doc/cppinternals.texi: Ditto.
13928 * doc/gcc.texi: Ditto.
13929 * doc/gccint.texi: Ditto.
13930 * doc/gcov.texi: Ditto.
13931 * doc/install.texi: Ditto.
13932 * doc/invoke.texi: Ditto.
13934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13936 * vector-builder.h (vector_builder::m_full_nelts): Change from
13937 unsigned int to poly_uint64.
13938 (vector_builder::full_nelts): Update prototype accordingly.
13939 (vector_builder::new_vector): Likewise.
13940 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13941 (vector_builder::operator ==): Likewise.
13942 (vector_builder::finalize): Likewise.
13943 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13944 Take the number of elements as a poly_uint64 rather than an
13946 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13947 from unsigned int to poly_uint64.
13948 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13949 (vec_perm_indices::new_vector): Likewise.
13950 (vec_perm_indices::length): Likewise.
13951 (vec_perm_indices::nelts_per_input): Likewise.
13952 (vec_perm_indices::input_nelts): Likewise.
13953 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13954 number of elements per input as a poly_uint64 rather than an
13955 unsigned int. Use the original encoding for variable-length
13956 vectors, rather than clamping each individual element.
13957 For the second and subsequent elements in each pattern,
13958 clamp the step and base before clamping their sum.
13959 (vec_perm_indices::series_p): Handle polynomial element counts.
13960 (vec_perm_indices::all_in_range_p): Likewise.
13961 (vec_perm_indices_to_tree): Likewise.
13962 (vec_perm_indices_to_rtx): Likewise.
13963 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13964 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13965 (tree_vector_builder::new_binary_operation): Handle polynomial
13966 element counts. Return false if we need to know the number
13967 of elements at compile time.
13968 * fold-const.c (fold_vec_perm): Punt if the number of elements
13969 isn't known at compile time.
13971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13973 * vec-perm-indices.h (vec_perm_builder): Change element type
13974 from HOST_WIDE_INT to poly_int64.
13975 (vec_perm_indices::element_type): Update accordingly.
13976 (vec_perm_indices::clamp): Handle polynomial element_types.
13977 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13978 (vec_perm_indices::all_in_range_p): Likewise.
13979 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13981 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13982 polynomial vec_perm_indices element types.
13983 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13984 * fold-const.c (fold_vec_perm): Likewise.
13985 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13986 * tree-vect-generic.c (lower_vec_perm): Likewise.
13987 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13988 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13989 element type to HOST_WIDE_INT.
13991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13992 Alan Hayward <alan.hayward@arm.com>
13993 David Sherwood <david.sherwood@arm.com>
13995 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13996 rather than an int. Use plus_constant.
13997 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13998 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14001 Alan Hayward <alan.hayward@arm.com>
14002 David Sherwood <david.sherwood@arm.com>
14004 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14005 a HOST_WIDE_INT to a poly_int64.
14007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14008 Alan Hayward <alan.hayward@arm.com>
14009 David Sherwood <david.sherwood@arm.com>
14011 * calls.c (load_register_parameters): Cope with polynomial
14012 mode sizes. Require a constant size for BLKmode parameters
14013 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14014 forces a parameter to be padded at the lsb end in order to
14015 fill a complete number of words, require the parameter size
14016 to be ordered wrt UNITS_PER_WORD.
14018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14019 Alan Hayward <alan.hayward@arm.com>
14020 David Sherwood <david.sherwood@arm.com>
14022 * reload1.c (spill_stack_slot_width): Change element type
14023 from unsigned int to poly_uint64_pod.
14024 (alter_reg): Treat mode sizes as polynomial.
14026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14027 Alan Hayward <alan.hayward@arm.com>
14028 David Sherwood <david.sherwood@arm.com>
14030 * reload.c (complex_word_subreg_p): New function.
14031 (reload_inner_reg_of_subreg, push_reload): Use it.
14033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14034 Alan Hayward <alan.hayward@arm.com>
14035 David Sherwood <david.sherwood@arm.com>
14037 * lra-constraints.c (process_alt_operands): Reject matched
14038 operands whose sizes aren't ordered.
14039 (match_reload): Refer to this check here.
14041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14042 Alan Hayward <alan.hayward@arm.com>
14043 David Sherwood <david.sherwood@arm.com>
14045 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14046 that the mode size is in the set {1, 2, 4, 8, 16}.
14048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14049 Alan Hayward <alan.hayward@arm.com>
14050 David Sherwood <david.sherwood@arm.com>
14052 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14053 Use plus_constant instead of gen_rtx_PLUS.
14055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14056 Alan Hayward <alan.hayward@arm.com>
14057 David Sherwood <david.sherwood@arm.com>
14059 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14060 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14061 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14062 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14063 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14064 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14065 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14066 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14067 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14068 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14070 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14071 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14072 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14073 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14074 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14075 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14076 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14077 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14078 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14079 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14081 * expr.c (emit_move_resolve_push): Treat the input and result
14082 of PUSH_ROUNDING as a poly_int64.
14083 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14084 (emit_push_insn): Likewise.
14085 * lra-eliminations.c (mark_not_eliminable): Likewise.
14086 * recog.c (push_operand): Likewise.
14087 * reload1.c (elimination_effects): Likewise.
14088 * rtlanal.c (nonzero_bits1): Likewise.
14089 * calls.c (store_one_arg): Likewise. Require the padding to be
14090 known at compile time.
14092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14093 Alan Hayward <alan.hayward@arm.com>
14094 David Sherwood <david.sherwood@arm.com>
14096 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14097 Use plus_constant instead of gen_rtx_PLUS.
14099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14100 Alan Hayward <alan.hayward@arm.com>
14101 David Sherwood <david.sherwood@arm.com>
14103 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14104 rather than an int.
14106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14107 Alan Hayward <alan.hayward@arm.com>
14108 David Sherwood <david.sherwood@arm.com>
14110 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14111 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14112 via stack temporaries. Treat the mode size as polynomial too.
14114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14115 Alan Hayward <alan.hayward@arm.com>
14116 David Sherwood <david.sherwood@arm.com>
14118 * expr.c (expand_expr_real_2): When handling conversions involving
14119 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14120 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14121 as a poly_uint64 too.
14123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14124 Alan Hayward <alan.hayward@arm.com>
14125 David Sherwood <david.sherwood@arm.com>
14127 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * combine.c (can_change_dest_mode): Handle polynomial
14134 REGMODE_NATURAL_SIZE.
14135 * expmed.c (store_bit_field_1): Likewise.
14136 * expr.c (store_constructor): Likewise.
14137 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14138 and polynomial REGMODE_NATURAL_SIZE.
14139 (gen_lowpart_common): Likewise.
14140 * reginfo.c (record_subregs_of_mode): Likewise.
14141 * rtlanal.c (read_modify_subreg_p): Likewise.
14143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14144 Alan Hayward <alan.hayward@arm.com>
14145 David Sherwood <david.sherwood@arm.com>
14147 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14148 numbers of elements.
14150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14151 Alan Hayward <alan.hayward@arm.com>
14152 David Sherwood <david.sherwood@arm.com>
14154 * match.pd: Cope with polynomial numbers of vector elements.
14156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14157 Alan Hayward <alan.hayward@arm.com>
14158 David Sherwood <david.sherwood@arm.com>
14160 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14161 in a POINTER_PLUS_EXPR.
14163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14164 Alan Hayward <alan.hayward@arm.com>
14165 David Sherwood <david.sherwood@arm.com>
14167 * omp-simd-clone.c (simd_clone_subparts): New function.
14168 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14169 (ipa_simd_modify_function_body): Likewise.
14171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14172 Alan Hayward <alan.hayward@arm.com>
14173 David Sherwood <david.sherwood@arm.com>
14175 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14176 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14177 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14178 (expand_vector_condition, vector_element): Likewise.
14179 (subparts_gt): New function.
14180 (get_compute_type): Use subparts_gt.
14181 (count_type_subparts): Delete.
14182 (expand_vector_operations_1): Use subparts_gt instead of
14183 count_type_subparts.
14185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14186 Alan Hayward <alan.hayward@arm.com>
14187 David Sherwood <david.sherwood@arm.com>
14189 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14190 (vect_compile_time_alias): ...this new function. Do the calculation
14191 on poly_ints rather than trees.
14192 (vect_prune_runtime_alias_test_list): Update call accordingly.
14194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14195 Alan Hayward <alan.hayward@arm.com>
14196 David Sherwood <david.sherwood@arm.com>
14198 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14200 (vect_schedule_slp_instance): Likewise.
14202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14203 Alan Hayward <alan.hayward@arm.com>
14204 David Sherwood <david.sherwood@arm.com>
14206 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14207 constant and extern definitions for variable-length vectors.
14208 (vect_get_constant_vectors): Note that the number of units
14209 is known to be constant.
14211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14212 Alan Hayward <alan.hayward@arm.com>
14213 David Sherwood <david.sherwood@arm.com>
14215 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14216 of units as polynomial. Choose between WIDE and NARROW based
14219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14220 Alan Hayward <alan.hayward@arm.com>
14221 David Sherwood <david.sherwood@arm.com>
14223 * tree-vect-stmts.c (simd_clone_subparts): New function.
14224 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14227 Alan Hayward <alan.hayward@arm.com>
14228 David Sherwood <david.sherwood@arm.com>
14230 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14231 vectors as polynomial. Use build_index_vector for
14232 IFN_GOMP_SIMD_LANE.
14234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14235 Alan Hayward <alan.hayward@arm.com>
14236 David Sherwood <david.sherwood@arm.com>
14238 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14239 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14240 for variable-length vectors.
14241 (vectorizable_mask_load_store): Treat the number of units as
14242 polynomial, asserting that it is constant if the condition has
14243 already been enforced.
14244 (vectorizable_store, vectorizable_load): Likewise.
14246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14247 Alan Hayward <alan.hayward@arm.com>
14248 David Sherwood <david.sherwood@arm.com>
14250 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14251 of units as polynomial. Punt if we can't tell at compile time
14252 which vector contains the final result.
14254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14255 Alan Hayward <alan.hayward@arm.com>
14256 David Sherwood <david.sherwood@arm.com>
14258 * tree-vect-loop.c (vectorizable_induction): Treat the number
14259 of units as polynomial. Punt on SLP inductions. Use an integer
14260 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14261 cast of such a series for variable-length floating-point
14264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14265 Alan Hayward <alan.hayward@arm.com>
14266 David Sherwood <david.sherwood@arm.com>
14268 * tree.h (build_index_vector): Declare.
14269 * tree.c (build_index_vector): New function.
14270 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14271 of units as polynomial, forcibly converting it to a constant if
14272 vectorizable_reduction has already enforced the condition.
14273 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14274 to create a {1,2,3,...} vector.
14275 (vectorizable_reduction): Treat the number of units as polynomial.
14276 Choose vectype_in based on the largest scalar element size rather
14277 than the smallest number of units. Enforce the restrictions
14280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14281 Alan Hayward <alan.hayward@arm.com>
14282 David Sherwood <david.sherwood@arm.com>
14284 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14285 number of units as polynomial.
14287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14288 Alan Hayward <alan.hayward@arm.com>
14289 David Sherwood <david.sherwood@arm.com>
14291 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14292 * target.def (autovectorize_vector_sizes): Return the vector sizes
14293 by pointer, using vector_sizes rather than a bitmask.
14294 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14295 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14296 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14298 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14299 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14300 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14301 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14302 * omp-general.c (omp_max_vf): Likewise.
14303 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14304 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14305 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14306 * tree-vect-slp.c (vect_slp_bb): Likewise.
14307 * doc/tm.texi: Regenerate.
14308 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14310 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14311 the vector size as a poly_uint64 rather than an unsigned int.
14312 (current_vector_size): Change from an unsigned int to a poly_uint64.
14313 (get_vectype_for_scalar_type): Update accordingly.
14314 * tree.h (build_truth_vector_type): Take the size and number of
14315 units as a poly_uint64 rather than an unsigned int.
14316 (build_vector_type): Add a temporary overload that takes
14317 the number of units as a poly_uint64 rather than an unsigned int.
14318 * tree.c (make_vector_type): Likewise.
14319 (build_truth_vector_type): Take the number of units as a poly_uint64
14320 rather than an unsigned int.
14322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14323 Alan Hayward <alan.hayward@arm.com>
14324 David Sherwood <david.sherwood@arm.com>
14326 * target.def (get_mask_mode): Take the number of units and length
14327 as poly_uint64s rather than unsigned ints.
14328 * targhooks.h (default_get_mask_mode): Update accordingly.
14329 * targhooks.c (default_get_mask_mode): Likewise.
14330 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14331 * doc/tm.texi: Regenerate.
14333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14334 Alan Hayward <alan.hayward@arm.com>
14335 David Sherwood <david.sherwood@arm.com>
14337 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14338 * omp-general.c (omp_max_vf): Likewise.
14339 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14340 (expand_omp_simd): Handle polynomial safelen.
14341 * omp-low.c (omplow_simd_context): Add a default constructor.
14342 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14343 (lower_rec_simd_input_clauses): Update accordingly.
14344 (lower_rec_input_clauses): Likewise.
14346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14347 Alan Hayward <alan.hayward@arm.com>
14348 David Sherwood <david.sherwood@arm.com>
14350 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14351 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14352 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14353 (vect_analyze_slp_cost): Likewise.
14354 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14355 (vect_model_load_cost): Likewise.
14357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14358 Alan Hayward <alan.hayward@arm.com>
14359 David Sherwood <david.sherwood@arm.com>
14361 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14362 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14363 from an unsigned int * to a poly_uint64_pod *.
14364 (calculate_unrolling_factor): New function.
14365 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14368 Alan Hayward <alan.hayward@arm.com>
14369 David Sherwood <david.sherwood@arm.com>
14371 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14372 from an unsigned int to a poly_uint64.
14373 (_loop_vec_info::slp_unrolling_factor): Likewise.
14374 (_loop_vec_info::vectorization_factor): Change from an int
14376 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14377 (vect_get_num_vectors): New function.
14378 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14379 (vect_get_num_copies): Use vect_get_num_vectors.
14380 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14381 to an unsigned int *.
14382 (vect_analyze_data_refs): Change min_vf from an int * to a
14384 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14385 than an unsigned HOST_WIDE_INT.
14386 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14387 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14388 to an unsigned int *.
14389 (vect_analyze_data_ref_dependences): Likewise.
14390 (vect_compute_data_ref_alignment): Handle polynomial vf.
14391 (vect_enhance_data_refs_alignment): Likewise.
14392 (vect_prune_runtime_alias_test_list): Likewise.
14393 (vect_shift_permute_load_chain): Likewise.
14394 (vect_supportable_dr_alignment): Likewise.
14395 (dependence_distance_ge_vf): Take the vectorization factor as a
14396 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14397 (vect_analyze_data_refs): Change min_vf from an int * to a
14399 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14400 vfm1 as a poly_uint64 rather than an int. Make the same change
14401 for the returned bound_scalar.
14402 (vect_gen_vector_loop_niters): Handle polynomial vf.
14403 (vect_do_peeling): Likewise. Update call to
14404 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14405 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14407 * tree-vect-loop.c (vect_determine_vectorization_factor)
14408 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14409 (vect_get_known_peeling_cost): Likewise.
14410 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14411 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14412 (vect_transform_loop): Likewise. Use the lowest possible VF when
14413 updating the upper bounds of the loop.
14414 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14415 rather than an int.
14416 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14417 polynomial unroll factors.
14418 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14419 (vect_make_slp_decision): Likewise.
14420 (vect_supported_load_permutation_p): Likewise, and polynomial
14422 (vect_analyze_slp_cost): Handle polynomial vf.
14423 (vect_slp_analyze_node_operations): Likewise.
14424 (vect_slp_analyze_bb_1): Likewise.
14425 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14426 than an unsigned HOST_WIDE_INT.
14427 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14428 (vectorizable_load): Handle polynomial vf.
14429 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14431 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14434 Alan Hayward <alan.hayward@arm.com>
14435 David Sherwood <david.sherwood@arm.com>
14437 * match.pd: Handle bit operations involving three constants
14438 and try to fold one pair.
14440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14442 * tree-vect-loop-manip.c: Include gimple-fold.h.
14443 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14444 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14445 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14446 Add a path that uses a step of VF instead of 1, but disable it
14448 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14449 and niters_no_overflow parameters. Update calls to
14450 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14451 Create a new SSA name if the latter choses to use a ste other
14452 than zero, and return it via niters_vector_mult_vf_var.
14453 * tree-vect-loop.c (vect_transform_loop): Update calls to
14454 vect_do_peeling, vect_gen_vector_loop_niters and
14455 slpeel_make_loop_iterate_ntimes.
14456 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14457 (vect_gen_vector_loop_niters): Update declarations after above changes.
14459 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14461 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14462 128-bit round to integer instructions.
14463 (ceil<mode>2): Likewise.
14464 (btrunc<mode>2): Likewise.
14465 (round<mode>2): Likewise.
14467 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14469 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14470 unaligned VSX load/store on P8/P9.
14471 (expand_block_clear): Allow the use of unaligned VSX
14472 load/store on P8/P9.
14474 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14476 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14478 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14479 swap associated with both a load and a store.
14481 2018-01-02 Andrew Waterman <andrew@sifive.com>
14483 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14484 * config/riscv/riscv.md (clear_cache): Use it.
14486 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14488 * web.c: Remove out-of-date comment.
14490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14492 * expr.c (fixup_args_size_notes): Check that any existing
14493 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14494 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14495 (emit_single_push_insn): ...here.
14497 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14499 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14500 (const_vector_encoded_nelts): New function.
14501 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14502 (const_vector_int_elt, const_vector_elt): Declare.
14503 * emit-rtl.c (const_vector_int_elt_1): New function.
14504 (const_vector_elt): Likewise.
14505 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14506 of CONST_VECTOR_ELT.
14508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14510 * expr.c: Include rtx-vector-builder.h.
14511 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14512 directly on the tree encoding.
14513 (const_vector_from_tree): Likewise.
14514 * optabs.c: Include rtx-vector-builder.h.
14515 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14516 sequence of "u" values.
14517 * vec-perm-indices.c: Include rtx-vector-builder.h.
14518 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14519 directly on the vec_perm_indices encoding.
14521 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14523 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14524 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14525 * rtx-vector-builder.h: New file.
14526 * rtx-vector-builder.c: Likewise.
14527 * rtl.h (rtx_def::u2): Add a const_vector field.
14528 (CONST_VECTOR_NPATTERNS): New macro.
14529 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14530 (CONST_VECTOR_DUPLICATE_P): Likewise.
14531 (CONST_VECTOR_STEPPED_P): Likewise.
14532 (CONST_VECTOR_ENCODED_ELT): Likewise.
14533 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14534 (unwrap_const_vec_duplicate): Likewise.
14535 (const_vec_series_p): Check for a non-duplicated vector encoding.
14536 Say that the function only returns true for integer vectors.
14537 * emit-rtl.c: Include rtx-vector-builder.h.
14538 (gen_const_vec_duplicate_1): Delete.
14539 (gen_const_vector): Call gen_const_vec_duplicate instead of
14540 gen_const_vec_duplicate_1.
14541 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14542 (gen_const_vec_duplicate): Use rtx_vector_builder.
14543 (gen_const_vec_series): Likewise.
14544 (gen_rtx_CONST_VECTOR): Likewise.
14545 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14546 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14547 Build a new vector rather than modifying a CONST_VECTOR in-place.
14548 (handle_special_swappables): Update call accordingly.
14549 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14550 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14551 Build a new vector rather than modifying a CONST_VECTOR in-place.
14552 (handle_special_swappables): Update call accordingly.
14554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14556 * simplify-rtx.c (simplify_const_binary_operation): Use
14557 CONST_VECTOR_ELT instead of XVECEXP.
14559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14561 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14562 the selector elements to be different from the data elements
14563 if the selector is a VECTOR_CST.
14564 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14565 ssizetype for the selector.
14567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14569 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14570 before testing each element individually.
14571 * tree-vect-generic.c (lower_vec_perm): Likewise.
14573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14575 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14576 * selftest-run-tests.c (selftest::run_tests): Call it.
14577 * vector-builder.h (vector_builder::operator ==): New function.
14578 (vector_builder::operator !=): Likewise.
14579 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14580 (vec_perm_indices::all_from_input_p): New function.
14581 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14582 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14583 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14584 instead of reading the VECTOR_CST directly. Detect whether both
14585 vector inputs are the same before constructing the vec_perm_indices,
14586 and update the number of inputs argument accordingly. Use the
14587 utility functions added above. Only construct sel2 if we need to.
14589 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14591 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14592 the broadcast of the low byte.
14593 (expand_mult_highpart): Use an explicit encoding for the permutes.
14594 * optabs-query.c (can_mult_highpart_p): Likewise.
14595 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14596 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14597 (vectorizable_bswap): Likewise.
14598 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14599 explicit encoding for the power-of-2 permutes.
14600 (vect_permute_store_chain): Likewise.
14601 (vect_grouped_load_supported): Likewise.
14602 (vect_permute_load_chain): Likewise.
14604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14606 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14607 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14608 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14609 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14610 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14611 (vect_gen_perm_mask_any): Likewise.
14613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14615 * int-vector-builder.h: New file.
14616 * vec-perm-indices.h: Include int-vector-builder.h.
14617 (vec_perm_indices): Redefine as an int_vector_builder.
14618 (auto_vec_perm_indices): Delete.
14619 (vec_perm_builder): Redefine as a stand-alone class.
14620 (vec_perm_indices::vec_perm_indices): New function.
14621 (vec_perm_indices::clamp): Likewise.
14622 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14623 (vec_perm_indices::new_vector): New function.
14624 (vec_perm_indices::new_expanded_vector): Update for new
14625 vec_perm_indices class.
14626 (vec_perm_indices::rotate_inputs): New function.
14627 (vec_perm_indices::all_in_range_p): Operate directly on the
14628 encoded form, without computing elided elements.
14629 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14630 encoding. Update for new vec_perm_indices class.
14631 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14632 the given vec_perm_builder.
14633 (expand_vec_perm_var): Update vec_perm_builder constructor.
14634 (expand_mult_highpart): Use vec_perm_builder instead of
14635 auto_vec_perm_indices.
14636 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14637 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14638 or double series encoding as appropriate.
14639 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14640 vec_perm_indices instead of auto_vec_perm_indices.
14641 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14642 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14643 (vect_permute_store_chain): Likewise.
14644 (vect_grouped_load_supported): Likewise.
14645 (vect_permute_load_chain): Likewise.
14646 (vect_shift_permute_load_chain): Likewise.
14647 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14648 (vect_transform_slp_perm_load): Likewise.
14649 (vect_schedule_slp_instance): Likewise.
14650 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14651 (vectorizable_mask_load_store): Likewise.
14652 (vectorizable_bswap): Likewise.
14653 (vectorizable_store): Likewise.
14654 (vectorizable_load): Likewise.
14655 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14656 vec_perm_indices instead of auto_vec_perm_indices. Use
14657 tree_to_vec_perm_builder to read the vector from a tree.
14658 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14659 vec_perm_builder instead of a vec_perm_indices.
14660 (have_whole_vector_shift): Use vec_perm_builder and
14661 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14662 truncation to calc_vec_perm_mask_for_shift.
14663 (vect_create_epilog_for_reduction): Likewise.
14664 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14665 from auto_vec_perm_indices to vec_perm_indices.
14666 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14667 instead of changing individual elements.
14668 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14669 the vector in d.perm.
14670 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14671 from auto_vec_perm_indices to vec_perm_indices.
14672 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14673 instead of changing individual elements.
14674 (arm_vectorize_vec_perm_const): Use new_vector to install
14675 the vector in d.perm.
14676 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14677 Update vec_perm_builder constructor.
14678 (rs6000_expand_interleave): Likewise.
14679 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14680 (rs6000_expand_interleave): Likewise.
14682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14684 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14685 to qimode could truncate the indices.
14686 * optabs.c (expand_vec_perm_var): Likewise.
14688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14690 * Makefile.in (OBJS): Add vec-perm-indices.o.
14691 * vec-perm-indices.h: New file.
14692 * vec-perm-indices.c: Likewise.
14693 * target.h (vec_perm_indices): Replace with a forward class
14695 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14696 * optabs.h: Include vec-perm-indices.h.
14697 (expand_vec_perm): Delete.
14698 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14699 (expand_vec_perm_const): Declare.
14700 * target.def (vec_perm_const_ok): Replace with...
14701 (vec_perm_const): ...this new hook.
14702 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14703 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14704 * doc/tm.texi: Regenerate.
14705 * optabs.def (vec_perm_const): Delete.
14706 * doc/md.texi (vec_perm_const): Likewise.
14707 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14708 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14709 expand_vec_perm for constant permutation vectors. Assert that
14710 the mode of variable permutation vectors is the integer equivalent
14711 of the mode that is being permuted.
14712 * optabs-query.h (selector_fits_mode_p): Declare.
14713 * optabs-query.c: Include vec-perm-indices.h.
14714 (selector_fits_mode_p): New function.
14715 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14716 is defined, instead of checking whether the vec_perm_const_optab
14717 exists. Use targetm.vectorize.vec_perm_const instead of
14718 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14719 fit in the vector mode before using a variable permute.
14720 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14721 vec_perm_indices instead of an rtx.
14722 (expand_vec_perm): Replace with...
14723 (expand_vec_perm_const): ...this new function. Take the selector
14724 as a vec_perm_indices rather than an rtx. Also take the mode of
14725 the selector. Update call to shift_amt_for_vec_perm_mask.
14726 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14727 Use vec_perm_indices::new_expanded_vector to expand the original
14728 selector into bytes. Check whether the indices fit in the vector
14729 mode before using a variable permute.
14730 (expand_vec_perm_var): Make global.
14731 (expand_mult_highpart): Use expand_vec_perm_const.
14732 * fold-const.c: Includes vec-perm-indices.h.
14733 * tree-ssa-forwprop.c: Likewise.
14734 * tree-vect-data-refs.c: Likewise.
14735 * tree-vect-generic.c: Likewise.
14736 * tree-vect-loop.c: Likewise.
14737 * tree-vect-slp.c: Likewise.
14738 * tree-vect-stmts.c: Likewise.
14739 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14741 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14742 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14743 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14744 (aarch64_vectorize_vec_perm_const): ...this new function.
14745 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14746 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14747 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14748 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14749 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14750 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14751 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14753 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14754 check for NEON modes.
14755 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14756 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14757 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14758 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14760 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14761 the old VEC_PERM_CONST conditions.
14762 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14763 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14764 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14765 (ia64_vectorize_vec_perm_const_ok): Merge into...
14766 (ia64_vectorize_vec_perm_const): ...this new function.
14767 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14768 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14769 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14770 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14771 * config/mips/mips.c (mips_expand_vec_perm_const)
14772 (mips_vectorize_vec_perm_const_ok): Merge into...
14773 (mips_vectorize_vec_perm_const): ...this new function.
14774 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14775 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14776 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14777 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14778 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14779 (rs6000_expand_vec_perm_const): Delete.
14780 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14782 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14783 (altivec_expand_vec_perm_const_le): Take each operand individually.
14784 Operate on constant selectors rather than rtxes.
14785 (altivec_expand_vec_perm_const): Likewise. Update call to
14786 altivec_expand_vec_perm_const_le.
14787 (rs6000_expand_vec_perm_const): Delete.
14788 (rs6000_vectorize_vec_perm_const_ok): Delete.
14789 (rs6000_vectorize_vec_perm_const): New function.
14790 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14791 an element count and rtx array.
14792 (rs6000_expand_extract_even): Update call accordingly.
14793 (rs6000_expand_interleave): Likewise.
14794 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14795 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14796 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14797 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14798 (rs6000_expand_vec_perm_const): Delete.
14799 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14800 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14801 (altivec_expand_vec_perm_const_le): Take each operand individually.
14802 Operate on constant selectors rather than rtxes.
14803 (altivec_expand_vec_perm_const): Likewise. Update call to
14804 altivec_expand_vec_perm_const_le.
14805 (rs6000_expand_vec_perm_const): Delete.
14806 (rs6000_vectorize_vec_perm_const_ok): Delete.
14807 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14808 reference to the SPE evmerge intructions.
14809 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14810 an element count and rtx array.
14811 (rs6000_expand_extract_even): Update call accordingly.
14812 (rs6000_expand_interleave): Likewise.
14813 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14814 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14816 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14820 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14821 vector mode and that that mode matches the mode of the data
14823 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14824 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14825 directly using expand_vec_perm_1 when forcing selectors into
14827 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14831 * optabs-query.h (can_vec_perm_p): Delete.
14832 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14833 * optabs-query.c (can_vec_perm_p): Split into...
14834 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14835 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14836 particular selector is valid.
14837 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14838 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14839 (vect_grouped_load_supported): Likewise.
14840 (vect_shift_permute_load_chain): Likewise.
14841 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14842 (vect_transform_slp_perm_load): Likewise.
14843 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14844 (vectorizable_bswap): Likewise.
14845 (vect_gen_perm_mask_checked): Likewise.
14846 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14847 implementations of variable permutation vectors into account
14848 when deciding which selector to use.
14849 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14850 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14851 with a false third argument.
14852 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14853 to test whether the constant selector is valid and can_vec_perm_var_p
14854 to test whether a variable selector is valid.
14856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14858 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14859 * optabs-query.c (can_vec_perm_p): Likewise.
14860 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14861 instead of vec_perm_indices.
14862 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14863 (vect_gen_perm_mask_checked): Likewise,
14864 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14865 (vect_gen_perm_mask_checked): Likewise,
14867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14869 * optabs-query.h (qimode_for_vec_perm): Declare.
14870 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14871 (qimode_for_vec_perm): ...this new function.
14872 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14874 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14876 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14877 does not have a conditional at the top.
14879 2018-01-02 Richard Biener <rguenther@suse.de>
14881 * ipa-inline.c (big_speedup_p): Fix expression.
14883 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14886 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14889 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14893 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14894 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14895 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14896 cond_taken_branch_cost 3->4.
14898 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14900 PR tree-optimization/83581
14901 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14902 TODO_cleanup_cfg if any changes have been made.
14904 PR middle-end/83608
14905 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14906 convert_modes if target mode has the right side, but different mode
14909 PR middle-end/83609
14910 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14911 last argument when extracting from CONCAT. If either from_real or
14912 from_imag is NULL, use expansion through memory. If result is not
14913 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14914 the parts directly to inner mode, if even that fails, use expansion
14917 PR middle-end/83623
14918 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14919 check for bswap in mode rather than HImode and use that in expand_unop
14922 Copyright (C) 2018 Free Software Foundation, Inc.
14924 Copying and distribution of this file, with or without modification,
14925 are permitted in any medium without royalty provided the copyright
14926 notice and this notice are preserved.