1 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4 Fix subreg tests so that we only return a choice between
5 GENERAL_REGS and FP_REGS if the original classes included both.
7 2018-06-01 Richard Biener <rguenther@suse.de>
10 * tree-ssa-structalias.c (get_function_part_constraint):
12 (find_func_aliases_for_call): Properly handle indirect
14 (find_func_clobbers): Likewise.
15 (ipa_pta_execute): Likewise.
16 (create_variable_info_for): For functions that are ifunc_resolver
17 resolve to a varinfo that contains the result of the resolver
19 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
22 2018-05-31 Michael Collison <michael.collison@arm.com>
24 * config/aarch64/aarch64.md:
25 (*fix_to_zero_extenddfdi2): New pattern.
26 * gcc.target/aarch64/fix_extend1.c: New testcase.
28 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
32 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
33 and BUILT_IN_STRNCMP_EQ.
34 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
36 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
37 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
38 (gimple_fold_builtin): Likewise.
39 * tree-ssa-strlen.c (compute_string_length): New function.
40 (determine_min_obsize): New function.
41 (handle_builtin_string_cmp): New function to handle calls to
42 string compare functions.
43 (strlen_optimize_stmt): Add handling to builtin string compare
45 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
46 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
47 * tree.c (build_common_builtin_nodes): Add new defines of
48 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
50 2018-05-31 Jakub Jelinek <jakub@redhat.com>
53 * bb-reorder.c (pass_partition_blocks::gate): Return false for
54 functions with naked attribute.
56 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
58 * config/i386/sse.md (avx_vec_concat<mode>):
59 Substitute concat_tg_mode mode attribute with xtg_mode.
60 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
61 (concat_tg_mode): Remove mode attribute.
63 2018-05-31 Martin Sebor <msebor@redhat.com>
66 * calls.c (alloc_max_size): Correct a logic error/typo.
67 Treat excessive arguments as infinite. Warn for invalid arguments.
68 * doc/invoke.texi (-Walloc-size-larger-than): Update.
70 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
73 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
76 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
77 Cesar Philippidis <cesar@codesourcery.com>
80 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
81 when emitting error on private/firstprivate reductions.
82 * omp-low.c (lower_omp_target): Avoid reference-type processing
83 on pointers for firstprivate clause.
85 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
87 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
90 * config/aarch64/aarch64-simd.md
91 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
92 (aarch64_ld1_x3_<mode>): Likewise
93 (aarch64_st1x2<VALLDIF:mode>): Likewise
94 (aarch64_st1_x2_<mode>): Likewise
95 (aarch64_st1x3<VALLDIF:mode>): Likewise
96 (aarch64_st1_x3_<mode>): Likewise
97 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
98 (vld1_s8_x3): Likewise.
99 (vld1_u16_x3): Likewise.
100 (vld1_s16_x3): Likewise.
101 (vld1_u32_x3): Likewise.
102 (vld1_s32_x3): Likewise.
103 (vld1_u64_x3): Likewise.
104 (vld1_s64_x3): Likewise.
105 (vld1_f16_x3): Likewise.
106 (vld1_f32_x3): Likewise.
107 (vld1_f64_x3): Likewise.
108 (vld1_p8_x3): Likewise.
109 (vld1_p16_x3): Likewise.
110 (vld1_p64_x3): Likewise.
111 (vld1q_u8_x3): Likewise.
112 (vld1q_s8_x3): Likewise.
113 (vld1q_u16_x3): Likewise.
114 (vld1q_s16_x3): Likewise.
115 (vld1q_u32_x3): Likewise.
116 (vld1q_s32_x3): Likewise.
117 (vld1q_u64_x3): Likewise.
118 (vld1q_s64_x3): Likewise.
119 (vld1q_f16_x3): Likewise.
120 (vld1q_f32_x3): Likewise.
121 (vld1q_f64_x3): Likewise.
122 (vld1q_p8_x3): Likewise.
123 (vld1q_p16_x3): Likewise.
124 (vld1q_p64_x3): Likewise.
125 (vst1_s64_x2): Likewise.
126 (vst1_u64_x2): Likewise.
127 (vst1_f64_x2): Likewise.
128 (vst1_s8_x2): Likewise.
129 (vst1_p8_x2): Likewise.
130 (vst1_s16_x2): Likewise.
131 (vst1_p16_x2): Likewise.
132 (vst1_s32_x2): Likewise.
133 (vst1_u8_x2): Likewise.
134 (vst1_u16_x2): Likewise.
135 (vst1_u32_x2): Likewise.
136 (vst1_f16_x2): Likewise.
137 (vst1_f32_x2): Likewise.
138 (vst1_p64_x2): Likewise.
139 (vst1q_s8_x2): Likewise.
140 (vst1q_p8_x2): Likewise.
141 (vst1q_s16_x2): Likewise.
142 (vst1q_p16_x2): Likewise.
143 (vst1q_s32_x2): Likewise.
144 (vst1q_s64_x2): Likewise.
145 (vst1q_u8_x2): Likewise.
146 (vst1q_u16_x2): Likewise.
147 (vst1q_u32_x2): Likewise.
148 (vst1q_u64_x2): Likewise.
149 (vst1q_f16_x2): Likewise.
150 (vst1q_f32_x2): Likewise.
151 (vst1q_f64_x2): Likewise.
152 (vst1q_p64_x2): Likewise.
153 (vst1_s64_x3): Likewise.
154 (vst1_u64_x3): Likewise.
155 (vst1_f64_x3): Likewise.
156 (vst1_s8_x3): Likewise.
157 (vst1_p8_x3): Likewise.
158 (vst1_s16_x3): Likewise.
159 (vst1_p16_x3): Likewise.
160 (vst1_s32_x3): Likewise.
161 (vst1_u8_x3): Likewise.
162 (vst1_u16_x3): Likewise.
163 (vst1_u32_x3): Likewise.
164 (vst1_f16_x3): Likewise.
165 (vst1_f32_x3): Likewise.
166 (vst1_p64_x3): Likewise.
167 (vst1q_s8_x3): Likewise.
168 (vst1q_p8_x3): Likewise.
169 (vst1q_s16_x3): Likewise.
170 (vst1q_p16_x3): Likewise.
171 (vst1q_s32_x3): Likewise.
172 (vst1q_s64_x3): Likewise.
173 (vst1q_u8_x3): Likewise.
174 (vst1q_u16_x3): Likewise.
175 (vst1q_u32_x3): Likewise.
176 (vst1q_u64_x3): Likewise.
177 (vst1q_f16_x3): Likewise.
178 (vst1q_f32_x3): Likewise.
179 (vst1q_f64_x3): Likewise.
180 (vst1q_p64_x3): Likewise.
182 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
184 * config/msp430/msp430.c (msp430_output_labelref): Prepend
185 user_label_prefix to name.
187 * tree-core.h: Update comment about the format of NAME string
188 passed to handler in attribute_spec.
190 * config/msp430/msp430.md: Remove erroneous subreg expression from
191 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
192 zero_extend{q,h}isi2.
194 2018-05-30 Borislav Petkov <bp@suse.de>
196 * doc/extend.texi: Document some architecture specific
197 constraints and sort entries.
199 2018-05-30 Martin Sebor <msebor@redhat.com>
202 * builtins.c (expand_builtin_stpcpy_1): New function.
203 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
204 only if the former succeeds.
206 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
208 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
211 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
213 * doc/invoke.texi (-flinker-output): Document
215 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
217 * passes.c (ipa_write_summaries): Only modify statements if body
219 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
220 incrementally linking.
221 (ipa_passes): Likewise.
222 * lto-cgraph.c (lto_output_node): When incrementally linking do not
223 pass down resolution info.
224 * common.opt (flag_incremental_link): Update info.
225 * gcc.c (plugin specs): Turn flinker-output=* to
226 -plugin-opt=-linker-output-known
227 * toplev.c (compile_file): Also cut compilation when doing incremental
229 * flag-types. (enum lto_partition_model): Add
230 LTO_LINKER_OUTPUT_NOLTOREL.
231 (invoke.texi): Add -flinker-output docs.
232 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
233 link same way as WPA; do not stream in dead initializers.
235 * dwarf2out.c (dwarf2out_die_ref_for_decl,
236 darf2out_register_external_decl): Support incremental link.
238 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
240 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
242 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
244 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
245 it down to simple_object_copy_lto_debug_sections.
246 (run_gcc): Determine incremental LTO link time and configure
247 lto1 into non-wpa mode, disable renaming of debug sections.
249 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
251 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
252 descriptions of various incorrectly documented functions.
254 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
257 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
258 address check not strict.
260 2018-05-30 Richard Biener <rguenther@suse.de>
262 PR tree-optimization/85964
263 * tracer.c (better_p): Drop initialized count check, we only
264 call the function with initialized counts now.
265 (find_best_successor): Do find a best edge if one
266 has uninitialized count.
267 (find_best_predecessor): Likewise. Do BB frequency check only
268 if count is initialized.
270 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
272 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
273 (aarch64_ldrstr_offset_compare): New.
274 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
275 load/store orderings.
276 (aarch64_gen_adjusted_ldpstp): Likewise.
278 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
280 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
281 Check for subset of GENERAL_REGS and FP_REGS.
282 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
285 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
287 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
288 and wi::to_poly_offset. Add the current offset and then check
289 whether the sum fits, rather than using an unchecked addition of
290 a checked term. Check for a shwi rather than a uhwi.
291 * expr.c (get_bit_range): Use tree_to_poly_uint64.
292 (store_constructor): Use poly_int_tree_p.
293 (expand_expr_real_1): Likewise.
294 * function.c (assign_temp): Likewise.
295 * fold-const.c (const_binop): Use poly_int_tree_p and
297 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
299 * ipa-icf-gimple.c (func_checker::compare_operand): Use
300 to_poly_offset for MEM offsets.
301 * ipa-icf.c (sem_variable::equals): Likewise.
302 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
303 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
304 wi::to_poly_offset for BIT_FIELD_REF offsets.
305 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
307 * var-tracking.c (emit_note_insn_var_location): Use
310 2018-05-29 Jim Wilson <jimw@sifive.com>
312 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
314 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
317 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
318 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
320 (sse4_1_round<mode>2): Use nonimmediate_operand
321 for operand 1 predicate.
323 2018-05-29 Martin Sebor <msebor@redhat.com>
324 Richard Biener <rguenther@suse.de>
327 * calls.c (get_size_range): Call determine_value_range instead
329 * tree-vrp.h (determine_value_range): Declared new function.
330 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
332 2018-05-29 Richard Biener <rguenther@suse.de>
334 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
335 sure to use non-pattern stmts for get_earlier_stmt arguments.
336 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
337 called on pattern stmts.
338 (get_later_stmt): Likewise.
340 2018-05-29 Martin Liska <mliska@suse.cz>
342 PR gcov-profile/85759
343 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
346 2018-05-29 Jakub Jelinek <jakub@redhat.com>
348 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
350 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
354 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
355 VEC_PACK_FLOAT_EXPR): New tree codes.
356 * tree-pretty-print.c (op_code_prio): Handle
357 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
358 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
359 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
360 * tree-inline.c (estimate_operator_cost): Likewise.
361 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
362 * fold-const.c (const_binop): Likewise.
363 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
364 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
365 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
366 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
367 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
368 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
369 * expr.c (expand_expr_real_2): Likewise.
370 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
371 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
372 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
374 * optabs.c (expand_widen_pattern_expr): For
375 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
376 sign from result type rather than operand's type.
377 (expand_binop_directly): For vec_packu_float_optab and
378 vec_packs_float_optab allow result type to be different from operand's
380 * optabs-tree.c (optab_for_tree_code): Handle
381 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
382 VEC_PACK_FLOAT_EXPR. Formatting fixes.
383 * tree-vect-generic.c (expand_vector_operations_1): Handle
384 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
386 * tree-vect-stmts.c (supportable_widening_operation): Handle
388 (supportable_narrowing_operation): Handle FLOAT_EXPR.
389 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
390 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
391 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
392 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
394 (vec_pack<floatprefix>_float_<mode>): New expander.
395 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
397 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
398 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
399 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
400 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
401 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
403 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
404 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
405 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
406 VEC_PACK_FLOAT_EXPR): Document.
408 2018-05-29 Richard Biener <rguenther@suse.de>
410 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
412 (stmt_vec_info_vec): Make pointer.
413 (init_stmt_vec_info_vec): Remove.
414 (free_stmt_vec_info_vec): Likewise.
415 (set_stmt_vec_info_vec): New function.
416 (free_stmt_vec_infos): Likewise.
417 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
418 (set_vinfo_for_stmt): Likewise.
419 (get_earlier_stmt): Likewise.
420 (get_later_stmt): Likewise.
421 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
422 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
423 (vec_info::~vec_info): Free stmt_vec_infos.
424 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
425 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
426 (pass_slp_vectorize::execute): Likewise.
427 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
428 (free_stmt_vec_info_vec): Likewise.
429 (set_stmt_vec_info_vec): New function.
430 (free_stmt_vec_infos): Likewise.
431 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
432 the global stmt_vec_info_vec.
433 * tree-parloops.c (gather_scalar_reductions): Use
434 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
437 2018-05-29 Richard Biener <rguenther@suse.de>
439 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
441 2018-05-29 Martin Liska <mliska@suse.cz>
442 David Malcolm <dmalcolm@redhat.com>
444 * vec.c (test_reverse): New.
445 (vec_c_tests): Add new test.
446 * vec.h (vl_ptr>::reverse): New function.
448 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
450 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
452 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
455 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
457 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
459 2018-05-28 Richard Biener <rguenther@suse.de>
461 PR tree-optimization/85933
462 * tree-vect-data-refs.c (vect_record_base_alignments): Only
463 look at stmts marked as vectorizable.
465 2018-05-28 Richard Biener <rguenther@suse.de>
467 PR tree-optimization/85934
468 * tree-vect-generic.c (expand_vector_operations_1): Hoist
469 vector boolean check before scalar optimization.
471 2018-05-28 Jakub Jelinek <jakub@redhat.com>
473 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
476 2018-05-28 Mark Wielaard <mark@klomp.org>
478 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
479 if it is an expression containing a minus sign.
481 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
483 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
485 2018-05-27 Paul Koning <ni1d@arrl.net>
487 * config/pdp11/pdp11.md (truncsihi2): Remove.
489 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
490 Chung-Ju Wu <jasonwucj@gmail.com>
492 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
494 (unaligned_store_dw): Ditto.
495 * config/nds32/nds32-memory-manipulation.c
496 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
497 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
498 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
499 (emit_setmem_word_loop): Rename to ...
500 (emit_setmem_doubleword_loop): ... this.
501 (nds32_gen_dup_4_byte_to_word_value): New function.
502 (nds32_gen_dup_8_byte_to_double_word_value): New function.
503 (nds32_expand_setmem_loop): Refine implementation.
504 (nds32_expand_setmem_loop_v3m): Ditto.
505 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
508 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
510 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
512 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
514 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
515 (nds32_init_machine_status): Initialize machine->attr_naked_p and
516 machine->attr_no_prologue_p.
517 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
518 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
519 (nds32_expand_epilogue): Consider attr_naked_p.
520 (nds32_expand_epilogue_v3pop): Likewise.
521 (nds32_can_use_return_insn): Likewise.
522 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
523 attr_no_prologue_p fields.
524 * config/nds32/nds32.opt (mret-in-naked-func): New option.
526 2018-05-27 Jakub Jelinek <jakub@redhat.com>
529 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
532 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
534 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
536 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
538 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
540 (*<floatsuffix>floatv2div2sf2): Rename to ...
541 (*float<floatunssuffix>v2div2sf2): ... this.
542 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
543 (float<floatunssuffix>v2div2sf2_mask): ... this.
544 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
545 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
546 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
548 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
550 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
552 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
554 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
556 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
558 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
559 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
560 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
561 gen_ufix_truncv8dfv8si2.
562 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
563 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
564 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
565 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
566 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
567 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
568 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
569 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
571 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
575 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
577 2018-05-25 Jim Wilson <jimw@sifive.com>
579 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
580 * config/riscv/riscv.c (struct machine_function): Add
581 interrupt_handler_p and attribute_checked_p fields.
582 (riscv_attribute_table): Add interrupt.
583 (riscv_interrupt_type_p): New.
584 (riscv_save_reg_p): Save extra regs for interrupt handler.
585 (riscv_use_save_libcall): Return false for interrupt handler.
586 (riscv_first_stack_step): Add forward declaration.
587 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
588 for interrupt handler with large frame. Use it for saved reg list.
589 (riscv_expand_prologue): Move flag_stack_usage_info support to
590 eliminate duplication.
591 (riscv_expand_epilogue): Generate mret for interrupt handler.
592 (riscv_epilogue_uses): New.
593 (riscv_can_use_return_insn): Return false for interrupt handler.
594 (riscv_function_ok_for_sibcall): Likewise.
595 (riscv_set_current_function): Add interrupt handler support.
596 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
597 * config/riscv/riscv.md (UNSPECV_MRET): New.
599 (riscv_frflags, riscv_fsflags): Use tab after opcode.
601 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
603 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
605 PR tree-optimization/85712
606 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
607 this candidate has already been replaced in-situ by a copy.
609 2018-05-25 Jason Merrill <jason@redhat.com>
611 PR c++/80485 - inline function non-zero address.
612 * symtab.c (nonzero_address): Check DECL_COMDAT.
614 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
617 * config/alpha/alpha.md (ashlsi3): New insn pattern.
618 (*ashlsi_se): Rename from *ashldi_se. Define as sign
619 extension of SImode operation. Use const123_operand predicate.
621 (*saddl_se_1): Ditto.
623 (*ssubl_se_1): Ditto.
624 * config/alpha/predicates.md (const123_operand): New predicate.
625 * config/alpha/constraints.md (P): Use IN_RANGE.
627 2018-05-25 Richard Biener <rguenther@suse.de>
629 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
631 (ref_maybe_used_by_stmt_p): Likewise.
632 (stmt_may_clobber_ref_p): Likewise.
633 (stmt_may_clobber_ref_p_1): Likewise.
634 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
636 (ref_maybe_used_by_stmt_p): Likewise.
637 (stmt_may_clobber_ref_p): Likewise.
638 (stmt_may_clobber_ref_p_1): Likewise.
639 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
640 the alias oracle to disambiguate DRs with stmts DR analysis
642 (vect_analyze_data_refs): Do not give up on not analyzable
643 DRs for BB vectorization. Remove code truncating the dataref
646 2018-05-25 Jakub Jelinek <jakub@redhat.com>
649 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
650 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
651 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
653 2018-05-25 Richard Biener <rguenther@suse.de>
655 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
656 function, combining stmt data ref gathering and fatal analysis
658 (vect_analyze_data_refs): Remove now redudnant code and simplify.
659 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
660 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
661 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
662 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
664 2018-05-25 Bin Cheng <bin.cheng@arm.com>
666 PR tree-optimization/85720
667 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
668 SCC if all partitions are builtins.
669 (version_loop_by_alias_check): New parameter. Generate cancelable
670 runtime alias check if all partitions are builtins.
671 (distribute_loop): Update call to above function.
673 2018-05-25 Bin Cheng <bin.cheng@arm.com>
675 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
676 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
677 (parm_default_def_partition_arg): Ditto.
678 (set_parm_default_def_partition): Ditto.
679 (get_parm_default_def_partitions): Ditto and make it static.
680 (get_undefined_value_partitions): Ditto and make it static.
681 (remove_ssa_form): Refactor call to init_var_map here.
682 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
683 computation for loop region.
684 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
685 (register_default_def): Delete.
686 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
687 (parm_default_def_partition_arg): Ditto.
688 (set_parm_default_def_partition): Ditto.
689 (get_parm_default_def_partitions): Ditto and make it static.
690 (get_undefined_value_partitions): Ditto and make it static.
691 (coalesce_with_default, coalesce_with_default): Update comment.
692 (create_coalesce_list_for_region): New func factored out from
693 create_outofssa_var_map.
694 (populate_coalesce_list_for_outofssa): New func factored out from
695 create_outofssa_var_map and coalesce_ssa_name.
696 (create_outofssa_var_map): Delete.
697 (coalesce_ssa_name): Refactor to support live range computation.
698 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
699 (get_parm_default_def_partitions): Delete.
700 (get_undefined_value_partitions): Ditto.
701 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
702 computation for loop region.
703 (new_tree_live_info, loe_visit_block): Ditto.
704 (live_worklist, set_var_live_on_entry): Ditto.
705 (calculate_live_on_exit, verify_live_on_entry): Ditto.
706 * tree-ssa-live.h (struct _var_map): New fields.
707 (init_var_map): Change decl.
708 (region_contains_p): New.
710 2018-05-25 Bin Cheng <bin.cheng@arm.com>
712 * tree-ssa-live.h (live_merge_and_clear): Delete.
714 2018-05-25 Richard Biener <rguenther@suse.de>
717 * tree-dump.c (dequeue_and_dump): Remove access to removed
718 operand 2 of a SWITCH_EXPR.
720 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
722 * doc/sourcebuild.texi (vect_double_cond_arith): Include
723 multiplication and division.
724 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
725 (cond_udiv@var{m}, cond_umod@var{m}): Document.
726 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
727 (cond_udiv_optab, cond_umod_optab): New optabs.
728 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
729 (IFN_COND_RDIV): New internal functions.
730 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
731 TRUNC_MOD_EXPR and RDIV_EXPR.
732 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
733 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
735 (SVE_INT_BINARY): Include mult.
736 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
737 (optab, sve_int_op): Handle mult.
738 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
740 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
741 for SVE_INT_BINARY_SD.
743 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
745 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
746 (optab, sve_int_op): Handle div and udiv.
747 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
748 for SVE_INT_BINARY_SD.
749 (*<optab><mode>3): New insn for the same.
751 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
753 * tree-vect-patterns.c: Include predict.h.
754 (vect_recog_divmod_pattern): Restrict check for division support
755 to when optimizing for size.
757 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
759 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
760 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
761 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
762 (gimple_match_op::set_op): Likewise.
763 (gimple_resimplify4): Declare.
764 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
765 (expr::gen_transform): Likewise.
766 (decision_tree::gen): Generate a simplification routine for 4 operands.
767 * gimple-match-head.c (gimple_simplify): Add an overload for
768 4 operands. In the top-level function, handle up to 4 call
769 arguments and call gimple_resimplify4.
770 (gimple_resimplify4): New function.
771 (build_call_internal): Pass a fourth operand.
772 (maybe_push_to_seq): Likewise.
773 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
774 Fold VEC_COND_EXPRs of an operation and a default value into
775 an IFN_COND_* function if possible.
776 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
778 (SVE_COND_FP_BINARY): Include them.
779 (optab, sve_fp_op): Handle them.
780 (SVE_INT_BINARY_REV): New code iterator.
781 (SVE_COND_FP_BINARY_REV): New int iterator.
782 (commutative): New int attribute.
783 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
785 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
787 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
788 (*cond_<optab><mode>): New patterns for reversed operands.
790 2018-05-25 Richard Biener <rguenther@suse.de>
792 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
793 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
794 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
795 (STMT_VINFO_GROUPED_ACCESS): Adjust.
796 * tree-vect-data-refs.c (everywhere): Adjust users.
797 * tree-vect-loop.c (everywhere): Likewise.
798 * tree-vect-slp.c (everywhere): Likewise.
799 * tree-vect-stmts.c (everywhere): Likewise.
800 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
802 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
804 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
806 (gcc_cv_as_section_exclude): ... this.
807 Try Solaris as #exclude syntax.
808 * configure: Regenerate.
809 * config.in: Regenerate.
810 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
812 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
813 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
815 * varasm.c (default_elf_asm_named_section): Don't check if
816 HAVE_GAS_SECTION_EXCLUDE is defined.
818 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
820 * doc/md.texi: Update the documentation of the cond_* optabs
821 to mention the new final operand. Fix GET_MODE_NUNITS call.
822 Describe the scalar case too.
823 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
824 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
826 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
827 (get_conditional_internal_fn): Update comment.
828 * tree-vect-loop.c (vectorizable_reduction): Pass the original
829 accumulator value as a final argument to conditional functions.
830 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
831 a define_expand and add an "else" operand. Assert for now that
832 the else operand is equal to operand 2. Use SVE_INT_BINARY and
833 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
834 (*cond_<optab><mode>): New patterns.
835 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
836 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
837 (UNSPEC_COND_EOR): Delete.
838 (optab): Remove associated mappings.
839 (SVE_INT_BINARY): New code iterator.
840 (sve_int_op): Remove int attribute and add "minus" to the code
842 (SVE_COND_INT_OP): Delete.
843 (SVE_COND_FP_OP): Rename to...
844 (SVE_COND_FP_BINARY): ...this.
846 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
848 * optabs.c (can_reuse_operands_p): New function.
849 (maybe_legitimize_operands): Try to reuse the results for
852 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
854 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
855 Add {q} suffix to insn mnemonic.
857 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
859 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
860 (msp430_warn_func_return): New.
862 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
864 * fold-const.c (tree_nonzero_bits): New function.
865 * fold-const.h (tree_nonzero_bits): Likewise.
866 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
867 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
869 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
873 * varasm.c (assemble_alias): Check ifunc_resolver only on
876 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
879 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
880 when memory input operand is handled.
882 2018-05-24 Luis Machado <luis.machado@linaro.org>
884 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
886 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
888 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
890 * match.pd: Delay FMA folds until after vectorization.
892 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
895 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
896 address check not strict.
898 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
900 * gimple-match.h (gimple_match_op): New class.
901 (mprts_hook): Replace parameters with a gimple_match_op *.
902 (maybe_build_generic_op): Likewise.
903 (gimple_simplified_result_is_gimple_val): Replace parameters with
904 a const gimple_match_op *.
905 (gimple_simplify): Replace code_helper * and tree * parameters with
906 a gimple_match_op * parameter.
907 (gimple_resimplify1): Replace code_helper *, tree and tree *
908 parameters with a gimple_match_op * parameter.
909 (gimple_resimplify2): Likewise.
910 (gimple_resimplify3): Likewise.
911 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
912 parameters with a gimple_match_op * parameter.
913 * gimple-match-head.c (gimple_simplify): Change prototypes of
914 auto-generated functions to take a gimple_match_op * instead of
915 separate code_helper * and tree * parameters. Make the same
916 change in the top-level overload and update calls to the
917 gimple_resimplify routines. Update calls to the auto-generated
918 functions and to maybe_push_res_to_seq in the publicly-facing
919 operation-specific gimple_simplify overloads.
920 (gimple_match_op::MAX_NUM_OPS): Define.
921 (gimple_resimplify1): Replace rcode and ops with a single res_op
922 parameter. Update call to gimple_simplify.
923 (gimple_resimplify2): Likewise.
924 (gimple_resimplify3): Likewise.
925 (mprts_hook): Replace parameters with a gimple_match_op *.
926 (maybe_build_generic_op): Likewise.
927 (build_call_internal): Replace type, nargs and ops with
929 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
930 with a single gimple_match_op *. Update calls to mprts_hook,
931 build_call_internal and gimple_simplified_result_is_gimple_val.
932 Factor out code that is common to the tree_code and combined_fn cases.
933 * genmatch.c (expr::gen_transform): Replace tem_code and
934 tem_ops with a gimple_match_op called tem_op. Update calls
935 to the gimple_resimplify functions and maybe_push_res_to_seq.
936 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
937 res_ops. Update call to the gimple_resimplify functions.
938 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
939 (decision_tree::gen): Make the functions take a gimple_match_op *
940 called res_op instead of separate res_code and res_ops parameters.
941 Update call accordingly.
942 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
943 and ops with a single res_op parameter. Update calls to
944 maybe_build_generic_op and maybe_push_res_to_seq.
945 (fold_stmt_1): Update calls to gimple_simplify and
946 replace_stmt_with_simplification.
947 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
948 and gimple_simplified_result_is_gimple_val.
949 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
951 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
952 with a gimple_match_op *.
953 (vn_nary_build_or_lookup): Likewise. Update call to
954 vn_nary_build_or_lookup_1.
955 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
956 gimple_match_op *. Update calls to the gimple_resimplify routines
957 and to gimple_simplified_result_is_gimple_val.
958 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
959 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
960 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
961 (visit_nary_op): Likewise.
962 (visit_reference_op_load): Likewise.
964 2018-05-23 Luis Machado <luis.machado@linaro.org>
966 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
967 modifier for printing the step amount.
969 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
972 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
975 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
977 * doc/sourcebuild.texi (Endianness): New subsubsection.
979 2018-05-23 Luis Machado <luis.machado@linaro.org>
981 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
982 <prefetch_dynamic_strides>: New const bool field.
983 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
984 prefetch_dynamic_strides.
985 (exynosm1_prefetch_tune): Likewise.
986 (thunderxt88_prefetch_tune): Likewise.
987 (thunderx_prefetch_tune): Likewise.
988 (thunderx2t99_prefetch_tune): Likewise.
989 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
991 (aarch64_override_options_internal): Update to set
992 PARAM_PREFETCH_DYNAMIC_STRIDES.
993 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
994 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
995 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
996 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
997 prefetch-dynamic-strides setting.
999 2018-05-23 Luis Machado <luis.machado@linaro.org>
1001 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1002 <minimum_stride>: New const int field.
1003 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1004 minimum_stride field defaulting to -1.
1005 (exynosm1_prefetch_tune): Likewise.
1006 (thunderxt88_prefetch_tune): Likewise.
1007 (thunderx_prefetch_tune): Likewise.
1008 (thunderx2t99_prefetch_tune): Likewise.
1009 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1010 <default_opt_level>: Set to 3.
1011 (aarch64_override_options_internal): Update to set
1012 PARAM_PREFETCH_MINIMUM_STRIDE.
1013 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1014 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1015 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1016 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1017 stride is constant and is below the minimum stride threshold.
1019 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1021 * config/arm/arm-cpus.in (mode26): Delete.
1022 (armv4): Delete mode26 reference.
1023 * config/arm/arm.c (arm_configure_build_target): Delete use of
1026 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1028 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1030 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1031 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1033 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1034 (floatunsdidf2): Ditto.
1036 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1038 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1039 (fixuns_trunc<mode>si2_avx512f): Ditto.
1040 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1041 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1042 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1044 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1046 PR rtl-optimization/79985
1047 * df-scan.c (df_insn_refs_collect): Remove special case for
1048 global registers and asm statements.
1050 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1052 * extend.texi (Global Register Variables): Rewrite the bullet list.
1053 Note that the register is available for allocation. Note that access
1054 via inline asm must use constraints. Add note about async-signal
1055 handlers. Remove paragraph about automagic register selection.
1057 2018-05-23 Richard Biener <rguenther@suse.de>
1059 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1060 of fixed offset from memset VN.
1062 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1064 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1066 (alloc_cand_and_find_basis): Initialize first_interp field.
1067 (slsr_process_mul): Modify first_interp field.
1068 (slsr_process_add): Likewise.
1069 (slsr_process_cast): Modify first_interp field for each new
1071 (slsr_process_copy): Likewise.
1072 (dump_candidate): Dump first_interp field.
1073 (replace_mult_candidate): Process all interpretations, not just
1075 (replace_rhs_if_not_dup): Likewise.
1076 (replace_one_candidate): Likewise.
1078 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1080 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1082 (aarch64_needs_frame_chain): New function.
1083 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1085 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1088 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1089 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1090 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1091 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1092 as true for strict-align.
1093 (aarch64_can_inline_p): Perform checks even when callee has no
1094 attributes to check for strict alignment.
1095 * doc/extend.texi (AArch64 Function Attributes): Document
1097 * doc/invoke.texi: (AArch64 Options): Likewise.
1099 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1101 PR tree-optimization/85853
1102 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1103 the handling of the root of the node to...
1104 (vect_slp_analyze_node_operations_1): ...this new function,
1105 and run the whole thing with the child nodes' def types
1106 set according to their SLP node's def type.
1108 2018-05-23 Richard Biener <rguenther@suse.de>
1111 * tree-data-ref.c (create_runtime_alias_checks): Defer
1112 and ignore overflow warnings.
1114 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1116 PR tree-optimization/85822
1117 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1120 2018-05-23 Richard Biener <rguenther@suse.de>
1122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1123 memset constants via native_interpret_expr.
1125 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1128 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1130 (cgraph_node::create_alias): Likewise.
1131 (cgraph_node::get_availability): Check ifunc_resolver instead
1132 of looking up ifunc attribute.
1133 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1134 * varasm.c (do_assemble_alias): Likewise.
1135 (assemble_alias): Likewise.
1136 (default_binds_local_p_3): Likewise.
1137 * cgraph.h (cgraph_node): Add ifunc_resolver.
1138 (cgraph_node::only_called_directly_or_aliased_p): Return false
1140 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1142 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1143 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1144 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1145 instead of looking up ifunc attribute.
1147 2018-05-22 Luis Machado <luis.machado@linaro.org>
1149 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1151 2018-05-22 Martin Sebor <msebor@redhat.com>
1154 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1155 only when expasion succeeds.
1156 (expand_builtin_strcmp): Same.
1157 (expand_builtin_strncmp): Same.
1159 2018-05-22 Martin Sebor <msebor@redhat.com>
1161 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1163 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1164 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1166 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1167 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1169 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1170 Add check for legitimate_address.
1171 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1172 (aarch64_swap_ldrstr_operands): New.
1173 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1176 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1177 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1179 * config/aarch64/aarch64.md: New patterns to generate stp
1181 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1182 single words and double words.
1183 (load_pair_sw, load_pair_dw): Likewise.
1184 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1186 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1188 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1189 for different mode ldpstp and add peephole for merged zero stores.
1191 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1193 (aarch64_gen_store_pair): Rename calls to match new patterns.
1194 (aarch64_gen_load_pair): Rename calls to match new patterns.
1195 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1196 (load_pair<DREG:mode><DREG2:mode>): ... This.
1197 (store_pair<mode>): Rename to...
1198 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1199 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1201 (V_INT_EQUIV): Handle SImode.
1202 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1205 2018-05-22 Martin Sebor <msebor@redhat.com>
1208 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1209 or ajust the presumed bound on an operation to avoid unnecessary
1212 2018-05-22 Martin Sebor <msebor@redhat.com>
1214 PR tree-optimization/85826
1215 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1216 assuming that a DECL necesarily has a constant size.
1218 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1221 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1223 2018-05-22 Richard Biener <rguenther@suse.de>
1225 PR tree-optimization/85834
1226 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1227 non-constant and non-zero memset arguments.
1229 2018-05-22 Martin Liska <mliska@suse.cz>
1232 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1234 2018-05-22 Richard Biener <rguenther@suse.de>
1236 PR tree-optimization/85863
1237 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1238 comparisons when vectype is specified.
1239 (vectorizable_condition): Do not specify vectype for
1240 vect_is_simple_cond when SLP vectorizing.
1242 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1245 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1246 define __ibm128 as long double.
1247 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1248 as a distinct type when IEEE 128-bit support is enabled.
1249 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1250 128-bit types to use the correct functions.
1251 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1252 convert between 128-bit floating point types that have different
1253 modes but the same representation, instead of using gen_lowpart to
1255 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1257 (IFKF_reg): New attributes to give the register constraints for
1259 (extend<mode>tf2_internal): New insns to mark an explicit
1260 conversion between 128-bit floating point types that have a
1261 different mode but share the same representation.
1263 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1265 PR tree-optimization/85814
1266 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1267 a null return from get_strinfo when unsharing the next
1268 strinfo in the chain.
1270 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1273 * varasm.c (weak_finish): Clean up weak_decls.
1275 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1277 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1278 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1279 UNSPEC_UADALP values.
1280 * config/aarch64/iterators.md (ABAL): New int iterator.
1283 (sur): Add mappings for the above.
1284 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1286 (aarch64_<sur>abal<mode>_4): Likewise.
1287 (aarch64_<sur>adalp<mode>_3): Likewise.
1288 (<sur>sadv16qi): New define_expand.
1290 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1292 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1293 (*movdf_internal): Ditto.
1294 (*rcpsf2_sse): Ditto.
1295 (*rsqrtsf2_sse): Ditto.
1296 (*sqrt<mode>2_sse): Ditto.
1298 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1300 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1302 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1303 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1304 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1305 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1307 * config/aarch64/arm_neon.h: Likewise.
1308 * config/aarch64/iterators.md (VQ_I): New.
1310 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1312 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1314 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1317 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1319 * config/nds32/constraints.md (S): New constraint.
1320 * config/nds32/nds32.md (call_internal): Use constraint S.
1321 (call_value_internal): Likewise.
1322 (sibcall_internal): Likewise.
1323 (sibcall_value_internal): Likewise.
1325 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1326 Chung-Ju Wu <jasonwucj@gmail.com>
1328 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1331 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1332 Chung-Ju Wu <jasonwucj@gmail.com>
1334 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1335 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1336 (nds32_rtx_costs_impl): Simplify.
1337 (nds32_address_cost_impl): Simplify.
1338 (nds32_init_rtx_costs): New function.
1339 (nds32_rtx_costs_speed_prefer): Likewise.
1340 (nds32_rtx_costs_size_prefer): Likewise.
1341 (nds32_address_cost_speed_prefer): Likewise.
1342 (nds32_address_cost_speed_fwprop): Likewise.
1343 (nds32_address_cost_size_prefer): Likewise.
1344 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1345 * config/nds32/nds32.c (nds32_option_override): Use
1346 nds32_init_rtx_costs function.
1348 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1350 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1351 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1352 (TARGET_PIPELINE_N8): Likewise.
1353 (TARGET_PIPELINE_N10): Likewise.
1354 (TARGET_PIPELINE_N13): Likewise.
1355 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1357 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1359 * config/nds32/nds32-fpu.md: Update copyright year.
1361 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1363 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1365 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1367 * config/nds32/nds32.c
1368 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1369 * config/nds32/nds32.opt (minline-asm-r15): New option.
1371 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1373 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1375 * config/nds32/nds32.md (abssi2): New pattern.
1377 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1379 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1380 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1381 Merge insn pattern from sse_cvtsi2ss<round_name> and
1382 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1383 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1384 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1385 using SWI48 mode iterator.
1386 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1387 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1388 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1389 pattern from sse_cvttss2si<round_saeonly_name>
1390 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1391 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1392 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1393 using SWI48 mode iterator.
1394 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1395 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1396 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1397 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1398 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1399 using SWI48 mode iterator.
1400 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1401 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1402 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1403 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1404 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1405 SWI48 mode iterator.
1406 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1407 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1408 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1409 pattern from sse_cvttsd2si<round_saeonly_name>
1410 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1412 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1414 * config/nds32/nds32-md-auxiliary.c
1415 (nds32_valid_smw_lwm_base_p): Refine.
1416 (nds32_output_smw_single_word): Refine.
1417 (nds32_output_smw_double_word): New.
1418 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1420 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1422 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1423 (nds32_output_stack_pop): Refine.
1424 (nds32_expand_unaligned_load): Refine.
1425 (nds32_expand_unaligned_store): Refine.
1427 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1428 Chung-Ju Wu <jasonwucj@gmail.com>
1430 * config/nds32/constants.md: Add TP_REGNUM constant.
1431 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1432 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1434 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1435 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1436 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1437 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1438 and PIC code generation.
1439 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1441 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1443 * config/nds32/nds32.md: Support TLS and PIC.
1444 * config/nds32/nds32.c: Support TLS and PIC.
1445 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1446 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1449 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1451 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1452 mode with E_ prefix.
1454 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1455 Chung-Ju Wu <jasonwucj@gmail.com>
1457 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1458 * config/nds32/nds32-md-auxiliary.c
1459 (symbolic_reference_mentioned_p): New.
1460 (nds32_legitimize_ict_address): New.
1461 (nds32_expand_ict_move): New.
1462 (nds32_indirect_call_referenced_p): New.
1463 (nds32_symbol_binds_local_p): Delete.
1464 (nds32_long_call_p): Modify.
1465 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1466 * config/nds32/nds32-protos.h
1467 (symbolic_reference_mentioned_p): Declare.
1468 (nds32_legitimize_ict_address): Declare.
1469 (nds32_expand_ict_move): Declare.
1470 (nds32_indirect_call_referenced_p): Declare.
1471 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1472 (nds32_relax_group): Use nds32_ict_const_p as condition.
1473 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1474 (nds32_asm_file_start): Output ict_model directive in asm code.
1475 (nds32_legitimate_address_p): Consider indirect call.
1476 (nds32_print_operand): Consider indirect call.
1477 (nds32_print_operand_address): Consider indirect call.
1478 (nds32_insert_attributes): Handle "indirect_call" attribute.
1479 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1480 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1481 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1482 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1483 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1484 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1485 (TARGET_ICT_MODEL_SMALL): Define.
1486 (TARGET_ICT_MODEL_LARGE): Define.
1487 * config/nds32/nds32.md (movsi): Consider ict model.
1488 (call, call_value): Consider ict model.
1489 (sibcall, sibcall_value): Consider ict model.
1490 * config/nds32/nds32.opt (mict-model): New option.
1491 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1494 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1495 Monk Chiang <sh.chiang04@gmail.com>
1496 Jim Wilson <jimw@sifive.com>
1498 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1499 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1500 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1501 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1502 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1503 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1504 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1505 compute save_libcall_adjustment properly.
1506 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1507 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1508 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1509 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1510 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1511 (ABI_SPEC): Handle mabi=ilp32e.
1512 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1513 (RVE): Add RVE mask.
1514 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1515 <-march>: Add rv32e as an example.
1517 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1520 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1521 (intra_create_variable_infos): Handle C++ constructors.
1523 2018-05-18 Martin Liska <mliska@suse.cz>
1525 * passes.def: Remove a redundant pass.
1527 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1530 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1532 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1534 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1536 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1538 (armv2, armv2a, armv3, armv3m): Delete architectures.
1539 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1540 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1541 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1543 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1544 (*mulsidi3adddi): Likewise.
1545 (mulsidi3): Likewise.
1546 (*mulsidi3_nov6): Likewise.
1547 (umulsidi3): Likewise.
1548 (umulsidi3_nov6): Likewise.
1549 (umaddsidi4): Likewise.
1550 (*umulsidi3adddi): Likewise.
1551 (smulsi3_highpart): Likewise.
1552 (*smulsi3_highpart_nov6): Likewise.
1553 (umulsi3_highpart): Likewise.
1554 (*umulsi3_highpart_nov6): Likewise.
1555 * config/arm/arm.h (arm_arch3m): Delete.
1556 * config/arm/arm.c (arm_arch3m): Delete.
1557 (arm_option_override_internal): Update armv3-related comment.
1558 (arm_configure_build_target): Delete use of isa_bit_mode32.
1559 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1560 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1561 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1563 (mulusa3): Likewise.
1564 * config/arm/arm-protos.h (arm_arch3m): Delete.
1565 * config/arm/arm-tables.opt: Regenerate.
1566 * config/arm/arm-tune.md: Likewise.
1567 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1568 deleted architectures.
1570 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1572 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1573 (armv5t, armv5te): New features.
1574 (ARMv5, ARMv5e): Delete fgroups.
1575 (ARMv5t, ARMv5te): Adjust for above changes.
1577 (armv5, armv5e): Delete arches.
1578 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1580 (*call_reg_arm): Likewise.
1581 (*call_value_reg_armv5): Likewise.
1582 (*call_value_reg_arm): Likewise.
1583 (*call_symbol): Likewise.
1584 (*call_value_symbol): Likewise.
1585 (*sibcall_insn): Likewise.
1586 (*sibcall_value_insn): Likewise.
1588 (prefetch): Likewise.
1589 (define_split and define_peephole2 dependent on arm_arch5):
1591 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1593 (TARGET_ARM_QBIT): Likewise.
1594 (TARGET_DSP_MULTIPLY): Likewise.
1595 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1596 (arm_arch5, arm_arch5e): Delete.
1597 (arm_arch5t, arm_arch5te): Declare.
1598 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1599 (arm_arch5t): Declare.
1600 (arm_option_reconfigure_globals): Update for the above.
1601 (arm_options_perform_arch_sanity_checks): Update comment, replace
1602 use of arm_arch5 with arm_arch5t.
1603 (use_return_insn): Likewise.
1604 (arm_emit_call_insn): Likewise.
1605 (output_return_instruction): Likewise.
1606 (arm_final_prescan_insn): Likewise.
1607 (arm_coproc_builtin_available): Likewise.
1608 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1609 arm_arch5e with arm_arch5t and arm_arch5te.
1610 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1611 (arm_arch5t, arm_arch5te): Declare.
1612 * config/arm/arm-tables.opt: Regenerate.
1613 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1614 * config/arm/t-multilib: Likewise.
1615 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1616 instead of arm_arch5.
1617 (*call_reg_thumb1): Likewise.
1618 (*call_value_reg_thumb1_v5): Likewise.
1619 (*call_value_reg_thumb1): Likewise.
1620 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1622 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1624 2018-05-18 Martin Liska <mliska@suse.cz>
1626 PR gcov-profile/84846
1627 * doc/gcov.texi: Document -t option of gcov tool.
1629 2018-05-18 Martin Liska <mliska@suse.cz>
1631 PR gcov-profile/84846
1632 * gcov.c (print_usage): Add new -t option.
1633 (process_args): Handle the option.
1634 (generate_results): Use stdout as output when requested by
1637 2018-05-18 Martin Liska <mliska@suse.cz>
1639 PR gcov-profile/84846
1640 * coverage.c (coverage_init): Write PWD to .gcno file.
1641 * doc/gcov.texi: Document how working directory is printed.
1642 * gcov-dump.c (dump_gcov_file): Print PWD.
1643 * gcov.c (output_intermediate_file): Likewise.
1644 (read_graph_file): Read PWD string.
1645 (output_lines): Print PWD.
1647 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1650 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1651 for retval and return false if all args to phi are zero.
1653 2018-05-18 Richard Biener <rguenther@suse.de>
1655 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1657 (evrp_dom_walker::before_dom_children): Call it.
1659 2018-05-18 Richard Biener <rguenther@suse.de>
1661 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1662 results when processing array refs with variable index.
1664 2018-05-18 Toon Moene <toon@moene.org>
1666 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1667 directly after that of -floop-interchange. Indicate that both
1668 options are enabled by default when specifying -O3.
1670 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1672 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1673 iterator. Delete separate integer-mode vec_set<mode> expander.
1674 (aarch64_simd_vec_setv2di): Delete.
1675 (vec_setv2di): Delete.
1676 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1677 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1678 the "w, r" alternative.
1680 2018-05-18 Martin Liska <mliska@suse.cz>
1682 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1683 * tree-pass.h (make_pass_lower_switch_O0): New function.
1684 * tree-switch-conversion.c (node_has_low_bound): Remove.
1685 (node_has_high_bound): Likewise.
1686 (node_is_bounded): Likewise.
1687 (class pass_lower_switch): Make it a template type and create
1689 (pass_lower_switch::execute): Add template argument.
1690 (make_pass_lower_switch): New function.
1691 (make_pass_lower_switch_O0): New function.
1692 (do_jump_if_equal): Remove.
1693 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1694 all the hard work to tree optimization passes.
1696 2018-05-18 Martin Liska <mliska@suse.cz>
1698 * dbgcnt.c (limit_low): Renamed from limit.
1699 (limit_high): New variable.
1700 (dbg_cnt_is_enabled): Check for upper limit.
1701 (dbg_cnt): Adjust dumping.
1702 (dbg_cnt_set_limit_by_index): Add new argument for high
1704 (dbg_cnt_set_limit_by_name): Likewise.
1705 (dbg_cnt_process_single_pair): Parse new format.
1706 (dbg_cnt_process_opt): Use strtok.
1707 (dbg_cnt_list_all_counters): Remove 'value' and add
1709 * doc/invoke.texi: Document changes.
1711 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1713 * doc/sourcebuild.texi (scalar_all_fma): Document.
1714 * tree.def (FMA_EXPR): Delete.
1715 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1716 * internal-fn.c (ternary_direct): New macro.
1717 (expand_ternary_optab_fn): Likewise.
1718 (direct_ternary_optab_supported_p): Likewise.
1719 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1720 * builtins.c (fold_builtin_fma): Delete.
1721 (fold_builtin_3): Don't call it.
1722 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1723 * expr.c (expand_expr_real_2): Likewise.
1724 * fold-const.c (operand_equal_p): Likewise.
1725 (fold_ternary_loc): Likewise.
1726 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1727 * gimple.c (DEFTREECODE): Likewise.
1728 * gimplify.c (gimplify_expr): Likewise.
1729 * optabs-tree.c (optab_for_tree_code): Likewise.
1730 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1731 * tree-eh.c (operation_could_trap_p): Likewise.
1732 (stmt_could_throw_1_p): Likewise.
1733 * tree-inline.c (estimate_operator_cost): Likewise.
1734 * tree-pretty-print.c (dump_generic_node): Likewise.
1735 (op_code_prio): Likewise.
1736 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1737 * tree-ssa-operands.c (get_expr_operands): Likewise.
1738 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1739 * fold-const-call.h (fold_fma): Delete.
1740 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1741 CFN_FNMA and CFN_FNMS.
1743 * genmatch.c (combined_fn): New enum.
1744 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1745 (commutative_op): New function.
1746 (commutate): Use it. Handle more than 2 operands.
1747 (dt_operand::gen_gimple_expr): Use commutative_op.
1748 (parser::parse_expr): Allow :c to be used with non-binary
1749 operators if the commutative operand is known.
1750 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1751 CFN_FMS, CFN_FNMA and CFN_FNMS.
1752 (backprop::process_assign_use): Remove FMA_EXPR handling.
1753 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1754 (gen_hsa_fma): New function.
1755 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1756 IFN_FNMA and IFN_FNMS.
1757 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1758 * gimple-fold.h (follow_all_ssa_edges): Declare.
1759 * gimple-fold.c (follow_all_ssa_edges): New function.
1760 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1761 gimple_build interface and use follow_all_ssa_edges to fold the result.
1762 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1763 instead of checking for optabs directly.
1764 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1765 rather than FMA_EXPRs.
1766 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1767 call to IFN_FMA instead of an FMA_EXPR.
1769 2018-05-17 Jim Wilson <jimw@sifive.com>
1771 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1772 sign extended promoted subreg, and the range does not have the sign bit
1773 set, then do a sign extend.
1775 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1776 test, check for sign extended subreg and/or constant operands, and
1777 do a sign extend in that case.
1779 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1781 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1782 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1784 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1785 Change logics_shift_reg to logics_shift_imm.
1786 (thunderx2t99_fp_loadpair_basic): Delete.
1787 (thunderx2t99_fp_storepair_basic): Delete.
1788 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1789 (thunderx2t99_asimd_polynomial): Delete.
1790 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1791 and neon_fp_mul_d_scalar_q.
1792 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1793 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1794 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1795 (thunderx2t99_asimd_lut): Add missing tbl types.
1796 (thunderx2t99_asimd_ext): Delete.
1797 (thunderx2t99_asimd_load1_1_mult): Delete.
1798 (thunderx2t99_asimd_load1_2_mult): Delete.
1799 (thunderx2t99_asimd_load1_ldp): New.
1800 (thunderx2t99_asimd_load1): New.
1801 (thunderx2t99_asimd_load2): Add missing *load2* types.
1802 (thunderx2t99_asimd_load3): New.
1803 (thunderx2t99_asimd_load4): New.
1804 (thunderx2t99_asimd_store1_1_mult): Delete.
1805 (thunderx2t99_asimd_store1_2_mult): Delete.
1806 (thunderx2t99_asimd_store2_mult): Delete.
1807 (thunderx2t99_asimd_store2_onelane): Delete.
1808 (thunderx2t99_asimd_store_stp): New.
1809 (thunderx2t99_asimd_store1): New.
1810 (thunderx2t99_asimd_store2): New.
1811 (thunderx2t99_asimd_store3): New.
1812 (thunderx2t99_asimd_store4): New.
1814 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1816 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1817 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1819 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1820 Segher Boessenkool <segher@kernel.crashing.org>
1823 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1826 2018-05-17 Richard Biener <rguenther@suse.de>
1828 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1829 for pruning loop and prune defs feeding only already visited PHIs.
1831 2018-05-17 Richard Biener <rguenther@suse.de>
1833 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1835 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1836 Richard Biener <rguenther@suse.de>
1838 PR tree-optimization/85793
1839 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1840 for VMAT_ELEMENTWISE.
1842 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1844 * internal-fn.h (lookup_internal_fn): Declare
1845 * internal-fn.c (lookup_internal_fn): New function.
1846 * gimple.c (gimple_build_call_from_tree): Handle calls to
1848 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1849 internal function names.
1850 * tree-pretty-print.c (dump_generic_node): Likewise.
1851 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1853 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1855 * gimple-fold.h (gimple_build): Make the function forms take
1856 combined_fn rather than built_in_function.
1857 (gimple_simplify): Likewise.
1858 * gimple-match-head.c (gimple_simplify): Likewise.
1859 * gimple-fold.c (gimple_build): Likewise.
1860 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1861 rather than gimple_build_call_internal.
1862 (get_initial_defs_for_reduction): Likewise.
1863 (vect_create_epilog_for_reduction): Likewise.
1864 (vectorizable_live_operation): Likewise.
1866 2018-05-17 Martin Liska <mliska@suse.cz>
1868 * gimple-ssa-sprintf.c (format_directive): Do not use
1869 space in between 'G_' and '('.
1871 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1874 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1875 even if the mask is not all ones.
1878 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1880 (ix86_gimple_fold_builtin): Likewise.
1883 * config/i386/i386.c: Include tree-vector-builder.h.
1884 (ix86_vector_shift_count): New function.
1885 (ix86_fold_builtin): Fold shift builtins by scalar count.
1886 (ix86_gimple_fold_builtin): Likewise.
1888 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1889 _mm512_setzero): New intrinsics.
1891 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1892 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1894 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1895 code generation for cases where splatting a value is not useful.
1896 * simplify-rtx.c (simplify_ternary_operation): Simplify
1897 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1898 mode to a vec_concat.
1900 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1902 * config.gcc: Support "goldmont-plus".
1903 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1905 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1906 PROCESSOR_GOLDMONT_PLUS.
1907 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1908 (processor_target_table): Add "goldmont-plus".
1909 (PTA_GOLDMONT_PLUS): Define.
1910 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1911 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1912 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1913 (fold_builtin_cpu): Add "goldmont-plus".
1914 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1915 (ix86_option_override_internal): Add "goldmont-plus".
1916 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1917 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1918 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1919 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1921 2018-05-17 Richard Biener <rguenther@suse.de>
1923 PR tree-optimization/85757
1924 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1925 remove defs that only feed that PHI from further processing.
1927 2018-05-16 Jim Wilson <jimw@sifive.com>
1929 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1931 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1933 2018-05-16 Mark Wielaard <mark@klomp.org>
1935 * dwarf2out.c (count_index_strings): New function.
1936 (output_indirect_strings): Call count_index_strings and generate
1937 header for dwarf_version >= 5.
1939 2018-05-16 Mark Wielaard <mark@klomp.org>
1941 * dwarf2out.c (dwarf_FORM): New function.
1942 (set_indirect_string): Use dwarf_FORM.
1943 (reset_indirect_string): Likewise.
1944 (size_of_die): Likewise.
1945 (value_format): Likewise.
1946 (output_die): Likewise.
1947 (add_skeleton_AT_string): Likewise.
1948 (output_macinfo_op): Likewise.
1949 (index_string): Likewise.
1950 (output_index_string_offset): Likewise.
1951 (output_index_string): Likewise.
1952 (count_index_strings): Likewise.
1954 2018-05-16 Carl Love <cel@us.ibm.com>
1956 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1957 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1959 2018-05-16 Martin Jambor <mjambor@suse.cz>
1961 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1962 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1964 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1966 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1967 (fnma<mode>4): Likewise.
1968 (fms<mode>4): Likewise.
1969 (fnms<mode>4): Likewise.
1970 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1971 (aarch64_fnma<mode>4): Likewise.
1972 (aarch64_fms<mode>4): Likewise.
1973 (aarch64_fnms<mode>4): Likewise.
1974 (aarch64_fnmadd<mode>4): Likewise.
1976 2018-05-16 Jason Merrill <jason@redhat.com>
1978 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1980 2018-05-16 Richard Biener <rguenther@suse.de>
1982 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1983 (dump_stmt_cost): Declare.
1984 (add_stmt_cost): Dump cost we add.
1985 (add_stmt_costs): New function.
1986 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1988 (vect_analyze_stmt): Adjust prototype.
1989 (vectorizable_condition): Likewise.
1990 (vectorizable_live_operation): Likewise.
1991 (vectorizable_reduction): Likewise.
1992 (vectorizable_induction): Likewise.
1993 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1994 cost vector to pass to vectorizable_ and record afterwards.
1995 (vect_model_reduction_cost): Take cost vector argument and adjust.
1996 (vect_model_induction_cost): Likewise.
1997 (vectorizable_reduction): Likewise.
1998 (vectorizable_induction): Likewise.
1999 (vectorizable_live_operation): Likewise.
2000 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2001 SLP_TREE_NUMBER_OF_VEC_STMTS.
2002 (vect_analyze_slp_cost_1): Remove.
2003 (vect_analyze_slp_cost): Likewise.
2004 (vect_slp_analyze_node_operations): Take visited args and
2005 a target cost vector. Avoid processing already visited stmt sets.
2006 (vect_slp_analyze_operations): Use a local cost vector to gather
2007 costs and register those of non-discarded instances.
2008 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2009 (vect_schedule_slp_instance): Remove copying of
2010 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2012 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2013 adding cost. Record cost entry location.
2014 (vect_prologue_cost_for_slp_op): Function to compute cost of
2015 a constant or invariant generated for SLP vect in the prologue,
2016 split out from vect_analyze_slp_cost_1.
2017 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2018 (vect_model_promotion_demotion_cost): Likewise.
2019 (vect_model_store_cost): Likewise, make static.
2020 (vect_model_load_cost): Likewise.
2021 (vectorizable_bswap): Add cost vector arg and adjust.
2022 (vectorizable_call): Likewise.
2023 (vectorizable_simd_clone_call): Likewise.
2024 (vectorizable_conversion): Likewise.
2025 (vectorizable_assignment): Likewise.
2026 (vectorizable_shift): Likewise.
2027 (vectorizable_operation): Likewise.
2028 (vectorizable_store): Likewise.
2029 (vectorizable_load): Likewise.
2030 (vectorizable_condition): Likewise.
2031 (vectorizable_comparison): Likewise.
2032 (can_vectorize_live_stmts): Likewise.
2033 (vect_analyze_stmt): Likewise.
2034 (vect_transform_stmt): Adjust calls to vectorizable_*.
2035 * tree-vectorizer.c: Include gimple-pretty-print.h.
2036 (dump_stmt_cost): New function.
2038 2018-05-16 Richard Biener <rguenther@suse.de>
2040 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2041 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2042 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2043 (check_name): New callback.
2044 (dse_classify_store): Track cycles via a visited bitmap of PHI
2045 defs and simplify handling of in-loop and across loop dead stores
2046 and properly fail for loop-variant refs. Handle byte-tracking with
2047 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2050 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2052 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2053 (vect_get_mask_type_for_stmt): Likewise.
2054 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2056 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2057 to determine the statement's vector type and the vector type that
2058 should be used for calculating nunits. Deal with cases in which
2059 the type has to be deferred.
2060 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2061 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2062 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2063 (vect_determine_vf_for_stmt): New functions, split out from...
2064 (vect_determine_vectorization_factor): ...here.
2065 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2066 (vect_get_mask_type_for_stmt): New functions, split out from
2067 vect_determine_vectorization_factor.
2069 2018-05-16 Richard Biener <rguenther@suse.de>
2071 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2072 verify the [VEC_]COND_EXPR embedded comparison.
2074 2018-05-15 Martin Sebor <msebor@redhat.com>
2076 PR tree-optimization/85753
2077 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2078 RECORD_TYPE in addition to ARRAY_TYPE.
2080 2018-05-15 Martin Sebor <msebor@redhat.com>
2083 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2085 2018-05-15 Richard Biener <rguenther@suse.de>
2087 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2088 add by_clobber_p one. Change algorithm to collect all defs
2089 representing uses we need to walk and try reducing them to
2090 a single one before failing.
2091 (dse_dom_walker::dse_optimize_stmt): Adjust.
2093 2018-05-13 Mark Wielaard <mark@klomp.org>
2095 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2096 (size_of_loc_descr): Likewise.
2097 (output_loc_operands): Likewise.
2098 (output_loc_operands_raw): Likewise.
2099 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2100 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2101 (hash_loc_operands): Likewise.
2102 (compare_loc_operands): Likewise.
2104 2018-05-14 Mark Wielaard <mark@klomp.org>
2106 * dwarf2out.c (count_index_addrs): New function.
2107 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2109 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2111 PR tree-optimization/83648
2112 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2113 return value as malloc candidate.
2115 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2118 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2119 as true in call to suggest_attribute.
2121 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2123 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2124 -mreadonly-in-sdata.
2126 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2128 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2130 (aarch64_crypto_aesd_fused): Likewise.
2132 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2134 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2135 (movsi_aarch64): Likewise.
2136 (load_pairsi): Likewise.
2137 (load_pairdi): Likewise.
2138 (store_pairsi): Likewise.
2139 (store_pairdi): Likewise.
2140 (load_pairsf): Likewise.
2141 (load_pairdf): Likewise.
2142 (store_pairsf): Likewise.
2143 (store_pairdf): Likewise.
2144 (zero_extend): Likewise.
2145 (trunc): Swap alternatives.
2146 (fcvt_target): Add '?' to prefer w over r.
2148 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2151 * config/i386/i386.md: Disallow non-commutative arithmetics in
2152 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2153 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2154 in the peephole2 before it.
2156 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2158 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2159 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2160 (ix86_handle_option): Handle -mcldemote.
2161 * config.gcc: New header.
2162 * config/i386/cldemoteintrin.h: New file.
2163 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2164 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2166 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2167 OPTION_MASK_ISA_CLDEMOTE.
2168 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2169 (ix86_valid_target_attribute_inner_p): Ditto.
2170 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2171 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2172 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2173 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2174 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2176 * config/i386/i386.opt: Add -mcldemote.
2177 * config/i386/x86intrin.h: New header.
2178 * doc/invoke.texi: Add -mcldemote.
2180 2018-05-14 Richard Biener <rguenther@suse.de>
2182 * doc/match-and-simplify.texi: Adjust :s documentation.
2184 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2186 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2187 intended memcpy size.
2188 (REORDER_45): Likewise.
2190 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2192 * sort.cc: New file.
2193 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2194 * vec.c (qsort_chk): Use gcc_qsort.
2195 * Makefile.in (OBJS-libcommon): Add sort.o.
2196 (build/sort.o): New target. Use it...
2197 (BUILD_RTL): ... here, and...
2198 (build/gencfn-macros): ... here, and...
2199 (build/genmatch): ... here.
2201 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2202 Chung-Ju Wu <jasonwucj@gmail.com>
2204 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2205 * config/nds32/nds32-graywolf.md: New file.
2206 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2207 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2209 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2210 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2211 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2212 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2213 * config/nds32/pipelines.md: Include n15 settings.
2215 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2216 Chung-Ju Wu <jasonwucj@gmail.com>
2218 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2219 * config/nds32/nds32-n13.md: New file.
2220 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2221 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2223 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2224 * config/nds32/nds32.md (pipeline_model): Add n13.
2225 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2226 * config/nds32/pipelines.md: Include n13 settings.
2228 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2229 Chung-Ju Wu <jasonwucj@gmail.com>
2231 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2232 * config/nds32/nds32-n10.md: New file.
2233 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2234 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2236 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2237 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2238 * config/nds32/nds32.md (pipeline_model): Add n10.
2239 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2240 * config/nds32/pipelines.md: Include n10 settings.
2242 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2243 Kito Cheng <kito.cheng@gmail.com>
2244 Chung-Ju Wu <jasonwucj@gmail.com>
2246 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2247 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2248 Add enum values for DSP extension instructions.
2249 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2251 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2252 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2254 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2255 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2256 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2257 * config/nds32/nds32-intrinsic.md: Likewise.
2258 * config/nds32/nds32_intrinsic.h: Likewise.
2259 * config/nds32/nds32-md-auxiliary.c: Likewise.
2260 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2261 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2262 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2263 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2264 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2265 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2266 TYPE_DMAC in switch statement.
2267 * config/nds32/nds32.c: New checking and implementation for DSP
2268 extension instructions.
2269 * config/nds32/nds32.h: Likewise.
2270 * config/nds32/nds32.md: Likewise.
2271 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2272 * config/nds32/predicates.md: Implement new predicates for DSP
2275 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2277 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2278 Reformat alternatives and attributes so it is easier to identify
2279 which constraints/attributes go with which instruction.
2280 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2281 (mov<mode>_softfloat32, FMOVE64): Likewise.
2282 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2283 (mov<mode>_softfloat64, FMOVE64): Likewise.
2285 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2287 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2289 (Basic PowerPC Built-in Functions): The new name of the
2290 subsection previously known as "PowerPC Built-in Functions".
2291 (Basic PowerPC Built-in Functions Available on all Configurations):
2293 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2294 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2295 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2296 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2298 2018-05-11 Martin Jambor <mjambor@suse.cz>
2301 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2304 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2307 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2309 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2311 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2312 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2313 (ix86_handle_option): Handle -mwaitpkg.
2314 * config.gcc: New header.
2315 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2316 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2317 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2319 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2320 OPTION_MASK_ISA_WAITPKG.
2321 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2322 (ix86_option_override_internal): Add PTA_WAITPKG.
2323 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2324 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2325 IX86_BUILTIN_TPAUSE.
2326 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2327 __builtin_ia32_umwait and __builtin_ia32_tpause.
2328 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2329 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2330 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2331 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2332 UNSPECV_TPAUSE): New.
2333 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2334 * config/i386/i386.opt: Add -mwaitpkg.
2335 * config/i386/waitpkgintrin.h: New file.
2336 * config/i386/x86intrin.h: New header.
2337 * doc/invoke.texi: Add -mwaitpkg.
2339 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2342 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2344 (cortex-m0): Use armv6s-m isa.
2345 (cortex-m0plus): Likewise.
2346 (cortex-m1): Likewise.
2347 (cortex-m0.small-multiply): Likewise.
2348 (cortex-m0plus.small-multiply): Likewise.
2349 (cortex-m1.small-multiply): Likewise.
2351 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2352 Jakub Jelinek <jakub@redhat.com>
2354 PR tree-optimization/85692
2355 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2356 source permute as well.
2358 2018-05-11 Martin Liska <mliska@suse.cz>
2361 * doc/extend.texi: Document LLVM style format for no_sanitize
2364 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2366 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2367 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2368 (mode_supports_vsx_dform_quad): Likewise.
2369 (mode_supports_vmx_dform): Move these functions to be next to the
2370 other mode_supports functions.
2371 (mode_supports_dq_form): Likewise.
2372 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2373 mode_supports_dq_form.
2374 (reg_offset_addressing_ok_p): Likewise.
2375 (offsettable_ok_by_alignment): Likewise.
2376 (rs6000_legitimate_offset_address_p): Likewise.
2377 (legitimate_lo_sum_address_p): Likewise.
2378 (rs6000_legitimize_address): Likewise.
2379 (rs6000_legitimize_reload_address): Likewise.
2380 (rs6000_secondary_reload_inner): Likewise.
2381 (rs6000_preferred_reload_class): Likewise.
2382 (rs6000_output_move_128bit): Likewise.
2384 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2386 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2387 Generate SImode target register for null target.
2388 <case IX86_BUILTIN_XGETBV>: Ditto.
2389 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2390 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2392 2018-05-10 Carl Love <cel@us.ibm.com>
2394 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2395 dcbtt and dcbtstt if operands[2] is 0.
2397 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2400 * config/i386/sse.md (usadv64qi): New expander.
2402 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2404 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2405 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2406 -maltivec=be support.
2407 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2408 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2409 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2410 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2411 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2412 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2413 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2414 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2415 altivec_vsumsws): Adjust.
2416 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2417 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2418 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2420 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2421 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2422 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2423 (altivec_lve<VI_char>x): Delete expand.
2424 (*altivec_lve<VI_char>x_internal): Rename to...
2425 (altivec_lve<VI_char>x): ... this.
2426 (altivec_lvxl_<mode>): Delete expand.
2427 (*altivec_lvxl_<mode>_internal): Rename to ...
2428 (altivec_lvxl_<mode>): ... this.
2429 (altivec_stvxl_<mode>): Delete expand.
2430 (*altivec_stvxl_<mode>_internal): Rename to ...
2431 (altivec_stvxl_<mode>): ... this.
2432 (altivec_stve<VI_char>x): Delete expand.
2433 (*altivec_stve<VI_char>x_internal): Rename to ...
2434 (altivec_stve<VI_char>x): ... this.
2435 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2436 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2437 reduc_plus_scal_<mode>): Adjust.
2438 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2440 (rs6000_cpu_cpp_builtins): Adjust.
2441 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2442 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2443 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2444 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2445 -maltivec=be support.
2446 (rs6000_split_vec_extract_var): Adjust.
2447 (rs6000_split_v4si_init): Adjust.
2448 (swap_selector_for_mode): Delete.
2449 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2450 altivec_expand_stvex_be): Delete.
2451 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2452 -maltivec=be support.
2453 (rs6000_gimple_fold_builtin): Ditto.
2454 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2456 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2457 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2458 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2459 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2460 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2461 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2462 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2463 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2464 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2465 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2466 anonymous split): Adjust.
2467 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2468 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2470 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2472 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2473 when --with-gxx-include-dir is also specified.
2474 * configure: Regenerate.
2476 2018-05-09 Jim Wilson <jimw@sifive.com>
2479 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2480 * config/riscv/t-withmultilib: New.
2481 * config/riscv/withmultilib.h: New.
2482 * doc/install.texi: Document RISC-V --with-multilib-list support.
2484 2018-05-09 Richard Biener <rguenther@suse.de>
2486 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2488 (vect_bb_vectorization_profitable_p): Adjust. Compute
2489 actual scalar cost using the cost vector and the add_stmt_cost
2492 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2494 PR rtl-optimization/85645
2495 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2496 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2498 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2500 PR rtl-optimization/85645
2501 * shrink-wrap.c (spread_components): Return a boolean saying if
2502 anything was changed.
2503 (try_shrink_wrapping_separate): Iterate spread_components until
2504 nothing changes anymore.
2506 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2508 PR rtl-optimization/85645
2509 * regrename.c (build_def_use): Also kill the chains that include the
2510 destination of a REG_CFA_REGISTER note.
2512 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2514 PR rtl-optimization/85645
2515 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2516 insn that has a REG_CFA_REGISTER note.
2518 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2520 * cfgexpand.c (expand_clobber): New function.
2521 (expand_gimple_stmt_1): Use it.
2522 * tree-vect-stmts.c (vect_clobber_variable): New function,
2524 (vectorizable_simd_clone_call): ...here.
2525 (vectorizable_store): Emit a clobber either side of an
2526 IFN_STORE_LANES sequence.
2527 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2529 2018-05-09 Tom de Vries <tom@codesourcery.com>
2532 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2533 (define_insn "trap_if_false"): Add exit after trap.
2535 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2537 PR rtl-optimization/85638
2538 * bb-reorder.c: Include common/common-target.h.
2539 (create_forwarder_block): New function extracted from...
2540 (fix_up_crossing_landing_pad): ...here. Rename into...
2541 (dw2_fix_up_crossing_landing_pad): ...this.
2542 (sjlj_fix_up_crossing_landing_pad): New function.
2543 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2544 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2545 from both partitions and exit the loop after one iteration.
2547 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2550 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2552 (Basic PowerPC Built-in Functions): The new name of the
2553 subsection previously known as "PowerPC Built-in Functions".
2554 (Basic PowerPC Built-in Functions Available on all Configurations):
2556 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2558 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2560 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2562 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2565 2018-05-08 Jim Wilson <jimw@sifive.com>
2567 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2568 (LD_EMUL_SUFFIX): New.
2569 (LINK_SPEC): Use it.
2571 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2573 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2575 (Basic PowerPC Built-in Functions): The new name of the
2576 subsection previously known as "PowerPC Built-in Functions".
2577 (Basic PowerPC Built-in Functions Available on all Configurations):
2579 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2581 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2583 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2585 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2588 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2591 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2592 after cmpelim optimization.
2594 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2596 * config.gcc: Support "goldmont".
2597 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2598 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2600 * config/i386/i386.c (m_GOLDMONT): Define.
2601 (processor_target_table): Add "goldmont".
2602 (PTA_GOLDMONT): Define.
2603 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2604 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2605 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2606 (fold_builtin_cpu): Add "goldmont".
2607 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2608 (ix86_option_override_internal): Add "goldmont".
2609 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2610 (processor_type): Add PROCESSOR_GOLDMONT.
2611 * config/i386/i386.md: Add CPU "glm".
2612 * config/i386/glm.md: New file.
2613 * config/i386/x86-tune.def: Add m_GOLDMONT.
2614 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2616 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2619 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2621 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2622 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2623 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2626 * config/i386/i386.c (ix86_fold_builtin): Handle
2627 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2630 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2631 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2633 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2636 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2637 (check_arch): Likewise.
2638 (check_fpu): Return the result rather than printing it.
2639 (end arch): Fix operator precedence.
2640 (end cpu): Likewise.
2641 (END): Print the result from check_fpu.
2643 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2644 Alan Hayward <alan.hayward@arm.com>
2645 David Sherwood <david.sherwood@arm.com>
2647 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2648 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2649 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2650 (*fcmuo<mode>_and): New patterns.
2652 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2654 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2655 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2656 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2657 (cmp_op, sve_imm_con): New code attributes.
2658 (SVE_COND_INT_CMP, imm_con): Delete.
2659 (cmp_op): Remove above unspecs from int attribute.
2660 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2662 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2663 comparison-specific unspecs.
2664 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2665 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2666 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2667 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2668 (*vec_fcm<cmp_op><mode>): Rename to...
2669 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2670 (*vec_fcmuo<mode>): Rename to...
2671 (*fcmuo<mode>): ...this and adjust likewise.
2672 (*pred_fcm<cmp_op><mode>): New pattern.
2673 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2674 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2676 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2678 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2679 (aarch64_emit_sve_predicated_cond): New function.
2680 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2681 (aarch64_emit_unspec_cond_or): Replace with...
2682 (aarch64_emit_sve_or_conds): ...this new function. Use
2683 aarch64_emit_sve_ptrue_op for the individual comparisons and
2684 aarch64_emit_binop to OR them together.
2685 (aarch64_emit_inverted_unspec_cond): Replace with...
2686 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2687 aarch64_emit_sve_ptrue_op for the comparison and
2688 aarch64_emit_unop to invert the result.
2689 (aarch64_expand_sve_vec_cmp_float): Update after the above
2690 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2692 2018-05-07 Nathan Sidwell <nathan@acm.org>
2694 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2695 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2696 (Backwards Compatibility): Likewise.
2698 2018-05-07 Luis Machado <luis.machado@linaro.org>
2702 2018-05-07 Luis Machado <luis.machado@linaro.org>
2704 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2705 <prefetch_dynamic_strides>: New const bool field.
2706 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2707 prefetch_dynamic_strides.
2708 (exynosm1_prefetch_tune): Likewise.
2709 (thunderxt88_prefetch_tune): Likewise.
2710 (thunderx_prefetch_tune): Likewise.
2711 (thunderx2t99_prefetch_tune): Likewise.
2712 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2713 (aarch64_override_options_internal): Update to set
2714 PARAM_PREFETCH_DYNAMIC_STRIDES.
2715 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2716 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2717 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2718 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2719 prefetch-dynamic-strides setting.
2721 2018-05-07 Luis Machado <luis.machado@linaro.org>
2723 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2724 <minimum_stride>: New const int field.
2725 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2726 minimum_stride field.
2727 (exynosm1_prefetch_tune): Likewise.
2728 (thunderxt88_prefetch_tune): Likewise.
2729 (thunderx_prefetch_tune): Likewise.
2730 (thunderx2t99_prefetch_tune): Likewise.
2731 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2732 (aarch64_override_options_internal): Update to set
2733 PARAM_PREFETCH_MINIMUM_STRIDE.
2734 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2735 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2736 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2737 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2738 stride is constant and is below the minimum stride threshold.
2740 2018-05-07 Luis Machado <luis.machado@linaro.org>
2742 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2745 2018-05-07 Luis Machado <luis.machado@linaro.org>
2747 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2748 <prefetch_dynamic_strides>: New const bool field.
2749 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2750 prefetch_dynamic_strides.
2751 (exynosm1_prefetch_tune): Likewise.
2752 (thunderxt88_prefetch_tune): Likewise.
2753 (thunderx_prefetch_tune): Likewise.
2754 (thunderx2t99_prefetch_tune): Likewise.
2755 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2756 (aarch64_override_options_internal): Update to set
2757 PARAM_PREFETCH_DYNAMIC_STRIDES.
2758 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2759 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2760 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2761 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2762 prefetch-dynamic-strides setting.
2764 2018-05-07 Luis Machado <luis.machado@linaro.org>
2766 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2767 <minimum_stride>: New const int field.
2768 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2769 minimum_stride field.
2770 (exynosm1_prefetch_tune): Likewise.
2771 (thunderxt88_prefetch_tune): Likewise.
2772 (thunderx_prefetch_tune): Likewise.
2773 (thunderx2t99_prefetch_tune): Likewise.
2774 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2775 (aarch64_override_options_internal): Update to set
2776 PARAM_PREFETCH_MINIMUM_STRIDE.
2777 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2778 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2779 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2780 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2781 stride is constant and is below the minimum stride threshold.
2783 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2786 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2787 the type is addressable. Don't force op into register if it has
2790 2018-05-05 Roland McGrath <mcgrathr@google.com>
2793 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2794 any section for which we don't know a specific type it should have,
2795 regardless of name. Previously this was done only for the exact
2796 names ".init_array", ".fini_array", and ".preinit_array".
2797 (default_elf_asm_named_section): Add comment about
2798 relationship with default_section_type_flags and SECTION_NOTYPE.
2799 (get_section): Don't consider it a type conflict if one side has
2800 SECTION_NOTYPE and the other doesn't, as long as neither has the
2801 SECTION_BSS et al used in the default_section_type_flags logic.
2803 2018-05-05 Tom de Vries <tom@codesourcery.com>
2806 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2807 (workaround_barsyncs): New function.
2808 (nvptx_reorg): Use workaround_barsyncs.
2809 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2810 (define_expand "nvptx_membar_cta"): New define_expand.
2811 (define_insn "*nvptx_membar_cta"): New insn.
2813 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2815 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2816 To improve optimization opportunities.
2817 * builtin-types.def: The new needed builtin types for the above.
2819 2018-05-04 Richard Biener <rguenther@suse.de>
2821 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2822 * gimple-ssa-store-merging.c
2823 (imm_store_chain_info::output_merged_store): Remove redundant create,
2824 release split_store vector contents on failure.
2825 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2826 scalar stmt vector on cache hit.
2828 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2830 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2832 * config.gcc (powerpc-xilinx-eabi*): Remove.
2833 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2835 (fusion_addis_mem_combo_load): Ditto.
2836 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2838 (rs6000_cpu_cpp_builtins): Ditto.
2839 * config/rs6000/rs6000-linux.c
2840 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2841 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2842 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2844 (rs6000_setup_reg_addr_masks): Ditto.
2845 (rs6000_init_hard_regno_mode_ok): Ditto.
2846 (rs6000_option_override_internal): Ditto.
2847 (legitimate_lo_sum_address_p): Ditto.
2848 (rs6000_legitimize_address): Ditto.
2849 (rs6000_legitimize_reload_address): Ditto.
2850 (rs6000_legitimate_address_p): Ditto.
2851 (abi_v4_pass_in_fpr): Ditto.
2852 (setup_incoming_varargs): Ditto.
2853 (rs6000_gimplify_va_arg): Ditto.
2854 (rs6000_split_multireg_move): Ditto.
2855 (rs6000_savres_strategy): Ditto.
2856 (rs6000_emit_prologue_components): Ditto.
2857 (rs6000_emit_epilogue_components): Ditto.
2858 (rs6000_emit_prologue): Ditto.
2859 (rs6000_emit_epilogue): Ditto.
2860 (rs6000_elf_file_end): Ditto.
2861 (rs6000_function_value): Ditto.
2862 (rs6000_libcall_value): Ditto.
2863 * config/rs6000/rs6000.h: Ditto.
2864 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2865 (TARGET_MINMAX): ... this. New.
2866 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2867 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2868 (*movsi_internal1_single): Delete.
2869 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2870 mfpu=, mxilinx-fpu): Delete.
2871 * config/rs6000/singlefp.h: Delete.
2872 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2873 * config/rs6000/t-rs6000: Ditto.
2874 * config/rs6000/t-xilinx: Delete.
2875 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2876 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2877 (VStype_simple): Delete.
2878 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2879 * config/rs6000/xfpu.h: Delete.
2880 * config/rs6000/xfpu.md: Delete.
2881 * config/rs6000/xilinx.h: Delete.
2882 * config/rs6000/xilinx.opt: Delete.
2883 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2884 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2886 2018-05-04 Tom de Vries <tom@codesourcery.com>
2889 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2892 2018-05-04 Richard Biener <rguenther@suse.de>
2895 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2896 (expand_complex_div_wide): Likewise.
2897 (expand_complex_operations_1): Likewise.
2898 (expand_complex_libcall): Preserve EH info of the original stmt.
2899 (tree_lower_complex): Handle removed blocks.
2900 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2901 on complex multiplication and division libcall builtins.
2903 2018-05-04 Richard Biener <rguenther@suse.de>
2906 * fold-const.c (negate_expr_p): Restrict negation of operand
2907 zero of a division to when we know that can happen without
2909 (fold_negate_expr_1): Likewise.
2911 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2914 * real.h (real_nextafter): Declare.
2915 * real.c (real_nextafter): New function.
2916 * fold-const-call.c (fold_const_nextafter): New function.
2917 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2918 CASE_CFN_NEXTTOWARD.
2919 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2920 even when arg1_mode is different from arg0_mode.
2922 2018-05-03 Nathan Sidwell <nathan@acm.org>
2924 * doc/extend.texi (Deprecated Features): Remove
2926 (Backwards Compatibility): Likewise.
2927 * doc/invoke.texi (C++ Language Options): Likewise.
2928 (C++ Dialect Options): Likewise.
2930 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2933 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2934 _mm512_mask_mullox_epi64): New intrinsics.
2936 2018-05-03 Tom de Vries <tom@codesourcery.com>
2939 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2940 dump files): Add offload-tree.
2942 2018-05-03 Richard Biener <rguenther@suse.de>
2944 PR tree-optimization/85615
2945 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2946 to loops not nested in BBs loop father to avoid creating multi-entry
2949 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2951 PR tree-optimization/70291
2952 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2953 arguments. Change return type to tree. Emit libcall as a new
2954 statement rather than replacing existing one when inplace_p is true.
2955 (expand_complex_multiplication_components): New function.
2956 (expand_complex_multiplication): Expand floating-point complex
2957 multiplication using the above.
2958 (expand_complex_division): Rename inner_type parameter to type.
2959 Update expand_complex_libcall call-site.
2960 (expand_complex_operations_1): Update expand_complex_multiplication
2961 and expand_complex_division call-sites.
2963 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2966 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2967 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2968 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2969 the highest significant bit of the shift count mask is clear. In
2970 check whether and[sq]i3 is needed verify that all significant bits
2971 of the shift count other than the highest are set.
2973 2018-05-02 Tom de Vries <tom@codesourcery.com>
2976 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2977 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2978 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2979 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2980 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2981 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2982 __builtin_goacc_parlevel_size.
2984 2018-05-02 Richard Biener <rguenther@suse.de>
2986 PR tree-optimization/85597
2987 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2988 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
2997 2018-05-02 Tom de Vries <tom@codesourcery.com>
3000 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3001 dump files): Add wpa-ipa.
3003 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3005 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3006 powerpc*-*-linux*paired* target.
3007 * config/rs6000/750cl.h: Delete.
3008 * config/rs6000/paired.h: Delete.
3009 * config/rs6000/paired.md: Delete.
3010 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3012 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3013 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3014 comment. Remove paired float support.
3015 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3016 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3018 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3019 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3021 * config/rs6000/rs6000.c: Remove paired float support.
3022 (paired_expand_vector_init, paired_expand_vector_move,
3023 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3024 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3025 paired_expand_builtin, paired_expand_predicate_builtin,
3026 paired_init_builtins): Delete.
3027 * config/rs6000/rs6000.h: Remove paired float support.
3028 * config/rs6000/rs6000.md: Remove paired float support.
3029 (move_from_CR_ov_bit): Delete.
3030 * config/rs6000/rs6000.opt (mpaired): Delete.
3031 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3032 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3034 2018-05-02 Richard Biener <rguenther@suse.de>
3037 * gimplify.c (gimplify_save_expr): When in SSA form allow
3038 SAVE_EXPRs to compute to SSA vars.
3040 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3043 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3044 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3045 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3046 clobber operands[2], instead use a new pseudo. Formatting fixes.
3048 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3050 PR tree-optimization/85586
3051 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3052 exit early for statements in the same group if the accesses are
3055 2018-05-02 Tom de Vries <tom@codesourcery.com>
3058 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3061 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3063 PR tree-optimization/85143
3064 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3066 2018-05-01 Tom de Vries <tom@codesourcery.com>
3069 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3070 not found" error message.
3072 2018-05-01 Tom de Vries <tom@codesourcery.com>
3075 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3076 * vec.c (test_ordered_remove_if): New function.
3077 (vec_c_tests): Call test_ordered_remove_if.
3078 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3079 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3080 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3081 VEC_ORDERED_REMOVE_IF.
3083 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3085 PR tree-optimization/82665
3086 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3087 pointer subtraction where arguments come from a memchr call.
3089 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3091 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3092 --push-state --as-needed and --pop-state instead of --as-needed and
3093 --no-as-needed if ld supports it.
3094 * configure: Regenerated.
3097 * doc/install.texi2html: Replace _002d with - and _002a with * in
3098 generated html files using sed.
3100 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3103 * gcc-rich-location.c (blank_line_before_p): New function.
3104 (use_new_line): New function.
3105 (gcc_rich_location::add_fixit_insert_formatted): New function.
3106 * gcc-rich-location.h
3107 (gcc_rich_location::add_fixit_insert_formatted): New function.
3109 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3111 * selftest.c (assert_streq): Rename "expected" and "actual" to
3112 "val1" and "val2". Extend NULL-handling to cover both inputs
3113 symmetrically, while still requiring both to be non-NULL for a pass.
3114 * selftest.h (assert_streq): Rename "expected" and "actual" to
3116 (ASSERT_EQ): Likewise.
3117 (ASSERT_EQ_AT): Likewise.
3118 (ASSERT_KNOWN_EQ): Likewise.
3119 (ASSERT_KNOWN_EQ_AT): Likewise.
3120 (ASSERT_NE): Likewise.
3121 (ASSERT_MAYBE_NE): Likewise.
3122 (ASSERT_MAYBE_NE_AT): Likewise.
3123 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3124 the assertion to pass.
3125 (ASSERT_STREQ_AT): Likewise.
3127 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3129 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3130 interaction with -pie.
3132 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3134 * selftest.h: Fix alphabetization of per-source-file selftest
3137 2018-04-30 Jason Merrill <jason@redhat.com>
3139 PR c++/61982 - dead stores to destroyed objects.
3140 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3143 2018-04-30 Jason Merrill <jason@redhat.com>
3145 * tree.c (build_clobber): New.
3146 * tree.h: Declare it.
3147 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3149 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3151 * diagnostic-show-locus.c (layout::layout): Update for
3152 location_get_source_line returning a char_span.
3153 (struct char_span): Move to input.h.
3154 (struct correction): Update for fields in char_span becoming
3156 (struct source_line): Update for location_get_source_line
3157 returning a char_span.
3158 (layout::print_line): Likewise.
3159 * edit-context.c (edited_file::print_content): Likewise.
3160 (edited_file::print_diff_hunk): Likewise.
3161 (edited_file::print_run_of_changed_lines): Likewise.
3162 (edited_file::get_num_lines): Likewise.
3163 (edited_line::edited_line): Likewise.
3164 * final.c (asm_show_source): Likewise.
3165 * input.c (location_get_source_line): Convert return type
3166 from const char * to char_span, losing the final "line_len"
3168 (dump_location_info): Update for the above.
3169 (get_substring_ranges_for_loc): Likewise. Use a char_span
3170 when handling the literal within the line.
3171 (test_reading_source_line): Update for location_get_source_line
3172 returning a char_span.
3173 * input.h (class char_span): Move here from
3174 diagnostic-show-locus.c, converting from a struct to a class.
3175 Make data members private.
3176 (char_span::operator bool): New.
3177 (char_span::length): New.
3178 (char_span::get_buffer): New.
3179 (char_span::operator[]): New.
3180 (char_span::subspan): Make const.
3181 (char_span::xstrdup): New.
3182 (location_get_source_line): Convert return type from const char *
3183 to char_span, losing the final "line_size" param.
3185 2018-04-30 Jan Hubicka <jh@suse.cz>
3187 * lto-wrapper.c (ltrans_priorities): New static var.
3188 (cmp_priority): New.
3189 (run_gcc): Read priorities and if doing parallel build order
3190 the Makefile by them.
3192 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3194 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3196 2018-04-30 Richard Biener <rguenther@suse.de>
3198 * tree-cfg.c (verify_address): Remove base argument, add
3199 flag whether to check TREE_ADDRESSABLE and do that.
3200 (verify_expr): Remove.
3201 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3202 (verify_gimple_assign_single): Likewise.
3203 (verify_gimple_switch): Likewise.
3204 (verify_expr_location_1): Dereference tp once. Add (disabled)
3205 piece from verify_expr.
3206 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3208 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3210 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3212 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3214 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3215 (small_data_pattern): Likewise.
3216 (arc_rewrite_small_data): Likewise.
3217 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3218 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3219 (get_symbol_alignment): New function.
3220 (legitimate_small_data_address_p): Likewise.
3221 (legitimate_scaled_address): Update, call
3222 legitimate_small_data_address_p.
3223 (output_sdata): New static variable.
3224 (arc_print_operand): Update how we handle small data operands.
3225 (arc_print_operand_address): Likewise.
3226 (arc_legitimate_address_p): Update, use
3227 legitimate_small_data_address_p.
3228 (arc_rewrite_small_data_p): Remove.
3229 (arc_rewrite_small_data_1): Likewise.
3230 (arc_rewrite_small_data): Likewise.
3231 (small_data_pattern): Likewise.
3232 (compact_sda_memory_operand): Update to use
3233 legitimate_small_data_address_p and get_symbol_alignment.
3234 (prepare_move_operands): Don't rewite sdata pattern.
3235 (prepare_extend_operands): Remove.
3236 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3238 (zero_extendqisi2): Likewise.
3239 (zero_extendhisi2): Likewise.
3240 (extendqihi2): Likewise.
3241 (extendqisi2): Likewise.
3242 (extendhisi2): Likewise.
3246 * config/arc/constraints.md (Usd): Change it to memory constraint.
3248 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3250 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3251 as source of std instructions.
3252 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3253 allow 6-bit constants as source for store instructions.
3254 (movdi_insn): Update instruction pattern to allow 6-bit constants
3255 as source for store instructions.
3257 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3259 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3261 2018-04-30 Nathan Sidwell <nathan@acm.org>
3262 Sandra Loosemore <sandra@codesourcery.com>
3264 * dumpfile.c (dump_open): Allow '-' for stdout.
3265 * doc/invoke.texi (Developer Options): Document dump filename
3266 determination early. Document stdin/stdout selection.
3268 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3270 Microblaze Target: PIC data text relative
3272 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3273 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3275 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3276 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3277 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3278 New addressing mode for data-text relative position indepenedent code.
3279 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3280 'ADDRESS_SYMBOLIC_TXT_REL'.
3281 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3282 (microblaze_legitimate_pic_operand): Exclude function calls from
3283 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3284 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3286 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3287 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3288 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3289 for 'address + offset'.
3290 (microblaze_expand_prologue): Add new function prologue call for
3292 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3293 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3294 table in case of TARGET_PIC_DATA_TEXT_REL.
3295 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3296 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3297 Add new macros 'UNSPEC_TEXT',
3298 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3299 + exclude function calls from 'UNSPEC_PLT' in case of data text
3301 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3302 new target hook for generating address diff vector tables in case of
3304 * doc/tm.texi : Regenerate.
3305 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3306 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3307 of addr diff vector generation.
3308 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3309 target hook definition.
3310 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3311 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3312 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3313 option for data text relative.
3315 2018-04-30 Richard Biener <rguenther@suse.de>
3317 * tree-chrec.h (evolution_function_is_constant_p): Remove
3319 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3321 2018-04-30 Richard Biener <rguenther@suse.de>
3324 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3326 2018-04-30 Richard Biener <rguenther@suse.de>
3328 PR tree-optimization/28364
3329 PR tree-optimization/85275
3330 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3331 copying first exit test.
3333 2018-04-28 Mark Wielaard <mark@klomp.org>
3335 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3337 (dwarf_AT): Handle DW_AT_addr_base.
3338 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3340 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3343 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3344 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3345 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3346 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3348 2018-04-28 Richard Biener <rguenther@suse.de>
3350 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3351 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3352 to reflect use. Only add interesting stmts.
3354 2018-04-27 Martin Jambor <mjambor@suse.cz>
3357 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3358 the jump function allows for passing through aggregate values.
3360 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3362 * input.h (in_system_header_at): Convert from macro to inline
3364 (from_macro_expansion_at): Likewise.
3365 (from_macro_definition_at): Likewise.
3367 2018-04-27 Jeff Law <law@redhat.com>
3369 * config.gcc: Mark tile* targets as deprecated/obsolete.
3371 2018-04-27 Richard Biener <rguenther@suse.de>
3373 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3376 2018-04-27 Richard Biener <rguenther@suse.de>
3378 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3380 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3382 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3383 with Yd constraint. Set "preferred_for_speed" attribute from
3384 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3386 (*movdi_internal): Ditto.
3387 (movti_interunit splitters): Remove
3388 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3389 (movdi_interunit splitters): Ditto.
3390 * config/i386/constraints.md (Ye): Remove.
3391 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3393 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3396 * config/aarch64/constraints.md (Usg): Limit to 31.
3399 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3401 PR tree-optimization/85529
3402 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3403 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3404 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3405 zero extension or masking of the MSB bit.
3406 (optimize_range_tests): Add FIRST_BB argument, pass it through
3407 to optimize_range_tests_var_bound.
3408 (maybe_optimize_range_tests, reassociate_bb): Adjust
3409 optimize_range_tests callers.
3411 2018-04-26 Richard Biener <rguenther@suse.de>
3412 Jakub Jelinek <jakub@redhat.com>
3414 * cgraph.h (symbol_table): Just declare debug method here.
3415 * symtab.c (symbol_table::debug): Define.
3417 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3419 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3421 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3423 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3424 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3425 (*movdi_internal): Substitute Yi and Yj constraint with x
3426 and Ym and Yn constraint with y constraint. Update "isa"
3427 attribute and set "preferred_for_speed" attribute from
3428 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3429 (*movsi_internal): Ditto.
3430 (*movdf_internal): Ditto.
3431 (*movsf_internal): Ditto.
3432 (*zero_extendsidi2): Ditto.
3433 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3434 (sse2_loadld): Ditto.
3435 (*vec_extract<ssevecmodelower>_0): Ditto.
3436 (*vec_extractv4si_0_zext_sse4): Ditto.
3437 (vec_concatv2di): Ditto.
3438 (*vec_dup<mode>): Ditto.
3439 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3440 * config/i386/constraints.md (Yi): Remove.
3445 2018-04-26 Nathan Sidwell <nathan@acm.org>
3447 * dumpfile.c (dump_open): New.
3448 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3449 (dump_finish): Detect stdio/stderr by value not name.
3451 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3453 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3455 2018-04-26 Tom de Vries <tom@codesourcery.com>
3458 * config/nvptx/nvptx.c (verify_neutering_jumps)
3459 (verify_neutering_labels): New function
3460 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3462 2018-04-26 Tom de Vries <tom@codesourcery.com>
3465 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3466 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3469 2018-04-26 Richard Biener <rguenther@suse.de>
3470 Tom de Vries <tom@codesourcery.com>
3473 * lto-streamer-out.c (output_function): Fixup loops if required to match
3474 discovery done in the reader.
3476 2018-04-26 Richard Biener <rguenther@suse.de>
3478 PR tree-optimization/85116
3479 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3480 have a loop exit from the single latch predecessor. Remove
3481 case of header with just condition.
3482 (ch_base::copy_headers): Exclude infinite loops from any
3484 (pass_ch::execute): Record exits.
3486 2018-04-26 Richard Biener <rguenther@suse.de>
3488 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3489 prologue cost vector and pass it to vect_get_load_cost.
3490 (vect_get_peeling_costs_all_drs): Likewise.
3491 (vect_peeling_hash_get_lowest_cost): Likewise.
3492 (vect_enhance_data_refs_alignment): Likewise.
3494 2018-04-26 Richard Biener <rguenther@suse.de>
3497 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3498 checking of integer<->pointer conversions.
3499 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3500 sign-/zero-extending pointer types.
3501 (expand_omp_for_static_chunk): Likewise.
3503 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3504 Jean Lee <xiaoyur347@gmail.com>
3506 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3507 (TARGET_ASAN_SHADOW_OFFSET): Define.
3508 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3509 true for -fsanitize=address.
3511 2018-04-25 Mark Wielaard <mark@klomp.org>
3513 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3516 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3518 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3519 than "alu", remove explicit "memory" and "imm_disp" attributes.
3520 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3523 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3524 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3525 gen_lowpart_no_emit.
3527 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3530 * config/i386/i386.c (ix86_expand_builtin): Change memory
3531 operand to XI, extend p0 to Pmode.
3532 * config/i386/i386.md: Change unspec volatile and operand
3533 1 mode to XI, change operand 0 mode to P.
3535 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3537 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3538 GET_MODE_MASK before any checking.
3539 (nds32_can_use_bset_p): Likewise.
3540 (nds32_can_use_btgl_p): Likewise.
3542 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3544 * config/nds32/nds32-doubleword.md: New define_split pattern for
3545 illegal register number.
3547 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3549 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3551 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3553 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3555 2018-04-25 Richard Biener <rguenther@suse.de>
3557 * lto-streamer.h (LTO_major_version): Bump to 8.
3559 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3561 * BASE-VER: Set to 9.0.0.
3563 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3565 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3566 in __abskf2 and __powikf2.
3568 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3571 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3572 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3573 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3574 Use the above on operand 2. Reindent.
3575 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3577 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3580 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3582 * config/i386/i386.opt (mcet): Removed.
3583 * doc/install.texi: Remove -mcet documentation.
3584 * doc/invoke.texi: Likewise.
3586 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3589 * doc/install.texi: Remove -mcet from bootstrap-cet.
3591 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3594 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3595 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3599 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3600 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3601 containing a CONST_VECTOR.
3603 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3605 * doc/install.texi: Update newlib dependency for nvptx.
3607 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3610 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3611 instead of INTVAL when shifting x left.
3613 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3615 PR tree-optimization/85478
3616 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3617 vect_grouped_store_supported for single element vectors.
3619 2018-04-24 Richard Biener <rguenther@suse.de>
3622 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3623 load cost increase to the case of non-constant step.
3625 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3628 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3629 destination if any_malformed_asm.
3631 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3634 * expr.c (store_field): In the bitfield case, if the value comes from
3635 a function call and is returned in registers by means of a PARALLEL,
3636 do not change the mode of the temporary unless BLKmode and VOIDmode.
3638 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3640 PR rtl-optimization/85423
3641 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3642 dependencies to debug insns when the previous insn is non-debug.
3644 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3646 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3647 enums into a single definition.
3648 (fls): Fix predicates and printing.
3651 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3653 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3654 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3655 and short u6 immediate.
3656 (check_if_valid_sleep_operand): Remove.
3657 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3659 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3661 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3662 flag_always_save_lp condition.
3663 * config/nds32/nds32.opt (malways-save-lp): New option.
3665 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3667 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3668 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3669 * config/nds32/nds32.h
3670 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3671 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3673 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3675 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3676 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3678 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3679 Chung-Ju Wu <jasonwucj@gmail.com>
3681 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3683 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3684 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3686 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3688 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3690 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32-protos.h (nds32_data_alignment,
3693 nds32_local_alignment): Declare.
3694 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3695 nds32_local_alignment): New functions.
3696 (TARGET_CONSTANT_ALIGNMENT): Define.
3697 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3699 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3701 * config/nds32/nds32.c
3702 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3703 (TARGET_MODES_TIEABLE_P): Likewise.
3705 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3707 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3710 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3711 Chung-Ju Wu <jasonwucj@gmail.com>
3713 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3714 for unaligned access.
3715 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3717 * config/nds32/nds32-intrinsic.md: Likewise.
3718 * config/nds32/nds32_intrinsic.h: Likewise.
3719 * config/nds32/nds32.h (nds32_builtins): Likewise.
3720 * config/nds32/nds32.opt (munaligned-access): New option.
3721 * config/nds32/nds32.c (nds32_asm_file_start): Display
3722 flag_unaligned_access status.
3724 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3726 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3727 -mno-relax is present.
3728 * config/riscv/linux.h (LINK_SPEC): Ditto.
3730 2018-04-20 Martin Sebor <msebor@redhat.com>
3733 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3735 (gimple_fold_builtin_stxcpy_chk): Same.
3736 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3738 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3741 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3742 __powikf2 when long double is IEEE 128-bit.
3744 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3746 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3747 step to make sure stack always aligned.
3749 2018-04-20 Carl Love <cel@us.ibm.com>
3752 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3753 size check for arg0.
3755 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3756 Tom de Vries <tom@codesourcery.com>
3759 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3760 Emit insns for calls too.
3761 (nvptx_find_par): Always look for worker-level predecessor insn.
3762 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3764 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3765 (nvptx_process_pars): Propagate frames for calls.
3767 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3770 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3772 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3773 (ix86_handle_option): Don't handle OPT_mibt.
3774 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3776 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3778 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3779 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3780 (ix86_target_macros): Define __CET__ with flag_cf_protection
3781 for -fcf-protection.
3782 * config/i386/i386.c (isa2_opts): Remove -mibt.
3783 * config/i386/i386.h (TARGET_IBT): Removed.
3784 (TARGET_IBT_P): Likewise.
3785 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3786 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3787 * config/i386/i386.opt (mcet): Update help message.
3790 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3791 -mcet as an alias for -mshstk.
3793 2018-04-20 Richard Biener <rguenther@suse.de>
3796 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3797 complexity by forcing a single use of the multiply operand.
3799 2018-04-20 Martin Jambor <mjambor@suse.cz>
3802 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3803 recursion dependency to only apply to non-clones.
3805 2018-04-20 Martin Jambor <mjambor@suse.cz>
3808 * ipa-cp.c (create_specialized_node): Check that clones of
3809 self-recursive edges exist during IPA-CP.
3811 2018-04-19 Toon Moene <toon@moene.org>
3813 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3816 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3818 PR tree-optimization/85467
3819 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3820 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3821 VECTOR_CST element to type.
3823 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3826 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3827 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3828 (builtin_longjmp): Likewise.
3829 (save_stack_nonlocal): New pattern.
3830 (restore_stack_nonlocal): Likewise.
3832 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3835 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3836 Replace ASM_OUTPUT_LABEL with fprintf.
3838 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3841 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3842 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3843 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3844 define __IBT__ and __SHSTK__ for -fcf-protection.
3845 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3847 (ix86_trampoline_init): Likewise.
3848 (x86_output_mi_thunk): Likewise.
3849 (ix86_notrack_prefixed_insn_p): Likewise.
3850 (ix86_option_override_internal): Don't disallow -fcf-protection.
3851 * config/i386/i386.md (rdssp<mode>): Also enable for
3853 (incssp<mode>): Likewise.
3854 (nop_endbr): Likewise.
3855 * config/i386/i386.opt (mcet): Change help message to built-in
3859 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3860 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3861 enable CET built-in functions.
3863 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3865 * common/config/i386/i386-common.c
3866 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3867 OPTION_MASK_ISA_MOVDIRI_UNSET,
3868 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3869 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3870 * config.gcc (movdirintrin.h): New header.
3871 * config/i386/cpuid.h (bit_MOVDIRI,
3872 bit_MOVDIR64B): New bits.
3873 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3875 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3876 (VOID, PVOID, PCVOID)): New function types.
3877 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3878 __builtin_ia32_directstoreu_u64,
3879 __builtin_ia32_movdir64b): New builtins.
3880 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3881 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3883 (ix86_valid_target_attribute_inner_p): Ditto.
3884 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3885 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3886 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3887 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3888 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3889 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3890 (movdiri<mode>, movdir64b_<mode>): New.
3891 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3892 * config/i386/immintrin.h: Include movdirintrin.h.
3893 * config/i386/movdirintrin.h: New file.
3894 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3896 2018-04-19 Richard Biener <rguenther@suse.de>
3899 * cfg.c (clear_bb_flags): When loop state says we have
3900 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3902 2018-04-19 Richard Biener <rguenther@suse.de>
3904 PR tree-optimization/84737
3905 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3906 copying restrict info.
3907 (vect_setup_realignment): Use it.
3908 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3909 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3910 the first DR to all generated stores.
3911 (vectorizable_load): Likewise for loads.
3913 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3915 PR tree-optimization/85446
3916 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3917 the integral and pointer types to have the same precision.
3919 * doc/install.texi: Document --disable-cet being the default and
3922 2018-04-18 Martin Liska <mliska@suse.cz>
3924 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3927 2018-04-18 Martin Liska <mliska@suse.cz>
3930 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3933 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3934 arguments if they are comparable.
3936 2018-04-18 Martin Liska <mliska@suse.cz>
3939 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3942 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3945 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3948 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3949 ENDBR after calling __morestack.
3951 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3954 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3955 by using gcc_base_ver to generate a gcc_driver_version, and use
3956 it when generating GCC_DRIVER_NAME.
3957 * configure: Regenerate.
3959 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3962 * config.gcc: Obsolete powerpc*-*-*spe*.
3964 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3967 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3968 (stabstr_D): Change type of unum from unsigned int to
3969 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3972 2018-04-17 Jim Wilson <jimw@sifive.com>
3975 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3976 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3977 Set arg_pointer_offset after using pretend_args_size.
3979 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3981 PR rtl-optimization/85431
3982 * dse.c (record_store): Ignore zero width stores.
3985 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3986 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3987 __builtin_stack_restore rather than after it.
3988 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3989 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3990 argument instead of virtual_dynamic_stack_rtx.
3992 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3994 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3996 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3997 Add note to error message to explain internal mapping of overloaded
3998 built-in function name to non-overloaded built-in function name.
3999 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4002 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4005 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4006 where the inputs overlap with the output.
4008 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4010 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4011 (=v, v) alternative and explicit "memory" attribute.
4012 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4013 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4015 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4016 "sselog1" type instead of "sselog".
4017 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4018 "sselog". Remove explicit "memory" attribute.
4019 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4020 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4022 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4023 "sselog1" type instead of "sselog". Remove explicit "memory"
4025 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4026 use "sselog1" type instead of "sselog". Remove explicit "memory"
4028 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4029 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4031 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4032 "sselog1" type instead of "sselog". Remove explicit "memory"
4034 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4035 use "sselog1" type instead of "sselog". Remove explicit "memory"
4039 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4042 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4045 2018-04-17 Martin Jambor <mjambor@suse.cz>
4048 * ipa-cp.c (create_specialized_node): Call
4049 expand_all_artificial_thunks if necessary.
4051 2018-04-17 Martin Liska <mliska@suse.cz>
4054 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4055 in message, remote space in between '_G' and '('.
4057 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4060 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4061 avx512f_vmcmp<mode>3<round_saeonly_name>,
4062 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4063 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4064 avx512f_rndscale<mode><round_saeonly_name>,
4065 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4066 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4067 Use %<iptr>2 instead of %2 for -masm=intel.
4068 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4069 avx512f_vcvttss2usi<round_saeonly_name>,
4070 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4072 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4073 avx512f_vcvttsd2usi<round_saeonly_name>,
4074 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4075 Use %q1 instead of %1 for -masm=intel.
4076 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4077 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4078 of %3 for -masm=intel.
4079 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4081 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4083 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4085 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4086 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4088 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4090 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4091 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4092 %g1 and one with %0 and %1.
4093 (avx512er_vmrcp28<mode><round_saeonly_name>,
4094 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4096 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4097 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4098 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4099 of %0 and %{%4%} for -masm=intel.
4100 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4101 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4102 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4103 order of %0 and %{%5%}%{z%} for -masm=intel.
4105 2018-04-17 Jan Hubicka <jh@suse.cz>
4108 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4110 2018-04-17 Martin Liska <mliska@suse.cz>
4113 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4114 for target_clone error message. Make default implementation
4115 clone to be a local declaration.
4116 (separate_attrs): Add new argument and check for an empty
4118 (expand_target_clones): Handle it.
4119 (ipa_target_clone): Make redirection just for target_clones
4122 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4123 Tom de Vries <tom@codesourcery.com>
4126 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4127 tiled basic blocks without omp continue statements.
4129 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4132 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4133 vec_extract expression as having side effects to make sure it gets
4136 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4139 * config/i386/i386.c (get_builtin_code_for_version): Check
4142 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4145 * gcc/config.gcc: Support "skylake".
4146 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4148 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4149 (processor_target_table): Add "skylake".
4150 (ix86_option_override_internal): Add "skylake".
4151 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4152 PROCESSOR_CANNONLAKE.
4153 (get_builtin_code_for_version): Fix priority for
4154 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4155 PROCESSOR_SKYLAKE-AVX512.
4156 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4157 (processor_type): Add PROCESSOR_SKYLAKE.
4159 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4160 Jason Merrill <jason@redhat.com>
4163 * convert.c (convert_to_integer_1): Use direct recursion for
4164 enumeral types and types with a precision less than the number
4165 of bits in their mode.
4167 2018-04-16 Julia Koval <julia.koval@intel.com>
4170 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4171 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4173 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4176 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4177 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4178 and -mno-direct-move.
4180 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4183 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4184 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4185 Ensure negative shifts result in {0}.
4187 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4189 PR rtl-optimization/79916
4190 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4191 regs (if any) to define how to gnerate SD moves when LRA is in
4194 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4196 PR rtl-optimization/85393
4197 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4198 * except.c (expand_dw2_landing_pad_for_region): Make static.
4199 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4200 a label and unconditional jump to old_bb, rather than
4201 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4204 PR rtl-optimization/85376
4205 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4206 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4207 instead of a specific value.
4209 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4210 Bin Cheng <bin.cheng@arm.com>
4212 PR tree-optimization/82965
4213 PR tree-optimization/83991
4214 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4215 by_profile_only parameter.
4216 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4217 information if the loop was predicted to iterate too many times.
4218 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4220 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4223 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4226 2018-04-13 Martin Liska <mliska@suse.cz>
4227 Jakub Jelinek <jakub@redhat.com>
4230 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4231 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4232 * builtins.c (expand_builtin_memory_copy_args): Use
4233 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4234 handle dest_addr == pc_rtx.
4236 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4239 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4240 asked to not generate direct moves.
4241 (fix_trunc<mode>si2_stfiwx): Similar.
4242 (fix_trunc<mode>si2_internal): Similar.
4244 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4247 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4248 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4249 lookup if dest in some wider mode is known to be const0_rtx and
4250 if so, record permanent equivalence for it to be ZERO_EXTEND of
4251 the narrower mode destination.
4253 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4255 * lto-streamer-out.c (output_function): Revert 259346.
4256 * omp-expand.c (expand_oacc_for): Likewise.
4258 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4260 PR rtl-optimization/85354
4261 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4262 * sel-sched.c (sel_global_init): ... here.
4264 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4267 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4268 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4269 mode for PE-COFF targets.
4270 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4271 (i386_pe_asm_lto_end): Likewise.
4272 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4273 (TARGET_ASM_LTO_END): Likewise.
4274 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4275 (i386_pe_asm_lto_start): New function.
4276 (i386_pe_asm_lto_end): Likewise.
4278 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4279 Richard Biener <rguenther@suse.de>
4282 * lto-streamer-out.c (output_function): Fix CFG loop state before
4284 * omp-expand.c (expand_oacc_for): Handle calls to internal
4285 functions like regular functions.
4287 2018-04-12 Richard Biener <rguenther@suse.de>
4290 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4291 for the early LTO debug to properly generate references to it
4292 during DIE emission. Do not re-use that for the skeleton for
4294 (dwarf2out_early_finish): Likewise.
4296 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4299 * config/i386/sse.md
4300 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4301 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4302 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4303 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4304 and output is a reg, avoid creating invalid lowpart subreg, but
4305 instead split into a 512-bit move. Don't split if not AVX512VL,
4306 input is xmm16+ reg and output is a mem.
4307 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4308 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4309 xmm16+ reg and output is a mem.
4311 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4313 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4314 also for flag_dwarf2_cfi_asm.
4316 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4318 PR rtl-optimization/85342
4319 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4320 a bool scalar var inside of the loop instead. Don't try to update
4321 recog_data.operand after failed apply_change_group.
4323 2018-04-12 Tom de Vries <tom@codesourcery.com>
4326 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4327 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4328 array with flexible array member as array without given dimension.
4329 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4330 argument for undefined param to true.
4332 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4335 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4336 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4337 from PowerPC section.
4338 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4339 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4340 help text that is too long.
4341 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4342 help text that is too long.
4343 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4344 help text that is too long.
4346 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4348 * config/alpha/alpha.md (stack_probe_internal): Rename
4349 from "probe_stack". Update all callers.
4351 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4353 PR rtl-optimization/84566
4354 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4355 sched_macro_fuse_insns.
4357 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4360 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4361 (compute_block_dependences): ... from here.
4363 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4365 PR tree-optimization/85331
4366 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4367 from int to HOST_WIDE_INT.
4369 2018-04-11 Martin Jambor <mjambor@suse.cz>
4372 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4373 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4374 not the same as the source val.
4375 (cgraph_edge_brings_value_p): New parameter.
4376 (gather_edges_for_value): Pass destination value to
4377 cgraph_edge_brings_value_p.
4378 (perhaps_add_new_callers): Likewise.
4379 (get_info_about_necessary_edges): Likewise and exclude values brought
4380 only by self-recursive edges.
4381 (create_specialized_node): Redirect only clones of self-calling edges.
4382 (+self_recursive_pass_through_p): New function.
4383 (find_more_scalar_values_for_callers_subset): Use it.
4384 (find_aggregate_values_for_callers_subset): Likewise.
4385 (known_aggs_to_agg_replacement_list): Removed.
4386 (decide_whether_version_node): Re-calculate known constants for all
4387 remaining context clones.
4389 2018-04-11 Richard Biener <rguenther@suse.de>
4392 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4393 from early DWARF output.
4394 (dwarf2out_early_finish): Output line info unconditionally into
4395 early DWARF and add reference to it.
4397 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4400 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4401 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4402 other than V2DFmode using iptr mode attribute.
4403 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4405 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4407 PR rtl-optimization/84659
4408 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4410 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4413 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4415 (output_loc_list): Pass address of a dummy size variable even in the
4416 locview handling loop.
4417 (index_location_lists): Add comment on why skip_loc_list_entry can't
4420 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4423 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4426 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4429 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4430 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4431 and -mstring-compare-inline-limit.
4433 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4436 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4437 for stack clash protection in a register whenever we need it to be in
4440 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4442 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4443 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4445 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4448 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4450 (mlong-double-): Ditto.
4451 * config/rs6000/sysv4.opt (msdata=): Ditto.
4452 (mtls-size=): Ditto.
4454 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4456 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4457 erroneous entries for
4458 "vector int vec_ldl (int, long int *)", and
4459 "vector unsigned int vec_ldl (int, unsigned long int *)".
4460 Add comments and entries for
4461 "vector bool char vec_ldl (int, bool char *)",
4462 "vector bool short vec_ldl (int, bool short *)",
4463 "vector bool int vec_ldl (int, bool int *)",
4464 "vector bool long long vec_ldl (int, bool long long *)",
4465 "vector pixel vec_ldl (int, pixel *)",
4466 "vector long long vec_ldl (int, long long *)",
4467 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4468 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4469 type tree bool_long_long_type_node and correct definition of
4470 bool_V2DI_type_node to make reference to this new type tree.
4471 (rs6000_mangle_type): Replace erroneous reference to
4472 bool_long_type_node with bool_long_long_type_node.
4473 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4474 comments to emphasize sign distinctions for char and int types and
4475 replace RS6000_BTI_bool_long constant with
4476 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4477 use of RS6000_BTI_pixel.
4478 (bool_long_type_node): Remove this macro definition.
4479 (bool_long_long_type_node): New macro definition
4481 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4483 PR rtl-optimization/85300
4484 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4485 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4486 simplify_unary_operation fails.
4488 2018-04-10 Martin Liska <mliska@suse.cz>
4490 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4491 cgraph_edge and ipa_ref.
4493 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4497 * config/i386/sse.md
4498 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4499 computation of the VEC_MERGE selector from mask.
4500 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4501 Fix decoding of the VEC_MERGE selector into mask.
4503 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4505 PR tree-optimization/85286
4506 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4508 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4510 * final.c (final_1): Set insn_last_address as well as
4511 insn_current_address.
4513 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4516 * explow.c (emit_stack_probe): Call validize_mem on memory location
4517 before passing it to gen_probe_stack. Create address operand and
4518 legitimize it for the probe_stack_address case.
4520 2018-04-09 Jan Hubicka <jh@suse.cz>
4523 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4524 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4525 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4527 (free_lang_data): Rebuild type inheritance graph.
4529 2018-04-09 Martin Sebor <msebor@redhat.com>
4531 * invoke.texi (-finline-small-functions): Mention other optimization
4533 (-findirect-inlining, -fpartial-inlining): Same.
4534 (-finline-functions-called-once): Same.
4535 (-freorder-blocks-and-partition): Same.
4537 2018-04-09 Jan Hubicka <jh@suse.cz>
4540 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4541 jumps; choose last target that matches the criteria (i.e.
4542 no partition changes for non-crossing jumps).
4543 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4544 support for redirecting crossing jumps to non-crossing.
4546 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4548 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4549 also for naked functions.
4551 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4553 * config/arc/arc.md (add_shift): New pattern.
4554 (add_shift2): Likewise.
4555 (sub_shift): Likewise.
4556 (sub_shift_cmp0_noout): Likewise.
4557 (compare_si_ashiftsi): Likewise.
4558 (xbfu_cmp0_noout): New combine pattern.
4559 (xbfu_cmp0"): Likewise.
4560 (movsi_set_cc_insn): Place the predicable variant first.
4561 (commutative_binary_cmp0_noout): Remove clobber.
4562 (commutative_binary_cmp0): New pattern.
4563 (noncommutative_binary_cmp0): Likewise.
4564 (noncommutative_binary_cmp0_noout): Likewise.
4565 (noncommutative_binary_comparison_result_used): Removed.
4566 (rsub_cmp0): New pattern.
4567 (rsub_cmp0_noout): Likewise.
4568 (extzvsi): Changed, keep only meaningful variants.
4569 (SQH, SEZ): New iterators.
4570 (SQH_postfix): New mode attribute.
4571 (SEZ_prefix): New code attribute.
4572 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4573 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4574 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4576 (noncommutative_operator): Check the availability of barrel
4579 2018-04-09 Richard Biener <rguenther@suse.de>
4581 PR tree-optimization/85284
4582 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4583 Only use the niter constraining form of simple_iv when the exit
4586 2018-04-09 Tom de Vries <tom@codesourcery.com>
4589 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4590 (define_expand "*memory_barrier"): New define_expand.
4591 (define_insn "memory_barrier"): New insn.
4593 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4595 PR rtl-optimization/80463
4596 PR rtl-optimization/83972
4597 PR rtl-optimization/83480
4599 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4600 correct producer for the insn.
4601 (tidy_control_flow): Fixup seqnos in case of debug insns.
4603 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4605 PR rtl-optimization/83913
4607 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4608 different sched-times when merging exprs.
4610 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4612 PR rtl-optimization/83962
4614 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4615 tidy_fallthru_edge and tidy_control_flow.
4617 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4619 PR rtl-optimization/83530
4621 * sel-sched.c (force_next_insn): New global variable.
4622 (remove_insn_for_debug): When force_next_insn is true, also leave only
4623 next insn in the ready list.
4624 (sel_sched_region): When the region wasn't scheduled, make another pass
4625 over it with force_next_insn set to 1.
4627 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4629 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4631 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4632 for interrupt control.
4633 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4634 functions for interrupt control.
4635 * config/nds32/nds32-intrinsic.md: Likewise.
4636 * config/nds32/nds32_intrinsic.h: Likewise.
4637 * config/nds32/nds32.h (nds32_builtins): Likewise.
4639 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4641 * config/nds32/nds32.c (nds32_init_machine_status,
4642 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4643 strict_aligned_p field.
4644 (nds32_expand_to_rtl_hook): New function.
4645 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4646 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4648 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4649 Chung-Ju Wu <jasonwucj@gmail.com>
4651 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4652 * config/nds32/nds32-n7.md: New file.
4653 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4654 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4656 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4657 * config/nds32/nds32.md (pipeline_model): Add n7.
4658 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4659 * config/nds32/pipelines.md: Include n7 settings.
4661 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4662 Chung-Ju Wu <jasonwucj@gmail.com>
4664 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4665 * config/nds32/nds32-e8.md: New file.
4666 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4667 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4669 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4670 * config/nds32/nds32.md (pipeline_model): Add e8.
4671 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4672 * config/nds32/pipelines.md: Include e8 settings.
4674 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4675 Chung-Ju Wu <jasonwucj@gmail.com>
4677 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4678 * config/nds32/nds32-n8.md: New file.
4679 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4680 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4682 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4683 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4684 * config/nds32/nds32.md (pipeline_model): Add n8.
4685 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4686 * config/nds32/pipelines.md: Include n8 settings.
4688 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4689 Chung-Ju Wu <jasonwucj@gmail.com>
4691 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4692 * config/nds32/nds32-n9-2r1w.md: New file.
4693 * config/nds32/nds32-n9-3r2w.md: New file.
4694 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4695 nds32_register_ports): New or modify for cpu n9.
4696 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4698 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4699 * config/nds32/nds32-utils.c: New file.
4700 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4701 TARGET_MUL_SLOW): Define.
4702 * config/nds32/nds32.md (pipeline_model): New attribute.
4703 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4704 New options that support cpu n9.
4705 * config/nds32/pipelines.md: Include n9 settings.
4706 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4708 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4711 information if necessary.
4712 (output_cond_branch_compare_zero): Likewise.
4713 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4714 (nds32_target_alignment): Refine for alignment.
4715 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4716 (FUNCTION_BOUNDARY): Modify.
4717 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4719 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4721 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4723 * config/nds32/constants.md (unspec_volatile_element): Add values for
4724 TLB operation and data prefetch.
4725 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4726 functions for TLB operation and data prefetch.
4727 * config/nds32/nds32-intrinsic.md: Likewise.
4728 * config/nds32/nds32_intrinsic.h: Likewise.
4729 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4730 (nds32_print_operand): Likewise.
4731 * config/nds32/nds32.h (nds32_builtins): Likewise.
4733 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4734 Andrew Pinski <pinsika@gcc.gnu.org>
4737 * match.pd: Use constant_boolean_node of correct type instead of
4738 boolean_true_node or boolean_false_node for simplifying
4739 pointer comparisons to zero.
4741 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4743 PR tree-optimization/80021
4744 * tree.c (verify_type_variant): Make error call in verify_variant_match
4745 translatable and remove final full stop.
4747 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4749 * config/nds32/constants.md (unspec_volatile_element): Add
4750 UNSPEC_VOLATILE_EH_RETURN.
4751 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4752 nds32_output_stack_pop): Support dwarf exception handling process.
4753 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4754 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4755 exception handling process.
4756 (nds32_compute_stack_frame): Likewise.
4757 (nds32_return_addr_rtx): Likewise.
4758 (nds32_initial_elimination_offset): Likewise.
4759 (nds32_expand_prologue): Likewise.
4760 (nds32_expand_epilogue): Likewise.
4761 (nds32_dynamic_chain_address): New function.
4762 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4764 (DYNAMIC_CHAIN_ADDRESS): Define.
4765 (EH_RETURN_DATA_REGNO): Define.
4766 (EH_RETURN_STACKADJ_RTX): Define.
4767 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4768 patterns for dwarf exception handling.
4770 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4772 * config/nds32/nds32.h: Clean up obsolete macros.
4774 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4776 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4777 Add enum values for particular instructions.
4778 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4779 particular intrinsic functions.
4780 * config/nds32/nds32-intrinsic.md: Likewise.
4781 * config/nds32/nds32_intrinsic.h: Likewise.
4782 * config/nds32/nds32.h (nds32_builtins): Likewise.
4783 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4784 (btst, ave): New patterns for particular instructions.
4786 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4788 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4789 Add enum values for atomic load/store and memory sync.
4790 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4792 * config/nds32/nds32-intrinsic.md: Likewise.
4793 * config/nds32/nds32_intrinsic.h: Likewise.
4794 * config/nds32/nds32.h (nds32_builtins): Likewise.
4796 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4798 PR tree-optimization/85257
4799 * fold-const.c (native_encode_vector): If not all elts could fit
4800 and off is -1, return 0 rather than offset.
4801 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4802 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4803 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4804 adjust buffer in native_interpret_expr call.
4806 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4808 * config/nds32/constants.md (unspec_volatile_element): Add cache
4809 control enum values.
4810 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4811 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4812 * config/nds32/nds32.c (nds32_cctl_names): New.
4813 (nds32_print_operand): Handle cache control register names.
4814 * config/nds32/nds32.h (nds32_builtins): New enum values.
4815 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4817 * config/nds32/nds32.md (type): Add mmu.
4818 * config/nds32/pipelines.md (simple_insn): Add mmu.
4820 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4822 * config/nds32/nds32.md (type): Remove call.
4823 * config/nds32/pipelines.md (simple_insn): Likewise.
4825 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4827 * config/nds32/constants.md (unspec_volatile_element): Add
4828 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4829 UNSPEC_VOLATILE_FMFCFG.
4830 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4831 description for fmfcfg and fmfcsr.
4832 (bdesc_1arg): Add fmtcsr.
4833 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4834 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4835 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4836 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4837 unspec_fmfcfg): New patterns.
4838 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4839 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4840 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4841 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4842 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4843 __nds32__fmfcfg): Define.
4845 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4847 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4848 intrinsic register names.
4849 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4850 intrinsic register enum values and macros.
4852 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4854 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4855 for load/store addressing form.
4856 (nds32_print_operand_address): Likewise.
4858 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4861 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4862 based on LABEL_REF. Remove useless assertion.
4863 (pic_address_needs_scratch): Fix formatting.
4864 (sparc_legitimize_pic_address): Minor tweaks.
4865 (sparc_delegitimize_address): Adjust assertion accordingly.
4866 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4867 into symbolic_operand.
4868 (movsi_high_pic_label_ref): Likewise.
4869 (movsi_lo_sum_pic_label_ref): Likewise.
4870 (movdi_pic_label_ref): Likewise.
4871 (movdi_high_pic_label_ref): Likewise.
4872 (movdi_lo_sum_pic_label_ref): Likewise.
4874 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4876 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4877 custom LIB_SPEC setup.
4879 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4880 Kito Cheng <kito.cheng@gmail.com>
4882 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4883 * config/riscv/freebsd.h: New.
4885 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4887 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4888 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4891 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4892 Kito Cheng <kito.cheng@gmail.com>
4894 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4895 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4896 * config/nds32/nds32-protos.h (nds32_output_call,
4897 nds32_output_return): Declare.
4898 * config/nds32/nds32.md: Refine all the call and return patterns.
4900 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4903 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4904 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4906 PR rtl-optimization/84872
4907 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4908 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4911 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4913 * expr.c (copy_blkmode_to_reg): Revert 254862.
4914 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4916 2018-04-06 Richard Biener <rguenther@suse.de>
4919 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4920 after seeing a component reference with an adjacent field. Treat
4921 refs to arrays at struct end of external decls similar to
4922 refs to unconstrained commons.
4924 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4927 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4928 look through SAVE_EXPRs with non-side-effects argument. Adjust
4930 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4933 2018-04-06 Richard Biener <rguenther@suse.de>
4936 * alias.c (find_base_term): New wrapper around find_base_term
4937 unwinding CSELIB_VAL_PTR changes.
4938 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4941 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4943 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4945 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4946 constant definitions.
4947 ("nop"): lr 0,0 -> nopr r0
4948 ("nop_lr0", "nop_lr1"): New insn definitions.
4950 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4952 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4953 NDS32_V3PUSH_AVAILABLE_P macro.
4955 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4956 Chung-Ju Wu <jasonwucj@gmail.com>
4958 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4959 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4960 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4961 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4962 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4963 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4964 * config/nds32/constraints.md: New constraints and checking for hard
4965 float configuration.
4966 * config/nds32/iterators.md: New mode iterator and attribute for hard
4967 float configuration.
4968 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4970 * config/nds32/nds32-fpu.md: New file.
4971 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4972 deal with hard float code generation.
4973 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4975 (abi_type, float_reg_number): New enum type.
4976 * config/nds32/nds32-predicates.c: New predicates for hard float.
4977 * config/nds32/nds32-protos.h: Declare functions for hard float.
4978 * config/nds32/nds32.c: Implementation for hard float configuration.
4979 * config/nds32/nds32.h: Definitions for hard float configuration.
4980 * config/nds32/nds32.md: Include hard float machine description and
4981 modify patterns for hard float configuration.
4982 * config/nds32/nds32.opt: New options for hard float configuration.
4983 * config/nds32/predicates.md: New predicates for hard float
4986 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4988 * common/config/nds32/nds32-common.c
4989 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4991 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4994 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4995 CONSTRUCTOR_ELT (ctor, ...)->value.
4997 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5000 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5002 2018-04-05 Tom de Vries <tom@codesourcery.com>
5005 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5008 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5009 Kito Cheng <kito.cheng@gmail.com>
5011 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5012 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5013 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5014 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5016 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5017 Kito Cheng <kito.cheng@gmail.com>
5019 * config/nds32/constraints.md (Ufe): New memory constraint.
5020 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5021 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5022 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5024 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5025 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5027 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5029 * config/nds32/nds32.md: Use optimize_size in the condition for
5030 alu-shift instructions.
5032 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5034 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5036 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5038 * config/nds32/nds32.md (negsi2): Refine pattern.
5040 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5041 Chung-Ju Wu <jasonwucj@gmail.com>
5043 * config/nds32/iterators.md (shift_rotate): New code iterator.
5044 (shift): New code attribute.
5045 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5046 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5047 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5048 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5049 bit-wise operations.
5050 (andsi3, *andsi3): Ditto.
5051 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5052 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5053 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5054 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5055 nds32_ior_operand, nds32_xor_operand): New predicates.
5057 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5059 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5060 (addsi3, subsi3): ... this.
5062 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5064 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5066 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5068 * config/nds32/nds32.md: Adjust indention.
5070 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5072 * config/nds32/nds32.md (feature): New attribute.
5074 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5076 * config/nds32/nds32.md (subtype): New attribute.
5078 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5081 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5082 expansion to perform a bitwise AND of the argument followed by a
5083 boolean negation of the result.
5085 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5087 PR rtl-optimization/84878
5088 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5089 the basic block. Assert the use reference is not artificial and that
5090 it has an associated insn.
5092 2018-04-04 Michael Matz <matz@suse.de>
5094 * builtins.c (compute_objsize): Pass correct operand
5095 to array_at_struct_end_p.
5097 2018-04-04 Richard Biener <rguenther@suse.de>
5100 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5101 from contexts for DINFO_LEVEL_TERSE and below.
5103 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5105 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5106 resiter_operand condition.
5107 * config/nds32/nds32.md (*move<mode>): Ditto.
5109 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5110 Monk Chiang <sh.chiang04@gmail.com>
5112 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5114 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5116 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5118 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5119 Kito Cheng <kito.cheng@gmail.com>
5121 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5122 nds32_cond_code_str, output_cond_branch,
5123 output_cond_branch_compare_zero, 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: New functions.
5129 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5130 nds32_expand_cstore, nds32_expand_movcc,
5131 nds32_output_cbranchsi4_equality_zero,
5132 nds32_output_cbranchsi4_equality_reg,
5133 nds32_output_cbranchsi4_equality_reg_or_const_int,
5134 nds32_output_cbranchsi4_greater_less_zero): Declare.
5135 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5136 nds32_rimm11s_operand): New predicates.
5137 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5138 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5141 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5143 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5144 * config/nds32/nds32.md: Ditto.
5145 * config/nds32/pipelines.md: Ditto.
5147 2018-04-04 Richard Biener <rguenther@suse.de>
5149 PR tree-optimization/85168
5150 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5151 propagating abnormals.
5153 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5155 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5157 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5158 Kito Cheng <kito.cheng@gmail.com>
5160 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5161 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5162 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5163 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5164 * config/nds32/nds32.md (sibcall_internal): New.
5165 (sibcall_register): Remove.
5166 (sibcall_immediate): Remove.
5167 (sibcall_value_internal): New.
5168 (sibcall_value_register): Remove.
5169 (sibcall_value_immediate): Remove.
5170 * config/nds32/predicates.md (nds32_general_register_operand): New.
5171 (nds32_call_address_operand): New.
5173 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5175 PR rtl-optimization/85167
5176 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5177 bb_defs if *split_p, instead preinitialize it to NULL.
5179 PR tree-optimization/85156
5180 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5181 evaluating the argument multiple times.
5183 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5185 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5187 (_mm_cvtpd_ps): Likewise.
5188 (_mm_cvttpd_epi32): Likewise.
5189 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5190 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5191 vector, pixel, and bool following altivec.h include.
5193 2018-04-03 Martin Sebor <msebor@redhat.com>
5195 * doc/extend.texi (Common Function Attributes): Clarify.
5196 (const attribute): Likewise.
5197 (pure attribute): Likewise.
5199 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5202 * config/i386/i386.c (ix86_expand_vector_set): Use
5203 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5205 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5207 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5208 instructions when changing rounding bits to preserve precision bits
5209 in the x87 control word.
5211 2018-04-03 Martin Liska <mliska@suse.cz>
5213 PR tree-optimization/82491
5214 * rtl.h (strip_offset_and_add): Replace += suboffset with
5215 poly_uint64 () + suboffset.
5217 2018-03-29 Martin Liska <mliska@suse.cz>
5218 Martin Jambor <mjambor@suse.cz>
5221 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5222 param_type is not an integral or pointer type.
5224 2018-04-03 Richard Biener <rguenther@suse.de>
5226 * sese.h (recompute_all_dominators): Remove.
5228 2018-04-02 Martin Sebor <msebor@redhat.com>
5230 * doc/invoke.texi (-Wrestrict): Fix typos.
5232 2018-04-02 Jim Wilson <jimw@sifive.com>
5234 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5235 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5236 (<optab>di3, <optab>si3_extend): Likewise.
5237 (<optab>si3_mask, <optab>si3_mask_1): New.
5238 (<optab>di3_mask, <optab>di3_mask_1): New.
5239 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5240 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5241 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5243 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5245 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5248 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5250 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5251 (nds32_canonicalize_comparison): New function.
5253 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5254 Kito Cheng <kito.cheng@gmail.com>
5255 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5257 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5258 * config/nds32/constants.md (unspec_volatile_element): Add
5259 UNSPEC_VOLATILE_RELAX_GROUP.
5260 * config/nds32/nds32-relax-opt.c: New file.
5261 * config/nds32/nds32-predicates.c
5262 (nds32_symbol_load_store_p): New function.
5263 * config/nds32/nds32-protos.h
5264 (nds32_symbol_load_store_p): Declare function.
5265 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5266 * config/nds32/nds32.c
5267 (nds32_register_pass): New function to register pass.
5268 (nds32_register_passes): New function to register passes.
5269 * config/nds32/nds32.md (relax_group): New pattern.
5270 * config/nds32/nds32.opt (mrelax-hint): New option.
5271 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5273 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5275 * config/nds32/t-nds32: Modify files dependency.
5277 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5279 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5280 (PROFILE_HOOK): Define its implementation.
5282 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5284 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5285 type and 32-bit size.
5287 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5290 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5291 (V_128_256): New mode iterator.
5292 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5293 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5294 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5296 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5297 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5299 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5302 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5303 NaN inputs correctly.
5305 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5308 * config/rs6000/vsx.md (??r): New mode attribute.
5309 (*vsx_mov<mode>_64bit): Use it.
5310 (*vsx_mov<mode>_32bit): Likewise.
5312 2018-03-30 Martin Sebor <msebor@redhat.com>
5314 PR tree-optimization/84818
5315 * builtins.c (check_access): Use warning_n.
5317 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5320 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5322 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5325 2018-03-30 Julia Koval <julia.koval@intel.com>
5328 * x86-tune.def (movx, partial_reg_dependency): Enable for
5331 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5334 * lra-constraints.c (process_alt_operands): Move setting
5335 this_alternative_matches below.
5337 2018-03-29 Martin Liska <mliska@suse.cz>
5340 * doc/invoke.texi: Document how LTO works with debug info.
5341 Describe auto-load support of binutils. Mention 'x86-64'
5342 as valid option value of -march option.
5344 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5346 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5349 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5350 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5351 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5354 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5357 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5358 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5359 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5360 for RS6000_BTM_POWERPC64.
5361 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5362 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5363 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5368 2018-03-28 Carl Love <cel@us.ibm.com>
5371 2017-09-27 Carl Love <cel@us.ibm.com>
5373 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5374 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5375 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5378 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5380 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5381 instead of __vector bool.
5382 (_mm_max_pu8): Likewise.
5383 (_mm_min_pi16): Likewise.
5385 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5388 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5389 (DIVWEUO): Likewise.
5391 (DIVDEUO): Likewise.
5392 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5393 DIVWEUO and DIVDEUO.
5394 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5395 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5396 (div_extend): Likewise.
5397 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5399 (__builtin_divweuo): Likewise.
5400 (__builtin_divdeo): Likewise.
5401 (__builtin_divdeuo): Likewise.
5403 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5406 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5407 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5409 PR tree-optimization/82004
5410 * gimple-match-head.c (optimize_pow_to_exp): New function.
5411 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5412 Don't fold to exp if optimize_pow_to_exp is false.
5414 2018-03-28 Martin Liska <mliska@suse.cz>
5417 * calls.c (initialize_argument_information): Fix trailing space.
5418 * common.opt: Fix typo and provide better explanation for
5419 -fsanitize-coverage option.
5420 * config/i386/i386.opt: Fix typo.
5422 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5423 Martin Liska <mliska@suse.cz>
5426 * gimplify.c (asan_poison_variable): Don't do the check for
5427 gimplify_omp_ctxp here.
5428 (gimplify_decl_expr): Do it here.
5429 (gimplify_target_expr): Likewise.
5431 2018-03-28 Martin Liska <mliska@suse.cz>
5434 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5435 chkp_type_bounds_count if MPX is not enabled.
5437 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5439 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5441 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5444 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5445 function to create the function decl for complex long double
5446 multiply and divide for -mabi=ieeelongdouble.
5447 (init_float128_ieee): Call it.
5449 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5452 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5453 -fcf-protection=branch -mibt.
5454 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5456 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5459 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5461 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5464 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5465 extern array declarations.
5467 2018-03-27 Richard Biener <rguenther@suse.de>
5470 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5471 explicit single_use checks.
5473 2018-03-27 Richard Biener <rguenther@suse.de>
5475 PR tree-optimization/85082
5476 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5479 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5481 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5482 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5483 Turn on fasynchronous-unwind-tables and funwind-tables.
5485 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5488 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5489 (*bmi_blsr_<mode>_ccz): Ditto.
5491 2018-03-26 Tom de Vries <tom@codesourcery.com>
5493 PR tree-optimization/85063
5494 * omp-general.c (offloading_function_p): New function. Factor out
5496 * omp-offload.c (pass_omp_target_link::gate): ... here.
5497 * omp-general.h (offloading_function_p): Declare.
5498 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5499 with attribute omp declare target for offloading functions.
5501 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5503 PR tree-optimization/84005
5504 * tree-data-ref.h (get_base_for_alignment): Declare.
5505 * tree-data-ref.c (get_base_for_alignment_1): New function.
5506 (get_base_for_alignment): Likewise.
5507 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5508 get_base_for_alignment to find a suitable base object, instead
5509 of always using drb->base_address.
5511 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5514 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5515 known size by default.
5517 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5520 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5521 and non BLKmode operands.
5523 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5526 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5527 Clean up attributes.
5529 2018-03-23 Richard Biener <rguenther@suse.de>
5532 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5533 we are going to emit early debug for LTO.
5535 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5538 * function.c (match_asm_constraints_1): Don't optimize if input
5539 doesn't satisfy general_operand predicate for output's mode.
5542 * alias.c (write_dependence_p): Don't require for x_canonicalized
5543 non-VOIDmode if x has VOIDmode.
5546 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5547 just don't try to optimize it rather than assert it never happens.
5549 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5551 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5552 macro expansions for definition of ST_INTERNAL_<mode> and
5553 LD_INTERNAL_<mode> builtins.
5554 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5556 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5558 (altivec_expand_st_builtin): Likewise.
5559 (altivec_expand_builtin): Remove calls to deleted functions.
5560 (rs6000_address_for_altivec): Delete this function.
5561 * config/rs6000/vector.md: Remove expands for
5562 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5564 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5567 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5568 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5569 re-computing once computed.
5570 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5571 (arm_init_machine_status): Initialize
5572 machine->static_chain_stack_bytes.
5574 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5577 * doc/extend.texi: Add four new prototypes for vec_ld.
5578 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5579 definitions for more logical presentation.
5580 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5581 entries for V1TI variants of __builtin_altivec_ld builtin.
5582 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5583 handling of V1TI variant of LVX icode pattern.
5584 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5585 (rs6000_gimple_fold_builtin): Likewise.
5586 (altivec_init_builtins): Add code to define
5587 __builtin_altivec_lvx_v1ti function.
5589 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5592 * function.c (match_asm_constraints_1): Don't do the optimization
5593 if input isn't a REG, SUBREG, MEM or constant.
5595 2018-03-22 Tom de Vries <tom@codesourcery.com>
5597 PR tree-optimization/84956
5598 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5599 bb_has_abnormal_pred.
5601 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5604 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5605 DECL_INITIAL (decl) to decl at the end.
5606 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5609 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5611 * doc/extend.texi (__builtin_tgmath): Document when complex
5612 integer types are treated as _Complex _Float64.
5614 2018-03-21 Tom de Vries <tom@codesourcery.com>
5616 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5618 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5620 PR tree-optimization/84960
5621 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5622 if it is ENTRY block, move them into single succ of ENTRY in that case.
5624 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5626 PR tree-optimization/84811
5627 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5628 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5629 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5630 (known_subrange_p): Update accordingly. Cast each value involved
5631 in the size comparison, rather than casting the result of the
5634 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5636 PR tree-optimization/84982
5637 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5638 by flipping the least significant bit rather than all bits from
5639 bitpos to bitpos + bitsize - 1.
5641 2018-03-21 Nathan Sidwell <nathan@acm.org>
5643 * doc/extend.texi (Deprecated Features): Remove mention of
5644 long-deleted deprecations.
5646 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5649 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5650 * configure: Regenerate.
5652 2018-03-21 Tom de Vries <tom@codesourcery.com>
5654 PR tree-optimization/83126
5655 * tree-parloops.c (num_phis): New function.
5656 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5658 2018-03-21 Nathan Sidwell <nathan@acm.org>
5660 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5661 mention anon-struct/union members and trailing attributes.
5663 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5665 PR tree-optimization/84969
5666 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5667 builtin memset partitions if they set different rhs values.
5669 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5671 PR rtl-optimization/84989
5672 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5673 VEC_DUPLICATE with scalar result mode.
5675 2018-03-21 Martin Liska <mliska@suse.cz>
5678 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5679 not intended return statement.
5681 2018-03-21 Martin Liska <mliska@suse.cz>
5684 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5685 (chkp_find_bound_slots_1): Limit number of iterations.
5687 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5690 * Minor grammar fixes for x86 options.
5692 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5695 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5696 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5698 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5701 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5702 (altivec_lvx_<mode>_1op): Likewise.
5703 (altivec_stvx_<mode>_2op): Likewise.
5704 (altivec_stvx_<mode>_1op): Likewise.
5705 (altivec_lvx_<VM2:mode>): New define_expand.
5706 (altivec_stvx_<VM2:mode>): Likewise.
5707 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5708 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5709 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5710 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5711 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5712 (rs6000_gen_lvx): Likewise.
5713 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5714 (altivec_expand_stv_builtin): Likewise.
5715 (altivec_expand_builtin): Likewise.
5716 * config/rs6000/vector.md: Likewise.
5718 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5721 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5724 2018-03-20 Richard Biener <rguenther@suse.de>
5727 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5728 sign-conversions as zero, fall back to standard scalar_stmt
5731 2018-03-20 Martin Liska <mliska@suse.cz>
5734 * predict.c (rebuild_frequencies): Handle case when we have
5735 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5737 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5740 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5741 flag_section_anchors.
5742 * varasm.c (use_blocks_for_decl_p): Remove hack for
5743 dw2_force_const_mem.
5746 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5748 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5749 be created, use lowpart_subreg of operands[0] rather than operands[0]
5751 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5752 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5753 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5754 and n constraint instead of aarch64_shift_imm_di and Usd.
5755 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5756 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5758 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5761 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5762 to favor GPR over NEON registers.
5763 (<shift>di3_neon): Likewise.
5765 2018-03-20 Tom de Vries <tom@codesourcery.com>
5768 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5769 (nvptx_process_pars): Emit bar.sync asap and alap.
5771 2018-03-20 Tom de Vries <tom@codesourcery.com>
5774 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5775 seen_label if seen_label is already set.
5777 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5780 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5781 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5782 Use 1U instead of 1. Formatting fixes.
5785 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5786 instead of TREE_TYPE (s1) for the return value.
5788 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5790 PR tree-optimization/84946
5791 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5792 bitsize + bitsize in poly_uint64 rather than poly_int64.
5795 * dwarf2asm.c: Include fold-const.c.
5796 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5797 of decl rather than decl itself.
5799 PR rtl-optimization/84643
5800 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5802 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5805 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5806 calling assemble_variable.
5808 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5811 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5812 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5814 2018-03-19 Jim Wilson <jimw@sifive.com>
5817 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5818 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5819 (riscv_first_stack_step): Likewise.
5820 (riscv_option_override): Use STACK_BOUNDARY instead of
5822 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5824 (BIGGEST_ALIGNMENT): Set to 128.
5825 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5826 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5829 2018-03-19 Richard Biener <rguenther@suse.de>
5831 PR tree-optimization/84933
5832 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5833 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5835 2018-03-19 Richard Biener <rguenther@suse.de>
5837 PR tree-optimization/84859
5838 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5839 (cond_if_else_store_replacement): Perform sinking operation on
5840 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5841 Generalize what a BB with a single eligible store is.
5843 2018-03-19 Richard Biener <rguenther@suse.de>
5845 PR tree-optimization/84929
5846 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5847 chrec_is_positive against non-chrec arg.
5849 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5852 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5854 2018-03-18 Martin Liska <mliska@suse.cz>
5856 PR rtl-optimization/84635
5857 * regrename.c (build_def_use): Use matches_mode only when
5860 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5862 PR tree-optimization/84913
5863 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5864 vectorize chains of COND_EXPRs.
5866 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5868 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5870 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5872 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5874 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5876 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5878 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5879 Kito Cheng <kito.cheng@gmail.com>
5881 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5882 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5883 (nds32_adjust_reg_alloc_order): New function.
5884 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5886 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5888 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5889 nds32_print_operand, nds32_print_operand_address): Use
5890 HOST_WIDE_INT_PRINT_DEC instead.
5892 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5894 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5896 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5899 * config/i386/i386.c (initial_ix86_tune_features,
5900 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5902 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5903 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5904 rather than 1u << ix86_tune. Formatting fix.
5905 (ix86_option_override_internal): Change ix86_arch_mask from
5906 unsigned int to unsigned HOST_WIDE_INT, initialize to
5907 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5908 (ix86_function_specific_restore): Likewise.
5910 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5913 * postreload.c (reload_combine_recognize_pattern): Perform
5914 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5915 truncate_int_for_mode the result for the destination's mode.
5918 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5919 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5921 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5924 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5927 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5928 regno_allocno_class_array and sorted_pseudos.
5929 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5930 insns where regno is used.
5932 2018-03-16 Martin Liska <mliska@suse.cz>
5935 * multiple_target.c (create_dispatcher_calls): Redirect
5936 reference in the symbol table.
5938 2018-03-16 Martin Liska <mliska@suse.cz>
5941 * multiple_target.c (create_dispatcher_calls): Redirect also
5944 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5948 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5949 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5950 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5952 2018-03-16 Julia Koval <julia.koval@intel.com>
5954 * doc/invoke.texi (Skylake Server): Add CLWB.
5955 Cannonlake): Remove CLWB.
5957 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5959 PR tree-optimization/84841
5960 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5962 (FLOAT_ONE_CONST_TYPE): Define.
5963 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5964 (sort_by_operand_rank): Put entries with higher constant_type last
5965 rather than first to match comments.
5967 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5969 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5972 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5975 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5976 check and use address of target always.
5978 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5981 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5982 (indirect_thunk_bnd_needed): Likewise.
5983 (indirect_thunks_used): Likewise.
5984 (indirect_thunks_bnd_used): Likewise.
5985 (indirect_return_needed): New.
5986 (indirect_return_bnd_needed): Likewise.
5987 (output_indirect_thunk_function): Add a bool argument for
5989 (output_indirect_thunk_function): Don't generate alias for
5990 function return thunk.
5991 (ix86_code_end): Call output_indirect_thunk_function to generate
5992 function return thunks.
5993 (ix86_output_function_return): Set indirect_return_bnd_needed
5994 and indirect_return_needed instead of indirect_thunk_bnd_needed
5995 and indirect_thunk_needed.
5997 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5999 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6000 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6001 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6003 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6004 Paul Hua <paul.hua.gm@gmail.com>
6007 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6009 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6011 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6012 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6015 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6018 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6019 instead of GET_MODE_SIZE when comparing Units.
6021 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6024 * varasm.c (hash_section): Return an unchangeble hash value
6025 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6026 Return !aarch64_can_use_per_function_literal_pools_p ().
6028 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6031 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6032 rather than address of cmode as last argument to prepare_cmp_insn.
6034 2018-03-15 Julia Koval <julia.koval@intel.com>
6036 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6037 F_AVX512VNNI, F_AVX512BITALG): New.
6039 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6042 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6043 insn for floating-point loads and stores.
6045 2018-03-14 Carl Love <cel@us.ibm.com>
6047 * config/rs6000/rs6000-c.c: Add macro definitions for
6048 ALTIVEC_BUILTIN_VEC_PERMXOR.
6049 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6050 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6051 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6052 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6054 * config/doc/extend.texi: Add prototypes for vec_permxor.
6056 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6059 * diagnostic-show-locus.c (class layout_point): Convert m_line
6060 from int to linenum_type.
6061 (line_span::comparator): Use linenum "compare" function when
6062 comparing line numbers.
6063 (test_line_span): New function.
6064 (layout_range::contains_point): Convert param "row" from int to
6066 (layout_range::intersects_line_p): Likewise.
6067 (layout::will_show_line_p): Likewise.
6068 (layout::print_source_line): Likewise.
6069 (layout::should_print_annotation_line_p): Likewise.
6070 (layout::print_annotation_line): Likewise.
6071 (layout::print_leading_fixits): Likewise.
6072 (layout::annotation_line_showed_range_p): Likewise.
6073 (struct line_corrections): Likewise for field m_row.
6074 (line_corrections::line_corrections): Likewise for param "row".
6075 (layout::print_trailing_fixits): Likewise.
6076 (layout::get_state_at_point): Likewise.
6077 (layout::get_x_bound_for_row): Likewise.
6078 (layout::print_line): Likewise.
6079 (diagnostic_show_locus): Likewise for locals "last_line" and
6081 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6082 * input.c (selftest::test_linenum_comparisons): New function.
6083 (selftest::input_c_tests): Call it.
6084 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6085 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6086 * selftest.h (ASSERT_GT): New macro.
6087 (ASSERT_GT_AT): New macro.
6088 (ASSERT_LT): New macro.
6089 (ASSERT_LT_AT): New macro.
6091 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6093 PR rtl-optimization/84780
6094 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6096 2018-03-14 Martin Liska <mliska@suse.cz>
6098 * tree.c (record_node_allocation_statistics): Use
6099 get_stats_node_kind.
6100 (get_stats_node_kind): New function extracted from
6101 record_node_allocation_statistics.
6102 (free_node): Use get_stats_node_kind.
6104 2018-03-14 Richard Biener <rguenther@suse.de>
6106 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6107 that the value-set of ANTIC_IN doesn't grow.
6110 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6112 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6113 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6115 2018-03-14 Julia Koval <julia.koval@intel.com>
6117 * config.gcc (icelake-client, icelake-server): New.
6119 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6120 (initial_ix86_arch_features): Ditto.
6121 (PTA_SKYLAKE): Add SGX.
6122 (PTA_ICELAKE): Remove.
6123 (PTA_ICELAKE_CLIENT): New.
6124 (PTA_ICELAKE_SERVER): New.
6125 (ix86_option_override_internal): Split up icelake on icelake client and
6127 (get_builtin_code_for_version): Ditto.
6128 (fold_builtin_cpu): Ditto.
6129 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6130 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6131 * config/i386/i386.h (processor_type): Ditto.
6132 * doc/invoke.texi: Ditto.
6134 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6137 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6138 INTEGER_CST offset, add it together with bitpos / 8 and
6139 sign extend based on POINTER_SIZE.
6143 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6146 * config/i386/constraints.md (Yc): New register constraint.
6147 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6148 Use Yc constraint for alternative 2 of operand 0. Remove
6149 preferred_for_speed attribute.
6151 2018-03-14 Richard Biener <rguenther@suse.de>
6153 PR tree-optimization/84830
6154 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6155 with the old one to avoid oscillations.
6157 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6160 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6162 (assign_by_spills): Return a flag of reload assignment failure.
6163 Do not process the reload assignment failures. Do not spill other
6164 reload pseudos if they has the same reg class. Update n if
6166 (lra_assign): Add a return arg. Set up from the result of
6167 assign_by_spills call.
6168 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6169 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6170 usage_insns if it is not NULL.
6171 (spill_hard_reg_in_range): New function.
6172 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6173 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6174 function prototypes.
6175 (lra_assign): Change prototype.
6176 * lra.c (lra): Add code to deal with fails by splitting hard reg
6179 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6181 * config/riscv/riscv.opt (mrelax): New option.
6182 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6183 "norelax" when riscv_mrelax is disabled.
6184 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6186 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6189 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6190 reassociation for int modes.
6192 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6194 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6195 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6197 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6198 * config/aarch64/aarch64-sve.md
6199 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6200 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6201 (*extend<mode><Vwide>2): Rename to...
6202 (aarch64_sve_extend<mode><Vwide>2): ...this.
6203 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6204 renaming the old pattern to...
6205 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6207 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6208 define_expand, renaming the old pattern to...
6209 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6210 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6211 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6212 account when deciding which SVE instruction the optab should use.
6213 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6215 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6217 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6218 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6219 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6220 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6221 (tlsdesc_small_<mode>): Turn a define_expand and use
6222 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6223 (tlsdesc_small_advsimd_<mode>): ...this.
6224 (tlsdesc_small_sve_<mode>): New pattern.
6226 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6228 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6229 (UNSPEC_UMUL_HIGHPART): New constants.
6230 (MUL_HIGHPART): New int iteraor.
6231 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6232 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6234 (*<su>mul<mode>3_highpart): New define_insn.
6236 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6239 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6242 2018-03-13 Martin Liska <mliska@suse.cz>
6245 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6247 (sem_item_optimizer::~sem_item_optimizer): Release it.
6248 (sem_item_optimizer::merge_classes): Register variable aliases.
6249 (sem_item_optimizer::fixup_pt_set): New function.
6250 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6251 * ipa-icf.h: Declare new variables and functions.
6253 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6256 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6257 integer_pow2p@2 and test integer_pow2p in condition.
6258 (A < 0 ? C : 0): Similarly for @1.
6261 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6262 characters starting at p contain '\0' character, don't look beyond
6266 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6267 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6270 * reg-stack.c (change_stack): Change update_end var from int to
6271 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6272 also call set_block_for_insn on the newly added insns and rescan.
6275 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6276 on the last operand.
6279 * tree.c (stabilize_reference_1): Return save_expr (e) for
6280 STATEMENT_LIST even if it doesn't have side-effects.
6282 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6284 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6286 2018-03-12 Renlin Li <renlin.li@arm.com>
6288 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6289 aarch64_output_scalar_simd_mov_immediate.
6291 2018-03-12 Martin Sebor <msebor@redhat.com>
6293 PR tree-optimization/83456
6294 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6295 for perfectly overlapping calls to memcpy.
6296 (gimple_fold_builtin_memory_chk): Same.
6297 (gimple_fold_builtin_strcpy): Handle no-warning.
6298 (gimple_fold_builtin_stxcpy_chk): Same.
6299 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6301 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6303 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6304 parameter. Use it for SFmode.
6305 (rs6000_function_arg_advance_1): Adjust.
6306 (rs6000_function_arg): Adjust.
6307 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6309 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6311 PR rtl-optimization/84169
6312 PR rtl-optimization/84780
6313 * combine.c (can_combine_p): Check for a 2-insn combination whether
6314 the destination register is used between the two insns, too.
6316 2018-03-12 Richard Biener <rguenther@suse.de>
6318 PR tree-optimization/84803
6319 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6320 for refs DR analysis didn't process.
6322 2018-03-12 Richard Biener <rguenther@suse.de>
6324 PR tree-optimization/84777
6325 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6326 force-vectorize loops ignore whether we are optimizing for size.
6328 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6330 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6331 (TARGET_MD_ASM_ADJUST): Define.
6333 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6334 Kito Cheng <kito.cheng@gmail.com>
6335 Chung-Ju Wu <jasonwucj@gmail.com>
6337 * config/nds32/nds32.c (nds32_compute_stack_frame,
6338 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6339 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6340 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6341 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6342 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6343 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6344 * config/nds32/nds32.md (prologue, epilogue): Use macro
6345 NDS32_V3PUSH_AVAILABLE_P to do checking.
6347 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6350 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6351 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6352 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6353 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6354 addition of most attributes on !orig_type_die or the attribute not
6355 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6357 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6358 Chung-Ju Wu <jasonwucj@gmail.com>
6360 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6362 * config/nds32/nds32.opt (mvh): New option.
6364 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6365 Chung-Ju Wu <jasonwucj@gmail.com>
6367 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6369 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6370 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6373 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6374 Chung-Ju Wu <jasonwucj@gmail.com>
6376 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6378 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6379 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6381 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6382 Kito Cheng <kito.cheng@gmail.com>
6383 Chung-Ju Wu <jasonwucj@gmail.com>
6385 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6386 UNSPEC_FFMISM and UNSPEC_FLMISM.
6387 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6388 for ffb, ffmism and flmism.
6389 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6390 (unspec_ffmism): Ditto.
6391 (unspec_flmism): Ditto.
6392 (nds32_expand_builtin_impl): Check if string extension is available.
6393 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6394 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6396 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6399 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6402 * lra-assigns.c (assign_by_spills): Return a flag of reload
6403 assignment failure. Do not process the reload assignment
6404 failures. Do not spill other reload pseudos if they has the same
6406 (lra_assign): Add a return arg. Set up from the result of
6407 assign_by_spills call.
6408 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6409 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6410 usage_insns if it is not NULL.
6411 (spill_hard_reg_in_range): New function.
6412 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6413 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6414 function prototypes.
6415 (lra_assign): Change prototype.
6416 * lra.c (lra): Add code to deal with fails by splitting hard reg
6419 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6422 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6424 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6427 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6428 (dw_val_node): Add val_symbolic_view.
6429 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6430 (symview_upper_bound): New.
6431 (new_line_info_table): Initialize symviews_since_reset.
6432 (dwarf2out_source_line): Count symviews_since_reset and set
6433 symview_upper_bound.
6434 (dw_val_equal_p): Handle symview.
6435 (add_AT_symview): New.
6436 (print_dw_val): Handle symview.
6437 (attr_checksum, attr_checksum_ordered): Likewise.
6438 (same_dw_val_p, size_of_die): Likewise.
6439 (value_format, output_die): Likewise.
6440 (add_high_low_attributes): Use add_AT_symview for entry_view.
6441 (dwarf2out_finish): Reset symview_upper_bound, clear
6444 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6447 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6448 Add strict argument and use it.
6449 (rs6000_split_multireg_move): Update for new strict argument.
6450 (mem_operand_gpr): Disallow all non-offsettable addresses.
6451 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6453 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6456 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6457 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6458 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6461 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6462 decl, use remap_type if we want to use the type.
6464 2018-03-09 Martin Sebor <msebor@redhat.com>
6466 PR tree-optimization/84526
6467 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6469 (builtin_access::generic_overlap): Be prepared to handle non-array
6472 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6474 PR rtl-optimization/84682
6475 * lra-constraints.c (process_address_1): Check is_address flag
6476 for address constraints.
6477 (process_alt_operands): Likewise.
6478 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6479 preprocess_constraints.
6480 * recog.h (preprocess_constraints): Add oploc parameter.
6482 * recog.c (preprocess_constraints): Test address_operand for
6483 CT_ADDRESS constraints.
6485 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6488 * lra-assigns.c (assign_by_spills): Return a flag of reload
6489 assignment failure. Do not process the reload assignment
6490 failures. Do not spill other reload pseudos if they has the same
6492 (lra_assign): Add a return arg. Set up from the result of
6493 assign_by_spills call.
6494 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6495 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6496 usage_insns if it is not NULL.
6497 (spill_hard_reg_in_range): New function.
6498 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6499 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6500 function prototypes.
6501 (lra_assign): Change prototype.
6502 * lra.c (lra): Add code to deal with fails by splitting hard reg
6505 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6508 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6509 Accept complain bool parameter. Only emit errors if it is true.
6510 (arm_parse_cpu_option_name): Likewise.
6511 (arm_target_thumb_only): Adjust callers of the above.
6512 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6513 prototype to take a default true bool parameter.
6514 (arm_parse_arch_option_name): Likewise.
6516 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6517 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6521 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6522 * configure: Regenerate.
6523 * configure.ac ("linker --version-script option"): New.
6524 ("linker soname option"): New.
6526 2018-03-09 Richard Biener <rguenther@suse.de>
6528 PR tree-optimization/84775
6529 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6530 immediate uses of predicate stmts and mark them modified.
6533 PR tree-optimization/84178
6534 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6536 (version_loop_for_if_conversion): Delay update_ssa call.
6537 (tree_if_conversion): Delay update_ssa until after predicate
6540 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6543 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6544 when the function accesses prior frames.
6546 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6549 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6550 gen_llsym, otherwise call maybe_gen_llsym.
6553 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6554 has ',' character inside of it.
6556 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6559 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6560 as clobbering CC_REGNUM.
6562 2018-03-08 Richard Biener <rguenther@suse.de>
6565 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6566 (follow_copies_to_constant): Do not follow SSA names registered
6569 2018-03-08 Richard Biener <rguenther@suse.de>
6571 PR tree-optimization/84178
6572 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6574 (version_loop_for_if_conversion): Delay update_ssa call.
6575 (tree_if_conversion): Delay update_ssa until after predicate
6578 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6580 PR tree-optimization/84178
6581 * tree-if-conv.c (release_bb_predicate): Remove the
6582 the assertion that the stmts have NULL use_ops.
6583 Discard the statements, asserting that they haven't
6584 yet been added to a BB.
6586 2018-03-08 Richard Biener <rguenther@suse.de>
6588 PR tree-optimization/84746
6589 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6590 (phi_translate): Pass in destination ANTIC_OUT set.
6591 (phi_translate_1): Likewise. For a simplified result lookup
6592 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6593 (phi_translate_set): Adjust.
6594 (do_pre_regular_insertion): Likewise.
6595 (do_pre_partial_partial_insertion): Likewise.
6597 2018-03-08 Martin Liska <mliska@suse.cz>
6599 PR gcov-profile/84735
6600 * doc/gcov.texi: Document usage of profile files.
6601 * gcov-io.h: Document changes in the format.
6603 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6607 * dwarf2out.c (struct dw_line_info_table): Update comments for
6609 (FORCE_RESET_NEXT_VIEW): New.
6610 (FORCE_RESETTING_VIEW_P): New.
6611 (RESETTING_VIEW_P): Check for -1 too.
6612 (ZERO_VIEW_P): Likewise.
6613 (new_line_info_table): Force-reset next view.
6614 (dwarf2out_begin_function): Likewise.
6615 (dwarf2out_source_line): Simplify zero_view_p initialization.
6616 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6617 view directly. Omit view when omitting .loc at line 0.
6619 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6621 PR tree-optimization/84740
6622 * tree-switch-conversion.c (process_switch): Call build_constructors
6623 only if info.phi_count is non-zero.
6625 PR tree-optimization/84739
6626 * tree-tailcall.c (find_tail_calls): Check call arguments against
6627 DECL_ARGUMENTS (current_function_decl) rather than
6628 DECL_ARGUMENTS (func) when checking for tail recursion.
6630 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6632 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6633 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6634 Volker Reichelt's entry and add entries for people that perform
6635 GCC fuzzy testing and report numerous bugs.
6637 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6640 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6641 readonly data in sdata, if that is disabled.
6642 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6643 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6644 -mreadonly-in-sdata option.
6646 2018-03-07 Martin Sebor <msebor@redhat.com>
6648 PR tree-optimization/84468
6649 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6650 basic block when looking for nul assignment.
6652 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6655 * except.h (output_function_exception_table): Adjust prototype.
6656 * except.c (output_function_exception_table): Remove FNNAME parameter
6657 and add SECTION parameter. Ouput one part of the table at a time.
6658 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6659 the first part of the exception table and emit unwind directives.
6660 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6661 (i386_pe_seh_cold_init): Likewise.
6662 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6663 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6664 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6665 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6666 * config/i386/winnt.c: Include except.h.
6667 (struct seh_frame_state): Add reg_offset, after_prologue and
6668 in_cold_section fields.
6669 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6670 (i386_pe_seh_cold_init): New function.
6671 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6672 to seh->in_cold_section.
6673 (seh_emit_push): Record the offset of the push.
6674 (seh_emit_save): Record the offet of the save.
6675 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6676 Test seh->after_prologue to disregard the epilogue.
6677 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6678 (i386_pe_end_cold_function): New function.
6680 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6683 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6684 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6687 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6689 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6690 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6693 * multiple_target.c: Include tree-inline.h and intl.h.
6694 (expand_target_clones): Diagnose and fail if node->definition and
6695 !tree_versionable_function_p (node->decl).
6697 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6699 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6701 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6702 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6703 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6705 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6708 * combine.c (try_combine): Use reg_or_subregno instead of handling
6709 just paradoxical SUBREGs and REGs.
6711 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6713 * config/arc/arc.c (arc_finalize_pic): Remove function.
6714 (arc_must_save_register): We use single base PIC register, remove
6715 checks to save/restore the PIC register.
6716 (arc_expand_prologue): Likewise.
6717 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6719 (arc_verify_short): Likewise.
6720 (arc_attr_type): Likewise.
6721 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6722 (walk_stores): Likewise.
6723 (arc_address_cost): Make it static.
6724 (arc_verify_short): Likewise.
6725 (branch_dest): Likewise.
6726 (arc_attr_type): Likewise.
6727 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6728 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6729 (arc_final_prescan_insn): Remove inserting the nops due to
6730 hardware hazards. It is done in reorg step.
6731 (insn_length_variant_t): Remove.
6732 (insn_length_parameters_t): Likewise.
6733 (arc_insn_length_parameters): Likewise.
6734 (arc_get_insn_variants): Likewise.
6735 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6737 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6740 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6743 PR tree-optimization/84687
6744 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6746 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6748 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6750 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6751 Rename to ppc_speculation_barrier.
6752 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6753 __builtin_ppc_speculation_barrier.
6755 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6758 * combine.c (combine_simplify_rtx): Don't try to simplify if
6759 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6762 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6764 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6765 to 32 bytes when compiling for POWER9.
6767 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6770 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6771 regparm >= 3 with no arg reg available also for calls with
6772 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6775 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6777 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6779 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6782 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6784 2018-03-05 Richard Biener <rguenther@suse.de>
6786 PR tree-optimization/84486
6787 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6788 When inserting a __builtin_assume_aligned call set the LHS
6789 SSA name alignment info accordingly.
6791 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6793 PR tree-optimization/84114
6794 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6795 Avoid reassociation of FLOAT_MODE addition.
6797 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6799 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6800 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6801 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6802 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6803 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6804 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6805 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6807 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6808 __builtin_ia32_wbinvd): New builtins.
6809 (SPECIAL_ARGS2): New.
6810 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6811 (SPECIAL_ARGS2): New.
6812 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6813 (ix86_valid_target_attribute_inner_p): Ditto.
6814 (ix86_init_mmx_sse_builtins): Add special_args2.
6815 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6816 TARGET_WBNOINVD_P): New.
6817 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6818 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6819 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6820 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6821 * config/i386/pconfigintrin.h: New file.
6822 * config/i386/wbnoinvdintrin.h: Ditto.
6823 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6824 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6826 2018-03-05 Richard Biener <rguenther@suse.de>
6828 PR tree-optimization/84670
6829 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6831 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6832 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6833 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6834 doesn't grow if all successors (recursively) were visited at least
6837 2018-03-05 Richard Biener <rguenther@suse.de>
6839 PR tree-optimization/84650
6840 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6841 if executed in the loop pipeline.
6843 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6845 * doc/configfiles.texi (Configuration Files): Move info about
6846 conditionalizing $target-protos.h to...
6847 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6848 differs from $target-protos.h.
6850 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6851 Chung-Ju Wu <jasonwucj@gmail.com>
6853 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6854 * config/nds32/nds32-multiple.md (setmemsi): Define.
6855 * config/nds32/nds32-memory-manipulation.c
6856 (nds32_gen_dup_4_byte_to_word_value): New.
6857 (emit_setmem_word_loop): New.
6858 (emit_setmem_byte_loop): New.
6859 (nds32_expand_setmem_loop): New.
6860 (nds32_expand_setmem_loop_v3m): New.
6861 (nds32_expand_setmem_unroll): New.
6862 (nds32_expand_setmem): New.
6864 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6865 Chung-Ju Wu <jasonwucj@gmail.com>
6867 * config/nds32/nds32-memory-manipulation.c
6868 (nds32_emit_load_store): New.
6869 (nds32_emit_post_inc_load_store): New.
6870 (nds32_emit_mem_move): New.
6871 (nds32_emit_mem_move_block): New.
6872 (nds32_expand_movmemsi_loop_unknown_size): New.
6873 (nds32_expand_movmemsi_loop_known_size): New.
6874 (nds32_expand_movmemsi_loop): New.
6875 (nds32_expand_movmemsi_unroll): New.
6876 (nds32_expand_movmemqi): Rename ...
6877 (nds32_expand_movmemsi): ... to this.
6878 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6879 (movmemsi): ... to this.
6880 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6881 (nds32_expand_movmemsi): ... to this.
6883 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6884 Monk Chiang <sh.chiang04@gmail.com>
6885 Chung-Ju Wu <jasonwucj@gmail.com>
6887 * config/nds32/nds32-protos.h
6888 (nds32_expand_load_multiple): New arguments.
6889 (nds32_expand_store_multiple): Ditto.
6890 (nds32_valid_multiple_load_store): Rename ...
6891 (nds32_valid_multiple_load_store_p): ... to this.
6892 * config/nds32/nds32-memory-manipulation.c
6893 (nds32_expand_load_multiple): Refine implementation.
6894 (nds32_expand_store_multiple): Ditto.
6895 * config/nds32/nds32-multiple.md
6896 (load_multiple): Update nds32_expand_load_multiple interface.
6897 (store_multiple): Update nds32_expand_store_multiple interface.
6898 * config/nds32/nds32-predicates.c
6899 (nds32_valid_multiple_load_store): Rename ...
6900 (nds32_valid_multiple_load_store_p): ... to this and refine
6902 * config/nds32/predicates.md
6903 (nds32_load_multiple_and_update_address_operation): New predicate.
6904 (nds32_store_multiple_and_update_address_operation): New predicate.
6906 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6907 Chung-Ju Wu <jasonwucj@gmail.com>
6909 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6910 (combo): New attribute.
6911 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6913 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6915 * config/nds32/nds32.opt: Change -mcmodel= default value.
6917 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6918 Monk Chiang <sh.chiang04@gmail.com>
6919 Chung-Ju Wu <jasonwucj@gmail.com>
6921 * config/nds32/constants.md (unspec_element): New enum.
6922 * config/nds32/constraints.md (Umw): New constraint.
6923 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6924 * config/nds32/nds32-intrinsic.md: Likewise.
6925 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6926 (nds32_valid_smw_lwm_base_p): New.
6927 (nds32_output_smw_single_word): New.
6928 (nds32_output_lmw_single_word): New.
6929 (nds32_expand_unaligned_load): New.
6930 (nds32_expand_unaligned_store): New.
6931 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6932 (nds32_output_smw_single_word): Declare.
6933 (nds32_output_lmw_single_word): Declare.
6934 (nds32_expand_unaligned_load): Declare.
6935 (nds32_expand_unaligned_store): Declare.
6936 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6937 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6938 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6939 NDS32_BUILTIN_UASTORE_DW.
6940 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6943 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6944 Kito Cheng <kito.cheng@gmail.com>
6945 Chung-Ju Wu <jasonwucj@gmail.com>
6947 * config/nds32/nds32-intrinsic.c
6948 (nds32_expand_builtin_null_ftype_reg): Delete.
6949 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6950 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6951 (nds32_read_argument): New.
6952 (nds32_legitimize_target): Ditto.
6953 (nds32_legitimize_argument): Ditto.
6954 (nds32_check_constant_argument): Ditto.
6955 (nds32_expand_unop_builtin): Ditto.
6956 (nds32_expand_unopimm_builtin): Ditto.
6957 (nds32_expand_binop_builtin): Ditto.
6958 (nds32_builtin_decl_impl): Ditto.
6959 (builtin_description): Ditto.
6960 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6961 (nds32_init_builtins_impl): Ditto.
6962 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6963 (nds32_builtin_decl): New.
6964 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6965 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6967 2018-03-02 Jeff Law <law@redhat.com>
6969 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6970 (redundant_insn, fill_simple_delay_slots): Likewise.
6971 (fill_slots_from_thread): Likewise.
6972 * resource.c (mark_referenced_resources): Likewise.
6973 (mark_set_resources, find_dead_or_set_registers): Likewise.
6975 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6977 * substring-locations.h (format_warning_va): Formatting fix for
6979 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6981 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6982 * substring-locations.c: Include intl.h.
6983 (format_warning_va): Turned into small wrapper around
6984 format_warning_n_va, renamed to ...
6985 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6986 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6988 (format_warning_at_substring_n): New function.
6989 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6990 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6991 format_warning_at_substring with just a shorter name instead of
6992 const function pointer.
6993 (fmtwarn_n): New function.
6994 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6995 appropriate, get rid of all the fmtstr temporaries, move conditionals
6996 with G_() wrapped string literals directly into fmtwarn arguments,
6997 cast dir.len to (int), formatting fixes.
6999 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7001 * doc/invoke.texi: Remove "Cilk Plus" references.
7003 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7004 Richard Biener <rguenther@suse.de>
7007 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7008 for error or warning attributes if CALL_FROM_THUNK_P is set.
7011 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7014 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7015 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7018 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7019 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7021 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7022 instead of -1U in last predictors element's probability member.
7024 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7027 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7028 arguments if they are comparable.
7030 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7032 PR tree-optimization/84634
7033 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7034 masks and masked_loop_p with a single loop_masks, making sure it's
7035 null for bb vectorization.
7037 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7039 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7040 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7041 loop->force_vectorize to check whether there is no alias.
7043 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7046 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7048 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7050 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7051 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7052 instead of a loop around prev_real_insn.
7053 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7057 * config/i386/i386.c (ix86_print_operand): Use conditional
7058 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7061 2018-03-02 Richard Biener <rguenther@suse.de>
7063 PR tree-optimization/84427
7064 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7065 (bitmap_set_subtract_values): Rewrite to handle multiple
7068 (prune_clobbered_mems): Likewise.
7069 (phi_translate): Take edge instead of pred/phiblock.
7070 (phi_translate_1): Likewise.
7071 (phi_translate_set): Likewise. Insert all translated
7072 exprs for a value into the set, keeping possibly multiple
7073 expressions per value.
7074 (compute_antic_aux): Adjust for phi_translate changes.
7075 When intersecting union the expressions and prune those
7076 not in the final value set, keeping possibly multiple
7077 expressions per value. Do not use value-insertion
7078 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7079 all expressions. Add verification that the value-sets
7080 only shrink during iteration.
7081 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7082 (do_pre_regular_insertion): Likewise.
7083 (do_pre_partial_partial_insertion): Likewise.
7085 2018-03-02 Richard Biener <rguenther@suse.de>
7088 * config/darwin.c (saved_debug_info_level): New static global.
7089 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7090 (darwin_asm_lto_end): Restore debug info generation settings.
7092 2018-03-01 Martin Liska <mliska@suse.cz>
7095 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7098 2018-03-01 Richard Biener <rguenther@suse.de>
7101 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7102 type annotation with LTO when debug was disabled at compile-time.
7104 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7106 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7108 (mips_final_postscan_insn): Likewise.
7110 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7112 PR rtl-optimization/84528
7113 * alias.c (init_alias_target): Add commentary.
7114 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7115 a unique base value if the frame pointer is not eliminated
7116 to the stack pointer.
7118 2018-03-01 Tom de Vries <tom@codesourcery.com>
7120 PR rtl-optimization/83327
7121 * lra-int.h (hard_regs_spilled_into): Declare.
7122 * lra.c (hard_regs_spilled_into): Define.
7123 (init_reg_info): Init hard_regs_spilled_into.
7124 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7125 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7126 (process_bb_lives): Handle hard_regs_spilled_into.
7127 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7128 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7130 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7132 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7133 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7134 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7135 * config/rs6000/aix72.h: New file.
7137 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7139 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7140 instead of warning_at with conditional singular and plural messages
7144 * stor-layout.c (update_alignment_for_field): For
7145 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7146 && !DECL_PACKED (field), do the alignment update, just use
7147 only desired_align instead of MAX (type_align, desired_align)
7149 (place_field): Don't do known_align < desired_align handling
7150 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7151 is non-NULL, instead do it after rli->prev_field handling and
7152 only if not within a bitfield word. For DECL_PACKED (field)
7153 use type_align of BITS_PER_UNIT.
7155 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7157 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7158 superfluous parentheses and trailing spaces.
7160 2018-02-28 Richard Biener <rguenther@suse.de>
7162 PR tree-optimization/84584
7163 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7164 SCoPs with fake exit edge.
7166 2018-02-28 Martin Liska <mliska@suse.cz>
7169 * timevar.c (timer::print): Fix format to properly print 100%
7172 2018-02-28 Richard Biener <rguenther@suse.de>
7175 * genmatch.c (capture_info::walk_match): Do not mark
7176 captured expressions without operands as expr_p given
7177 they act more like predicates and should be subject to
7178 "lost tail" side-effect preserving.
7180 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7182 PR rtl-optimization/81611
7183 * auto-inc-dec.c (attempt_change): Move dead note from
7184 mem_insn if it's the next use of regno
7185 (find_address): Take address use of reg holding
7186 non-incremented value. Add parm to limit search to the named
7188 (merge_in_block): Attempt to use a mem insn that is the next
7189 use of the original regno.
7191 2018-02-27 Martin Sebor <msebor@redhat.com>
7194 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7195 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7197 2018-02-27 Martin Sebor <msebor@redhat.com>
7199 PR translation/84207
7200 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7201 n argument to unsigned HOST_WIDE_INT.
7202 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7203 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7204 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7205 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7207 2018-02-27 Richard Biener <rguenther@suse.de>
7209 PR tree-optimization/84512
7210 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7211 Do not use the estimate returned from record_stmt_cost for
7212 the scalar iteration cost but sum properly using add_stmt_cost.
7214 2018-02-27 Richard Biener <rguenther@suse.de>
7216 PR tree-optimization/84466
7217 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7218 Adjust last change to less strictly validate use operands.
7220 2018-02-27 Martin Liska <mliska@suse.cz>
7222 PR gcov-profile/84548
7223 * gcov.c (process_file): Allow partial overlap and consider it
7224 also as group functions.
7225 (output_lines): Properly calculate range of lines for a group.
7227 2018-02-27 Martin Liska <mliska@suse.cz>
7229 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7230 'ggc' suffixes. Change first column width.
7231 (timer::print): Fix formatting of the column.
7233 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7235 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7236 preserve inline entry blocks for the sake of debug inline
7237 entry point markers alone.
7238 (remove_unused_locals): Suggest in comments a better place to
7239 force the preservation of inline entry blocks that are
7240 otherwise unused, but do not preserve them.
7242 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7244 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7246 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7249 * config/i386/constraints.md (Bs): Replace
7250 ix86_indirect_branch_register with
7251 TARGET_INDIRECT_BRANCH_REGISTER.
7253 * config/i386/i386.md (indirect_jump): Likewise.
7254 (tablejump): Likewise.
7255 (*sibcall_memory): Likewise.
7256 (*sibcall_value_memory): Likewise.
7257 Peepholes of indirect call and jump via memory: Likewise.
7258 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7259 (*sibcall_value_GOT_32): Likewise.
7260 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7261 (GOT_memory_operand): Likewise.
7262 (call_insn_operand): Likewise.
7263 (sibcall_insn_operand): Likewise.
7264 (GOT32_symbol_operand): Likewise.
7265 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7267 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7269 PR rtl-optimization/83496
7270 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7271 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7272 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7273 redundant insn, if any.
7274 (relax_delay_slots): Likewise.
7275 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7277 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7279 PR tree-optimization/83965
7280 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7281 that grouped statements are part of a reduction chain. Return
7282 true if the statement is not marked as a reduction itself but
7284 (vect_recog_dot_prod_pattern): Don't check whether the statement
7285 is part of a group here.
7286 (vect_recog_sad_pattern): Likewise.
7287 (vect_recog_widen_sum_pattern): Likewise.
7289 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7292 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7294 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7297 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7299 (ix86_output_indirect_function_return): New prototype.
7300 (ix86_split_simple_return_pop_internal): Likewise.
7301 * config/i386/i386.c (indirect_return_via_cx): New.
7302 (indirect_return_via_cx_bnd): Likewise.
7303 (indirect_thunk_name): Handle return va CX_REG.
7304 (output_indirect_thunk_function): Create alias for
7305 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7306 (ix86_output_indirect_jmp): Remove the bool argument.
7307 (ix86_output_indirect_function_return): New function.
7308 (ix86_split_simple_return_pop_internal): Likewise.
7309 * config/i386/i386.md (*indirect_jump): Don't pass false
7310 to ix86_output_indirect_jmp.
7311 (*tablejump_1): Likewise.
7312 (simple_return_pop_internal): Change it to define_insn_and_split.
7313 Call ix86_split_simple_return_pop_internal to split it for
7315 (simple_return_indirect_internal): Call
7316 ix86_output_indirect_function_return instead of
7317 ix86_output_indirect_jmp.
7319 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7322 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7323 memset and value initialization afterwards.
7325 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7327 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7329 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7332 * common/config/aarch64/aarch64-common.c
7333 (aarch_option_optimization_table[]): Switch
7334 off fomit-frame-pointer
7336 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7337 Chung-Ju Wu <jasonwucj@gmail.com>
7339 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7341 (store_multiple): Ditto.
7343 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7345 * config.gcc: Add --with-cpu support for nds32 target.
7346 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7347 * config/nds32/nds32.opt: Add -mcpu= option.
7349 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7351 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7352 isel=yes): Warn for these deprecated options.
7354 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7356 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7357 ISA_2_5_MASKS_EMBEDDED.
7359 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7361 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7362 p->max as pointers rather than using iterative_hash_expr.
7364 2018-02-23 Carl Love <cel@us.ibm.com>
7366 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7367 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7369 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7370 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7371 P8V_BUILTIN_VEC_VUNSIGNED2.
7373 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7376 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7377 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7379 * lra-constraints.c (curr_insn_transform): Set up
7380 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7382 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7383 * lra-eliminations.c (spill_pseudos): Ditto.
7384 (process_insn_for_elimination): Ditto.
7385 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7386 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7387 LRA_NON_CLOBBERED_ALT.
7389 2018-02-22 Martin Sebor <msebor@redhat.com>
7391 PR tree-optimization/84480
7392 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7393 to maybe_diag_stxncpy_trunc. Call it.
7394 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7395 from gimple_fold_builtin_strcpy. Print inlining stack.
7396 (handle_builtin_stxncpy): Print inlining stack.
7397 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7399 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7402 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7403 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7404 and -fcheck-pointer-bounds are used together.
7405 (indirect_thunk_prefix): New enum.
7406 (indirect_thunk_need_prefix): New function.
7407 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7408 "_nt" instead of "_bnd" for NOTRACK prefix.
7409 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7410 (output_indirect_thunk_function): Likewise.
7412 (ix86_code_end): Update output_indirect_thunk_function calls.
7413 (ix86_output_indirect_branch_via_reg): Replace
7414 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7415 (ix86_output_indirect_branch_via_push): Likewise.
7416 (ix86_output_function_return): Likewise.
7417 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7418 incompatible with -fcf-protection=branch and
7419 -fcheck-pointer-bounds.
7421 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7424 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7425 Change gcc_assert call to output_operand_lossage.
7427 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7429 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7431 2018-02-22 DJ Delorie <dj@redhat.com>
7432 Sebastian Perta <sebastian.perta@renesas.com>
7433 Oleg Endo <olegendo@gcc.gnu.org>
7435 * config/rx/rx.c (rx_rtx_costs): New function.
7436 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7438 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7440 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7442 2018-02-22 Martin Liska <mliska@suse.cz>
7445 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7446 Add "native" as a possible value.
7448 2018-02-22 Martin Liska <mliska@suse.cz>
7451 * config/i386/i386.c (ix86_option_override_internal):
7452 Add "native" as a possible value for -march and -mtune.
7454 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7457 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7458 to all type variants.
7460 PR tree-optimization/84503
7461 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7462 width as info->bitpos + info->bitsize - start.
7463 (merged_store_group::merge_overlapping): Simplify width computation.
7464 (check_no_overlap): New function.
7465 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7466 start + width and last_order of the group, fail if check_no_overlap
7468 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7469 to group if check_no_overlap fails.
7471 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7473 * config/rs6000/altivec.md: Delete contraint arguments to
7474 define_expand, define_split, and define_peephole2, and in
7475 define_insn_and_split if always unused.
7476 * config/rs6000/darwin.md: Ditto.
7477 * config/rs6000/dfp.md: Ditto.
7478 * config/rs6000/rs6000.md: Ditto.
7479 * config/rs6000/sync.md: Ditto.
7480 * config/rs6000/vector.md: Ditto.
7481 * config/rs6000/vsx.md: Ditto.
7483 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7485 * config/rs6000/altivec.md: Write output control strings as braced
7486 blocks instead of double-quoted strings.
7487 * config/rs6000/darwin.md: Ditto.
7488 * config/rs6000/rs6000.md: Ditto.
7489 * config/rs6000/vector.md: Ditto.
7490 * config/rs6000/vsx.md: Ditto.
7492 2018-02-21 Jason Merrill <jason@redhat.com>
7494 PR c++/84314 - ICE with templates and fastcall attribute.
7495 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7497 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7499 * ipa-cp.c (determine_versionability): Fix comment typos.
7501 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7504 * ipa-cp.c (determine_versionability): Do not version functions caling
7507 2018-02-21 Martin Liska <mliska@suse.cz>
7510 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7511 Add "native" as a possible value.
7512 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7513 the macro when native cpu detection is available.
7515 2018-02-21 Martin Liska <mliska@suse.cz>
7518 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7519 Add "native" as a possible value.
7520 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7521 when native cpu detection is available.
7523 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7524 Martin Sebor <msebor@redhat.com>
7526 PR tree-optimization/84478
7527 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7529 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7530 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7531 support which is conservatively correct, for 2 only stay conservative
7532 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7533 argument to the 2 argument get_range_strlen, adjust 6 arg
7534 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7536 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7537 (gimple_fold_builtin_strlen): Pass true as last argument to
7540 2018-02-20 Martin Sebor <msebor@redhat.com>
7543 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7544 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7545 (builtin_memref::builtin_memref): Factor out parts into
7546 set_base_and_offset and call it.
7548 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7551 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7552 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7553 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7554 search at the associated MODE_INT.
7556 2018-02-20 Jeff Law <law@redhat.com>
7559 PR tree-optimization/81592
7561 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7562 for range data rather than using global data.
7563 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7564 range data rather than using global data.
7565 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7566 pass it to children as needed.
7567 (struct directive::fmtresult): Similarly.
7568 (struct directive::set_width): Similarly.
7569 (struct directive::set_precision): Similarly.
7570 (format_integer, format_directive, parse_directive): Similarly.
7571 (format_none): Accept unnamed vr_values parameter.
7572 (format_percent, format_floating, format_character): Similarly.
7573 (format_string, format_plain): Similarly.
7574 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7575 the EVRP range analyzer for range data rather than using global data.
7576 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7577 gimple-ssa-evrp-analyze.h
7578 (class sprintf_dom_walker): Add after_dom_children member function.
7579 Add evrp_range_analyzer member.
7580 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7581 range analyzer as needed.
7582 (sprintf_dom_walker::after_dom_children): New member function.
7583 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7585 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7586 (evrp_range_analyzer::pop_to_marker): Likewise.
7588 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7590 PR tree-optimization/84419
7591 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7592 with the required type if its current type is compatible but
7595 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7598 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7599 after vectorization.
7601 2018-02-20 Martin Liska <mliska@suse.cz>
7604 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7605 possible values if we don't have a hint.
7607 2018-02-20 Martin Liska <mliska@suse.cz>
7611 * final.c (shorten_branches): Build align_tab array with one
7613 * opts.c (finish_options): Add alignment option limit check.
7614 (MAX_CODE_ALIGN): Likewise.
7615 (MAX_CODE_ALIGN_VALUE): Likewise.
7616 * doc/invoke.texi: Document maximum allowed option value for
7617 all -falign-* options.
7619 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7622 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7623 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7624 * var-tracking.c (emit_note_insn_var_location): Remove all references
7625 to NOTE_INSN_CALL_ARG_LOCATION.
7626 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7627 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7628 Use copy_rtx_if_shared.
7629 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7630 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7631 (dwarf2out_var_location): Remove handling of
7632 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7634 * final.c (final_scan_insn): Remove all references to
7635 NOTE_INSN_CALL_ARG_LOCATION.
7636 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7637 before dumping final insns.
7638 * except.c (emit_note_eh_region_end): Remove all references to
7639 NOTE_INSN_CALL_ARG_LOCATION.
7640 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7641 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7642 * config/arc/arc.c (hwloop_optimize): Likewise.
7643 * config/arm/arm.c (create_fix_barrier): Likewise.
7644 * config/s390/s390.c (s390_chunkify_start): Likewise.
7645 * config/sh/sh.c (find_barrier): Likewise.
7646 * config/i386/i386.c (rest_of_insert_endbranch,
7647 ix86_seh_fixup_eh_fallthru): Likewise.
7648 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7649 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7650 * config/frv/frv.c (frv_function_prologue): Likewise.
7651 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7653 (note_outside_basic_block_p): Remove all references to
7654 NOTE_INSN_CALL_ARG_LOCATION.
7655 * gengtype.c (adjust_field_rtx_def): Likewise.
7656 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7658 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7659 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7662 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7665 PR tree-optimization/84452
7666 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7667 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7670 2018-02-19 Martin Liska <mliska@suse.cz>
7673 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7675 2018-02-19 Martin Liska <mliska@suse.cz>
7676 Richard Sandiford <richard.sandiford@linaro.org>
7678 PR tree-optimization/82491
7679 * gimple-fold.c (get_base_constructor): Make earlier bail out
7682 2018-02-19 Carl Love <cel@us.ibm.com>
7684 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7685 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7687 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7688 P8V_BUILTIN_VEC_NEG.
7690 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7692 * config/rl78/rl78.md (movdf): New define expand.
7694 2018-02-19 Martin Liska <mliska@suse.cz>
7697 * doc/invoke.texi: Fix typo.
7698 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7700 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7702 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7703 handle rs6000_single_float and rs6000_double_float specially for
7706 2018-02-16 Jeff Law <law@redhat.com>
7708 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7711 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7714 * ipa-inline.c (inline_small_functions): Fix a typo.
7716 2018-02-16 Nathan Sidwell <nathan@acm.org>
7718 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7720 2018-02-16 Carl Love <cel@us.ibm.com>
7722 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7723 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7724 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7725 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7726 expansion to P8V_BUILTIN_VEC_FLOAT2.
7728 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7730 PR rtl-optimization/70023
7731 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7732 src_regno into account.
7734 2018-02-16 Carl Love <cel@us.ibm.com>
7736 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7737 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7738 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7739 * config/rs6000/rs6000.c: Remove case statements for
7740 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7741 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7742 and P9V_BUILTIN_VEC_VINSERT4B.
7743 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7744 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7745 * config/rs6000/vsx.md:
7746 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7749 2018-02-16 Carl Love <cel@us.ibm.com>
7751 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7753 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7755 * config/rs6000/rs6000-c.c: Add the definitions for
7756 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7757 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7758 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7759 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7760 definition for insert4b and define insn *insert3b_internal.
7761 * doc/extend.texi: Add documentation for vec_extract4b.
7763 2018-02-16 Nathan Sidwell <nathan@acm.org>
7765 * doc/extend.texi (Backwards Compatibility): Mention friend
7766 injection. Note for-scope is deprecated.
7767 * doc/invoke.texi (-ffriend-injection): Deprecate.
7769 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7771 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7772 that moved to I2, also allow destinations that are a paradoxical
7773 subreg (instead of a normal reg).
7775 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7778 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7781 2018-02-16 Richard Biener <rguenther@suse.de>
7783 PR tree-optimization/84037
7784 PR tree-optimization/84016
7786 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7787 Adjust vec_construct for the fact we need additional higher latency
7788 128bit inserts for AVX256 and AVX512 vector builds.
7789 (ix86_add_stmt_cost): Scale vector construction cost for
7792 2018-02-16 Richard Biener <rguenther@suse.de>
7794 PR tree-optimization/84417
7795 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7796 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7797 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7799 2018-02-16 Martin Liska <mliska@suse.cz>
7802 * internal-fn.def (ASAN_CHECK): Set proper flags.
7803 (ASAN_MARK): Likewise.
7805 2018-02-16 Julia Koval <julia.koval@intel.com>
7807 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7808 from PTA_CANNONLAKE.
7810 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7813 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7814 Use ++iter rather than iter++ for std::list iterators.
7815 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7816 defer deleting them until all nodes in the forest are processed. Do
7817 free even leaf nodes. Change to_process into auto_vec.
7820 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7821 * vec.h (vec_default_construct): Use memset instead of placement new
7822 if BROKEN_VALUE_INITIALIZATION is defined.
7823 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7824 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7827 PR rtl-optimization/83723
7828 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7829 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7830 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7832 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7834 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7836 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7838 PR rtl-optimization/81443
7839 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7840 from inner REGs to paradoxical SUBREGs.
7842 2018-02-16 Richard Biener <rguenther@suse.de>
7844 PR tree-optimization/84399
7845 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7846 For operands we can analyze at their definition make sure we can
7847 analyze them at each use as well.
7849 2018-02-16 Richard Biener <rguenther@suse.de>
7851 PR tree-optimization/84190
7852 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7853 volatile accesses if the decl isn't volatile.
7855 2018-02-15 Jason Merrill <jason@redhat.com>
7857 PR c++/84314 - ICE with templates and fastcall attribute.
7858 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7859 TYPE_CANONICAL on an existing type.
7861 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7863 PR tree-optimization/84383
7864 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7865 dstoff nor call operand_equal_p if dstbase is NULL.
7867 PR tree-optimization/84334
7868 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7869 also a CONSTANT_CLASS_P, punt.
7871 2018-02-14 Jim Wilson <jimw@sifive.com>
7873 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7874 first SMALL_OPERAND check. New local min_second_step. Move assert
7875 to where locals are set. Add TARGET_RVC support.
7876 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7878 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7880 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7882 2018-02-14 Martin Sebor <msebor@redhat.com>
7884 PR tree-optimization/83698
7885 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7886 arrays constrain the offset range to their bounds.
7887 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7888 (builtin_access::overlap): Avoid setting the size of overlap if it's
7890 (maybe_diag_overlap): Also consider arrays when deciding what values
7891 of offsets to include in diagnostics.
7893 2018-02-14 Martin Sebor <msebor@redhat.com>
7896 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7897 that correspond to the kind of a declaration.
7899 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7902 * config/pa/pa.md: Load address of PIC label using the linkage table
7903 if the label is nonlocal.
7905 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7907 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7908 warning message if user requests -maltivec=be.
7909 * doc/invoke.texi: Document deprecation of -maltivec=be.
7911 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7914 * config/rs6000/rs6000-c.c: Update definitions for
7915 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7916 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7918 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7921 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7922 add _get_ssp intrinsics. Remove argument from
7923 __builtin_ia32_rdssp[d|q].
7924 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7925 * config/i386/i386-builtin.def: Remove argument from
7926 __builtin_ia32_rdssp[d|q].
7927 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7928 ix86_expand_special_args_builtin for _rdssp[d|q].
7929 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7930 Clear register before usage.
7931 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7932 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7934 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7936 PR tree-optimization/84357
7937 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7938 operand 1 of an ARRAY_REF too.
7940 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7943 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7944 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7946 (set_of_reg): New struct.
7947 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7948 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7949 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7951 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7952 Split into bitclr, bitset, bitinvert patterns if appropriate.
7953 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7954 use rx_fuse_in_memory_bitop.
7955 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7956 to named insn, correct maximum insn length.
7958 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7961 * machmode.def: Define a complex mode for PARTIAL_INT.
7962 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7964 * doc/rtl.texi: Document CSPImode.
7965 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7967 (msp430_hard_regno_nregs_with_padding): Likewise.
7969 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7972 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7974 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7976 PR rtl-optimization/84169
7977 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7978 we generated a parallel as new i3 and we split that to new i2 and i3
7979 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7980 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7981 those to i2, not i1. Partially rewrite this scan code.
7983 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7986 * stor-layout.c (place_field): For variable length fields, adjust
7987 offset_align afterwards not just based on the field's alignment,
7988 but also on the size.
7991 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7992 of exps and logs in the use_exp2 case.
7994 2018-02-13 Jeff Law <law@redhat.com>
7996 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7999 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8002 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8006 * common.opt (gas-loc-support, gas-locview-support): New.
8007 (ginline-points, ginternal-reset-location-views): New.
8008 * doc/invoke.texi: Document them. Use @itemx where intended.
8009 (gvariable-location-views): Adjust.
8010 * target.def (reset_location_view): New.
8011 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8012 (TARGET_RESET_LOCATION_VIEW): New.
8013 * doc/tm.texi: Rebuilt.
8014 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8015 (dwarf2out_default_as_locview_support): New.
8016 (output_asm_line_debug_info): Use option variables.
8017 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8018 (output_loc_list): Likewise.
8019 (add_high_low_attributes): Check option variables.
8020 Don't output entry view attribute in strict mode.
8021 (gen_inlined_subroutine_die): Check option variables.
8022 (dwarf2out_inline_entry): Likewise.
8023 (init_sections_and_labels): Likewise.
8024 (dwarf2out_early_finish): Likewise.
8025 (maybe_reset_location_view): New, from...
8026 (dwarf2out_var_location): ... here. Call it.
8027 * debug.h (dwarf2out_default_as_loc_support): Declare.
8028 (dwarf2out_default_as_locview_support): Declare.
8029 * hooks.c (hook_int_rtx_insn_0): New.
8030 * hooks.h (hook_int_rtx_insn_0): Declare.
8031 * toplev.c (process_options): Take -gas-loc-support and
8032 -gas-locview-support from dwarf2out. Enable
8033 -gvariable-location-views by default only with locview
8034 assembler support. Enable -ginternal-reset-location-views by
8035 default only if the target defines the corresponding hook.
8036 Enable -ginline-points by default if location views are
8037 enabled; force it disabled if statement frontiers are
8039 * tree-inline.c (expand_call_inline): Check option variables.
8040 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8042 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8044 PR tree-optimization/84321
8045 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8046 handling. Also check whether the anti-range contains any values
8047 that satisfy the mask; switch to a VR_RANGE if not.
8049 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8052 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8054 2018-02-13 Martin Jambor <mjambor@suse.cz>
8057 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8058 of call statements, also set location of a load to a temporary.
8060 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8062 * config/rl78/rl78.c (add_vector_labels): New function.
8063 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8064 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8065 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8066 which checks that no arguments are passed.
8067 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8068 * doc/extend.texi: Documentation for the new attribute.
8070 2018-02-13 Andreas Schwab <schwab@suse.de>
8072 * config/riscv/linux.h (CPP_SPEC): Define.
8074 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8077 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8078 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8079 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8080 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8081 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8082 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8083 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8084 temporarily for AES and PCLMUL builtins.
8086 PR tree-optimization/84339
8087 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8088 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8092 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8093 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8094 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8096 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8098 * omp-simd-clone.h: New file.
8099 * omp-simd-clone.c: Include omp-simd-clone.h.
8100 (expand_simd_clones): No longer static.
8101 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8102 cgraph.h and omp-simd-clone.h.
8103 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8104 (vect_recog_widen_shift_pattern): Formatting fix.
8105 (vect_pattern_recog_1): Don't check optab for calls.
8108 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8109 operands[2] into a REG before using gen_lowpart on it.
8111 2018-02-12 Jeff Law <law@redhat.com>
8114 * config/sh/sh.c (find_barrier): Consider a sibling call
8117 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8118 successfully back substituting a reg.
8120 2018-02-12 Richard Biener <rguenther@suse.de>
8122 PR tree-optimization/84037
8123 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8124 parameter, move visited init to caller.
8125 (vect_slp_analyze_operations): Separate cost from validity
8126 check, initialize visited once for all instances.
8127 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8129 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8131 (vect_model_promotion_demotion_cost): Likewise.
8132 (vectorizable_bswap): Guard cost modeling with !slp_node
8133 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8135 (vectorizable_call): Likewise.
8136 (vectorizable_conversion): Likewise.
8137 (vectorizable_assignment): Likewise.
8138 (vectorizable_shift): Likewise.
8139 (vectorizable_operation): Likewise.
8140 (vectorizable_store): Likewise.
8141 (vectorizable_load): Likewise.
8142 (vectorizable_condition): Likewise.
8143 (vectorizable_comparison): Likewise.
8145 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8148 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8149 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8150 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8152 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8155 * params.def (inline-min-speedup): Increase from 8 to 15.
8156 (max-inline-insns-auto): Decrease from 40 to 30.
8157 * ipa-split.c (consider_split): Add some buffer for function to
8158 be considered inlining candidate.
8159 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8162 2018-02-12 Richard Biener <rguenther@suse.de>
8164 PR tree-optimization/84037
8165 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8166 matched stmts if we cannot swap the non-matched ones.
8168 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8170 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8171 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8172 _mm_maskz_scalef_round_ss): New intrinsics.
8173 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8174 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8175 __builtin_ia32_scalefss_round): Remove.
8176 (__builtin_ia32_scalefsd_mask_round,
8177 __builtin_ia32_scalefss_mask_round): New intrinsics.
8178 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8179 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8180 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8181 "<round_constraint>")): Changed to ...
8182 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8183 "<round_scalar_constraint>")): ... this.
8184 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8185 %0, %1, %2<round_op3>}"): Changed to ...
8186 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8187 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8188 %2<round_scalar_mask_op3>}"): ... this.
8189 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8191 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8193 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8194 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8195 (_mm_maskz_sqrt_round_ss): New intrinsics.
8196 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8197 (__builtin_ia32_sqrtsd_mask_round)
8198 (__builtin_ia32_sqrtss_mask_round): New builtins.
8199 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8200 (__builtin_ia32_sqrtss_round): Remove.
8201 (__builtin_ia32_sqrtsd_mask_round)
8202 (__builtin_ia32_sqrtss_mask_round): New builtins.
8203 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8204 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8205 ((match_operand:VF_128 1 "vector_operand"
8206 "xBm,<round_constraint>")): Changed to ...
8207 ((match_operand:VF_128 1 "vector_operand"
8208 "xBm,<round_scalar_constraint>")): ... this.
8209 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8210 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8211 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8212 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8213 %<iptr>1<round_scalar_mask_op3>}): ... this.
8214 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8215 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8217 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8220 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8221 Cast vec_cmpeq result to correct type.
8222 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8223 Cast vec_cmpgt result to correct type.
8225 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8227 * final.c (final_scan_insn_1): Renamed from...
8228 (final_scan_insn): ... this. New wrapper, to recover
8229 seen from the outermost call in recursive ones.
8230 * config/sparc/sparc.c (output_return): Drop seen from call.
8231 (output_sibcall): Likewise.
8232 * config/visium/visium.c (output_branch): Likewise.
8234 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8236 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8239 2018-02-10 Alan Modra <amodra@gmail.com>
8242 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8243 Specify LR as an input.
8245 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8248 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8249 remove_member_access_dummy_vars): New functions.
8250 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8251 lower_omp_1, execute_lower_omp): Use them.
8253 PR rtl-optimization/84308
8254 * shrink-wrap.c (spread_components): Release todo vector.
8256 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8258 PR rtl-optimization/57193
8259 * ira-color.c (struct allocno_color_data): Add member
8260 conflict_allocno_hard_prefs.
8261 (update_conflict_allocno_hard_prefs): New.
8262 (bucket_allocno_compare_func): Add a preference based on
8263 conflict_allocno_hard_prefs.
8264 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8265 (color_allocnos): Remove a dead code. Initiate
8266 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8268 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8271 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8272 constraint from =wa to wa. Avoid a subreg on the output operand,
8273 instead use a pseudo and subreg it in a move.
8274 (p9_xxbrd_<mode>): Changed to ...
8275 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8276 (p9_xxbrd_v2df): New expander.
8277 (p9_xxbrw_<mode>): Changed to ...
8278 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8279 (p9_xxbrw_v4sf): New expander.
8281 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8283 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8285 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8288 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8289 multiply in 32-bit mode.
8290 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8291 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8294 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8296 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8297 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8298 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8299 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8301 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8304 * dwarf2out.c (is_trivial_indirect_ref): New function.
8305 (dwarf2out_late_global_decl): Do not generate a location
8306 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8307 and that are not defined in the current unit.
8309 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8311 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8312 instead of a libcall for UNORDERED.
8314 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8317 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8318 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8320 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8323 * config/s390/s390.c (s390_set_current_function): Invoke
8324 s390_indirect_branch_settings also if fndecl didn't change.
8326 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8328 * config/rs6000/rs6000.md (blockage): Set length to zero.
8330 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8332 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8334 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8337 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8338 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8342 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8343 PARALLEL incoming that failed vt_get_decl_and_offset check.
8346 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8347 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8349 (get_variable_section): For decls in named .bss* sections pass true as
8350 second argument to bss_initializer_p.
8352 2018-02-09 Marek Polacek <polacek@redhat.com>
8353 Jakub Jelinek <jakub@redhat.com>
8356 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8357 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8358 Sync some changes from cxx_fold_indirect_ref.
8360 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8362 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8364 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8365 (BLOCK_INLINE_ENTRY_LABEL): New.
8366 (dwarf2out_var_location): Disregard inline entry markers.
8367 (inline_entry_data): New struct.
8368 (inline_entry_data_hasher): New hashtable type.
8369 (inline_entry_data_hasher::hash): New.
8370 (inline_entry_data_hasher::equal): New.
8371 (inline_entry_data_table): New variable.
8372 (add_high_low_attributes): Add DW_AT_entry_pc and
8373 DW_AT_GNU_entry_view attributes if a pending entry is found
8374 in inline_entry_data_table. Add old entry_pc attribute only
8375 if debug nonbinding markers are disabled.
8376 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8377 markers are enabled.
8378 (block_within_block_p, dwarf2out_inline_entry): New.
8379 (dwarf2out_finish): Check that no entries remained in
8380 inline_entry_data_table.
8381 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8382 (final_scan_insn, notice_source_line): Likewise.
8383 (rest_of_clean_state): Skip inline entry markers.
8384 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8386 * gimple.c (gimple_build_debug_inline_entry): New.
8387 * gimple.h (enum gimple_debug_subcode): Add
8388 GIMPLE_DEBUG_INLINE_ENTRY.
8389 (gimple_build_debug_inline_entry): Declare.
8390 (gimple_debug_inline_entry_p): New.
8391 (gimple_debug_nonbind_marker_p): Adjust.
8392 * insn-notes.def (INLINE_ENTRY): New.
8393 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8394 inline entry marker notes.
8395 (print_insn): Likewise.
8396 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8397 (INSN_DEBUG_MARKER_KIND): Likewise.
8398 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8399 * tree-inline.c (expand_call_inline): Build and insert
8400 debug_inline_entry stmt.
8401 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8402 inline entry blocks early, if nonbind markers are enabled.
8403 (dump_scope_block): Dump fragment info.
8404 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8405 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8406 (gimple_build_debug_inline_entry): New.
8407 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8408 Enable/disable inline entry points too.
8409 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8410 (DEBUG_INSN): Describe inline entry markers.
8412 * common.opt (gvariable-location-views): New.
8413 (gvariable-location-views=incompat5): New.
8414 * config.in: Rebuilt.
8415 * configure: Rebuilt.
8416 * configure.ac: Test assembler for view support.
8417 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8418 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8419 * dwarf2out.c (var_loc_view): New typedef.
8420 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8421 (dwarf2out_locviews_in_attribute): New.
8422 (dwarf2out_locviews_in_loclist): New.
8423 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8424 (enum dw_line_info_opcode): Add LI_adv_address.
8425 (struct dw_line_info_table): Add view.
8426 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8427 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8428 (zero_view_p): New variable.
8429 (ZERO_VIEW_P): New macro.
8430 (output_asm_line_debug_info): New.
8431 (struct var_loc_node): Add view.
8432 (add_AT_view_list, AT_loc_list): New.
8433 (add_var_loc_to_decl): Add view param. Test it against last.
8434 (new_loc_list): Add view params. Record them.
8435 (AT_loc_list_ptr): Handle loc and view lists.
8436 (view_list_to_loc_list_val_node): New.
8437 (print_dw_val): Handle dw_val_class_view_list.
8438 (size_of_die): Likewise.
8439 (value_format): Likewise.
8440 (loc_list_has_views): New.
8441 (gen_llsym): Set vl_symbol too.
8442 (maybe_gen_llsym, skip_loc_list_entry): New.
8443 (dwarf2out_maybe_output_loclist_view_pair): New.
8444 (output_loc_list): Output view list or entries too.
8445 (output_view_list_offset): New.
8446 (output_die): Handle dw_val_class_view_list.
8447 (output_dwarf_version): New.
8448 (output_compilation_unit_header): Use it.
8449 (output_skeleton_debug_sections): Likewise.
8450 (output_rnglists, output_line_info): Likewise.
8451 (output_pubnames, output_aranges): Update version comments.
8452 (output_one_line_info_table): Output view numbers in asm comments.
8453 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8454 Call maybe_gen_llsym.
8455 (loc_list_from_tree_1): Adjust.
8456 (add_AT_location_description): Create view list attribute if
8457 needed, check it's absent otherwise.
8458 (convert_cfa_to_fb_loc_list): Adjust.
8459 (maybe_emit_file): Call output_asm_line_debug_info for test.
8460 (dwarf2out_var_location): Reset views as needed. Precompute
8461 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8462 attribute. Set view.
8463 (new_line_info_table): Reset next view.
8464 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8465 (dwarf2out_source_line): Likewise. Output view resets and labels to
8466 the assembler, or select appropriate line info opcodes.
8467 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8468 (optimize_string_length): Catch it. Adjust.
8469 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8470 dw_val_class_view_list, and remove it if no longer needed.
8471 (hash_loc_list): Hash view numbers.
8472 (loc_list_hasher::equal): Compare them.
8473 (optimize_location_lists): Check whether a view list symbol is
8474 needed, and whether the locview attribute is present, and
8475 whether they match. Remove the locview attribute if no longer
8477 (index_location_lists): Call skip_loc_list_entry for test.
8478 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8479 Use output_dwarf_version.
8480 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8481 (struct dw_val_node): Add val_view_list.
8482 * final.c (SEEN_NEXT_VIEW): New.
8483 (set_next_view_needed): New.
8484 (clear_next_view_needed): New.
8485 (maybe_output_next_view): New.
8486 (final_start_function): Rename to...
8487 (final_start_function_1): ... this. Take pointer to FIRST,
8488 add SEEN parameter. Emit param bindings in the initial view.
8489 (final_start_function): Reintroduce SEEN-less interface.
8490 (final): Rename to...
8491 (final_1): ... this. Take SEEN parameter. Output final pending
8492 next view at the end.
8493 (final): Reintroduce seen-less interface.
8494 (final_scan_insn): Output pending next view before switching
8495 sections or ending a block. Mark the next view as needed when
8496 outputting variable locations. Notify debug backend of section
8497 changes, and of location view changes.
8498 (rest_of_handle_final): Adjust.
8499 * toplev.c (process_options): Autodetect value for debug variable
8500 location views option. Warn on incompat5 without -gdwarf-5.
8501 * doc/invoke.texi (gvariable-location-views): New.
8502 (gvariable-location-views=incompat5): New.
8503 (gno-variable-location-views): New.
8505 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8507 PR tree-optimization/84136
8508 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8509 that the result of find_edge is non-NULL.
8511 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8514 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8515 storing integer register in SImode. Fix cost of 256 and 512
8516 byte aligned SSE register store.
8518 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8520 * config/i386/i386.c (ix86_multiplication_cost): Fix
8521 multiplication cost for TARGET_AVX512DQ.
8523 2018-02-08 Marek Polacek <polacek@redhat.com>
8525 PR tree-optimization/84238
8526 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8529 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8531 PR tree-optimization/84265
8532 * tree-vect-stmts.c (vectorizable_store): Don't treat
8533 VMAT_CONTIGUOUS accesses as grouped.
8534 (vectorizable_load): Likewise.
8536 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8538 PR tree-optimization/81635
8539 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8540 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8541 (test_round_for_mask): New functions.
8542 (wide_int_cc_tests): Call test_round_for_mask.
8543 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8544 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8545 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8546 range returned by get_range_info.
8548 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8551 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8552 * symtab.c: Include builtins.h
8553 (symtab_node::output_to_lto_symbol_table_p): Move here
8554 from lto-streamer-out.c:output_symbol_p.
8555 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8556 (output_symbol_p): Move all logic to symtab.c
8557 (produce_symtab): Update.
8559 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8561 * config/s390/s390-opts.h (enum indirect_branch): Define.
8562 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8563 (s390_indirect_branch_via_thunk)
8564 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8565 (enum s390_indirect_branch_type): Define.
8566 * config/s390/s390.c (struct s390_frame_layout, struct
8567 machine_function): Remove.
8568 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8569 (indirect_branch_table_label_no, indirect_branch_table_name):
8571 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8572 (enum s390_indirect_branch_option): Define.
8573 (s390_return_addr_from_memory): New function.
8574 (s390_handle_string_attribute): New function.
8575 (s390_attribute_table): Add new attribute handler.
8576 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8577 (s390_indirect_branch_via_thunk): New function.
8578 (s390_indirect_branch_via_inline_thunk): New function.
8579 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8580 sibling call optimization for non z10 compiles.
8581 (s390_emit_call): Force indirect branch target to be a single
8582 register. Add r1 clobber for non-z10 compiles.
8583 (s390_emit_epilogue): Emit return jump via return_use expander.
8584 (s390_reorg): Handle JUMP_INSNs as execute targets.
8585 (s390_option_override_internal): Perform validity checks for the
8586 new command line options.
8587 (s390_indirect_branch_attrvalue): New function.
8588 (s390_indirect_branch_settings): New function.
8589 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8590 (s390_output_indirect_thunk_function): New function.
8591 (s390_code_end): Implement target hook.
8592 (s390_case_values_threshold): Implement target hook.
8593 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8595 * config/s390/s390.h (struct s390_frame_layout)
8596 (struct machine_function): Move here from s390.c.
8597 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8598 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8599 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8600 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8601 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8602 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8603 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8604 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8605 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8606 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8607 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8608 (mnemonic attribute): Add values which aren't recognized
8610 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8611 pattern for branch conversion. Fix mnemonic attribute.
8612 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8613 indirect branch via thunk if requested.
8614 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8615 ("*indirect_jump"): Disable for branch conversion using out of
8617 ("indirect_jump_via_thunk<mode>_z10")
8618 ("indirect_jump_via_thunk<mode>")
8619 ("indirect_jump_via_inlinethunk<mode>_z10")
8620 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8621 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8622 ("casesi_jump_via_inlinethunk<mode>_z10")
8623 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8624 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8625 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8626 ("*indirect2_jump"): Disable for branch conversion.
8627 ("casesi_jump"): Turn into expander and expand patterns for branch
8629 ("return_use"): New expander.
8630 ("*return"): Emit return via thunk and rename it to ...
8631 ("*return<mode>"): ... this one.
8632 * config/s390/s390.opt: Add new options and and enum for the
8635 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8637 * lra-constraints.c (match_reload): Unconditionally use
8638 gen_lowpart_SUBREG, rather than selecting between that
8639 and equivalent gen_rtx_SUBREG code.
8641 2018-02-08 Richard Biener <rguenther@suse.de>
8643 PR tree-optimization/84233
8644 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8645 changed flag instead of boguously re-using phi_inserted.
8647 2018-02-08 Martin Jambor <mjambor@suse.cz>
8649 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8650 static local variables.
8652 2018-02-08 Richard Biener <rguenther@suse.de>
8654 PR tree-optimization/84278
8655 * tree-vect-stmts.c (vectorizable_store): When looking for
8656 smaller vector types to perform grouped strided loads/stores
8657 make sure the mode is supported by the target.
8658 (vectorizable_load): Likewise.
8660 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8662 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8663 Increase LDP/STP opportunities by adding adjacent callee-saves.
8665 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8667 PR rtl-optimization/84068
8668 PR rtl-optimization/83459
8669 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8671 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8673 PR tree-optimization/84224
8674 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8675 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8678 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8681 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8682 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8683 count, remove one USE.
8685 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8687 * doc/install.texi (Configuration): Document the
8688 --with-long-double-format={ibm,ieee} PowerPC configuration
8692 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8693 Convert from define_expand to be define_insn_and_split. Rework
8694 float/double/_Float128 conversions to QI/HI/SImode to work with
8695 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8696 conversions to QI/HImode types did a store and then a load to
8697 truncate the value. For conversions to VSX registers, don't split
8698 the insn, instead emit the code directly. Use the code iterator
8699 any_fix to combine signed and unsigned conversions.
8700 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8701 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8702 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8703 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8704 (fix_<mode>di2_hw): Likewise.
8705 (fixuns_<mode>di2_hw): Likewise.
8706 (fix_<mode>si2_hw): Likewise.
8707 (fixuns_<mode>si2_hw): Likewise.
8708 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8709 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8710 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8711 fix<uns>_trunc<SFDF:mode>si2_p8.
8712 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8714 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8715 (fix<uns>_<mode>_mem): Likewise.
8716 (fctiw<u>z_<mode>_mem): Likewise.
8717 (fix<uns>_<mode>_mem): Likewise.
8718 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8719 the register allocator from doing a direct move to the GPRs to do
8720 a store, and instead use the ISA 3.0 store byte/half-word from
8721 vector register instruction. For IEEE 128-bit floating point,
8722 also optimize stores of 32-bit ints.
8723 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8725 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8727 * genextract.c (push_pathstr_operand): New function to support
8729 (walk_rtx): Call push_pathstr_operand.
8730 (print_path): Support [a-zA-Z].
8732 2018-02-07 Richard Biener <rguenther@suse.de>
8734 PR tree-optimization/84037
8735 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8736 (cse_and_gimplify_to_preheader): Declare.
8737 (vect_get_place_in_interleaving_chain): Likewise.
8738 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8740 (_loop_vec_info::~_loop_vec_info): Delete it.
8741 (cse_and_gimplify_to_preheader): New function.
8742 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8743 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8744 (vectorizable_load): Likewise. For grouped stores always base
8745 the IV on the first element.
8746 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8747 condition before gimplifying.
8749 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8751 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8752 *DIV_EXPR and *MOD_EXPR.
8754 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8757 * config/i386/i386.c (ix86_option_override_internal): Mask out
8758 the CF_SET bit when checking -fcf-protection.
8760 2018-02-07 Tom de Vries <tom@codesourcery.com>
8763 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8766 2018-02-07 Richard Biener <rguenther@suse.de>
8768 PR tree-optimization/84204
8769 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8772 PR tree-optimization/84205
8773 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8774 special-case isl_ast_op_zdiv_r.
8776 PR tree-optimization/84223
8777 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8778 Only add conditions from within the region.
8779 (gather_bbs::after_dom_children): Adjust.
8781 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8784 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8785 * config/avr/avr.md: Only post-reload split REG-REG moves if
8786 either register is GENERAL_REG_P.
8788 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8790 PR tree-optimization/84235
8791 * tree-ssa-scopedtables.c
8792 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8793 if the subtraction is performed in floating point type where NaNs are
8794 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8795 build 1. Formatting fix.
8797 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8800 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8801 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8802 and skip it regardless of bb boundaries. Use CALL_P macro,
8803 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8804 unnecessarily, formatting fix.
8806 2018-02-06 Michael Collison <michael.collison@arm.com>
8808 * config/arm/thumb2.md:
8809 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8810 (*thumb_mov_notscc): Ditto.
8812 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8815 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8816 unsigned_fix, not "s".
8818 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8820 * configure.ac (gcc_fn_eh_frame_ro): New function.
8821 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8822 correct .eh_frame permissions.
8823 * configure: Regenerate.
8825 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8827 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8830 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8832 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8833 Display warning message for -mno-speculate-indirect-jumps.
8835 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8837 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8839 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8841 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8843 PR tree-optimization/84225
8844 * tree-eh.c (find_trapping_overflow): Only call
8845 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8847 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8850 * config/i386/i386.c: Reimplement the check of possible options
8851 -mibt/-mshstk conbination. Change error messages.
8852 * doc/invoke.texi: Fix a typo: remove extra '='.
8854 2018-02-06 Marek Polacek <polacek@redhat.com>
8856 PR tree-optimization/84228
8857 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8859 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8862 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8863 emitted arch directives.
8864 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8865 __ARM_FEATURE_COPROC before changing architectures.
8867 2018-02-06 Richard Biener <rguenther@suse.de>
8869 * config/i386/i386.c (print_reg): Fix typo.
8870 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8872 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8874 * configure: Regenerate.
8876 2018-02-05 Martin Sebor <msebor@redhat.com>
8878 PR tree-optimization/83369
8879 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8882 2018-02-05 Martin Liska <mliska@suse.cz>
8884 * doc/invoke.texi: Cherry-pick upstream r323995.
8886 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8888 * ira.c (ira_init_register_move_cost): Adjust comment.
8890 2018-02-05 Martin Liska <mliska@suse.cz>
8892 PR gcov-profile/84137
8893 * doc/gcov.texi: Fix typo in documentation.
8895 2018-02-05 Martin Liska <mliska@suse.cz>
8897 PR gcov-profile/83879
8898 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8899 using dlopen functionality.
8901 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8903 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8904 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8905 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8906 _mm_maskz_range_round_ss): New intrinsics.
8907 (__builtin_ia32_rangesd128_round)
8908 (__builtin_ia32_rangess128_round): Remove.
8909 (__builtin_ia32_rangesd128_mask_round,
8910 __builtin_ia32_rangess128_mask_round): New builtins.
8911 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8912 __builtin_ia32_rangess128_round): Remove.
8913 (__builtin_ia32_rangesd128_mask_round,
8914 __builtin_ia32_rangess128_mask_round): New builtins.
8915 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8916 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8917 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8918 "<round_saeonly_constraint>")): Changed to ...
8919 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8920 "<round_saeonly_scalar_constraint>")): ... this.
8921 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8922 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8923 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8924 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8925 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8927 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8929 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8931 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8932 Remove all values except native, 8540 and 8548.
8934 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8936 * config/i386/i386.c (ix86_output_function_return): Pass
8937 INVALID_REGNUM, instead of -1, as invalid register number to
8938 indirect_thunk_name and output_indirect_thunk.
8940 2018-02-02 Julia Koval <julia.koval@intel.com>
8942 * config.gcc: Add -march=icelake.
8943 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8944 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8945 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8946 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8947 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8948 (processor_target_table): Add icelake.
8949 (ix86_option_override_internal): Handle new PTAs.
8950 (get_builtin_code_for_version): Handle icelake.
8951 (M_INTEL_COREI7_ICELAKE): New.
8952 (fold_builtin_cpu): Handle icelake.
8953 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8954 * doc/invoke.texi: Add -march=icelake.
8956 2018-02-02 Julia Koval <julia.koval@intel.com>
8958 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8959 to wide_int_bitmask.
8960 * wide-int-bitmask.h: New.
8962 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8965 * config/i386/i386.md: Replace Pmode with word_mode in
8966 builtin_setjmp_setup and builtin_longjmp to support x32.
8968 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8972 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8974 (rs6000_supported_cpu_names): New static variable.
8975 (linux_cpu_translation_table): Likewise.
8976 (elf_platform) <cpu>: Define new static variable and use it.
8977 Translate kernel AT_PLATFORM name to canonical name if needed.
8978 Error if platform name is unknown.
8980 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8983 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8985 2018-02-01 Jeff Law <law@redhat.com>
8988 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8989 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8990 the scratch if RELEASE_VIA_POP is false.
8991 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8992 If we have to save a temporary register, decrement SIZE appropriately.
8993 Pass new arguments to release_scratch_register_on_entry.
8994 (ix86_adjust_stack_and_probe): Likewise.
8995 (ix86_emit_probe_stack_range): Pass new arguments to
8996 release_scratch_register_on_entry.
8998 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9000 PR rtl-optimization/84157
9001 * combine.c (change_zero_ext): Use REG_P predicate in
9002 front of HARD_REGISTER_P predicate.
9004 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9006 * config/avr/avr.c (avr_option_override): Move disabling of
9007 -fdelete-null-pointer-checks to...
9008 * common/config/avr/avr-common.c (avr_option_optimization_table):
9011 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9013 PR tree-optimization/81635
9014 * tree-data-ref.c (split_constant_offset_1): For types that
9015 wrap on overflow, try to use range info to prove that wrapping
9018 2018-02-01 Renlin Li <renlin.li@arm.com>
9021 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9023 (aarch64_register_move_cost): Likewise.
9024 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9026 (REG_CLASS_NAMES): Likewise.
9027 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9028 TAILCALL_ADDR_REGS. Remove IP registers.
9029 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9031 2018-02-01 Richard Biener <rguenther@suse.de>
9033 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9034 for specifying RPO order and allow NULL for that.
9035 * domwalk.c (dom_walker::dom_walker): Likewise.
9036 (dom_walker::walk): Handle NULL RPO order.
9037 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9039 (rewrite_update_dom_walker): Likewise.
9040 (mark_def_dom_walker): Likewise.
9042 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9044 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9045 (aarch64_maybe_expand_sve_subreg_move): Declare.
9046 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9047 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9049 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9050 that are semantically a reverse operation.
9051 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9052 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9053 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9055 (aarch64_can_change_mode_class): For big-endian, forbid changes
9056 between two SVE modes if they have different element sizes.
9058 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9060 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9061 the TImode handling for big-endian targets.
9063 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9065 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9066 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9068 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9069 Remove BSWAP handing for big-endian targets and use the form of
9070 LD1RQ appropariate for the mode.
9072 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9074 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9075 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9078 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9081 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9082 check for operands that need to go through aarch64_sve_reload_be.
9084 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9086 PR tree-optimization/81661
9087 PR tree-optimization/84117
9088 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9089 * tree-eh.c: Include gimplify.h.
9090 (find_trapping_overflow, replace_trapping_overflow,
9091 rewrite_to_non_trapping_overflow): New functions.
9092 * tree-vect-loop.c: Include tree-eh.h.
9093 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9094 * tree-data-ref.c: Include tree-eh.h.
9095 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9097 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9099 PR rtl-optimization/84123
9100 * combine.c (change_zero_ext): Check if hard register satisfies
9101 can_change_dest_mode before calling gen_lowpart_SUBREG.
9103 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9106 * ira.c (ira_init_register_move_cost): Remove assert.
9108 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9110 PR rtl-optimization/84071
9111 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9112 * doc/tm.texi: Regenerate.
9114 2018-01-31 Richard Biener <rguenther@suse.de>
9116 PR tree-optimization/84132
9117 * tree-data-ref.c (analyze_miv_subscript): Properly
9118 check whether evolution_function_is_affine_multivariate_p
9119 before calling gcd_of_steps_may_divide_p.
9121 2018-01-31 Julia Koval <julia.koval@intel.com>
9124 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9125 * config/i386/i386.md (rdpid_rex64) New.
9126 (rdpid): Make 32bit only.
9128 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9131 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9132 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9134 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9137 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9139 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9141 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9143 PR rtl-optimization/84071
9144 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9145 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9147 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9149 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9150 (arc_attribute_table): Add 'aux' attribute.
9151 (arc_in_small_data_p): Consider aux like variables.
9152 (arc_is_aux_reg_p): New function.
9153 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9154 (arc_get_aux_arg): New function.
9155 (prepare_move_operands): Handle aux-register access.
9156 (arc_handle_aux_attribute): New function.
9157 * doc/extend.texi (ARC Variable attributes): Add subsection.
9159 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9161 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9162 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9163 (arc_attribute_table): Add 'uncached' attribute.
9164 (arc_print_operand): Print '.di' flag for uncached memory
9166 (arc_in_small_data_p): Do not consider for small data the uncached
9168 (arc_is_uncached_mem_p): New function.
9169 * config/arc/predicates.md (compact_store_memory_operand): Check
9170 for uncached memory accesses.
9171 (nonvol_nonimm_operand): Likewise.
9172 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9174 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9177 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9178 falign-loops=): Add Optimization flag.
9180 2018-01-30 Jeff Law <law@redhat.com>
9183 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9184 INT_REGISTERS_SAVED. Check it prior to calling
9185 get_scratch_register_on_entry.
9186 (ix86_adjust_stack_and_probe): Similarly.
9187 (ix86_emit_probe_stack_range): Similarly.
9188 (ix86_expand_prologue): Corresponding changes.
9190 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9193 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9194 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9196 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9199 * lra-constraints.c (curr_insn_transform): Process AND in the
9202 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9204 PR rtl-optimization/83986
9205 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9206 dependence against last_pending_memory_flush in addition to
9209 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9211 PR tree-optimization/81611
9212 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9215 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9218 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9221 2018-01-30 Richard Biener <rguenther@suse.de>
9222 Jakub Jelinek <jakub@redhat.com>
9224 PR tree-optimization/84111
9225 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9226 inner loops added during recursion, as they don't have up-to-date
9229 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9232 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9233 (can_inline_edge_by_limits_p): ... here.
9234 (can_early_inline_edge_p, check_callers,
9235 update_caller_keys, update_callee_keys, recursive_inlining,
9236 add_new_edges_to_heap, speculation_useful_p,
9237 inline_small_functions,
9238 inline_small_functions, flatten_function,
9239 inline_to_all_callers_1): Update.
9241 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9243 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9246 2018-01-30 Richard Biener <rguenther@suse.de>
9248 PR tree-optimization/83008
9249 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9250 invariant and constant vector uses in stmts when they need
9253 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9256 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9257 * configure: Regenerate.
9259 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9261 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9263 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9264 Use gen_rtx_REG rather than gen_lowpart.
9266 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9268 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9269 rather than 0 when creating partial subregs.
9271 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9273 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9276 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9279 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9280 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9281 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9282 flags. This restores the settings used before the 2017-07-24.
9283 Turning off pre increment/decrement/modify allows IVOPTS to
9284 optimize DF/SF loops where the index is an int.
9286 2018-01-29 Richard Biener <rguenther@suse.de>
9287 Kelvin Nilsen <kelvin@gcc.gnu.org>
9290 * tree-vect-stmts.c (vectorizable_call): Don't call
9291 targetm.vectorize_builtin_md_vectorized_function if callee is
9294 2018-01-22 Carl Love <cel@us.ibm.com>
9296 * doc/extend.tex: Fix typo in second arg in
9297 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9299 2018-01-29 Richard Biener <rguenther@suse.de>
9301 PR tree-optimization/84086
9302 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9303 (flush_ssaname_freelist): When SSA names were released reset
9304 the SCEV hash table.
9306 2018-01-29 Richard Biener <rguenther@suse.de>
9308 PR tree-optimization/84057
9309 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9310 removed paths when removing edges.
9312 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9314 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9315 -mfunction-return=@var{choice}.
9317 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9320 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9322 (layout::print_source_line): Likewise.
9323 (test_get_line_width_without_trailing_whitespace): Add test cases.
9325 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9328 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9331 2018-01-26 Jim Wilson <jimw@sifive.com>
9333 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9335 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9338 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9340 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9341 and CMP + SUB-immediate -> SUBS.
9343 2018-01-26 Martin Sebor <msebor@redhat.com>
9345 PR tree-optimization/83896
9346 * tree-ssa-strlen.c (get_string_len): Rename...
9347 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9348 Avoid assuming length is constant.
9349 (handle_char_store): Use HOST_WIDE_INT for string length.
9351 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9354 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9355 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9357 2018-01-26 Richard Biener <rguenther@suse.de>
9359 PR rtl-optimization/84003
9360 * dse.c (record_store): Only record redundant stores when
9361 the earlier store aliases at least all accesses the later one does.
9363 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9365 PR rtl-optimization/83985
9366 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9367 REG_CFA_RESTORE insns.
9368 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9369 REG_CFA_RESTORE insns here.
9372 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9373 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9375 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9377 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9378 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9379 (arc_init): Likewise.
9380 (arc_override_options): Likewise.
9381 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9383 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9385 * config/arc/arc.h (TARGET_DBNZ): Define.
9386 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9387 properly set the tune attribute.
9388 (dbnz): Use TARGET_DBNZ guard.
9389 * config/arc/arc.opt (mtune): Add core3 option.
9391 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9393 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9394 recognize new pic like addresses.
9395 (arc_delegitimize_address): Clean up.
9397 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9399 * config/arc/arc-arches.def: Option mrf16 valid for all
9401 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9402 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9403 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9404 * config/arc/arc-tables.opt: Regenerate.
9405 * config/arc/arc.c (arc_conditional_register_usage): Handle
9406 reduced register file case.
9407 (arc_file_start): Set must have build attributes.
9408 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9410 * config/arc/arc.opt (mrf16): Add new option.
9411 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9412 * config/arc/genmultilib.awk: Handle new mrf16 option.
9413 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9414 * config/arc/t-multilib: Regenerate.
9415 * doc/invoke.texi (ARC Options): Document mrf16 option.
9417 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9419 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9420 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9421 (arc_attribute_table): Add 'secure_call' attribute.
9422 (arc_print_operand): Print secure call operand.
9423 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9425 (arc_is_secure_call_p): New function. * config/arc/arc.md
9426 (call_i): Add support for sjli instruction.
9427 (call_value_i): Likewise.
9428 * config/arc/constraints.md (Csc): New constraint.
9430 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9431 John Eric Martin <John.Martin@emmicro-us.com>
9433 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9434 * config/arc/arc.c (_arc_jli_section): New struct.
9435 (arc_jli_section): New type.
9436 (rc_jli_sections): New static variable.
9437 (arc_handle_jli_attribute): New function.
9438 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9439 (arc_file_end): New function.
9440 (TARGET_ASM_FILE_END): Define.
9441 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9442 (arc_add_jli_section): New function.
9443 (jli_call_scan): Likewise.
9444 (arc_reorg): Call jli_call_scan.
9445 (arc_output_addsi): Remove 'S' from printing asm operand.
9446 (arc_is_jli_call_p): New function.
9447 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9449 (movhi_insn): Likewise.
9450 (movsi_insn): Likewise.
9451 (movsi_set_cc_insn): Likewise.
9452 (loadqi_update): Likewise.
9453 (load_zeroextendqisi_update): Likewise.
9454 (load_signextendqisi_update): Likewise.
9455 (loadhi_update): Likewise.
9456 (load_zeroextendhisi_update): Likewise.
9457 (load_signextendhisi_update): Likewise.
9458 (loadsi_update): Likewise.
9459 (loadsf_update): Likewise.
9460 (movsicc_insn): Likewise.
9461 (bset_insn): Likewise.
9462 (bxor_insn): Likewise.
9463 (bclr_insn): Likewise.
9464 (bmsk_insn): Likewise.
9465 (bicsi3_insn): Likewise.
9466 (cmpsi_cc_c_insn): Likewise.
9467 (movsi_ne): Likewise.
9468 (movsi_cond_exec): Likewise.
9469 (clrsbsi2): Likewise.
9480 (call_i): Remove 'S' asm letter, add jli instruction.
9481 (call_value_i): Likewise.
9482 * config/arc/arc.op (mjli-always): New option.
9483 * config/arc/constraints.md (Cji): New constraint.
9484 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9486 (subsf3_fpx): Likewise.
9487 (mulsf3_fpx): Likewise.
9488 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9490 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9492 * doc/invoke.texi (ARC): Document mjli-always option.
9494 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9496 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9497 avoid addition with 0 and use incw and decw where possible.
9499 2018-01-26 Richard Biener <rguenther@suse.de>
9501 PR tree-optimization/81082
9502 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9503 association if it requires casting to unsigned.
9504 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9505 from fold_plusminus_mult_expr to catch important cases late when
9506 range info is available.
9508 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9510 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9511 * configure.ac (hidden_linkonce): New test.
9512 * configure: Regenerate.
9513 * config.in: Regenerate.
9515 2018-01-26 Julia Koval <julia.koval@intel.com>
9517 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9518 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9519 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9520 _mm_mask_bitshuffle_epi64_mask): Fix type.
9521 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9522 USI_FTYPE_V4DI_V4DI_USI): Remove.
9523 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9524 __builtin_ia32_vpshufbitqmb256_mask,
9525 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9526 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9527 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9529 2018-01-26 Alan Modra <amodra@gmail.com>
9532 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9533 UNSPEC_VBPERMQ. Sort other unspecs.
9535 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9537 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9539 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9542 * predict.c (drop_profile): Do not push/pop cfun; update also
9544 (handle_missing_profiles): Fix logic looking for zero profiles.
9546 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9549 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9550 on functions with #pragma omp declare simd or functions with simd
9552 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9553 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9554 Remove trailing \n from warning_at calls.
9556 2018-01-25 Tom de Vries <tom@codesourcery.com>
9559 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9560 for neutered workers.
9562 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9565 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9566 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9568 2018-01-24 Jeff Law <law@redhat.com>
9571 * i386.c (get_probe_interval): Move to earlier point.
9572 (ix86_compute_frame_layout): If -fstack-clash-protection and
9573 the frame is larger than the probe interval, then use pushes
9574 to save registers rather than reg->mem moves.
9575 (ix86_expand_prologue): Remove conditional for int_registers_saved
9578 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9581 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9582 min/max for never referenced object.
9584 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9587 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9589 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9590 attributes from DECL_ATTRIBUTES (decl) without affecting
9591 DECL_ATTRIBUTES (current_function_decl).
9592 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9593 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9595 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9597 PR tree-optimization/83979
9598 * fold-const.c (fold_comparison): Use constant_boolean_node
9599 instead of boolean_{true,false}_node.
9601 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9603 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9606 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9608 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9609 Simplify the clause that sets the length attribute.
9610 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9611 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9612 clause that sets the length attribute.
9613 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9615 2018-01-24 Tom de Vries <tom@codesourcery.com>
9618 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9619 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9620 Add strict parameter.
9621 (prevent_branch_around_nothing): Insert dummy insn between branch to
9622 label and label with no ptx insn inbetween.
9623 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9625 2018-01-24 Tom de Vries <tom@codesourcery.com>
9628 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9629 for neutered threads in warp.
9630 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9632 2018-01-24 Richard Biener <rguenther@suse.de>
9634 PR tree-optimization/83176
9635 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9638 2018-01-24 Richard Biener <rguenther@suse.de>
9640 PR tree-optimization/82819
9641 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9642 code generating pluses that are no-ops in the target precision.
9644 2018-01-24 Richard Biener <rguenther@suse.de>
9647 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9649 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9651 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9652 to merge probabilities.
9653 * predict.c (probably_never_executed): Also mark as cold functions
9654 with global 0 profile and guessed local profile.
9655 * profile-count.c (profile_probability::combine_with_count): New
9657 * profile-count.h (profile_probability::operator*,
9658 profile_probability::operator*=, profile_probability::operator/,
9659 profile_probability::operator/=): Reduce precision to adjusted
9660 and set value to guessed on contradictory divisions.
9661 (profile_probability::combine_with_freq): Remove.
9662 (profile_probability::combine_wiht_count): Declare.
9663 (profile_count::force_nonzero):: Set to adjusted.
9664 (profile_count::probability_in):: Set quality to adjusted.
9665 * tree-ssa-tail-merge.c (replace_block_by): Use
9668 2018-01-23 Andrew Waterman <andrew@sifive.com>
9669 Jim Wilson <jimw@sifive.com>
9671 * config/riscv/riscv.c (riscv_stack_boundary): New.
9672 (riscv_option_override): Set riscv_stack_boundary. Handle
9673 riscv_preferred_stack_boundary_arg.
9674 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9675 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9676 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9677 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9678 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9679 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9681 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9684 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9685 of struct ix86_frame.
9686 (ix86_expand_epilogue): Likewise. Add a local variable for
9687 the reg_save_offset field in struct ix86_frame.
9689 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9691 PR tree-optimization/82604
9692 * tree-loop-distribution.c (enum partition_kind): New enum item
9693 PKIND_PARTIAL_MEMSET.
9694 (partition_builtin_p): Support above new enum item.
9695 (generate_code_for_partition): Ditto.
9696 (compute_access_range): Differentiate cases that equality can be
9697 proven at all loops, the innermost loops or no loops.
9698 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9699 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9700 (finalize_partitions, distribute_loop): Don't fuse partition of
9701 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9702 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9705 2018-01-23 Martin Liska <mliska@suse.cz>
9707 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9708 order to ignore the predictor.
9709 (PRED_POLYMORPHIC_CALL): Likewise.
9710 (PRED_RECURSIVE_CALL): Likewise.
9712 2018-01-23 Martin Liska <mliska@suse.cz>
9714 * tree-profile.c (tree_profiling): Print function header to
9715 aware reader which function we are working on.
9716 * value-prof.c (gimple_find_values_to_profile): Do not print
9717 not interesting value histograms.
9719 2018-01-23 Martin Liska <mliska@suse.cz>
9721 * profile-count.h (enum profile_quality): Add
9722 profile_uninitialized as the first value. Do not number values
9723 as they are zero based.
9724 (profile_count::verify): Update sanity check.
9725 (profile_probability::verify): Likewise.
9727 2018-01-23 Nathan Sidwell <nathan@acm.org>
9729 * doc/invoke.texi (ffor-scope): Deprecate.
9731 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9733 PR tree-optimization/83510
9734 * domwalk.c (set_all_edges_as_executable): New function.
9735 (dom_walker::dom_walker): Convert bool param
9736 "skip_unreachable_blocks" to enum reachability. Move setup of
9737 edge flags to set_all_edges_as_executable and only do it when
9738 reachability is REACHABLE_BLOCKS.
9739 * domwalk.h (enum dom_walker::reachability): New enum.
9740 (dom_walker::dom_walker): Convert bool param
9741 "skip_unreachable_blocks" to enum reachability.
9742 (set_all_edges_as_executable): New decl.
9743 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9744 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9746 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9747 but converting true to REACHABLE_BLOCKS.
9748 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9750 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9751 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9752 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9754 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9755 if check_all_array_refs will be called.
9757 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9759 * tree.c (selftest::test_location_wrappers): Add more test
9762 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9764 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9765 (selftest::test_bit_in_range): Likewise.
9767 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9770 * doc/sourcebuild.texi (vect_float): Say that the selector
9771 only describes the situation when -funsafe-math-optimizations is on.
9772 (vect_float_strict): Document.
9774 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9776 PR tree-optimization/83965
9777 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9778 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9779 instead of checking only for a reduction.
9780 (vect_recog_widen_sum_pattern): Likewise.
9782 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9784 * predict.c (probably_never_executed): Only use precise profile info.
9785 (compute_function_frequency): Skip after inlining hack since we now
9786 have quality checking.
9788 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9790 * profile-count.h (profile_probability::very_unlikely,
9791 profile_probability::unlikely, profile_probability::even): Set
9792 precision to guessed.
9794 2018-01-23 Richard Biener <rguenther@suse.de>
9796 PR tree-optimization/83963
9797 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9798 Properly terminate dominator walk when crossing the exit edge not
9799 when visiting its source block.
9801 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9804 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9805 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9807 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9809 PR tree-optimization/83957
9810 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9811 semicolon after for body surrounded by braces.
9813 PR tree-optimization/83081
9814 * profile-count.h (profile_probability::split): New method.
9815 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9816 Use profile_probability::split.
9817 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9818 when splitting a single conditional jump into 2.
9820 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9822 PR tree-optimization/69452
9823 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9826 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9828 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9829 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9830 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9832 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9834 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9835 * config/rl78/rl78.md (movdi): New define_expand.
9836 * config/rl78/rl78.c (rl78_split_movdi): New function.
9838 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9841 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9843 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9844 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9845 128-bit to produce an UNSPEC move to get the double word with the
9846 signbit and then a shift directly to do signbit.
9847 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9848 implementation with a new version that just does either a direct
9849 move or a regular move. Move memory interface to separate insns.
9850 Move insns so they are next to the expander.
9851 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9852 with signbit move. Split big and little endian case.
9853 (signbit<mode>2_dm_mem_le): Likewise.
9854 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9855 (signbit<mode>2_dm2): Likewise.
9857 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9859 * config/rl78/rl78.md (anddi3): New define_expand.
9861 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9863 * config/rl78/rl78.md (umindi3): New define_expand.
9865 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9867 * config/rl78/rl78.md (smindi3): New define_expand.
9869 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9871 * config/rl78/rl78.md (smaxdi3): New define_expand.
9873 2018-01-22 Carl Love <cel@us.ibm.com>
9875 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9876 LVX_V1TI): Add macro expansion.
9877 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9878 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9879 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9880 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9881 Change check to determine if the instruction is a byte reversing
9882 entry. Fix typo in comment.
9883 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9884 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9885 Add def_builtin calls for new builtins.
9886 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9887 Add define_insn expansion.
9889 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9891 * config/rl78/rl78.md (umaxdi3): New define_expand.
9893 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9895 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9896 for non-QImode registers.
9898 2018-01-22 Richard Biener <rguenther@suse.de>
9900 PR tree-optimization/83963
9901 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9902 including the loop exit block.
9903 (scop_detection::merge_sese): Likewise.
9904 (scop_detection::add_scop): Do it here instead.
9906 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9908 * doc/sourcebuild.texi (arm_softfloat): Document.
9910 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9913 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9914 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9915 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9917 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9918 David Edelsohn <dje.gcc@gmail.com>
9921 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9922 Change "crset eq" to "crset 2".
9923 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9924 (*call_indirect_aix<mode>_nospec): Likewise.
9925 (*call_value_indirect_aix<mode>_nospec): Likewise.
9926 (*call_indirect_elfv2<mode>_nospec): Likewise.
9927 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9928 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9929 change assembly output from . to $.
9930 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9931 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9932 (*tablejump<mode>_internal1_nospec): Likewise.
9934 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9937 * config/sh/sh_optimize_sett_clrt.cc:
9938 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9940 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9942 PR tree-optimization/83940
9943 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9944 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9945 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9946 use it to pass back the definition type.
9947 (vect_check_store_rhs): Likewise rhs_dt_out.
9948 (vect_build_gather_load_calls): Add a mask_dt argument and use
9949 it instead of a call to vect_is_simple_use.
9950 (vectorizable_store): Update calls to vect_check_load_store_mask
9951 and vect_check_store_rhs. Use the dt returned by the latter instead
9952 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9953 instead of calls to vect_is_simple_use. Pass the scalar rather
9954 than the vector operand to vect_is_simple_use when handling
9955 second and subsequent copies of an rhs value.
9956 (vectorizable_load): Update calls to vect_check_load_store_mask
9957 and vect_build_gather_load_calls. Use the cached mask_dt and
9958 gs_info.offset_dt instead of calls to vect_is_simple_use.
9960 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9963 * tree-emutls.c: Include gimplify.h.
9964 (lower_emutls_2): New function.
9965 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9966 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9967 it before further processing.
9970 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9971 UINTVAL (trueop1) instead of INTVAL (op1).
9973 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9977 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9978 INCOMING_FRAME_SP_OFFSET if not defined.
9979 (scan_trace): Add ENTRY argument. If true and
9980 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9981 emit a note to adjust the CFA offset.
9982 (create_cfi_notes): Adjust scan_trace callers.
9983 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9984 INCOMING_FRAME_SP_OFFSET in the CIE.
9985 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9986 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9988 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9989 * doc/tm.texi: Regenerated.
9991 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9993 PR rtl-optimization/83147
9994 * lra-constraints.c (remove_inheritance_pseudos): Use
9995 lra_substitute_pseudo_within_insn.
9997 2018-01-19 Tom de Vries <tom@codesourcery.com>
9998 Cesar Philippidis <cesar@codesourcery.com>
10001 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10003 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10006 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10007 spaces for function labels.
10009 2018-01-19 Martin Liska <mliska@suse.cz>
10011 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10012 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10013 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10014 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10015 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10016 (PRED_CONST_RETURN): Change from 69 to 65.
10017 (PRED_NULL_RETURN): Change from 91 to 71.
10018 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10019 (PRED_LOOP_GUARD): Change from 66 to 73.
10021 2018-01-19 Martin Liska <mliska@suse.cz>
10023 * predict.c (predict_insn_def): Add new assert.
10024 (struct branch_predictor): Change type to signed integer.
10025 (test_prediction_value_range): Amend test to cover
10026 PROB_UNINITIALIZED.
10027 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10028 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10029 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10030 (PRED_LOOP_IV_COMPARE): Likewise.
10031 * predict.h (PROB_UNINITIALIZED): Define new constant.
10033 2018-01-19 Martin Liska <mliska@suse.cz>
10035 * predict.c (dump_prediction): Add new format for
10036 analyze_brprob.py script which is enabled with -details
10038 * profile-count.h (precise_p): New function.
10040 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10042 PR tree-optimization/83922
10043 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10044 there are no statements that need masking.
10045 (vect_active_double_reduction_p): New function.
10046 (vect_analyze_loop_operations): Use it when handling phis that
10047 are not in the loop header.
10049 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10051 PR tree-optimization/83914
10052 * tree-vect-loop.c (vectorizable_induction): Don't convert
10053 init_expr or apply the peeling adjustment for inductions
10054 that are nested within the vectorized loop.
10056 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10058 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10061 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10065 * function.h (gimplify_parameters): Add gimple_seq * argument.
10066 * function.c: Include gimple.h and options.h.
10067 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10068 for the added local temporaries if needed.
10069 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10070 if there are any parameter cleanups, wrap whole body into a
10071 try/finally with the cleanups.
10073 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10076 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10077 Use GET_MODE_CLASS for scalar floating point.
10079 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10083 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10084 Fix call of call_cgraph_insertion_hooks.
10086 2018-01-18 Martin Sebor <msebor@redhat.com>
10088 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10090 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10093 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10096 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10099 * common.opt: (-ffile-prefix-map): New option.
10100 * opts.c (common_handle_option): Defer it.
10101 * opts-global.c (handle_common_deferred_options): Handle it.
10102 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10103 * file-prefix-map.h: New file.
10104 (remap_debug_filename, add_debug_prefix_map): ...here.
10105 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10106 * final.c (debug_prefix_map, add_debug_prefix_map
10107 remap_debug_filename): Move to...
10108 * file-prefix-map.c: New file.
10109 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10110 generalize, get rid of alloca(), use strrchr() instead of strchr().
10111 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10112 Implement in terms of add_prefix_map().
10113 (remap_macro_filename, remap_debug_filename): Implement in term of
10115 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10116 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10117 * dbxout.c: Include file-prefix-map.h.
10118 * varasm.c: Likewise.
10119 * vmsdbgout.c: Likewise.
10120 * xcoffout.c: Likewise.
10121 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10122 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10123 * doc/invoke.texi (-ffile-prefix-map): Document.
10124 (-fdebug-prefix-map): Update description.
10126 2018-01-18 Martin Liska <mliska@suse.cz>
10128 * config/i386/i386.c (indirect_thunk_name): Document that also
10130 (output_indirect_thunk): Document why both instructions
10131 (pause and lfence) are generated.
10133 2018-01-18 Richard Biener <rguenther@suse.de>
10135 PR tree-optimization/83887
10136 * graphite-scop-detection.c
10137 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10138 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10139 (scop_detection::merge_sese): Re-implement with a flood-fill
10140 algorithm that properly finds a SESE region if it exists.
10142 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10145 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10146 pointer_diff optimizations use view_convert instead of convert.
10148 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10150 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10151 Generate different code for -mno-speculate-indirect-jumps.
10152 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10153 (*call_indirect_aix<mode>): Disable for
10154 -mno-speculate-indirect-jumps.
10155 (*call_indirect_aix<mode>_nospec): New define_insn.
10156 (*call_value_indirect_aix<mode>): Disable for
10157 -mno-speculate-indirect-jumps.
10158 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10159 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10160 -mno-speculate-indirect-jumps.
10161 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10163 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10165 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10166 long double type, set the flags for noting the default long double
10167 type, even if we don't pass or return a long double type.
10169 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10172 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10175 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10177 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10178 support for merge[hl].
10179 (fold_mergehl_helper): New helper function.
10180 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10181 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10182 (altivec_vmrglw_direct): Add xxmrglw insn.
10184 2018-01-17 Andrew Waterman <andrew@sifive.com>
10186 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10187 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10189 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10192 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10193 call the lto_location_cache before reading the
10194 DECL_SOURCE_LOCATION of the types.
10196 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10197 Richard Sandiford <richard.sandiford@linaro.org>
10199 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10200 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10201 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10202 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10203 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10205 * config/aarch64/constraints.md (aarch64_movti_operand):
10207 * config/aarch64/predicates.md (Uti): Add new constraint.
10209 2018-01-17 Carl Love <cel@us.ibm.com>
10211 * config/rs6000/vsx.md (define_expand xl_len_r,
10212 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10213 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10215 (define_expand, define_insn): Move the shift left from the
10216 define_insn to the define_expand for lxvl and stxvl instructions.
10217 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10218 and XL_LEN_R definitions to PURE.
10220 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10222 * config/i386/i386.c (indirect_thunk_name): Declare regno
10223 as unsigned int. Compare regno with INVALID_REGNUM.
10224 (output_indirect_thunk): Ditto.
10225 (output_indirect_thunk_function): Ditto.
10226 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10227 in the call to output_indirect_thunk_function.
10229 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10231 PR middle-end/83884
10232 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10233 rather than the size of inner_type to determine the stack slot size
10234 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10236 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10239 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10242 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10244 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10245 endian Linux systems to optionally enable multilibs for selecting
10246 the long double type if the user configured an explicit type.
10247 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10248 have no long double multilibs if not defined.
10249 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10250 warn if the user used -mabi={ieee,ibm}longdouble and we built
10251 multilibs for long double.
10252 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10253 appropriate multilib option.
10254 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10256 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10257 for building long double multilibs.
10258 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10260 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10262 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10265 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10267 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10270 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10273 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10276 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10278 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10279 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10281 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10283 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10284 different rtl trees depending on TARGET_64BIT.
10285 (rs6000_gen_lvx): Likewise.
10287 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10289 * config/visium/visium.md (nop): Tweak comment.
10290 (hazard_nop): Likewise.
10292 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10294 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10295 -mspeculate-indirect-jumps.
10296 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10297 for -mno-speculate-indirect-jumps.
10298 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10299 (*call_value_indirect_elfv2<mode>): Disable for
10300 -mno-speculate-indirect-jumps.
10301 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10302 (indirect_jump): Emit different RTL for
10303 -mno-speculate-indirect-jumps.
10304 (*indirect_jump<mode>): Disable for
10305 -mno-speculate-indirect-jumps.
10306 (*indirect_jump<mode>_nospec): New define_insn.
10307 (tablejump): Emit different RTL for
10308 -mno-speculate-indirect-jumps.
10309 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10310 (tablejumpsi_nospec): New define_expand.
10311 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10312 (tablejumpdi_nospec): New define_expand.
10313 (*tablejump<mode>_internal1): Disable for
10314 -mno-speculate-indirect-jumps.
10315 (*tablejump<mode>_internal1_nospec): New define_insn.
10316 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10319 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10321 * caller-save.c (insert_save): Drop unnecessary parameter. All
10324 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10325 Richard Biener <rguenth@suse.de>
10328 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10329 return early, inline manually is_gimple_sizepos. Make sure if we
10330 call gimplify_expr we don't end up with a gimple constant.
10331 * tree.c (variably_modified_type_p): Don't return true for
10332 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10333 * gimplify.h (is_gimple_sizepos): Remove.
10335 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10337 PR tree-optimization/83857
10338 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10339 vectorizable_live_operation for pure SLP statements.
10340 (vectorizable_live_operation): Handle PHIs.
10342 2018-01-16 Richard Biener <rguenther@suse.de>
10344 PR tree-optimization/83867
10345 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10346 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10348 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10351 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10352 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10353 If off is not INTEGER_CST, issue a may not be aligned warning
10354 rather than isn't aligned. Use isn%'t rather than isn't.
10355 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10357 <case MULT_EXPR>: Improve the case when bottom and one of the
10358 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10359 operand, in that case check if the other operand is multiple of
10360 bottom divided by the INTEGER_CST operand.
10362 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10365 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10366 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10367 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10368 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10369 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10370 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10371 (pa_function_arg_size): New function.
10373 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10375 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10376 in a separate statement.
10378 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10380 PR tree-optimization/83847
10381 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10382 group gathers and scatters.
10384 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10386 PR rtl-optimization/86620
10387 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10389 PR rtl-optimization/83213
10390 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10391 to last if both are JUMP_INSNs.
10393 PR tree-optimization/83843
10394 * gimple-ssa-store-merging.c
10395 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10396 store_immediate_info for bswap/nop orig_stores.
10398 2018-01-15 Andrew Waterman <andrew@sifive.com>
10400 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10402 <UDIV>: Increase cost if !TARGET_DIV.
10404 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10406 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10407 (define_attr "cr_logical_3op"): New.
10408 (cceq_ior_compare): Adjust.
10409 (cceq_ior_compare_complement): Adjust.
10410 (*cceq_rev_compare): Adjust.
10411 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10412 (is_cracked_insn): Adjust.
10413 (insn_must_be_first_in_group): Adjust.
10414 * config/rs6000/40x.md: Adjust.
10415 * config/rs6000/440.md: Adjust.
10416 * config/rs6000/476.md: Adjust.
10417 * config/rs6000/601.md: Adjust.
10418 * config/rs6000/603.md: Adjust.
10419 * config/rs6000/6xx.md: Adjust.
10420 * config/rs6000/7450.md: Adjust.
10421 * config/rs6000/7xx.md: Adjust.
10422 * config/rs6000/8540.md: Adjust.
10423 * config/rs6000/cell.md: Adjust.
10424 * config/rs6000/e300c2c3.md: Adjust.
10425 * config/rs6000/e500mc.md: Adjust.
10426 * config/rs6000/e500mc64.md: Adjust.
10427 * config/rs6000/e5500.md: Adjust.
10428 * config/rs6000/e6500.md: Adjust.
10429 * config/rs6000/mpc.md: Adjust.
10430 * config/rs6000/power4.md: Adjust.
10431 * config/rs6000/power5.md: Adjust.
10432 * config/rs6000/power6.md: Adjust.
10433 * config/rs6000/power7.md: Adjust.
10434 * config/rs6000/power8.md: Adjust.
10435 * config/rs6000/power9.md: Adjust.
10436 * config/rs6000/rs64.md: Adjust.
10437 * config/rs6000/titan.md: Adjust.
10439 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10441 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10442 ix86_indirect_branch_register logic.
10444 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10446 * config/i386/constraints.md (Bs): Update
10447 ix86_indirect_branch_register check. Don't check
10448 ix86_indirect_branch_register with GOT_memory_operand.
10450 * config/i386/predicates.md (GOT_memory_operand): Don't check
10451 ix86_indirect_branch_register here.
10452 (GOT32_symbol_operand): Likewise.
10454 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10456 * config/i386/predicates.md (constant_call_address_operand):
10457 Rewrite ix86_indirect_branch_register logic.
10458 (sibcall_insn_operand): Likewise.
10460 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10462 * config/i386/constraints.md (Bs): Replace
10463 ix86_indirect_branch_thunk_register with
10464 ix86_indirect_branch_register.
10466 * config/i386/i386.md (indirect_jump): Likewise.
10467 (tablejump): Likewise.
10468 (*sibcall_memory): Likewise.
10469 (*sibcall_value_memory): Likewise.
10470 Peepholes of indirect call and jump via memory: Likewise.
10471 * config/i386/i386.opt: Likewise.
10472 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10473 (GOT_memory_operand): Likewise.
10474 (call_insn_operand): Likewise.
10475 (sibcall_insn_operand): Likewise.
10476 (GOT32_symbol_operand): Likewise.
10478 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10480 PR middle-end/83837
10481 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10482 type rather than type addr's type points to.
10483 (expand_omp_atomic_mutex): Likewise.
10484 (expand_omp_atomic): Likewise.
10486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10489 * config/i386/i386.c (output_indirect_thunk_function): Use
10490 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10491 for __x86_return_thunk.
10493 2018-01-15 Richard Biener <rguenther@suse.de>
10495 PR middle-end/83850
10496 * expmed.c (extract_bit_field_1): Fix typo.
10498 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10501 * config/arm/iterators.md (VF): New mode iterator.
10502 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10503 Remove integer-related logic from pattern.
10504 (neon_vabd<mode>_3): Likewise.
10506 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10508 PR middle-end/82694
10509 * common.opt (fstrict-overflow): No longer an alias.
10510 (fwrapv-pointer): New option.
10511 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10512 also for pointer types based on flag_wrapv_pointer.
10513 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10514 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10515 opts->x_flag_wrapv got set.
10516 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10517 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10518 POINTER_TYPE_OVERFLOW_UNDEFINED.
10519 * match.pd: Likewise in address comparison pattern.
10520 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10522 2018-01-15 Richard Biener <rguenther@suse.de>
10525 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10526 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10527 Reset type names to their identifier if their TYPE_DECL doesn't
10528 have linkage (and thus is used for ODR and devirt).
10529 (save_debug_info_for_decl): Remove.
10530 (save_debug_info_for_type): Likewise.
10531 (add_tree_to_fld_list): Adjust.
10532 * tree-pretty-print.c (dump_generic_node): Make dumping of
10533 type names more robust.
10535 2018-01-15 Richard Biener <rguenther@suse.de>
10537 * BASE-VER: Bump to 8.0.1.
10539 2018-01-14 Martin Sebor <msebor@redhat.com>
10542 * builtins.c (check_access): Avoid warning when the no-warning bit
10545 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10547 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10548 * ira-color (allocno_hard_regs_compare): Likewise.
10550 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10553 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10554 Use .pushsection/.popsection.
10556 2018-01-14 Martin Sebor <msebor@redhat.com>
10559 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10561 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10563 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10564 entry from extra_headers.
10565 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10566 extra_headers, make the list bitwise identical to the i?86-*-* one.
10568 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10570 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10571 -mcmodel=large with -mindirect-branch=thunk,
10572 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10573 -mfunction-return=thunk-extern.
10574 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10575 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10576 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10578 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10580 * config/i386/i386.c (print_reg): Print the name of the full
10581 integer register without '%'.
10582 (ix86_print_operand): Handle 'V'.
10583 * doc/extend.texi: Document 'V' modifier.
10585 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10587 * config/i386/constraints.md (Bs): Disallow memory operand for
10588 -mindirect-branch-register.
10590 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10591 (GOT_memory_operand): Likewise.
10592 (call_insn_operand): Likewise.
10593 (sibcall_insn_operand): Likewise.
10594 (GOT32_symbol_operand): Likewise.
10595 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10596 for -mindirect-branch-register.
10597 (tablejump): Likewise.
10598 (*sibcall_memory): Likewise.
10599 (*sibcall_value_memory): Likewise.
10600 Disallow peepholes of indirect call and jump via memory for
10601 -mindirect-branch-register.
10602 (*call_pop): Replace m with Bw.
10603 (*call_value_pop): Likewise.
10604 (*sibcall_pop_memory): Replace m with Bs.
10605 * config/i386/i386.opt (mindirect-branch-register): New option.
10606 * doc/invoke.texi: Document -mindirect-branch-register option.
10608 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10610 * config/i386/i386-protos.h (ix86_output_function_return): New.
10611 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10612 set function_return_type.
10613 (indirect_thunk_name): Add ret_p to indicate thunk for function
10615 (output_indirect_thunk_function): Pass false to
10616 indirect_thunk_name.
10617 (ix86_output_indirect_branch_via_reg): Likewise.
10618 (ix86_output_indirect_branch_via_push): Likewise.
10619 (output_indirect_thunk_function): Create alias for function
10620 return thunk if regno < 0.
10621 (ix86_output_function_return): New function.
10622 (ix86_handle_fndecl_attribute): Handle function_return.
10623 (ix86_attribute_table): Add function_return.
10624 * config/i386/i386.h (machine_function): Add
10625 function_return_type.
10626 * config/i386/i386.md (simple_return_internal): Use
10627 ix86_output_function_return.
10628 (simple_return_internal_long): Likewise.
10629 * config/i386/i386.opt (mfunction-return=): New option.
10630 (indirect_branch): Mention -mfunction-return=.
10631 * doc/extend.texi: Document function_return function attribute.
10632 * doc/invoke.texi: Document -mfunction-return= option.
10634 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10636 * config/i386/i386-opts.h (indirect_branch): New.
10637 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10638 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10639 with local indirect jump when converting indirect call and jump.
10640 (ix86_set_indirect_branch_type): New.
10641 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10642 (indirectlabelno): New.
10643 (indirect_thunk_needed): Likewise.
10644 (indirect_thunk_bnd_needed): Likewise.
10645 (indirect_thunks_used): Likewise.
10646 (indirect_thunks_bnd_used): Likewise.
10647 (INDIRECT_LABEL): Likewise.
10648 (indirect_thunk_name): Likewise.
10649 (output_indirect_thunk): Likewise.
10650 (output_indirect_thunk_function): Likewise.
10651 (ix86_output_indirect_branch_via_reg): Likewise.
10652 (ix86_output_indirect_branch_via_push): Likewise.
10653 (ix86_output_indirect_branch): Likewise.
10654 (ix86_output_indirect_jmp): Likewise.
10655 (ix86_code_end): Call output_indirect_thunk_function if needed.
10656 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10658 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10659 (ix86_attribute_table): Add indirect_branch.
10660 * config/i386/i386.h (machine_function): Add indirect_branch_type
10661 and has_local_indirect_jump.
10662 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10664 (tablejump): Likewise.
10665 (*indirect_jump): Use ix86_output_indirect_jmp.
10666 (*tablejump_1): Likewise.
10667 (simple_return_indirect_internal): Likewise.
10668 * config/i386/i386.opt (mindirect-branch=): New option.
10669 (indirect_branch): New.
10672 (thunk-inline): Likewise.
10673 (thunk-extern): Likewise.
10674 * doc/extend.texi: Document indirect_branch function attribute.
10675 * doc/invoke.texi: Document -mindirect-branch= option.
10677 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10680 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10682 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10684 * ipa-inline.c (want_inline_small_function_p): Return false if
10685 inlining has already failed with CIF_FINAL_ERROR.
10686 (update_caller_keys): Call want_inline_small_function_p before
10688 (update_callee_keys): Likewise.
10690 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10692 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10694 (rs6000_quadword_masked_address_p): Likewise.
10695 (quad_aligned_load_p): Likewise.
10696 (quad_aligned_store_p): Likewise.
10697 (const_load_sequence_p): Add comment to describe the outer-most loop.
10698 (mimic_memory_attributes_and_flags): New function.
10699 (rs6000_gen_stvx): Likewise.
10700 (replace_swapped_aligned_store): Likewise.
10701 (rs6000_gen_lvx): Likewise.
10702 (replace_swapped_aligned_load): Likewise.
10703 (replace_swapped_load_constant): Capitalize argument name in
10704 comment describing this function.
10705 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10706 and stores that access quad-word aligned addresses and replace
10707 with stvx or lvx instructions when appropriate.
10708 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10709 New function prototype.
10710 (rs6000_quadword_masked_address_p): Likewise.
10711 (rs6000_gen_lvx): Likewise.
10712 (rs6000_gen_stvx): Likewise.
10713 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10714 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10715 when memory address is aligned.
10716 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10717 this split to select lvx instruction when memory address is aligned.
10718 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10719 instruction when memory address is aligned.
10720 (*vsx_le_perm_load_v16qi): Likewise.
10721 (four unnamed splitters): Modify to select the stvx instruction
10722 when memory is aligned.
10724 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10726 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10727 which appears in the queue multiple times.
10729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10730 Alan Hayward <alan.hayward@arm.com>
10731 David Sherwood <david.sherwood@arm.com>
10733 * tree-vectorizer.h (vec_lower_bound): New structure.
10734 (_loop_vec_info): Add check_nonzero and lower_bounds.
10735 (LOOP_VINFO_CHECK_NONZERO): New macro.
10736 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10737 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10738 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10739 fields. Make seg_len the distance travelled, not including the
10741 (dr_direction_indicator): Declare.
10742 (dr_zero_step_indicator): Likewise.
10743 (dr_known_forward_stride_p): Likewise.
10744 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10746 (runtime_alias_check_p): Allow runtime alias checks with
10748 (operator ==): Compare access_size and align.
10749 (prune_runtime_alias_test_list): Rework for new distinction between
10750 the access_size and seg_len.
10751 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10753 (get_segment_min_max): New function.
10754 (create_intersect_range_checks): Use it.
10755 (dr_step_indicator): New function.
10756 (dr_direction_indicator): Likewise.
10757 (dr_zero_step_indicator): Likewise.
10758 (dr_known_forward_stride_p): Likewise.
10759 * tree-loop-distribution.c (data_ref_segment_size): Return
10760 DR_STEP * (niters - 1).
10761 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10763 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10764 (vect_preserves_scalar_order_p): New function, split out from...
10765 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10766 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10767 (vect_vfa_access_size): New function.
10768 (vect_vfa_align): Likewise.
10769 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10770 (dump_lower_bound): New function.
10771 (vect_check_lower_bound): Likewise.
10772 (vect_small_gap_p): Likewise.
10773 (vectorizable_with_step_bound_p): Likewise.
10774 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10775 depencies if the vectorization factor is 1. Convert the checks
10776 for nonzero steps into checks on the bounds of DR_STEP. Try using
10777 a bunds check for variable steps if the minimum required step is
10778 relatively small. Update calls to the dr_with_seg_len
10779 constructor and to vect_compile_time_alias.
10780 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10782 (vect_loop_versioning): Call it.
10783 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10785 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10788 Alan Hayward <alan.hayward@arm.com>
10789 David Sherwood <david.sherwood@arm.com>
10791 * doc/sourcebuild.texi (vect_scatter_store): Document.
10792 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10794 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10796 * genopinit.c (main): Add supports_vec_scatter_store and
10797 supports_vec_scatter_store_cached to target_optabs.
10798 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10799 IFN_MASK_SCATTER_STORE.
10800 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10802 * internal-fn.h (internal_store_fn_p): Declare.
10803 (internal_fn_stored_value_index): Likewise.
10804 * internal-fn.c (scatter_store_direct): New macro.
10805 (expand_scatter_store_optab_fn): New function.
10806 (direct_scatter_store_optab_supported_p): New macro.
10807 (internal_store_fn_p): New function.
10808 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10809 IFN_MASK_SCATTER_STORE.
10810 (internal_fn_mask_index): Likewise.
10811 (internal_fn_stored_value_index): New function.
10812 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10813 for scatter stores.
10814 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10815 * optabs-query.c (supports_vec_scatter_store_p): New function.
10816 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10817 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10818 true for scatter stores.
10819 (vect_gather_scatter_fn_p): Handle scatter stores too.
10820 (vect_check_gather_scatter): Consider using scatter stores if
10821 supports_vec_scatter_store_p.
10822 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10823 scatter stores too.
10824 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10825 internal_fn_stored_value_index.
10826 (check_load_store_masking): Handle scatter stores too.
10827 (vect_get_store_rhs): Make public.
10828 (vectorizable_call): Use internal_store_fn_p.
10829 (vectorizable_store): Handle scatter store internal functions.
10830 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10831 when deciding whether the end of the group has been reached.
10832 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10833 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10834 (mask_scatter_store<mode>): New insns.
10836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10837 Alan Hayward <alan.hayward@arm.com>
10838 David Sherwood <david.sherwood@arm.com>
10840 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10841 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10842 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10844 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10845 Use vect_truncate_gather_scatter_offset if we can't treat the
10846 operation as a normal gather load or scatter store.
10847 (get_group_load_store_type): Take the gather_scatter_info
10848 as argument. Try using a gather load or scatter store for
10849 single-element groups.
10850 (get_load_store_type): Update calls to get_group_load_store_type
10851 and vect_use_strided_gather_scatters_p.
10853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10854 Alan Hayward <alan.hayward@arm.com>
10855 David Sherwood <david.sherwood@arm.com>
10857 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10858 optional tree argument.
10859 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10861 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10862 but continue to use the current value as a fallback.
10863 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10864 to compare the updates.
10865 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10866 (get_load_store_type): Use it when handling a strided access.
10867 (vect_get_strided_load_store_ops): New function.
10868 (vect_get_data_ptr_increment): Likewise.
10869 (vectorizable_load): Handle strided gather loads. Always pass
10870 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10873 Alan Hayward <alan.hayward@arm.com>
10874 David Sherwood <david.sherwood@arm.com>
10876 * doc/md.texi (gather_load@var{m}): Document.
10877 (mask_gather_load@var{m}): Likewise.
10878 * genopinit.c (main): Add supports_vec_gather_load and
10879 supports_vec_gather_load_cached to target_optabs.
10880 * optabs-tree.c (init_tree_optimization_optabs): Use
10881 ggc_cleared_alloc to allocate target_optabs.
10882 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10883 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10885 * internal-fn.h (internal_load_fn_p): Declare.
10886 (internal_gather_scatter_fn_p): Likewise.
10887 (internal_fn_mask_index): Likewise.
10888 (internal_gather_scatter_fn_supported_p): Likewise.
10889 * internal-fn.c (gather_load_direct): New macro.
10890 (expand_gather_load_optab_fn): New function.
10891 (direct_gather_load_optab_supported_p): New macro.
10892 (direct_internal_fn_optab): New function.
10893 (internal_load_fn_p): Likewise.
10894 (internal_gather_scatter_fn_p): Likewise.
10895 (internal_fn_mask_index): Likewise.
10896 (internal_gather_scatter_fn_supported_p): Likewise.
10897 * optabs-query.c (supports_at_least_one_mode_p): New function.
10898 (supports_vec_gather_load_p): Likewise.
10899 * optabs-query.h (supports_vec_gather_load_p): Declare.
10900 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10901 and memory_type field.
10902 (NUM_PATTERNS): Bump to 15.
10903 * tree-vect-data-refs.c: Include internal-fn.h.
10904 (vect_gather_scatter_fn_p): New function.
10905 (vect_describe_gather_scatter_call): Likewise.
10906 (vect_check_gather_scatter): Try using internal functions for
10907 gather loads. Recognize existing calls to a gather load function.
10908 (vect_analyze_data_refs): Consider using gather loads if
10909 supports_vec_gather_load_p.
10910 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10911 (vect_get_gather_scatter_offset_type): Likewise.
10912 (vect_convert_mask_for_vectype): Likewise.
10913 (vect_add_conversion_to_patterm): Likewise.
10914 (vect_try_gather_scatter_pattern): Likewise.
10915 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10916 (vect_vect_recog_func_ptrs): Add it.
10917 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10918 internal_fn_mask_index and internal_gather_scatter_fn_p.
10919 (check_load_store_masking): Take the gather_scatter_info as an
10920 argument and handle gather loads.
10921 (vect_get_gather_scatter_ops): New function.
10922 (vectorizable_call): Check internal_load_fn_p.
10923 (vectorizable_load): Likewise. Handle gather load internal
10925 (vectorizable_store): Update call to check_load_store_masking.
10926 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10927 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10928 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10929 (aarch64_gather_scale_operand_d): New predicates.
10930 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10931 (mask_gather_load<mode>): New insns.
10933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10934 Alan Hayward <alan.hayward@arm.com>
10935 David Sherwood <david.sherwood@arm.com>
10937 * optabs.def (fold_left_plus_optab): New optab.
10938 * doc/md.texi (fold_left_plus_@var{m}): Document.
10939 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10940 * internal-fn.c (fold_left_direct): Define.
10941 (expand_fold_left_optab_fn): Likewise.
10942 (direct_fold_left_optab_supported_p): Likewise.
10943 * fold-const-call.c (fold_const_fold_left): New function.
10944 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10945 * tree-parloops.c (valid_reduction_p): New function.
10946 (gather_scalar_reductions): Use it.
10947 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10948 (vect_finish_replace_stmt): Declare.
10949 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10950 (needs_fold_left_reduction_p): New function, split out from...
10951 (vect_is_simple_reduction): ...here. Accept reductions that
10952 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10953 (vect_force_simple_reduction): Also store the reduction type in
10954 the assignment's STMT_VINFO_REDUC_TYPE.
10955 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10956 (merge_with_identity): New function.
10957 (vect_expand_fold_left): Likewise.
10958 (vectorize_fold_left_reduction): Likewise.
10959 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10960 scalar phi in place for it. Check for target support and reject
10961 cases that would reassociate the operation. Defer the transform
10962 phase to vectorize_fold_left_reduction.
10963 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10964 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10965 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10969 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10970 call to ifc_temp_var.
10972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10973 Alan Hayward <alan.hayward@arm.com>
10974 David Sherwood <david.sherwood@arm.com>
10976 * target.def (legitimize_address_displacement): Take the original
10977 offset as a poly_int.
10978 * targhooks.h (default_legitimize_address_displacement): Update
10980 * targhooks.c (default_legitimize_address_displacement): Likewise.
10981 * doc/tm.texi: Regenerate.
10982 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10983 as an argument, moving assert of ad->disp == ad->disp_term to...
10984 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10985 Try calling targetm.legitimize_address_displacement before expanding
10986 the address rather than afterwards, and adjust for the new interface.
10987 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10988 Match the new hook interface. Handle SVE addresses.
10989 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10990 new hook interface.
10992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10994 * Makefile.in (OBJS): Add early-remat.o.
10995 * target.def (select_early_remat_modes): New hook.
10996 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10997 * doc/tm.texi: Regenerate.
10998 * targhooks.h (default_select_early_remat_modes): Declare.
10999 * targhooks.c (default_select_early_remat_modes): New function.
11000 * timevar.def (TV_EARLY_REMAT): New timevar.
11001 * passes.def (pass_early_remat): New pass.
11002 * tree-pass.h (make_pass_early_remat): Declare.
11003 * early-remat.c: New file.
11004 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11006 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11009 Alan Hayward <alan.hayward@arm.com>
11010 David Sherwood <david.sherwood@arm.com>
11012 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11013 vfm1 with a bound_epilog parameter.
11014 (vect_do_peeling): Update calls accordingly, and move the prologue
11015 call earlier in the function. Treat the base bound_epilog as 0 for
11016 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11017 this base when peeling for gaps.
11018 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11019 with fully-masked loops.
11020 (vect_estimate_min_profitable_iters): Handle the single peeled
11021 iteration in that case.
11023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11024 Alan Hayward <alan.hayward@arm.com>
11025 David Sherwood <david.sherwood@arm.com>
11027 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11028 single-element interleaving even if the size is not a power of 2.
11029 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11030 accesses for single-element interleaving if the group size is
11033 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11034 Alan Hayward <alan.hayward@arm.com>
11035 David Sherwood <david.sherwood@arm.com>
11037 * doc/md.texi (fold_extract_last_@var{m}): Document.
11038 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11039 * optabs.def (fold_extract_last_optab): New optab.
11040 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11041 * internal-fn.c (fold_extract_direct): New macro.
11042 (expand_fold_extract_optab_fn): Likewise.
11043 (direct_fold_extract_optab_supported_p): Likewise.
11044 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11045 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11046 EXTRACT_LAST_REDUCTION.
11047 (get_initial_def_for_reduction): Do not create an initial vector
11048 for EXTRACT_LAST_REDUCTION reductions.
11049 (vectorizable_reduction): Leave the scalar phi in place for
11050 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11051 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11052 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11053 transform phase to vectorizable_condition.
11054 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11056 (vect_finish_stmt_generation): ...here.
11057 (vect_finish_replace_stmt): New function.
11058 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11059 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11061 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11064 Alan Hayward <alan.hayward@arm.com>
11065 David Sherwood <david.sherwood@arm.com>
11067 * doc/md.texi (extract_last_@var{m}): Document.
11068 * optabs.def (extract_last_optab): New optab.
11069 * internal-fn.def (EXTRACT_LAST): New internal function.
11070 * internal-fn.c (cond_unary_direct): New macro.
11071 (expand_cond_unary_optab_fn): Likewise.
11072 (direct_cond_unary_optab_supported_p): Likewise.
11073 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11074 loops using EXTRACT_LAST.
11075 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11076 (extract_last_<mode>): ...this optab.
11077 (vec_extract<mode><Vel>): Update accordingly.
11079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11080 Alan Hayward <alan.hayward@arm.com>
11081 David Sherwood <david.sherwood@arm.com>
11083 * target.def (empty_mask_is_expensive): New hook.
11084 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11085 * doc/tm.texi: Regenerate.
11086 * targhooks.h (default_empty_mask_is_expensive): Declare.
11087 * targhooks.c (default_empty_mask_is_expensive): New function.
11088 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11089 if the target says that empty masks are expensive.
11090 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11092 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11095 Alan Hayward <alan.hayward@arm.com>
11096 David Sherwood <david.sherwood@arm.com>
11098 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11099 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11100 (vect_use_loop_mask_for_alignment_p): New function.
11101 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11102 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11103 niters_skip argument. Make sure that the first niters_skip elements
11104 of the first iteration are inactive.
11105 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11106 Update call to vect_set_loop_masks_directly.
11107 (get_misalign_in_elems): New function, split out from...
11108 (vect_gen_prolog_loop_niters): ...here.
11109 (vect_update_init_of_dr): Take a code argument that specifies whether
11110 the adjustment should be added or subtracted.
11111 (vect_update_init_of_drs): Likewise.
11112 (vect_prepare_for_masked_peels): New function.
11113 (vect_do_peeling): Skip prologue peeling if we're using a mask
11114 instead. Update call to vect_update_inits_of_drs.
11115 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11117 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11118 alignment. Do not include the number of peeled iterations in
11119 the minimum threshold in that case.
11120 (vectorizable_induction): Adjust the start value down by
11121 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11122 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11123 Take the number of skipped iterations into account when calculating
11125 * tree-vect-stmts.c (vect_gen_while_not): New function.
11127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11128 Alan Hayward <alan.hayward@arm.com>
11129 David Sherwood <david.sherwood@arm.com>
11131 * doc/sourcebuild.texi (vect_fully_masked): Document.
11132 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11133 default value to 0.
11134 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11136 (vect_analyze_loop_2): ...here. Don't check the vectorization
11137 factor against the number of loop iterations if the loop is
11140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11141 Alan Hayward <alan.hayward@arm.com>
11142 David Sherwood <david.sherwood@arm.com>
11144 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11145 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11146 (dump_groups): Update accordingly.
11147 (iv_use::mem_type): New member variable.
11148 (address_p): New function.
11149 (record_use): Add a mem_type argument and initialize the new
11151 (record_group_use): Add a mem_type argument. Use address_p.
11152 Remove obsolete null checks of base_object. Update call to record_use.
11153 (find_interesting_uses_op): Update call to record_group_use.
11154 (find_interesting_uses_cond): Likewise.
11155 (find_interesting_uses_address): Likewise.
11156 (get_mem_type_for_internal_fn): New function.
11157 (find_address_like_use): Likewise.
11158 (find_interesting_uses_stmt): Try find_address_like_use before
11159 calling find_interesting_uses_op.
11160 (addr_offset_valid_p): Use the iv mem_type field as the type
11161 of the addressed memory.
11162 (add_autoinc_candidates): Likewise.
11163 (get_address_cost): Likewise.
11164 (split_small_address_groups_p): Use address_p.
11165 (split_address_groups): Likewise.
11166 (add_iv_candidate_for_use): Likewise.
11167 (autoinc_possible_for_pair): Likewise.
11168 (rewrite_groups): Likewise.
11169 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11170 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11171 (get_alias_ptr_type_for_ptr_address): New function.
11172 (rewrite_use_address): Rewrite address uses in calls that were
11173 identified by find_address_like_use.
11175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11176 Alan Hayward <alan.hayward@arm.com>
11177 David Sherwood <david.sherwood@arm.com>
11179 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11181 * gimple-expr.h (is_gimple_addressable: Likewise.
11182 * gimple-expr.c (is_gimple_address): Likewise.
11183 * internal-fn.c (expand_call_mem_ref): New function.
11184 (expand_mask_load_optab_fn): Use it.
11185 (expand_mask_store_optab_fn): Likewise.
11187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11188 Alan Hayward <alan.hayward@arm.com>
11189 David Sherwood <david.sherwood@arm.com>
11191 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11192 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11193 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11194 (cond_umax@var{mode}): Document.
11195 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11196 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11197 (cond_umin_optab, cond_umax_optab): New optabs.
11198 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11199 (COND_IOR, COND_XOR): New internal functions.
11200 * internal-fn.h (get_conditional_internal_fn): Declare.
11201 * internal-fn.c (cond_binary_direct): New macro.
11202 (expand_cond_binary_optab_fn): Likewise.
11203 (direct_cond_binary_optab_supported_p): Likewise.
11204 (get_conditional_internal_fn): New function.
11205 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11206 Cope with reduction statements that are vectorized as calls rather
11208 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11209 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11210 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11211 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11212 (UNSPEC_COND_EOR): New unspecs.
11213 (optab): Add mappings for them.
11214 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11215 (sve_int_op, sve_fp_op): New int attributes.
11217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11218 Alan Hayward <alan.hayward@arm.com>
11219 David Sherwood <david.sherwood@arm.com>
11221 * optabs.def (while_ult_optab): New optab.
11222 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11223 * internal-fn.def (WHILE_ULT): New internal function.
11224 * internal-fn.h (direct_internal_fn_supported_p): New override
11225 that takes two types as argument.
11226 * internal-fn.c (while_direct): New macro.
11227 (expand_while_optab_fn): New function.
11228 (convert_optab_supported_p): Likewise.
11229 (direct_while_optab_supported_p): New macro.
11230 * wide-int.h (wi::udiv_ceil): New function.
11231 * tree-vectorizer.h (rgroup_masks): New structure.
11232 (vec_loop_masks): New typedef.
11233 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11234 and fully_masked_p.
11235 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11236 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11237 (vect_max_vf): New function.
11238 (slpeel_make_loop_iterate_ntimes): Delete.
11239 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11240 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11241 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11242 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11243 internal-fn.h, stor-layout.h and optabs-query.h.
11244 (vect_set_loop_mask): New function.
11245 (add_preheader_seq): Likewise.
11246 (add_header_seq): Likewise.
11247 (interleave_supported_p): Likewise.
11248 (vect_maybe_permute_loop_masks): Likewise.
11249 (vect_set_loop_masks_directly): Likewise.
11250 (vect_set_loop_condition_masked): Likewise.
11251 (vect_set_loop_condition_unmasked): New function, split out from
11252 slpeel_make_loop_iterate_ntimes.
11253 (slpeel_make_loop_iterate_ntimes): Rename to..
11254 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11255 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11256 (vect_do_peeling): Update call accordingly.
11257 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11259 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11260 mask_compare_type, can_fully_mask_p and fully_masked_p.
11261 (release_vec_loop_masks): New function.
11262 (_loop_vec_info): Use it to free the loop masks.
11263 (can_produce_all_loop_masks_p): New function.
11264 (vect_get_max_nscalars_per_iter): Likewise.
11265 (vect_verify_full_masking): Likewise.
11266 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11267 retries, and free the mask rgroups before retrying. Check loop-wide
11268 reasons for disallowing fully-masked loops. Make the final decision
11269 about whether use a fully-masked loop or not.
11270 (vect_estimate_min_profitable_iters): Do not assume that peeling
11271 for the number of iterations will be needed for fully-masked loops.
11272 (vectorizable_reduction): Disable fully-masked loops.
11273 (vectorizable_live_operation): Likewise.
11274 (vect_halve_mask_nunits): New function.
11275 (vect_double_mask_nunits): Likewise.
11276 (vect_record_loop_mask): Likewise.
11277 (vect_get_loop_mask): Likewise.
11278 (vect_transform_loop): Handle the case in which the final loop
11279 iteration might handle a partial vector. Call vect_set_loop_condition
11280 instead of slpeel_make_loop_iterate_ntimes.
11281 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11282 (check_load_store_masking): New function.
11283 (prepare_load_store_mask): Likewise.
11284 (vectorizable_store): Handle fully-masked loops.
11285 (vectorizable_load): Likewise.
11286 (supportable_widening_operation): Use vect_halve_mask_nunits for
11288 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11289 (vect_gen_while): New function.
11290 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11291 (aarch64_uqdec<mode>): New insn.
11293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11294 Alan Hayward <alan.hayward@arm.com>
11295 David Sherwood <david.sherwood@arm.com>
11297 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11298 (reduc_xor_scal_optab): New optabs.
11299 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11300 (reduc_xor_scal_@var{m}): Document.
11301 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11302 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11303 internal functions.
11304 * fold-const-call.c (fold_const_call): Handle them.
11305 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11306 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11307 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11308 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11309 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11310 (UNSPEC_XORV): New unspecs.
11311 (optab): Add entries for them.
11312 (BITWISEV): New int iterator.
11313 (bit_reduc_op): New int attributes.
11315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11316 Alan Hayward <alan.hayward@arm.com>
11317 David Sherwood <david.sherwood@arm.com>
11319 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11320 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11321 * optabs.def (vec_shl_insert_optab): New optab.
11322 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11323 (duplicate_and_interleave): Likewise.
11324 * tree-vect-loop.c: Include internal-fn.h.
11325 (neutral_op_for_slp_reduction): New function, split out from
11326 get_initial_defs_for_reduction.
11327 (get_initial_def_for_reduction): Handle option 2 for variable-length
11328 vectors by loading the neutral value into a vector and then shifting
11329 the initial value into element 0.
11330 (get_initial_defs_for_reduction): Replace the code argument with
11331 the neutral value calculated by neutral_op_for_slp_reduction.
11332 Use gimple_build_vector for constant-length vectors.
11333 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11334 but the first group_size elements have a neutral value.
11335 Use duplicate_and_interleave otherwise.
11336 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11337 Update call to get_initial_defs_for_reduction. Handle SLP
11338 reductions for variable-length vectors by creating one vector
11339 result for each scalar result, with the elements associated
11340 with other scalar results stubbed out with the neutral value.
11341 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11342 Require IFN_VEC_SHL_INSERT for double reductions on
11343 variable-length vectors, or SLP reductions that have
11344 a neutral value. Require can_duplicate_and_interleave_p
11345 support for variable-length unchained SLP reductions if there
11346 is no neutral value, such as for MIN/MAX reductions. Also require
11347 the number of vector elements to be a multiple of the number of
11348 SLP statements when doing variable-length unchained SLP reductions.
11349 Update call to vect_create_epilog_for_reduction.
11350 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11351 and remove initial values.
11352 (duplicate_and_interleave): Make public.
11353 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11354 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11357 Alan Hayward <alan.hayward@arm.com>
11358 David Sherwood <david.sherwood@arm.com>
11360 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11361 (can_duplicate_and_interleave_p): New function.
11362 (vect_get_and_check_slp_defs): Take the vector of statements
11363 rather than just the current one. Remove excess parentheses.
11364 Restriction rejectinon of vect_constant_def and vect_external_def
11365 for variable-length vectors to boolean types, or types for which
11366 can_duplicate_and_interleave_p is false.
11367 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11368 (duplicate_and_interleave): New function.
11369 (vect_get_constant_vectors): Use gimple_build_vector for
11370 constant-length vectors and suitable variable-length constant
11371 vectors. Use duplicate_and_interleave for other variable-length
11372 vectors. Don't defer the update when inserting new statements.
11374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11379 min_profitable_iters doesn't go negative.
11381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11382 Alan Hayward <alan.hayward@arm.com>
11383 David Sherwood <david.sherwood@arm.com>
11385 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11386 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11387 * optabs.def (vec_mask_load_lanes_optab): New optab.
11388 (vec_mask_store_lanes_optab): Likewise.
11389 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11390 (MASK_STORE_LANES): Likewise.
11391 * internal-fn.c (mask_load_lanes_direct): New macro.
11392 (mask_store_lanes_direct): Likewise.
11393 (expand_mask_load_optab_fn): Handle masked operations.
11394 (expand_mask_load_lanes_optab_fn): New macro.
11395 (expand_mask_store_optab_fn): Handle masked operations.
11396 (expand_mask_store_lanes_optab_fn): New macro.
11397 (direct_mask_load_lanes_optab_supported_p): Likewise.
11398 (direct_mask_store_lanes_optab_supported_p): Likewise.
11399 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11401 (vect_load_lanes_supported): Likewise.
11402 * tree-vect-data-refs.c (strip_conversion): New function.
11403 (can_group_stmts_p): Likewise.
11404 (vect_analyze_data_ref_accesses): Use it instead of checking
11405 for a pair of assignments.
11406 (vect_store_lanes_supported): Take a masked_p parameter.
11407 (vect_load_lanes_supported): Likewise.
11408 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11409 vect_store_lanes_supported and vect_load_lanes_supported.
11410 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11411 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11412 parameter. Don't allow gaps for masked accesses.
11413 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11414 and vect_load_lanes_supported.
11415 (get_load_store_type): Take a masked_p parameter and update
11416 call to get_group_load_store_type.
11417 (vectorizable_store): Update call to get_load_store_type.
11418 Handle IFN_MASK_STORE_LANES.
11419 (vectorizable_load): Update call to get_load_store_type.
11420 Handle IFN_MASK_LOAD_LANES.
11422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11423 Alan Hayward <alan.hayward@arm.com>
11424 David Sherwood <david.sherwood@arm.com>
11426 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11428 * config/aarch64/aarch64-protos.h
11429 (aarch64_sve_struct_memory_operand_p): Declare.
11430 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11431 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11432 (VPRED, vpred): Handle SVE structure modes.
11433 * config/aarch64/constraints.md (Utx): New constraint.
11434 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11435 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11436 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11437 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11438 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11439 structure modes. Split into pieces after RA.
11440 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11441 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11443 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11444 SVE structure modes.
11445 (aarch64_classify_address): Likewise.
11446 (sizetochar): Move earlier in file.
11447 (aarch64_print_operand): Handle SVE register lists.
11448 (aarch64_array_mode): New function.
11449 (aarch64_sve_struct_memory_operand_p): Likewise.
11450 (TARGET_ARRAY_MODE): Redefine.
11452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11453 Alan Hayward <alan.hayward@arm.com>
11454 David Sherwood <david.sherwood@arm.com>
11456 * target.def (array_mode): New target hook.
11457 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11458 * doc/tm.texi: Regenerate.
11459 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11460 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11461 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11462 targetm.array_mode.
11463 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11467 Alan Hayward <alan.hayward@arm.com>
11468 David Sherwood <david.sherwood@arm.com>
11470 * fold-const.c (fold_binary_loc): Check the argument types
11471 rather than the result type when testing for a vector operation.
11473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11475 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11476 * doc/tm.texi: Regenerate.
11478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11483 (sve): Document new AArch64 extension.
11484 * doc/md.texi (w): Extend the description of the AArch64
11485 constraint to include SVE vectors.
11486 (Upl, Upa): Document new AArch64 predicate constraints.
11487 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11489 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11490 (msve-vector-bits=): New option.
11491 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11492 SVE when these are disabled.
11493 (sve): New extension.
11494 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11495 modes. Adjust their number of units based on aarch64_sve_vg.
11496 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11497 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11498 aarch64_addr_query_type.
11499 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11500 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11501 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11502 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11503 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11504 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11505 (aarch64_simd_imm_zero_p): Delete.
11506 (aarch64_check_zero_based_sve_index_immediate): Declare.
11507 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11508 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11509 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11510 (aarch64_sve_float_mul_immediate_p): Likewise.
11511 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11512 rather than an rtx.
11513 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11514 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11515 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11516 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11517 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11518 (aarch64_regmode_natural_size): Likewise.
11519 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11520 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11522 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11523 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11524 for VG and the SVE predicate registers.
11525 (V_ALIASES): Add a "z"-prefixed alias.
11526 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11527 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11528 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11529 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11530 (REG_CLASS_NAMES): Add entries for them.
11531 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11532 and the predicate registers.
11533 (aarch64_sve_vg): Declare.
11534 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11535 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11536 (REGMODE_NATURAL_SIZE): Define.
11537 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11539 * config/aarch64/aarch64.c: Include cfgrtl.h.
11540 (simd_immediate_info): Add a constructor for series vectors,
11541 and an associated step field.
11542 (aarch64_sve_vg): New variable.
11543 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11544 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11545 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11546 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11547 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11548 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11549 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11550 (aarch64_get_mask_mode): New functions.
11551 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11552 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11553 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11554 predicate modes and predicate registers. Explicitly restrict
11555 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11556 to store a vector mode if it is recognized by
11557 aarch64_classify_vector_mode.
11558 (aarch64_regmode_natural_size): New function.
11559 (aarch64_hard_regno_caller_save_mode): Return the original mode
11561 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11562 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11563 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11564 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11566 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11567 does not overlap dest if the function is frame-related. Handle
11569 (aarch64_split_add_offset): New function.
11570 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11571 them aarch64_add_offset.
11572 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11573 and update call to aarch64_sub_sp.
11574 (aarch64_add_cfa_expression): New function.
11575 (aarch64_expand_prologue): Pass extra temporary registers to the
11576 functions above. Handle the case in which we need to emit new
11577 DW_CFA_expressions for registers that were originally saved
11578 relative to the stack pointer, but now have to be expressed
11579 relative to the frame pointer.
11580 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11582 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11583 IP0 and IP1 values for SVE frames.
11584 (aarch64_expand_vec_series): New function.
11585 (aarch64_expand_sve_widened_duplicate): Likewise.
11586 (aarch64_expand_sve_const_vector): Likewise.
11587 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11588 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11589 into the register, rather than emitting a SET directly.
11590 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11591 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11592 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11593 (offset_9bit_signed_scaled_p): New functions.
11594 (aarch64_replicate_bitmask_imm): New function.
11595 (aarch64_bitmask_imm): Use it.
11596 (aarch64_cannot_force_const_mem): Reject expressions involving
11597 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11598 (aarch64_classify_index): Handle SVE indices, by requiring
11599 a plain register index with a scale that matches the element size.
11600 (aarch64_classify_address): Handle SVE addresses. Assert that
11601 the mode of the address is VOIDmode or an integer mode.
11602 Update call to aarch64_classify_symbol.
11603 (aarch64_classify_symbolic_expression): Update call to
11604 aarch64_classify_symbol.
11605 (aarch64_const_vec_all_in_range_p): New function.
11606 (aarch64_print_vector_float_operand): Likewise.
11607 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11608 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11609 and the FP immediates 1.0 and 0.5.
11610 (aarch64_print_address_internal): Handle SVE addresses.
11611 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11612 (aarch64_regno_regclass): Handle predicate registers.
11613 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11615 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11616 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11617 (aarch64_convert_sve_vector_bits): New function.
11618 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11619 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11620 rather than an rtx.
11621 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11622 Handle SVE vector and predicate modes. Accept VL-based constants
11623 that need only one temporary register, and VL offsets that require
11624 no temporary registers.
11625 (aarch64_conditional_register_usage): Mark the predicate registers
11626 as fixed if SVE isn't available.
11627 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11628 Return true for SVE vector and predicate modes.
11629 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11630 rather than an unsigned int. Handle SVE modes.
11631 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11633 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11635 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11636 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11637 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11638 (aarch64_sve_float_mul_immediate_p): New functions.
11639 (aarch64_sve_valid_immediate): New function.
11640 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11641 Explicitly reject structure modes. Check for INDEX constants.
11642 Handle PTRUE and PFALSE constants.
11643 (aarch64_check_zero_based_sve_index_immediate): New function.
11644 (aarch64_simd_imm_zero_p): Delete.
11645 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11646 vector modes. Accept constants in the range of CNT[BHWD].
11647 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11648 ask for an Advanced SIMD mode.
11649 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11650 (aarch64_simd_vector_alignment): Handle SVE predicates.
11651 (aarch64_vectorize_preferred_vector_alignment): New function.
11652 (aarch64_simd_vector_alignment_reachable): Use it instead of
11654 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11655 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11657 (MAX_VECT_LEN): Delete.
11658 (expand_vec_perm_d): Add a vec_flags field.
11659 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11660 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11661 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11663 (aarch64_evpc_rev): Rename to...
11664 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11665 (aarch64_evpc_rev_global): New function.
11666 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11667 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11669 (aarch64_evpc_sve_tbl): New function.
11670 (aarch64_expand_vec_perm_const_1): Update after rename of
11671 aarch64_evpc_rev. Handle SVE permutes too, trying
11672 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11673 than aarch64_evpc_tbl.
11674 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11675 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11676 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11677 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11678 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11679 (aarch64_expand_sve_vcond): New functions.
11680 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11681 of aarch64_vector_mode_p.
11682 (aarch64_dwarf_poly_indeterminate_value): New function.
11683 (aarch64_compute_pressure_classes): Likewise.
11684 (aarch64_can_change_mode_class): Likewise.
11685 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11686 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11687 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11688 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11689 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11690 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11691 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11692 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11694 (Dn, Dl, Dr): Accept const as well as const_vector.
11695 (Dz): Likewise. Compare against CONST0_RTX.
11696 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11697 of "vector" where appropriate.
11698 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11699 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11700 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11701 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11702 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11703 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11704 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11705 (v_int_equiv): Extend to SVE modes.
11706 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11708 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11709 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11710 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11711 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11712 (SVE_COND_FP_CMP): New int iterators.
11713 (perm_hilo): Handle the new unpack unspecs.
11714 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11716 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11717 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11718 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11719 (aarch64_equality_operator, aarch64_constant_vector_operand)
11720 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11721 (aarch64_sve_nonimmediate_operand): Likewise.
11722 (aarch64_sve_general_operand): Likewise.
11723 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11724 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11725 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11726 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11727 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11728 (aarch64_sve_float_arith_immediate): Likewise.
11729 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11730 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11731 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11732 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11733 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11734 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11735 (aarch64_sve_float_arith_operand): Likewise.
11736 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11737 (aarch64_sve_float_mul_operand): Likewise.
11738 (aarch64_sve_vec_perm_operand): Likewise.
11739 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11740 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11741 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11742 as well as const_vector.
11743 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11744 in file. Use CONST0_RTX and CONSTM1_RTX.
11745 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11746 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11747 Use aarch64_simd_imm_zero.
11748 * config/aarch64/aarch64-sve.md: New file.
11749 * config/aarch64/aarch64.md: Include it.
11750 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11751 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11752 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11753 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11754 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11755 (sve): New attribute.
11756 (enabled): Disable instructions with the sve attribute unless
11758 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11759 aarch64_expand_mov_immediate.
11760 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11761 CNT[BHSD] immediates.
11762 (movti): Split CONST_POLY_INT moves into two halves.
11763 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11764 Split additions that need a temporary here if the destination
11765 is the stack pointer.
11766 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11767 (*add<mode>3_poly_1): New instruction.
11768 (set_clobber_cc): New expander.
11770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11772 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11773 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11774 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11775 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11776 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11777 Change innermode from fixed_mode_size to machine_mode.
11778 (simplify_subreg): Update call accordingly. Handle a constant-sized
11779 subreg of a variable-length CONST_VECTOR.
11781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11782 Alan Hayward <alan.hayward@arm.com>
11783 David Sherwood <david.sherwood@arm.com>
11785 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11786 (add_offset_to_base): New function, split out from...
11787 (create_mem_ref): ...here. When handling a scale other than 1,
11788 check first whether the address is valid without the offset.
11789 Add it into the base if so, leaving the index and scale as-is.
11791 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11794 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11795 fold_for_warn before checking if arg2 is INTEGER_CST.
11797 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11799 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11800 (store_multiple_operation): Delete.
11801 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11802 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11803 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11804 guarded by TARGET_STRING.
11805 (rs6000_output_load_multiple): Delete.
11806 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11807 OPTION_MASK_STRING / TARGET_STRING handling.
11808 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11809 (const rs6000_opt_masks) <"string">: Change mask to 0.
11810 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11811 (MASK_STRING): Delete.
11812 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11814 (load_multiple): Delete.
11821 (store_multiple): Delete.
11828 (movmemsi_8reg): Delete.
11829 (corresponding unnamed define_insn): Delete.
11830 (movmemsi_6reg): Delete.
11831 (corresponding unnamed define_insn): Delete.
11832 (movmemsi_4reg): Delete.
11833 (corresponding unnamed define_insn): Delete.
11834 (movmemsi_2reg): Delete.
11835 (corresponding unnamed define_insn): Delete.
11836 (movmemsi_1reg): Delete.
11837 (corresponding unnamed define_insn): Delete.
11838 * config/rs6000/rs6000.opt (mno-string): New.
11839 (mstring): Replace by deprecation warning stub.
11840 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11842 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11844 * regrename.c (regrename_do_replace): If replacing the same
11845 reg multiple times, try to reuse last created gen_raw_REG.
11848 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11849 main to workaround a bug in GDB.
11851 2018-01-12 Tom de Vries <tom@codesourcery.com>
11854 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11856 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11858 PR rtl-optimization/80481
11859 * ira-color.c (get_cap_member): New function.
11860 (allocnos_conflict_by_live_ranges_p): Use it.
11861 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11862 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11864 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11867 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11868 (*saddl_se_1): Ditto.
11869 (*ssubsi_1): Ditto.
11870 (*ssubl_se_1): Ditto.
11872 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11874 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11875 rather than wi::to_widest for DR_INITs.
11876 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11877 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11878 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11880 (vect_analyze_group_access_1): Note that here.
11882 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11884 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11885 polynomial type sizes.
11887 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11889 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11890 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11891 (gimple_add_tmp_var): Likewise.
11893 2018-01-12 Martin Liska <mliska@suse.cz>
11895 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11896 (gimple_alloc_sizes): Likewise.
11897 (dump_gimple_statistics): Use PRIu64 in printf format.
11898 * gimple.h: Change uint64_t to int.
11900 2018-01-12 Martin Liska <mliska@suse.cz>
11902 * tree-core.h: Use uint64_t instead of int.
11903 * tree.c (tree_node_counts): Likewise.
11904 (tree_node_sizes): Likewise.
11905 (dump_tree_statistics): Use PRIu64 in printf format.
11907 2018-01-12 Martin Liska <mliska@suse.cz>
11909 * Makefile.in: As qsort_chk is implemented in vec.c, add
11910 vec.o to linkage of gencfn-macros.
11911 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11912 passing the info to record_node_allocation_statistics.
11913 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11915 * ggc-common.c (struct ggc_usage): Add operator== and use
11916 it in operator< and compare function.
11917 * mem-stats.h (struct mem_usage): Likewise.
11918 * vec.c (struct vec_usage): Remove operator< and compare
11919 function. Can be simply inherited.
11921 2018-01-12 Martin Jambor <mjambor@suse.cz>
11924 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11925 * tree-ssa-math-opts.c: Include domwalk.h.
11926 (convert_mult_to_fma_1): New function.
11927 (fma_transformation_info): New type.
11928 (fma_deferring_state): Likewise.
11929 (cancel_fma_deferring): New function.
11930 (result_of_phi): Likewise.
11931 (last_fma_candidate_feeds_initial_phi): Likewise.
11932 (convert_mult_to_fma): Added deferring logic, split actual
11933 transformation to convert_mult_to_fma_1.
11934 (math_opts_dom_walker): New type.
11935 (math_opts_dom_walker::after_dom_children): New method, body moved
11936 here from pass_optimize_widening_mul::execute, added deferring logic
11938 (pass_optimize_widening_mul::execute): Moved most of code to
11939 math_opts_dom_walker::after_dom_children.
11940 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11941 * config/i386/i386.c (ix86_option_override_internal): Added
11942 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11944 2018-01-12 Richard Biener <rguenther@suse.de>
11947 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11948 inline instance vars.
11950 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11953 * config/rx/rx.c (rx_is_restricted_memory_address):
11954 Handle SUBREG case.
11956 2018-01-12 Richard Biener <rguenther@suse.de>
11958 PR tree-optimization/80846
11959 * target.def (split_reduction): New target hook.
11960 * targhooks.c (default_split_reduction): New function.
11961 * targhooks.h (default_split_reduction): Declare.
11962 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11963 target requests first reduce vectors by combining low and high
11965 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11966 (get_vectype_for_scalar_type_and_size): Export.
11967 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11968 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11969 * doc/tm.texi: Regenerate.
11970 * config/i386/i386.c (ix86_split_reduction): Implement
11971 TARGET_VECTORIZE_SPLIT_REDUCTION.
11973 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11976 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11977 in PIC mode except for TARGET_VXWORKS_RTP.
11978 * config/sparc/sparc.c: Include cfgrtl.h.
11979 (TARGET_INIT_PIC_REG): Define.
11980 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11981 (sparc_pic_register_p): New predicate.
11982 (sparc_legitimate_address_p): Use it.
11983 (sparc_legitimize_pic_address): Likewise.
11984 (sparc_delegitimize_address): Likewise.
11985 (sparc_mode_dependent_address_p): Likewise.
11986 (gen_load_pcrel_sym): Remove 4th parameter.
11987 (load_got_register): Adjust call to above. Remove obsolete stuff.
11988 (sparc_expand_prologue): Do not call load_got_register here.
11989 (sparc_flat_expand_prologue): Likewise.
11990 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11991 (sparc_use_pseudo_pic_reg): New function.
11992 (sparc_init_pic_reg): Likewise.
11993 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11994 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11996 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11998 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11999 Add item for branch_cost.
12001 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12003 PR rtl-optimization/83565
12004 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12005 not extend the result to a larger mode for rotate operations.
12006 (num_sign_bit_copies1): Likewise.
12008 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12011 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12013 Use values-Xc.o for -pedantic.
12014 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12016 2018-01-12 Martin Liska <mliska@suse.cz>
12019 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12021 (possible_polymorphic_call_targets): Use it.
12022 (ipa_devirt): Likewise.
12024 2018-01-12 Martin Liska <mliska@suse.cz>
12026 * profile-count.h (enum profile_quality): Use 0 as invalid
12027 enum value of profile_quality.
12029 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12031 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12032 -mext-string options.
12034 2018-01-12 Richard Biener <rguenther@suse.de>
12036 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12037 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12038 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12040 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12042 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12044 * configure.ac (--with-long-double-format): Add support for the
12045 configuration option to change the default long double format on
12047 * config.gcc (powerpc*-linux*-*): Likewise.
12048 * configure: Regenerate.
12049 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12050 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12051 used without modification.
12053 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12055 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12056 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12057 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12058 MISC_BUILTIN_SPEC_BARRIER.
12059 (rs6000_init_builtins): Likewise.
12060 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12062 (speculation_barrier): New define_insn.
12063 * doc/extend.texi: Document __builtin_speculation_barrier.
12065 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12068 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12069 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12070 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12072 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12073 integral modes instead of "ss" and "sd".
12074 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12075 vectors with 32-bit and 64-bit elements.
12076 (vecdupssescalarmodesuffix): New mode attribute.
12077 (vec_dup<mode>): Use it.
12079 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12082 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12083 frame if argument is passed on stack.
12085 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12088 * ree.c (combine_reaching_defs): Optimize also
12089 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12090 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12092 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12094 PR middle-end/83189
12095 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12097 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12099 PR middle-end/83718
12100 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12101 after they are computed.
12103 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12105 PR tree-optimization/83695
12106 * gimple-loop-linterchange.cc
12107 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12108 reset cached scev information after interchange.
12109 (pass_linterchange::execute): Remove call to scev_reset_htab.
12111 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12113 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12114 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12115 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12116 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12117 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12118 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12119 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12120 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12121 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12122 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12123 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12124 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12125 (V_lane_reg): Likewise.
12126 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12128 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12129 (vfmal_lane_low<mode>_intrinsic,
12130 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12131 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12132 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12133 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12134 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12135 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12137 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12139 * config/arm/arm-cpus.in (fp16fml): New feature.
12140 (ALL_SIMD): Add fp16fml.
12141 (armv8.2-a): Add fp16fml as an option.
12142 (armv8.3-a): Likewise.
12143 (armv8.4-a): Add fp16fml as part of fp16.
12144 * config/arm/arm.h (TARGET_FP16FML): Define.
12145 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12147 * config/arm/arm-modes.def (V2HF): Define.
12148 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12149 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12150 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12151 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12152 vfmsl_low, vfmsl_high): New set of builtins.
12153 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12154 (vfml_op): New code attribute.
12155 (VFMLHALVES): New int iterator.
12156 (VFML, VFMLSEL): New mode attributes.
12157 (V_reg): Define mapping for V2HF.
12158 (V_hi, V_lo): New mode attributes.
12159 (VF_constraint): Likewise.
12160 (vfml_half, vfml_half_selector): New int attributes.
12161 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12163 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12164 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12166 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12167 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12168 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12169 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12171 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12172 Document new effective target and option set.
12174 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12176 * config/arm/arm-cpus.in (armv8_4): New feature.
12177 (ARMv8_4a): New fgroup.
12178 (armv8.4-a): New arch.
12179 * config/arm/arm-tables.opt: Regenerate.
12180 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12181 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12182 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12183 Add matching rules for -march=armv8.4-a and extensions.
12184 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12186 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12189 * config/rx/rx.md (BW): New mode attribute.
12190 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12192 2018-01-11 Richard Biener <rguenther@suse.de>
12194 PR tree-optimization/83435
12195 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12196 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12197 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12199 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12200 Alan Hayward <alan.hayward@arm.com>
12201 David Sherwood <david.sherwood@arm.com>
12203 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12205 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12206 (aarch64_print_address_internal): Use it to check for a zero offset.
12208 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12209 Alan Hayward <alan.hayward@arm.com>
12210 David Sherwood <david.sherwood@arm.com>
12212 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12213 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12214 Return a poly_int64 rather than a HOST_WIDE_INT.
12215 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12216 rather than a HOST_WIDE_INT.
12217 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12218 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12219 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12220 final_offset from HOST_WIDE_INT to poly_int64.
12221 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12222 to_constant when getting the number of units in an Advanced SIMD
12224 (aarch64_builtin_vectorized_function): Check for a constant number
12226 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12228 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12229 attribute instead of GET_MODE_NUNITS.
12230 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12231 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12232 GET_MODE_SIZE for fixed-size registers.
12233 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12234 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12235 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12236 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12237 (aarch64_print_operand, aarch64_print_address_internal)
12238 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12239 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12240 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12241 Handle polynomial GET_MODE_SIZE.
12242 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12243 wider than SImode without modification.
12244 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12245 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12246 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12247 passing and returning SVE modes.
12248 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12249 rather than GEN_INT.
12250 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12251 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12252 (aarch64_allocate_and_probe_stack_space): Likewise.
12253 (aarch64_layout_frame): Cope with polynomial offsets.
12254 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12255 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12256 polynomial offsets.
12257 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12258 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12259 poly_int64 rather than a HOST_WIDE_INT.
12260 (aarch64_get_separate_components, aarch64_process_components)
12261 (aarch64_expand_prologue, aarch64_expand_epilogue)
12262 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12263 (aarch64_anchor_offset): New function, split out from...
12264 (aarch64_legitimize_address): ...here.
12265 (aarch64_builtin_vectorization_cost): Handle polynomial
12266 TYPE_VECTOR_SUBPARTS.
12267 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12269 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12270 number of elements from the PARALLEL rather than the mode.
12271 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12272 rather than GET_MODE_BITSIZE.
12273 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12274 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12275 (aarch64_expand_vec_perm_const_1): Handle polynomial
12276 d->perm.length () and d->perm elements.
12277 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12278 Apply to_constant to d->perm elements.
12279 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12280 polynomial CONST_VECTOR_NUNITS.
12281 (aarch64_move_pointer): Take amount as a poly_int64 rather
12283 (aarch64_progress_pointer): Avoid temporary variable.
12284 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12285 the mode attribute instead of GET_MODE.
12287 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12292 x exists before using it.
12293 (aarch64_add_constant_internal): Rename to...
12294 (aarch64_add_offset_1): ...this. Replace regnum with separate
12295 src and dest rtxes. Handle the case in which they're different,
12296 including when the offset is zero. Replace scratchreg with an rtx.
12297 Use 2 additions if there is no spare register into which we can
12298 move a 16-bit constant.
12299 (aarch64_add_constant): Delete.
12300 (aarch64_add_offset): Replace reg with separate src and dest
12301 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12302 Use aarch64_add_offset_1.
12303 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12304 an rtx rather than an int. Take the delta as a poly_int64
12305 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12306 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12307 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12308 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12309 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12310 and aarch64_add_sp.
12311 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12312 aarch64_add_constant.
12314 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12316 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12317 Use scalar_float_mode.
12319 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12321 * config/aarch64/aarch64-simd.md
12322 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12323 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12324 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12325 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12326 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12327 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12328 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12329 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12330 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12331 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12333 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12336 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12337 targ_options->x_arm_arch_string is non NULL.
12339 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12341 * config/aarch64/aarch64.h
12342 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12344 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12347 * expmed.c (emit_store_flag_force): Swap if const op0
12348 and change VOIDmode to mode of op0.
12350 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12352 PR rtl-optimization/83761
12353 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12354 than bytes to mode_for_size.
12356 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12358 PR middle-end/83189
12359 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12360 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12363 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12365 PR middle-end/83575
12366 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12367 when in layout mode.
12368 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12369 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12372 2018-01-10 Michael Collison <michael.collison@arm.com>
12374 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12375 * config/aarch64/aarch64-option-extension.def: Add
12376 AARCH64_OPT_EXTENSION of 'fp16fml'.
12377 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12378 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12379 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12380 * config/aarch64/constraints.md (Ui7): New constraint.
12381 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12382 (VFMLA_SEL_W): Ditto.
12385 (VFMLA16_LOW): New int iterator.
12386 (VFMLA16_HIGH): Ditto.
12387 (UNSPEC_FMLAL): New unspec.
12388 (UNSPEC_FMLSL): Ditto.
12389 (UNSPEC_FMLAL2): Ditto.
12390 (UNSPEC_FMLSL2): Ditto.
12391 (f16mac): New code attribute.
12392 * config/aarch64/aarch64-simd-builtins.def
12393 (aarch64_fmlal_lowv2sf): Ditto.
12394 (aarch64_fmlsl_lowv2sf): Ditto.
12395 (aarch64_fmlalq_lowv4sf): Ditto.
12396 (aarch64_fmlslq_lowv4sf): Ditto.
12397 (aarch64_fmlal_highv2sf): Ditto.
12398 (aarch64_fmlsl_highv2sf): Ditto.
12399 (aarch64_fmlalq_highv4sf): Ditto.
12400 (aarch64_fmlslq_highv4sf): Ditto.
12401 (aarch64_fmlal_lane_lowv2sf): Ditto.
12402 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12403 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12404 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12405 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12406 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12407 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12408 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12409 (aarch64_fmlal_lane_highv2sf): Ditto.
12410 (aarch64_fmlsl_lane_highv2sf): Ditto.
12411 (aarch64_fmlal_laneq_highv2sf): Ditto.
12412 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12413 (aarch64_fmlalq_lane_highv4sf): Ditto.
12414 (aarch64_fmlsl_lane_highv4sf): Ditto.
12415 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12416 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12417 * config/aarch64/aarch64-simd.md:
12418 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12419 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12420 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12421 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12422 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12423 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12424 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12425 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12426 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12427 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12428 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12429 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12430 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12431 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12432 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12433 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12434 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12435 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12436 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12437 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12438 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12439 (vfmlsl_low_u32): Ditto.
12440 (vfmlalq_low_u32): Ditto.
12441 (vfmlslq_low_u32): Ditto.
12442 (vfmlal_high_u32): Ditto.
12443 (vfmlsl_high_u32): Ditto.
12444 (vfmlalq_high_u32): Ditto.
12445 (vfmlslq_high_u32): Ditto.
12446 (vfmlal_lane_low_u32): Ditto.
12447 (vfmlsl_lane_low_u32): Ditto.
12448 (vfmlal_laneq_low_u32): Ditto.
12449 (vfmlsl_laneq_low_u32): Ditto.
12450 (vfmlalq_lane_low_u32): Ditto.
12451 (vfmlslq_lane_low_u32): Ditto.
12452 (vfmlalq_laneq_low_u32): Ditto.
12453 (vfmlslq_laneq_low_u32): Ditto.
12454 (vfmlal_lane_high_u32): Ditto.
12455 (vfmlsl_lane_high_u32): Ditto.
12456 (vfmlal_laneq_high_u32): Ditto.
12457 (vfmlsl_laneq_high_u32): Ditto.
12458 (vfmlalq_lane_high_u32): Ditto.
12459 (vfmlslq_lane_high_u32): Ditto.
12460 (vfmlalq_laneq_high_u32): Ditto.
12461 (vfmlslq_laneq_high_u32): Ditto.
12462 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12463 (AARCH64_FL_FOR_ARCH8_4): New.
12464 (AARCH64_ISA_F16FML): New ISA flag.
12465 (TARGET_F16FML): New feature flag for fp16fml.
12466 (doc/invoke.texi): Document new fp16fml option.
12468 2018-01-10 Michael Collison <michael.collison@arm.com>
12470 * config/aarch64/aarch64-builtins.c:
12471 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12472 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12473 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12474 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12475 (AARCH64_ISA_SHA3): New ISA flag.
12476 (TARGET_SHA3): New feature flag for sha3.
12477 * config/aarch64/iterators.md (sha512_op): New int attribute.
12478 (CRYPTO_SHA512): New int iterator.
12479 (UNSPEC_SHA512H): New unspec.
12480 (UNSPEC_SHA512H2): Ditto.
12481 (UNSPEC_SHA512SU0): Ditto.
12482 (UNSPEC_SHA512SU1): Ditto.
12483 * config/aarch64/aarch64-simd-builtins.def
12484 (aarch64_crypto_sha512hqv2di): New builtin.
12485 (aarch64_crypto_sha512h2qv2di): Ditto.
12486 (aarch64_crypto_sha512su0qv2di): Ditto.
12487 (aarch64_crypto_sha512su1qv2di): Ditto.
12488 (aarch64_eor3qv8hi): Ditto.
12489 (aarch64_rax1qv2di): Ditto.
12490 (aarch64_xarqv2di): Ditto.
12491 (aarch64_bcaxqv8hi): Ditto.
12492 * config/aarch64/aarch64-simd.md:
12493 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12494 (aarch64_crypto_sha512su0qv2di): Ditto.
12495 (aarch64_crypto_sha512su1qv2di): Ditto.
12496 (aarch64_eor3qv8hi): Ditto.
12497 (aarch64_rax1qv2di): Ditto.
12498 (aarch64_xarqv2di): Ditto.
12499 (aarch64_bcaxqv8hi): Ditto.
12500 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12501 (vsha512h2q_u64): Ditto.
12502 (vsha512su0q_u64): Ditto.
12503 (vsha512su1q_u64): Ditto.
12504 (veor3q_u16): Ditto.
12505 (vrax1q_u64): Ditto.
12506 (vxarq_u64): Ditto.
12507 (vbcaxq_u16): Ditto.
12508 * config/arm/types.md (crypto_sha512): New type attribute.
12509 (crypto_sha3): Ditto.
12510 (doc/invoke.texi): Document new sha3 option.
12512 2018-01-10 Michael Collison <michael.collison@arm.com>
12514 * config/aarch64/aarch64-builtins.c:
12515 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12516 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12517 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12518 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12519 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12520 (AARCH64_ISA_SM4): New ISA flag.
12521 (TARGET_SM4): New feature flag for sm4.
12522 * config/aarch64/aarch64-simd-builtins.def
12523 (aarch64_sm3ss1qv4si): Ditto.
12524 (aarch64_sm3tt1aq4si): Ditto.
12525 (aarch64_sm3tt1bq4si): Ditto.
12526 (aarch64_sm3tt2aq4si): Ditto.
12527 (aarch64_sm3tt2bq4si): Ditto.
12528 (aarch64_sm3partw1qv4si): Ditto.
12529 (aarch64_sm3partw2qv4si): Ditto.
12530 (aarch64_sm4eqv4si): Ditto.
12531 (aarch64_sm4ekeyqv4si): Ditto.
12532 * config/aarch64/aarch64-simd.md:
12533 (aarch64_sm3ss1qv4si): Ditto.
12534 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12535 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12536 (aarch64_sm4eqv4si): Ditto.
12537 (aarch64_sm4ekeyqv4si): Ditto.
12538 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12539 (sm3part_op): Ditto.
12540 (CRYPTO_SM3TT): Ditto.
12541 (CRYPTO_SM3PART): Ditto.
12542 (UNSPEC_SM3SS1): New unspec.
12543 (UNSPEC_SM3TT1A): Ditto.
12544 (UNSPEC_SM3TT1B): Ditto.
12545 (UNSPEC_SM3TT2A): Ditto.
12546 (UNSPEC_SM3TT2B): Ditto.
12547 (UNSPEC_SM3PARTW1): Ditto.
12548 (UNSPEC_SM3PARTW2): Ditto.
12549 (UNSPEC_SM4E): Ditto.
12550 (UNSPEC_SM4EKEY): Ditto.
12551 * config/aarch64/constraints.md (Ui2): New constraint.
12552 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12553 * config/arm/types.md (crypto_sm3): New type attribute.
12554 (crypto_sm4): Ditto.
12555 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12556 (vsm3tt1aq_u32): Ditto.
12557 (vsm3tt1bq_u32): Ditto.
12558 (vsm3tt2aq_u32): Ditto.
12559 (vsm3tt2bq_u32): Ditto.
12560 (vsm3partw1q_u32): Ditto.
12561 (vsm3partw2q_u32): Ditto.
12562 (vsm4eq_u32): Ditto.
12563 (vsm4ekeyq_u32): Ditto.
12564 (doc/invoke.texi): Document new sm4 option.
12566 2018-01-10 Michael Collison <michael.collison@arm.com>
12568 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12569 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12570 (AARCH64_FL_FOR_ARCH8_4): New.
12571 (AARCH64_FL_V8_4): New flag.
12572 (doc/invoke.texi): Document new armv8.4-a option.
12574 2018-01-10 Michael Collison <michael.collison@arm.com>
12576 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12577 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12578 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12579 * config/aarch64/aarch64-option-extension.def: Add
12580 AARCH64_OPT_EXTENSION of 'sha2'.
12581 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12582 (crypto): Disable sha2 and aes if crypto disabled.
12583 (crypto): Enable aes and sha2 if enabled.
12584 (simd): Disable sha2 and aes if simd disabled.
12585 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12587 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12588 (TARGET_SHA2): New feature flag for sha2.
12589 (TARGET_AES): New feature flag for aes.
12590 * config/aarch64/aarch64-simd.md:
12591 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12592 conditional on TARGET_AES.
12593 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12594 (aarch64_crypto_sha1hsi): Make pattern conditional
12596 (aarch64_crypto_sha1hv4si): Ditto.
12597 (aarch64_be_crypto_sha1hv4si): Ditto.
12598 (aarch64_crypto_sha1su1v4si): Ditto.
12599 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12600 (aarch64_crypto_sha1su0v4si): Ditto.
12601 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12602 (aarch64_crypto_sha256su0v4si): Ditto.
12603 (aarch64_crypto_sha256su1v4si): Ditto.
12604 (doc/invoke.texi): Document new aes and sha2 options.
12606 2018-01-10 Martin Sebor <msebor@redhat.com>
12608 PR tree-optimization/83781
12609 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12612 2018-01-11 Martin Sebor <msebor@gmail.com>
12613 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12615 PR tree-optimization/83501
12616 PR tree-optimization/81703
12618 * tree-ssa-strlen.c (get_string_cst): Rename...
12619 (get_string_len): ...to this. Handle global constants.
12620 (handle_char_store): Adjust.
12622 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12623 Jim Wilson <jimw@sifive.com>
12625 * config/riscv/riscv-protos.h (riscv_output_return): New.
12626 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12627 (riscv_attribute_table, riscv_output_return),
12628 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12629 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12630 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12632 (riscv_expand_prologue): Add early return for naked function.
12633 (riscv_expand_epilogue): Likewise.
12634 (riscv_function_ok_for_sibcall): Return false for naked function.
12635 (riscv_set_current_function): New.
12636 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12637 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12638 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12639 * doc/extend.texi (RISC-V Function Attributes): New.
12641 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12643 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12644 check for 128-bit long double before checking TCmode.
12645 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12646 128-bit long doubles before checking TFmode or TCmode.
12647 (FLOAT128_IBM_P): Likewise.
12649 2018-01-10 Martin Sebor <msebor@redhat.com>
12651 PR tree-optimization/83671
12652 * builtins.c (c_strlen): Unconditionally return zero for the empty
12654 Use -Warray-bounds for warnings.
12655 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12656 for non-constant array indices with COMPONENT_REF, arrays of
12657 arrays, and pointers to arrays.
12658 (gimple_fold_builtin_strlen): Determine and set length range for
12659 non-constant character arrays.
12661 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12663 PR middle-end/81897
12664 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12667 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12669 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12671 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12674 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12675 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12676 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12677 indexed_or_indirect_operand predicate.
12678 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12679 (*vsx_le_perm_load_v8hi): Likewise.
12680 (*vsx_le_perm_load_v16qi): Likewise.
12681 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12682 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12683 (*vsx_le_perm_store_v8hi): Likewise.
12684 (*vsx_le_perm_store_v16qi): Likewise.
12685 (eight unnamed splitters): Likewise.
12687 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12689 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12690 * config/rs6000/emmintrin.h: Likewise.
12691 * config/rs6000/mmintrin.h: Likewise.
12692 * config/rs6000/xmmintrin.h: Likewise.
12694 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12697 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12699 * tree.c (tree_nop_conversion): Return true for location wrapper
12701 (maybe_wrap_with_location): New function.
12702 (selftest::check_strip_nops): New function.
12703 (selftest::test_location_wrappers): New function.
12704 (selftest::tree_c_tests): Call it.
12705 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12706 (maybe_wrap_with_location): New decl.
12707 (EXPR_LOCATION_WRAPPER_P): New macro.
12708 (location_wrapper_p): New inline function.
12709 (tree_strip_any_location_wrapper): New inline function.
12711 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12714 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12715 stack_realign_offset for the largest alignment of stack slot
12717 (ix86_find_max_used_stack_alignment): New function.
12718 (ix86_finalize_stack_frame_flags): Use it. Set
12719 max_used_stack_alignment if we don't realign stack.
12720 * config/i386/i386.h (machine_function): Add
12721 max_used_stack_alignment.
12723 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12725 * config/arm/arm.opt (-mbranch-cost): New option.
12726 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12729 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12732 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12733 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12735 2018-01-10 Richard Biener <rguenther@suse.de>
12738 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12739 early out so it also covers the case where we have a non-NULL
12742 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12744 PR tree-optimization/83753
12745 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12746 for non-strided grouped accesses if the number of elements is 1.
12748 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12751 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12752 * i386.h (TARGET_USE_GATHER): Define.
12753 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12755 2018-01-10 Martin Liska <mliska@suse.cz>
12758 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12759 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12761 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12762 CLEANUP_NO_PARTITIONING is not set.
12764 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12766 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12767 for vectors, as a partial revert of r254296.
12768 * rtl.h (const_vec_p): Delete.
12769 (const_vec_duplicate_p): Don't test for vector CONSTs.
12770 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12771 * expmed.c (make_tree): Likewise.
12774 * common.md (E, F): Use CONSTANT_P instead of checking for
12776 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12777 checking for CONST_VECTOR.
12779 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12781 PR middle-end/83575
12782 * predict.c (force_edge_cold): Handle in more sane way edges
12783 with no prediction.
12785 2018-01-09 Carl Love <cel@us.ibm.com>
12787 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12789 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12790 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12791 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12792 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12793 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12794 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12795 * config/rs6000/rs6000-protos.h: Add extern defition for
12796 rs6000_generate_float2_double_code.
12797 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12799 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12800 (float2_v2df): Add define_expand.
12802 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12805 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12806 op_mode in the force_to_mode call.
12808 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12810 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12811 instead of checking each element individually.
12812 (aarch64_evpc_uzp): Likewise.
12813 (aarch64_evpc_zip): Likewise.
12814 (aarch64_evpc_ext): Likewise.
12815 (aarch64_evpc_rev): Likewise.
12816 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12817 instead of checking each element individually. Return true without
12819 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12820 whether all selected elements come from the same input, instead of
12821 checking each element individually. Remove calls to gen_rtx_REG,
12822 start_sequence and end_sequence and instead assert that no rtl is
12825 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12827 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12828 order of HIGH and CONST checks.
12830 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12832 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12833 if the destination isn't an SSA_NAME.
12835 2018-01-09 Richard Biener <rguenther@suse.de>
12837 PR tree-optimization/83668
12838 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12840 (canonicalize_loop_form): ... here, renamed from ...
12841 (canonicalize_loop_closed_ssa_form): ... this and amended to
12842 swap successor edges for loop exit blocks to make us use
12843 the RPO order we need for initial schedule generation.
12845 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12847 PR tree-optimization/64811
12848 * match.pd: When optimizing comparisons with Inf, avoid
12849 introducing or losing exceptions from comparisons with NaN.
12851 2018-01-09 Martin Liska <mliska@suse.cz>
12854 * asan.c (shadow_mem_size): Add gcc_assert.
12856 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12858 Don't save registers in main().
12861 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12862 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12863 * config/avr/avr.c (avr_set_current_function): Don't error if
12864 naked, OS_task or OS_main are specified at the same time.
12865 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12867 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12869 * common/config/avr/avr-common.c (avr_option_optimization_table):
12870 Switch on -mmain-is-OS_task for optimizing compilations.
12872 2018-01-09 Richard Biener <rguenther@suse.de>
12874 PR tree-optimization/83572
12875 * graphite.c: Include cfganal.h.
12876 (graphite_transform_loops): Connect infinite loops to exit
12877 and remove fake edges at the end.
12879 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12881 * ipa-inline.c (edge_badness): Revert accidental checkin.
12883 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12886 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12887 symbols; not inline clones.
12889 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12892 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12893 hard registers. Formatting fixes.
12895 PR preprocessor/83722
12896 * gcc.c (try_generate_repro): Pass
12897 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12898 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12901 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12902 Kito Cheng <kito.cheng@gmail.com>
12904 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12905 (riscv_leaf_function_p): Delete.
12906 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12908 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12910 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12912 (do_ifelse): New function.
12913 (do_isel): New function.
12914 (do_sub3): New function.
12915 (do_add3): New function.
12916 (do_load_mask_compare): New function.
12917 (do_overlap_load_compare): New function.
12918 (expand_compare_loop): New function.
12919 (expand_block_compare): Call expand_compare_loop() when appropriate.
12920 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12921 option description.
12922 (-mblock-compare-inline-loop-limit): New option.
12924 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12927 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12928 Reverse order of second and third operands in first alternative.
12929 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12930 of first and second elements in UNSPEC_VPERMR vector.
12931 (altivec_expand_vec_perm_le): Likewise.
12933 2018-01-08 Jeff Law <law@redhat.com>
12935 PR rtl-optimizatin/81308
12936 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12937 (process_switch): If group_case_labels makes a change, then set
12939 (pass_convert_switch::execute): If a switch is converted, then
12940 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12942 PR rtl-optimization/81308
12943 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12946 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12948 PR target/83663 - Revert r255946
12949 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12950 generation for cases where splatting a value is not useful.
12951 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12952 across a vec_duplicate and a paradoxical subreg forming a vector
12953 mode to a vec_concat.
12955 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12957 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12958 -march=armv8.3-a variants.
12959 * config/arm/t-multilib: Likewise.
12960 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12962 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12964 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12966 (cceq_ior_compare_complement): Give it a name so I can use it, and
12967 change boolean_or_operator predicate to boolean_operator so it can
12968 be used to generate a crand.
12969 (eqne): New code iterator.
12970 (bd/bd_neg): New code_attrs.
12971 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12972 a single define_insn.
12973 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12974 decrement (bdnzt/bdnzf/bdzt/bdzf).
12975 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12976 with the new names of the branch decrement patterns, and added the
12977 names of the branch decrement conditional patterns.
12979 2018-01-08 Richard Biener <rguenther@suse.de>
12981 PR tree-optimization/83563
12982 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12985 2018-01-08 Richard Biener <rguenther@suse.de>
12987 PR middle-end/83713
12988 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12990 2018-01-08 Richard Biener <rguenther@suse.de>
12992 PR tree-optimization/83685
12993 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12994 references to abnormals.
12996 2018-01-08 Richard Biener <rguenther@suse.de>
12999 * dwarf2out.c (output_indirect_strings): Handle empty
13000 skeleton_debug_str_hash.
13001 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13003 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13005 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13006 (emit_store_direct): Likewise.
13007 (arc_trampoline_adjust_address): Likewise.
13008 (arc_asm_trampoline_template): New function.
13009 (arc_initialize_trampoline): Use asm_trampoline_template.
13010 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13011 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13012 * config/arc/arc.md (flush_icache): Delete pattern.
13014 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13016 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13017 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13020 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13023 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13024 by not USED_FOR_TARGET.
13025 (make_pass_resolve_sw_modes): Likewise.
13027 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13029 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13032 2018-01-08 Richard Biener <rguenther@suse.de>
13034 PR middle-end/83580
13035 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13037 2018-01-08 Richard Biener <rguenther@suse.de>
13039 PR middle-end/83517
13040 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13042 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13044 PR middle-end/81897
13045 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13046 basic blocks with a small number of successors.
13047 (convert_control_dep_chain_into_preds): Improve handling of
13049 (dump_predicates): Split apart into...
13050 (dump_pred_chain): ...here...
13051 (dump_pred_info): ...and here.
13052 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13053 (can_chain_union_be_invalidated_p): Improve check for invalidation
13055 (uninit_uses_cannot_happen): Avoid unnecessary if
13056 convert_control_dep_chain_into_preds yielded nothing.
13058 2018-01-06 Martin Sebor <msebor@redhat.com>
13060 PR tree-optimization/83640
13061 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13062 subtracting negative offset from size.
13063 (builtin_access::overlap): Adjust offset bounds of the access to fall
13064 within the size of the object if possible.
13066 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13068 PR rtl-optimization/83699
13069 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13070 extract_bit_field_as_subreg to cases in which the extracted
13071 value is also a vector.
13073 * lra-constraints.c (process_alt_operands): Test for the equivalence
13074 substitutions when detecting a possible reload cycle.
13076 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13079 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13080 by default if flag_selective_schedling{,2}. Formatting fixes.
13082 PR rtl-optimization/83682
13083 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13084 if it has non-VECTOR_MODE element mode.
13085 (vec_duplicate_p): Likewise.
13087 PR middle-end/83694
13088 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13089 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13091 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13094 * config/i386/i386-builtin.def
13095 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13096 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13097 Require also OPTION_MASK_ISA_AVX512F in addition to
13098 OPTION_MASK_ISA_GFNI.
13099 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13100 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13101 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13102 to OPTION_MASK_ISA_GFNI.
13103 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13104 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13105 OPTION_MASK_ISA_AVX512BW.
13106 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13107 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13108 addition to OPTION_MASK_ISA_GFNI.
13109 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13110 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13111 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13112 to OPTION_MASK_ISA_GFNI.
13113 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13114 a requirement for all ISAs rather than any of them with a few
13116 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13118 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13119 bitmasks to be enabled with 3 exceptions, instead of requiring any
13120 enabled ISA with lots of exceptions.
13121 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13122 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13123 Change avx512bw in isa attribute to avx512f.
13124 * config/i386/sgxintrin.h: Add license boilerplate.
13125 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13126 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13127 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13128 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13130 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13131 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13132 temporarily sse2 rather than sse if not enabled already.
13135 * config/i386/sse.md (VI248_VLBW): Rename to ...
13136 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13137 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13138 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13139 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13140 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13141 mode iterator instead of VI248_VLBW.
13143 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13145 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13146 (record_modified): Skip clobbers; add debug output.
13147 (param_change_prob): Use sreal frequencies.
13149 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13151 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13152 punt for user-aligned variables.
13154 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13156 * tree-chrec.c (chrec_contains_symbols): Return true for
13159 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13162 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13163 of (x|y) == x for BICS pattern.
13165 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13167 PR tree-optimization/83605
13168 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13169 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13172 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13174 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13175 * config/epiphany/rtems.h: New file.
13177 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13178 Uros Bizjak <ubizjak@gmail.com>
13181 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13182 QIreg_operand instead of register_operand predicate.
13183 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13184 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13185 comments instead of -fmitigate[-_]rop.
13187 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13190 * cgraphunit.c (symbol_table::compile): Switch to text_section
13191 before calling assembly_start debug hook.
13192 * run-rtl-passes.c (run_rtl_passes): Likewise.
13195 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13197 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13198 range_int_cst_p rather than !symbolic_range_p before calling
13199 extract_range_from_multiplicative_op_1.
13201 2018-01-04 Jeff Law <law@redhat.com>
13203 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13204 redundant test in assertion.
13206 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13208 * doc/rtl.texi: Document machine_mode wrapper classes.
13210 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13212 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13213 using tree_to_uhwi.
13215 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13217 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13218 the VEC_PERM_EXPR fold to fail.
13220 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13223 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13224 to switched_sections.
13226 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13229 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13230 test for d.testing.
13232 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13235 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13236 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13238 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13241 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13242 is BLKmode and bitpos not zero or mode change is needed.
13244 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13247 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13250 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13253 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13254 instead of MULT rtx. Update all corresponding splitters.
13255 (*saddl_se): Ditto.
13256 (*ssub<modesuffix>): Ditto.
13257 (*ssubl_se): Ditto.
13258 (*cmp_sadd_di): Update split patterns.
13259 (*cmp_sadd_si): Ditto.
13260 (*cmp_sadd_sidi): Ditto.
13261 (*cmp_ssub_di): Ditto.
13262 (*cmp_ssub_si): Ditto.
13263 (*cmp_ssub_sidi): Ditto.
13264 * config/alpha/predicates.md (const23_operand): New predicate.
13265 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13266 Look for ASHIFT, not MULT inner operand.
13267 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13269 2018-01-04 Martin Liska <mliska@suse.cz>
13271 PR gcov-profile/83669
13272 * gcov.c (output_intermediate_file): Add version to intermediate
13274 * doc/gcov.texi: Document new field 'version' in intermediate
13275 file format. Fix location of '-k' option of gcov command.
13277 2018-01-04 Martin Liska <mliska@suse.cz>
13280 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13282 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13284 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13286 2018-01-03 Martin Sebor <msebor@redhat.com>
13288 PR tree-optimization/83655
13289 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13290 checking calls with invalid arguments.
13292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13294 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13295 (vectorizable_mask_load_store): Delete.
13296 (vectorizable_call): Return false for masked loads and stores.
13297 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13298 instead of gimple_assign_rhs1.
13299 (vectorizable_load): Handle IFN_MASK_LOAD.
13300 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13304 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13306 (vectorizable_mask_load_store): ...here.
13307 (vectorizable_load): ...and here.
13309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13311 * tree-vect-stmts.c (vect_build_all_ones_mask)
13312 (vect_build_zero_merge_argument): New functions, split out from...
13313 (vectorizable_load): ...here.
13315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13317 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13319 (vectorizable_mask_load_store): ...here.
13320 (vectorizable_store): ...and here.
13322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13324 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13326 (vectorizable_mask_load_store): ...here.
13328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13330 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13331 (vect_model_store_cost): Take a vec_load_store_type instead of a
13333 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13334 (vect_model_store_cost): Take a vec_load_store_type instead of a
13336 (vectorizable_mask_load_store): Update accordingly.
13337 (vectorizable_store): Likewise.
13338 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13342 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13343 IFN_MASK_LOAD calls here rather than...
13344 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13347 Alan Hayward <alan.hayward@arm.com>
13348 David Sherwood <david.sherwood@arm.com>
13350 * expmed.c (extract_bit_field_1): For vector extracts,
13351 fall back to extract_bit_field_as_subreg if vec_extract
13354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13355 Alan Hayward <alan.hayward@arm.com>
13356 David Sherwood <david.sherwood@arm.com>
13358 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13359 they are variable or constant sized.
13360 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13361 slots for constant-sized data.
13363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13364 Alan Hayward <alan.hayward@arm.com>
13365 David Sherwood <david.sherwood@arm.com>
13367 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13368 handling COND_EXPRs with boolean comparisons, try to find a better
13369 basis for the mask type than the boolean itself.
13371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13373 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13374 is calculated and how it can be overridden.
13375 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13376 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13378 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13382 Alan Hayward <alan.hayward@arm.com>
13383 David Sherwood <david.sherwood@arm.com>
13385 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13386 Remove the mode argument.
13387 (aarch64_simd_valid_immediate): Remove the mode and inverse
13389 * config/aarch64/iterators.md (bitsize): New iterator.
13390 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13391 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13392 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13393 aarch64_simd_valid_immediate.
13394 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13395 (aarch64_reg_or_bic_imm): Likewise.
13396 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13397 with an insn_type enum and msl with a modifier_type enum.
13398 Replace element_width with a scalar_mode. Change the shift
13399 to unsigned int. Add constructors for scalar_float_mode and
13400 scalar_int_mode elements.
13401 (aarch64_vect_float_const_representable_p): Delete.
13402 (aarch64_can_const_movi_rtx_p)
13403 (aarch64_simd_scalar_immediate_valid_for_move)
13404 (aarch64_simd_make_constant): Update call to
13405 aarch64_simd_valid_immediate.
13406 (aarch64_advsimd_valid_immediate_hs): New function.
13407 (aarch64_advsimd_valid_immediate): Likewise.
13408 (aarch64_simd_valid_immediate): Remove mode and inverse
13409 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13410 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13411 and aarch64_float_const_representable_p on the result.
13412 (aarch64_output_simd_mov_immediate): Remove mode argument.
13413 Update call to aarch64_simd_valid_immediate and use of
13414 simd_immediate_info.
13415 (aarch64_output_scalar_simd_mov_immediate): Update call
13418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13419 Alan Hayward <alan.hayward@arm.com>
13420 David Sherwood <david.sherwood@arm.com>
13422 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13423 (mode_nunits): Likewise CONST_MODE_NUNITS.
13424 * machmode.def (ADJUST_NUNITS): Document.
13425 * genmodes.c (mode_data::need_nunits_adj): New field.
13426 (blank_mode): Update accordingly.
13427 (adj_nunits): New variable.
13428 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13430 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13431 listed in adj_nunits.
13432 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13433 listed in adj_nunits. Don't emit case statements for such modes.
13434 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13435 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13436 nothing if adj_nunits is nonnull.
13437 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13438 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13439 (emit_mode_fbit): Update use of print_maybe_const_decl.
13440 (emit_move_size): Likewise. Treat the array as non-const
13442 (emit_mode_adjustments): Handle adj_nunits.
13444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13446 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13447 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13448 (VECTOR_MODES): Use it.
13449 (make_vector_modes): Take the prefix as an argument.
13451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13452 Alan Hayward <alan.hayward@arm.com>
13453 David Sherwood <david.sherwood@arm.com>
13455 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13456 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13457 for MODE_VECTOR_BOOL.
13458 * machmode.def (VECTOR_BOOL_MODE): Document.
13459 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13460 (make_vector_bool_mode): New function.
13461 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13463 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13464 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13466 * stor-layout.c (int_mode_for_mode): Likewise.
13467 * tree.c (build_vector_type_for_mode): Likewise.
13468 * varasm.c (output_constant_pool_2): Likewise.
13469 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13470 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13471 for MODE_VECTOR_BOOL.
13472 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13473 of mode class checks.
13474 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13475 instead of a list of mode class checks.
13476 (expand_vector_scalar_condition): Likewise.
13477 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13480 Alan Hayward <alan.hayward@arm.com>
13481 David Sherwood <david.sherwood@arm.com>
13483 * machmode.h (mode_size): Change from unsigned short to
13485 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13486 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13487 or if measurement_type is not polynomial.
13488 (fixed_size_mode::includes_p): Check for constant-sized modes.
13489 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13490 return a poly_uint16 rather than an unsigned short.
13491 (emit_mode_size): Change the type of mode_size from unsigned short
13492 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13493 (emit_mode_adjustments): Cope with polynomial vector sizes.
13494 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13496 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13498 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13499 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13500 * caller-save.c (setup_save_areas): Likewise.
13501 (replace_reg_with_saved_mem): Likewise.
13502 * calls.c (emit_library_call_value_1): Likewise.
13503 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13504 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13505 (gen_lowpart_for_combine): Likewise.
13506 * convert.c (convert_to_integer_1): Likewise.
13507 * cse.c (equiv_constant, cse_insn): Likewise.
13508 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13509 (cselib_subst_to_values): Likewise.
13510 * dce.c (word_dce_process_block): Likewise.
13511 * df-problems.c (df_word_lr_mark_ref): Likewise.
13512 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13513 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13514 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13515 (rtl_for_decl_location): Likewise.
13516 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13517 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13518 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13519 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13520 (expand_expr_real_1): Likewise.
13521 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13522 (pad_below): Likewise.
13523 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13524 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13525 * ira.c (get_subreg_tracking_sizes): Likewise.
13526 * ira-build.c (ira_create_allocno_objects): Likewise.
13527 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13528 (ira_sort_regnos_for_alter_reg): Likewise.
13529 * ira-costs.c (record_operand_costs): Likewise.
13530 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13531 (resolve_simple_move): Likewise.
13532 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13533 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13534 (lra_constraints): Likewise.
13535 (CONST_POOL_OK_P): Reject variable-sized modes.
13536 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13537 (add_pseudo_to_slot, lra_spill): Likewise.
13538 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13539 * optabs-query.c (get_best_extraction_insn): Likewise.
13540 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13541 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13542 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13543 * recog.c (offsettable_address_addr_space_p): Likewise.
13544 * regcprop.c (maybe_mode_change): Likewise.
13545 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13546 * regrename.c (build_def_use): Likewise.
13547 * regstat.c (dump_reg_info): Likewise.
13548 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13549 (find_reloads, find_reloads_subreg_address): Likewise.
13550 * reload1.c (eliminate_regs_1): Likewise.
13551 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13552 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13553 (simplify_binary_operation_1, simplify_subreg): Likewise.
13554 * targhooks.c (default_function_arg_padding): Likewise.
13555 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13556 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13557 (verify_gimple_assign_ternary): Likewise.
13558 * tree-inline.c (estimate_move_cost): Likewise.
13559 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13560 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13561 (get_address_cost_ainc): Likewise.
13562 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13563 (vect_supportable_dr_alignment): Likewise.
13564 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13565 (vectorizable_reduction): Likewise.
13566 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13567 (vectorizable_operation, vectorizable_load): Likewise.
13568 * tree.c (build_same_sized_truth_vector_type): Likewise.
13569 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13570 * var-tracking.c (emit_note_insn_var_location): Likewise.
13571 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13572 (ADDR_VEC_ALIGN): Likewise.
13574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13575 Alan Hayward <alan.hayward@arm.com>
13576 David Sherwood <david.sherwood@arm.com>
13578 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13580 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13581 or if measurement_type is polynomial.
13582 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13583 * combine.c (make_extraction): Likewise.
13584 * dse.c (find_shift_sequence): Likewise.
13585 * dwarf2out.c (mem_loc_descriptor): Likewise.
13586 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13587 (extract_bit_field, extract_low_bits): Likewise.
13588 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13589 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13590 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13591 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13592 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13593 * reload.c (find_reloads): Likewise.
13594 * reload1.c (alter_reg): Likewise.
13595 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13596 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13597 * tree-if-conv.c (predicate_mem_writes): Likewise.
13598 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13599 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13600 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13601 * valtrack.c (dead_debug_insert_temp): Likewise.
13602 * varasm.c (mergeable_constant_section): Likewise.
13603 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13606 Alan Hayward <alan.hayward@arm.com>
13607 David Sherwood <david.sherwood@arm.com>
13609 * expr.c (expand_assignment): Cope with polynomial mode sizes
13610 when assigning to a CONCAT.
13612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13613 Alan Hayward <alan.hayward@arm.com>
13614 David Sherwood <david.sherwood@arm.com>
13616 * machmode.h (mode_precision): Change from unsigned short to
13618 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13620 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13621 or if measurement_type is not polynomial.
13622 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13623 in which the mode is already known to be a scalar_int_mode.
13624 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13625 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13627 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13628 for GET_MODE_PRECISION.
13629 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13630 for GET_MODE_PRECISION.
13631 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13633 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13634 (expand_field_assignment, make_extraction): Likewise.
13635 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13636 (get_last_value): Likewise.
13637 * convert.c (convert_to_integer_1): Likewise.
13638 * cse.c (cse_insn): Likewise.
13639 * expr.c (expand_expr_real_1): Likewise.
13640 * lra-constraints.c (simplify_operand_subreg): Likewise.
13641 * optabs-query.c (can_atomic_load_p): Likewise.
13642 * optabs.c (expand_atomic_load): Likewise.
13643 (expand_atomic_store): Likewise.
13644 * ree.c (combine_reaching_defs): Likewise.
13645 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13646 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13647 * tree.h (type_has_mode_precision_p): Likewise.
13648 * ubsan.c (instrument_si_overflow): Likewise.
13650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13651 Alan Hayward <alan.hayward@arm.com>
13652 David Sherwood <david.sherwood@arm.com>
13654 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13655 polynomial numbers of units.
13656 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13657 (valid_vector_subparts_p): New function.
13658 (build_vector_type): Remove temporary shim and take the number
13659 of units as a poly_uint64 rather than an int.
13660 (build_opaque_vector_type): Take the number of units as a
13661 poly_uint64 rather than an int.
13662 * tree.c (build_vector_from_ctor): Handle polynomial
13663 TYPE_VECTOR_SUBPARTS.
13664 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13665 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13666 (build_vector_from_val): If the number of units is variable,
13667 use build_vec_duplicate_cst for constant operands and
13668 VEC_DUPLICATE_EXPR otherwise.
13669 (make_vector_type): Remove temporary is_constant ().
13670 (build_vector_type, build_opaque_vector_type): Take the number of
13671 units as a poly_uint64 rather than an int.
13672 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13674 * cfgexpand.c (expand_debug_expr): Likewise.
13675 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13676 (store_constructor, expand_expr_real_1): Likewise.
13677 (const_scalar_mask_from_tree): Likewise.
13678 * fold-const-call.c (fold_const_reduction): Likewise.
13679 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13680 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13681 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13682 (fold_relational_const): Likewise.
13683 (native_interpret_vector): Likewise. Change the size from an
13684 int to an unsigned int.
13685 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13686 TYPE_VECTOR_SUBPARTS.
13687 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13688 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13689 duplicating a non-constant operand into a variable-length vector.
13690 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13691 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13692 * ipa-icf.c (sem_variable::equals): Likewise.
13693 * match.pd: Likewise.
13694 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13695 * print-tree.c (print_node): Likewise.
13696 * stor-layout.c (layout_type): Likewise.
13697 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13698 * tree-cfg.c (verify_gimple_comparison): Likewise.
13699 (verify_gimple_assign_binary): Likewise.
13700 (verify_gimple_assign_ternary): Likewise.
13701 (verify_gimple_assign_single): Likewise.
13702 * tree-pretty-print.c (dump_generic_node): Likewise.
13703 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13704 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13705 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13706 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13707 (vect_shift_permute_load_chain): Likewise.
13708 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13709 (expand_vector_condition, optimize_vector_constructor): Likewise.
13710 (lower_vec_perm, get_compute_type): Likewise.
13711 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13712 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13713 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13714 (vect_recog_mask_conversion_pattern): Likewise.
13715 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13716 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13717 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13718 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13719 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13720 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13721 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13722 (supportable_widening_operation): Likewise.
13723 (supportable_narrowing_operation): Likewise.
13724 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13726 * varasm.c (output_constant): Likewise.
13728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13729 Alan Hayward <alan.hayward@arm.com>
13730 David Sherwood <david.sherwood@arm.com>
13732 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13733 so that both the length == 3 and length != 3 cases set up their
13734 own permute vectors. Add comments explaining why we know the
13735 number of elements is constant.
13736 (vect_permute_load_chain): Likewise.
13738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 Alan Hayward <alan.hayward@arm.com>
13740 David Sherwood <david.sherwood@arm.com>
13742 * machmode.h (mode_nunits): Change from unsigned char to
13744 (ONLY_FIXED_SIZE_MODES): New macro.
13745 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13746 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13747 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13749 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13750 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13751 or if measurement_type is not polynomial.
13752 * genmodes.c (ZERO_COEFFS): New macro.
13753 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13755 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13756 Use ZERO_COEFFS when emitting initializers.
13757 * data-streamer.h (bp_pack_poly_value): New function.
13758 (bp_unpack_poly_value): Likewise.
13759 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13760 for GET_MODE_NUNITS.
13761 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13762 for GET_MODE_NUNITS.
13763 * tree.c (make_vector_type): Remove temporary shim and make
13764 the real function take the number of units as a poly_uint64
13765 rather than an int.
13766 (build_vector_type_for_mode): Handle polynomial nunits.
13767 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13768 * emit-rtl.c (const_vec_series_p_1): Likewise.
13769 (gen_rtx_CONST_VECTOR): Likewise.
13770 * fold-const.c (test_vec_duplicate_folding): Likewise.
13771 * genrecog.c (validate_pattern): Likewise.
13772 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13773 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13774 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13775 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13776 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13777 * rtlanal.c (subreg_get_info): Likewise.
13778 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13779 (vect_grouped_load_supported): Likewise.
13780 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13781 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13782 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13783 (simplify_const_unary_operation, simplify_binary_operation_1)
13784 (simplify_const_binary_operation, simplify_ternary_operation)
13785 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13786 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13787 instead of CONST_VECTOR_NUNITS.
13788 * varasm.c (output_constant_pool_2): Likewise.
13789 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13790 explicit-encoded elements in the XVEC for variable-length vectors.
13792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13794 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * coretypes.h (fixed_size_mode): Declare.
13801 (fixed_size_mode_pod): New typedef.
13802 * builtins.h (target_builtins::x_apply_args_mode)
13803 (target_builtins::x_apply_result_mode): Change type to
13804 fixed_size_mode_pod.
13805 * builtins.c (apply_args_size, apply_result_size, result_vector)
13806 (expand_builtin_apply_args_1, expand_builtin_apply)
13807 (expand_builtin_return): Update accordingly.
13809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13811 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13812 * cselib.c (cselib_hash_rtx): Likewise.
13813 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13814 CONST_VECTOR encoding.
13816 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13817 Jeff Law <law@redhat.com>
13820 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13821 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13822 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13823 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13826 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13827 explicitly probe *sp in a noreturn function if there were any callee
13828 register saves or frame pointer is needed.
13830 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13833 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13834 BLKmode for ternary, binary or unary expressions.
13837 * var-tracking.c (delete_vta_debug_insn): New inline function.
13838 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13839 insns from get_insns () to NULL instead of each bb separately.
13840 Use delete_vta_debug_insn. No longer static.
13841 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13842 delete_vta_debug_insns callers.
13843 * rtl.h (delete_vta_debug_insns): Declare.
13844 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13845 instead of variable_tracking_main.
13847 2018-01-03 Martin Sebor <msebor@redhat.com>
13849 PR tree-optimization/83603
13850 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13851 arguments past the endof the argument list in functions declared
13852 without a prototype.
13853 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13854 Avoid checking when arguments are null.
13856 2018-01-03 Martin Sebor <msebor@redhat.com>
13859 * doc/extend.texi (attribute const): Fix a typo.
13860 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13861 issuing -Wsuggest-attribute for void functions.
13863 2018-01-03 Martin Sebor <msebor@redhat.com>
13865 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13866 offset_int::from instead of wide_int::to_shwi.
13867 (maybe_diag_overlap): Remove assertion.
13868 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13869 * gimple-ssa-sprintf.c (format_directive): Same.
13870 (parse_directive): Same.
13871 (sprintf_dom_walker::compute_format_length): Same.
13872 (try_substitute_return_value): Same.
13874 2018-01-03 Jeff Law <law@redhat.com>
13876 PR middle-end/83654
13877 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13878 non-constant residual for zero at runtime and avoid probing in
13879 that case. Reorganize code for trailing problem to mirror handling
13882 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13884 PR tree-optimization/83501
13885 * tree-ssa-strlen.c (get_string_cst): New.
13886 (handle_char_store): Call get_string_cst.
13888 2018-01-03 Martin Liska <mliska@suse.cz>
13890 PR tree-optimization/83593
13891 * tree-ssa-strlen.c: Include tree-cfg.h.
13892 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13893 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13894 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13896 (strlen_dom_walker::before_dom_children): Call
13897 gimple_purge_dead_eh_edges. Dump tranformation with details
13899 (strlen_dom_walker::before_dom_children): Update call by adding
13900 new argument cleanup_eh.
13901 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13903 2018-01-03 Martin Liska <mliska@suse.cz>
13906 * cif-code.def (VARIADIC_THUNK): New enum value.
13907 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13910 2018-01-03 Jan Beulich <jbeulich@suse.com>
13912 * sse.md (mov<mode>_internal): Tighten condition for when to use
13913 vmovdqu<ssescalarsize> for TI and OI modes.
13915 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13917 Update copyright years.
13919 2018-01-03 Martin Liska <mliska@suse.cz>
13922 * ipa-visibility.c (function_and_variable_visibility): Skip
13923 functions with noipa attribure.
13925 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13927 * gcc.c (process_command): Update copyright notice dates.
13928 * gcov-dump.c (print_version): Ditto.
13929 * gcov.c (print_version): Ditto.
13930 * gcov-tool.c (print_version): Ditto.
13931 * gengtype.c (create_file): Ditto.
13932 * doc/cpp.texi: Bump @copying's copyright year.
13933 * doc/cppinternals.texi: Ditto.
13934 * doc/gcc.texi: Ditto.
13935 * doc/gccint.texi: Ditto.
13936 * doc/gcov.texi: Ditto.
13937 * doc/install.texi: Ditto.
13938 * doc/invoke.texi: Ditto.
13940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13942 * vector-builder.h (vector_builder::m_full_nelts): Change from
13943 unsigned int to poly_uint64.
13944 (vector_builder::full_nelts): Update prototype accordingly.
13945 (vector_builder::new_vector): Likewise.
13946 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13947 (vector_builder::operator ==): Likewise.
13948 (vector_builder::finalize): Likewise.
13949 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13950 Take the number of elements as a poly_uint64 rather than an
13952 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13953 from unsigned int to poly_uint64.
13954 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13955 (vec_perm_indices::new_vector): Likewise.
13956 (vec_perm_indices::length): Likewise.
13957 (vec_perm_indices::nelts_per_input): Likewise.
13958 (vec_perm_indices::input_nelts): Likewise.
13959 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13960 number of elements per input as a poly_uint64 rather than an
13961 unsigned int. Use the original encoding for variable-length
13962 vectors, rather than clamping each individual element.
13963 For the second and subsequent elements in each pattern,
13964 clamp the step and base before clamping their sum.
13965 (vec_perm_indices::series_p): Handle polynomial element counts.
13966 (vec_perm_indices::all_in_range_p): Likewise.
13967 (vec_perm_indices_to_tree): Likewise.
13968 (vec_perm_indices_to_rtx): Likewise.
13969 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13970 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13971 (tree_vector_builder::new_binary_operation): Handle polynomial
13972 element counts. Return false if we need to know the number
13973 of elements at compile time.
13974 * fold-const.c (fold_vec_perm): Punt if the number of elements
13975 isn't known at compile time.
13977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13979 * vec-perm-indices.h (vec_perm_builder): Change element type
13980 from HOST_WIDE_INT to poly_int64.
13981 (vec_perm_indices::element_type): Update accordingly.
13982 (vec_perm_indices::clamp): Handle polynomial element_types.
13983 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13984 (vec_perm_indices::all_in_range_p): Likewise.
13985 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13987 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13988 polynomial vec_perm_indices element types.
13989 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13990 * fold-const.c (fold_vec_perm): Likewise.
13991 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13992 * tree-vect-generic.c (lower_vec_perm): Likewise.
13993 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13994 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13995 element type to HOST_WIDE_INT.
13997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13998 Alan Hayward <alan.hayward@arm.com>
13999 David Sherwood <david.sherwood@arm.com>
14001 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14002 rather than an int. Use plus_constant.
14003 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14004 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14007 Alan Hayward <alan.hayward@arm.com>
14008 David Sherwood <david.sherwood@arm.com>
14010 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14011 a HOST_WIDE_INT to a poly_int64.
14013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14014 Alan Hayward <alan.hayward@arm.com>
14015 David Sherwood <david.sherwood@arm.com>
14017 * calls.c (load_register_parameters): Cope with polynomial
14018 mode sizes. Require a constant size for BLKmode parameters
14019 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14020 forces a parameter to be padded at the lsb end in order to
14021 fill a complete number of words, require the parameter size
14022 to be ordered wrt UNITS_PER_WORD.
14024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14025 Alan Hayward <alan.hayward@arm.com>
14026 David Sherwood <david.sherwood@arm.com>
14028 * reload1.c (spill_stack_slot_width): Change element type
14029 from unsigned int to poly_uint64_pod.
14030 (alter_reg): Treat mode sizes as polynomial.
14032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14033 Alan Hayward <alan.hayward@arm.com>
14034 David Sherwood <david.sherwood@arm.com>
14036 * reload.c (complex_word_subreg_p): New function.
14037 (reload_inner_reg_of_subreg, push_reload): Use it.
14039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14040 Alan Hayward <alan.hayward@arm.com>
14041 David Sherwood <david.sherwood@arm.com>
14043 * lra-constraints.c (process_alt_operands): Reject matched
14044 operands whose sizes aren't ordered.
14045 (match_reload): Refer to this check here.
14047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14048 Alan Hayward <alan.hayward@arm.com>
14049 David Sherwood <david.sherwood@arm.com>
14051 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14052 that the mode size is in the set {1, 2, 4, 8, 16}.
14054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14055 Alan Hayward <alan.hayward@arm.com>
14056 David Sherwood <david.sherwood@arm.com>
14058 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14059 Use plus_constant instead of gen_rtx_PLUS.
14061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14062 Alan Hayward <alan.hayward@arm.com>
14063 David Sherwood <david.sherwood@arm.com>
14065 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14066 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14067 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14068 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14069 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14070 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14071 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14072 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14073 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14074 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14076 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14077 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14078 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14079 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14080 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14081 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14082 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14083 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14084 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14085 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14087 * expr.c (emit_move_resolve_push): Treat the input and result
14088 of PUSH_ROUNDING as a poly_int64.
14089 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14090 (emit_push_insn): Likewise.
14091 * lra-eliminations.c (mark_not_eliminable): Likewise.
14092 * recog.c (push_operand): Likewise.
14093 * reload1.c (elimination_effects): Likewise.
14094 * rtlanal.c (nonzero_bits1): Likewise.
14095 * calls.c (store_one_arg): Likewise. Require the padding to be
14096 known at compile time.
14098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14099 Alan Hayward <alan.hayward@arm.com>
14100 David Sherwood <david.sherwood@arm.com>
14102 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14103 Use plus_constant instead of gen_rtx_PLUS.
14105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14106 Alan Hayward <alan.hayward@arm.com>
14107 David Sherwood <david.sherwood@arm.com>
14109 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14110 rather than an int.
14112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14113 Alan Hayward <alan.hayward@arm.com>
14114 David Sherwood <david.sherwood@arm.com>
14116 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14117 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14118 via stack temporaries. Treat the mode size as polynomial too.
14120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14121 Alan Hayward <alan.hayward@arm.com>
14122 David Sherwood <david.sherwood@arm.com>
14124 * expr.c (expand_expr_real_2): When handling conversions involving
14125 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14126 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14127 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14136 Alan Hayward <alan.hayward@arm.com>
14137 David Sherwood <david.sherwood@arm.com>
14139 * combine.c (can_change_dest_mode): Handle polynomial
14140 REGMODE_NATURAL_SIZE.
14141 * expmed.c (store_bit_field_1): Likewise.
14142 * expr.c (store_constructor): Likewise.
14143 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14144 and polynomial REGMODE_NATURAL_SIZE.
14145 (gen_lowpart_common): Likewise.
14146 * reginfo.c (record_subregs_of_mode): Likewise.
14147 * rtlanal.c (read_modify_subreg_p): Likewise.
14149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14150 Alan Hayward <alan.hayward@arm.com>
14151 David Sherwood <david.sherwood@arm.com>
14153 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14154 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
14162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14163 Alan Hayward <alan.hayward@arm.com>
14164 David Sherwood <david.sherwood@arm.com>
14166 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14167 in a POINTER_PLUS_EXPR.
14169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14170 Alan Hayward <alan.hayward@arm.com>
14171 David Sherwood <david.sherwood@arm.com>
14173 * omp-simd-clone.c (simd_clone_subparts): New function.
14174 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14175 (ipa_simd_modify_function_body): Likewise.
14177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14178 Alan Hayward <alan.hayward@arm.com>
14179 David Sherwood <david.sherwood@arm.com>
14181 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14182 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14183 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14184 (expand_vector_condition, vector_element): Likewise.
14185 (subparts_gt): New function.
14186 (get_compute_type): Use subparts_gt.
14187 (count_type_subparts): Delete.
14188 (expand_vector_operations_1): Use subparts_gt instead of
14189 count_type_subparts.
14191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14192 Alan Hayward <alan.hayward@arm.com>
14193 David Sherwood <david.sherwood@arm.com>
14195 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14196 (vect_compile_time_alias): ...this new function. Do the calculation
14197 on poly_ints rather than trees.
14198 (vect_prune_runtime_alias_test_list): Update call accordingly.
14200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14201 Alan Hayward <alan.hayward@arm.com>
14202 David Sherwood <david.sherwood@arm.com>
14204 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14206 (vect_schedule_slp_instance): Likewise.
14208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14209 Alan Hayward <alan.hayward@arm.com>
14210 David Sherwood <david.sherwood@arm.com>
14212 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14213 constant and extern definitions for variable-length vectors.
14214 (vect_get_constant_vectors): Note that the number of units
14215 is known to be constant.
14217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14218 Alan Hayward <alan.hayward@arm.com>
14219 David Sherwood <david.sherwood@arm.com>
14221 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14222 of units as polynomial. Choose between WIDE and NARROW based
14225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14226 Alan Hayward <alan.hayward@arm.com>
14227 David Sherwood <david.sherwood@arm.com>
14229 * tree-vect-stmts.c (simd_clone_subparts): New function.
14230 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14233 Alan Hayward <alan.hayward@arm.com>
14234 David Sherwood <david.sherwood@arm.com>
14236 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14237 vectors as polynomial. Use build_index_vector for
14238 IFN_GOMP_SIMD_LANE.
14240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14241 Alan Hayward <alan.hayward@arm.com>
14242 David Sherwood <david.sherwood@arm.com>
14244 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14245 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14246 for variable-length vectors.
14247 (vectorizable_mask_load_store): Treat the number of units as
14248 polynomial, asserting that it is constant if the condition has
14249 already been enforced.
14250 (vectorizable_store, vectorizable_load): Likewise.
14252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14253 Alan Hayward <alan.hayward@arm.com>
14254 David Sherwood <david.sherwood@arm.com>
14256 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14257 of units as polynomial. Punt if we can't tell at compile time
14258 which vector contains the final result.
14260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14261 Alan Hayward <alan.hayward@arm.com>
14262 David Sherwood <david.sherwood@arm.com>
14264 * tree-vect-loop.c (vectorizable_induction): Treat the number
14265 of units as polynomial. Punt on SLP inductions. Use an integer
14266 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14267 cast of such a series for variable-length floating-point
14270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14271 Alan Hayward <alan.hayward@arm.com>
14272 David Sherwood <david.sherwood@arm.com>
14274 * tree.h (build_index_vector): Declare.
14275 * tree.c (build_index_vector): New function.
14276 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14277 of units as polynomial, forcibly converting it to a constant if
14278 vectorizable_reduction has already enforced the condition.
14279 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14280 to create a {1,2,3,...} vector.
14281 (vectorizable_reduction): Treat the number of units as polynomial.
14282 Choose vectype_in based on the largest scalar element size rather
14283 than the smallest number of units. Enforce the restrictions
14286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14287 Alan Hayward <alan.hayward@arm.com>
14288 David Sherwood <david.sherwood@arm.com>
14290 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14291 number of units as polynomial.
14293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14294 Alan Hayward <alan.hayward@arm.com>
14295 David Sherwood <david.sherwood@arm.com>
14297 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14298 * target.def (autovectorize_vector_sizes): Return the vector sizes
14299 by pointer, using vector_sizes rather than a bitmask.
14300 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14301 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14302 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14304 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14305 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14306 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14307 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14308 * omp-general.c (omp_max_vf): Likewise.
14309 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14310 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14311 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14312 * tree-vect-slp.c (vect_slp_bb): Likewise.
14313 * doc/tm.texi: Regenerate.
14314 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14316 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14317 the vector size as a poly_uint64 rather than an unsigned int.
14318 (current_vector_size): Change from an unsigned int to a poly_uint64.
14319 (get_vectype_for_scalar_type): Update accordingly.
14320 * tree.h (build_truth_vector_type): Take the size and number of
14321 units as a poly_uint64 rather than an unsigned int.
14322 (build_vector_type): Add a temporary overload that takes
14323 the number of units as a poly_uint64 rather than an unsigned int.
14324 * tree.c (make_vector_type): Likewise.
14325 (build_truth_vector_type): Take the number of units as a poly_uint64
14326 rather than an unsigned int.
14328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14329 Alan Hayward <alan.hayward@arm.com>
14330 David Sherwood <david.sherwood@arm.com>
14332 * target.def (get_mask_mode): Take the number of units and length
14333 as poly_uint64s rather than unsigned ints.
14334 * targhooks.h (default_get_mask_mode): Update accordingly.
14335 * targhooks.c (default_get_mask_mode): Likewise.
14336 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14337 * doc/tm.texi: Regenerate.
14339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14340 Alan Hayward <alan.hayward@arm.com>
14341 David Sherwood <david.sherwood@arm.com>
14343 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14344 * omp-general.c (omp_max_vf): Likewise.
14345 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14346 (expand_omp_simd): Handle polynomial safelen.
14347 * omp-low.c (omplow_simd_context): Add a default constructor.
14348 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14349 (lower_rec_simd_input_clauses): Update accordingly.
14350 (lower_rec_input_clauses): Likewise.
14352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14353 Alan Hayward <alan.hayward@arm.com>
14354 David Sherwood <david.sherwood@arm.com>
14356 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14357 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14358 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14359 (vect_analyze_slp_cost): Likewise.
14360 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14361 (vect_model_load_cost): Likewise.
14363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14364 Alan Hayward <alan.hayward@arm.com>
14365 David Sherwood <david.sherwood@arm.com>
14367 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14368 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14369 from an unsigned int * to a poly_uint64_pod *.
14370 (calculate_unrolling_factor): New function.
14371 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14378 from an unsigned int to a poly_uint64.
14379 (_loop_vec_info::slp_unrolling_factor): Likewise.
14380 (_loop_vec_info::vectorization_factor): Change from an int
14382 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14383 (vect_get_num_vectors): New function.
14384 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14385 (vect_get_num_copies): Use vect_get_num_vectors.
14386 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14387 to an unsigned int *.
14388 (vect_analyze_data_refs): Change min_vf from an int * to a
14390 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14391 than an unsigned HOST_WIDE_INT.
14392 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14393 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14394 to an unsigned int *.
14395 (vect_analyze_data_ref_dependences): Likewise.
14396 (vect_compute_data_ref_alignment): Handle polynomial vf.
14397 (vect_enhance_data_refs_alignment): Likewise.
14398 (vect_prune_runtime_alias_test_list): Likewise.
14399 (vect_shift_permute_load_chain): Likewise.
14400 (vect_supportable_dr_alignment): Likewise.
14401 (dependence_distance_ge_vf): Take the vectorization factor as a
14402 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14403 (vect_analyze_data_refs): Change min_vf from an int * to a
14405 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14406 vfm1 as a poly_uint64 rather than an int. Make the same change
14407 for the returned bound_scalar.
14408 (vect_gen_vector_loop_niters): Handle polynomial vf.
14409 (vect_do_peeling): Likewise. Update call to
14410 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14411 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14413 * tree-vect-loop.c (vect_determine_vectorization_factor)
14414 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14415 (vect_get_known_peeling_cost): Likewise.
14416 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14417 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14418 (vect_transform_loop): Likewise. Use the lowest possible VF when
14419 updating the upper bounds of the loop.
14420 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14421 rather than an int.
14422 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14423 polynomial unroll factors.
14424 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14425 (vect_make_slp_decision): Likewise.
14426 (vect_supported_load_permutation_p): Likewise, and polynomial
14428 (vect_analyze_slp_cost): Handle polynomial vf.
14429 (vect_slp_analyze_node_operations): Likewise.
14430 (vect_slp_analyze_bb_1): Likewise.
14431 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14432 than an unsigned HOST_WIDE_INT.
14433 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14434 (vectorizable_load): Handle polynomial vf.
14435 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14437 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14440 Alan Hayward <alan.hayward@arm.com>
14441 David Sherwood <david.sherwood@arm.com>
14443 * match.pd: Handle bit operations involving three constants
14444 and try to fold one pair.
14446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14448 * tree-vect-loop-manip.c: Include gimple-fold.h.
14449 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14450 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14451 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14452 Add a path that uses a step of VF instead of 1, but disable it
14454 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14455 and niters_no_overflow parameters. Update calls to
14456 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14457 Create a new SSA name if the latter choses to use a ste other
14458 than zero, and return it via niters_vector_mult_vf_var.
14459 * tree-vect-loop.c (vect_transform_loop): Update calls to
14460 vect_do_peeling, vect_gen_vector_loop_niters and
14461 slpeel_make_loop_iterate_ntimes.
14462 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14463 (vect_gen_vector_loop_niters): Update declarations after above changes.
14465 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14467 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14468 128-bit round to integer instructions.
14469 (ceil<mode>2): Likewise.
14470 (btrunc<mode>2): Likewise.
14471 (round<mode>2): Likewise.
14473 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14475 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14476 unaligned VSX load/store on P8/P9.
14477 (expand_block_clear): Allow the use of unaligned VSX
14478 load/store on P8/P9.
14480 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14482 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14484 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14485 swap associated with both a load and a store.
14487 2018-01-02 Andrew Waterman <andrew@sifive.com>
14489 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14490 * config/riscv/riscv.md (clear_cache): Use it.
14492 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14494 * web.c: Remove out-of-date comment.
14496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14498 * expr.c (fixup_args_size_notes): Check that any existing
14499 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14500 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14501 (emit_single_push_insn): ...here.
14503 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14505 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14506 (const_vector_encoded_nelts): New function.
14507 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14508 (const_vector_int_elt, const_vector_elt): Declare.
14509 * emit-rtl.c (const_vector_int_elt_1): New function.
14510 (const_vector_elt): Likewise.
14511 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14512 of CONST_VECTOR_ELT.
14514 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14516 * expr.c: Include rtx-vector-builder.h.
14517 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14518 directly on the tree encoding.
14519 (const_vector_from_tree): Likewise.
14520 * optabs.c: Include rtx-vector-builder.h.
14521 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14522 sequence of "u" values.
14523 * vec-perm-indices.c: Include rtx-vector-builder.h.
14524 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14525 directly on the vec_perm_indices encoding.
14527 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14529 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14530 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14531 * rtx-vector-builder.h: New file.
14532 * rtx-vector-builder.c: Likewise.
14533 * rtl.h (rtx_def::u2): Add a const_vector field.
14534 (CONST_VECTOR_NPATTERNS): New macro.
14535 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14536 (CONST_VECTOR_DUPLICATE_P): Likewise.
14537 (CONST_VECTOR_STEPPED_P): Likewise.
14538 (CONST_VECTOR_ENCODED_ELT): Likewise.
14539 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14540 (unwrap_const_vec_duplicate): Likewise.
14541 (const_vec_series_p): Check for a non-duplicated vector encoding.
14542 Say that the function only returns true for integer vectors.
14543 * emit-rtl.c: Include rtx-vector-builder.h.
14544 (gen_const_vec_duplicate_1): Delete.
14545 (gen_const_vector): Call gen_const_vec_duplicate instead of
14546 gen_const_vec_duplicate_1.
14547 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14548 (gen_const_vec_duplicate): Use rtx_vector_builder.
14549 (gen_const_vec_series): Likewise.
14550 (gen_rtx_CONST_VECTOR): Likewise.
14551 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14552 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14553 Build a new vector rather than modifying a CONST_VECTOR in-place.
14554 (handle_special_swappables): Update call accordingly.
14555 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14556 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14557 Build a new vector rather than modifying a CONST_VECTOR in-place.
14558 (handle_special_swappables): Update call accordingly.
14560 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14562 * simplify-rtx.c (simplify_const_binary_operation): Use
14563 CONST_VECTOR_ELT instead of XVECEXP.
14565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14567 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14568 the selector elements to be different from the data elements
14569 if the selector is a VECTOR_CST.
14570 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14571 ssizetype for the selector.
14573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14575 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14576 before testing each element individually.
14577 * tree-vect-generic.c (lower_vec_perm): Likewise.
14579 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14581 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14582 * selftest-run-tests.c (selftest::run_tests): Call it.
14583 * vector-builder.h (vector_builder::operator ==): New function.
14584 (vector_builder::operator !=): Likewise.
14585 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14586 (vec_perm_indices::all_from_input_p): New function.
14587 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14588 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14589 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14590 instead of reading the VECTOR_CST directly. Detect whether both
14591 vector inputs are the same before constructing the vec_perm_indices,
14592 and update the number of inputs argument accordingly. Use the
14593 utility functions added above. Only construct sel2 if we need to.
14595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14597 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14598 the broadcast of the low byte.
14599 (expand_mult_highpart): Use an explicit encoding for the permutes.
14600 * optabs-query.c (can_mult_highpart_p): Likewise.
14601 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14602 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14603 (vectorizable_bswap): Likewise.
14604 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14605 explicit encoding for the power-of-2 permutes.
14606 (vect_permute_store_chain): Likewise.
14607 (vect_grouped_load_supported): Likewise.
14608 (vect_permute_load_chain): Likewise.
14610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14612 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14613 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14614 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14615 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14616 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14617 (vect_gen_perm_mask_any): Likewise.
14619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14621 * int-vector-builder.h: New file.
14622 * vec-perm-indices.h: Include int-vector-builder.h.
14623 (vec_perm_indices): Redefine as an int_vector_builder.
14624 (auto_vec_perm_indices): Delete.
14625 (vec_perm_builder): Redefine as a stand-alone class.
14626 (vec_perm_indices::vec_perm_indices): New function.
14627 (vec_perm_indices::clamp): Likewise.
14628 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14629 (vec_perm_indices::new_vector): New function.
14630 (vec_perm_indices::new_expanded_vector): Update for new
14631 vec_perm_indices class.
14632 (vec_perm_indices::rotate_inputs): New function.
14633 (vec_perm_indices::all_in_range_p): Operate directly on the
14634 encoded form, without computing elided elements.
14635 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14636 encoding. Update for new vec_perm_indices class.
14637 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14638 the given vec_perm_builder.
14639 (expand_vec_perm_var): Update vec_perm_builder constructor.
14640 (expand_mult_highpart): Use vec_perm_builder instead of
14641 auto_vec_perm_indices.
14642 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14643 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14644 or double series encoding as appropriate.
14645 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14646 vec_perm_indices instead of auto_vec_perm_indices.
14647 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14648 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14649 (vect_permute_store_chain): Likewise.
14650 (vect_grouped_load_supported): Likewise.
14651 (vect_permute_load_chain): Likewise.
14652 (vect_shift_permute_load_chain): Likewise.
14653 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14654 (vect_transform_slp_perm_load): Likewise.
14655 (vect_schedule_slp_instance): Likewise.
14656 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14657 (vectorizable_mask_load_store): Likewise.
14658 (vectorizable_bswap): Likewise.
14659 (vectorizable_store): Likewise.
14660 (vectorizable_load): Likewise.
14661 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14662 vec_perm_indices instead of auto_vec_perm_indices. Use
14663 tree_to_vec_perm_builder to read the vector from a tree.
14664 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14665 vec_perm_builder instead of a vec_perm_indices.
14666 (have_whole_vector_shift): Use vec_perm_builder and
14667 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14668 truncation to calc_vec_perm_mask_for_shift.
14669 (vect_create_epilog_for_reduction): Likewise.
14670 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14671 from auto_vec_perm_indices to vec_perm_indices.
14672 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14673 instead of changing individual elements.
14674 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14675 the vector in d.perm.
14676 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14677 from auto_vec_perm_indices to vec_perm_indices.
14678 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14679 instead of changing individual elements.
14680 (arm_vectorize_vec_perm_const): Use new_vector to install
14681 the vector in d.perm.
14682 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14683 Update vec_perm_builder constructor.
14684 (rs6000_expand_interleave): Likewise.
14685 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14686 (rs6000_expand_interleave): Likewise.
14688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14690 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14691 to qimode could truncate the indices.
14692 * optabs.c (expand_vec_perm_var): Likewise.
14694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14696 * Makefile.in (OBJS): Add vec-perm-indices.o.
14697 * vec-perm-indices.h: New file.
14698 * vec-perm-indices.c: Likewise.
14699 * target.h (vec_perm_indices): Replace with a forward class
14701 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14702 * optabs.h: Include vec-perm-indices.h.
14703 (expand_vec_perm): Delete.
14704 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14705 (expand_vec_perm_const): Declare.
14706 * target.def (vec_perm_const_ok): Replace with...
14707 (vec_perm_const): ...this new hook.
14708 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14709 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14710 * doc/tm.texi: Regenerate.
14711 * optabs.def (vec_perm_const): Delete.
14712 * doc/md.texi (vec_perm_const): Likewise.
14713 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14714 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14715 expand_vec_perm for constant permutation vectors. Assert that
14716 the mode of variable permutation vectors is the integer equivalent
14717 of the mode that is being permuted.
14718 * optabs-query.h (selector_fits_mode_p): Declare.
14719 * optabs-query.c: Include vec-perm-indices.h.
14720 (selector_fits_mode_p): New function.
14721 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14722 is defined, instead of checking whether the vec_perm_const_optab
14723 exists. Use targetm.vectorize.vec_perm_const instead of
14724 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14725 fit in the vector mode before using a variable permute.
14726 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14727 vec_perm_indices instead of an rtx.
14728 (expand_vec_perm): Replace with...
14729 (expand_vec_perm_const): ...this new function. Take the selector
14730 as a vec_perm_indices rather than an rtx. Also take the mode of
14731 the selector. Update call to shift_amt_for_vec_perm_mask.
14732 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14733 Use vec_perm_indices::new_expanded_vector to expand the original
14734 selector into bytes. Check whether the indices fit in the vector
14735 mode before using a variable permute.
14736 (expand_vec_perm_var): Make global.
14737 (expand_mult_highpart): Use expand_vec_perm_const.
14738 * fold-const.c: Includes vec-perm-indices.h.
14739 * tree-ssa-forwprop.c: Likewise.
14740 * tree-vect-data-refs.c: Likewise.
14741 * tree-vect-generic.c: Likewise.
14742 * tree-vect-loop.c: Likewise.
14743 * tree-vect-slp.c: Likewise.
14744 * tree-vect-stmts.c: Likewise.
14745 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14747 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14748 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14749 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14750 (aarch64_vectorize_vec_perm_const): ...this new function.
14751 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14752 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14753 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14754 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14755 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14756 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14757 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14759 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14760 check for NEON modes.
14761 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14762 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14763 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14764 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14766 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14767 the old VEC_PERM_CONST conditions.
14768 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14769 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14770 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14771 (ia64_vectorize_vec_perm_const_ok): Merge into...
14772 (ia64_vectorize_vec_perm_const): ...this new function.
14773 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14774 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14775 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14776 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14777 * config/mips/mips.c (mips_expand_vec_perm_const)
14778 (mips_vectorize_vec_perm_const_ok): Merge into...
14779 (mips_vectorize_vec_perm_const): ...this new function.
14780 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14781 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14782 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14783 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14784 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14785 (rs6000_expand_vec_perm_const): Delete.
14786 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14788 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14789 (altivec_expand_vec_perm_const_le): Take each operand individually.
14790 Operate on constant selectors rather than rtxes.
14791 (altivec_expand_vec_perm_const): Likewise. Update call to
14792 altivec_expand_vec_perm_const_le.
14793 (rs6000_expand_vec_perm_const): Delete.
14794 (rs6000_vectorize_vec_perm_const_ok): Delete.
14795 (rs6000_vectorize_vec_perm_const): New function.
14796 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14797 an element count and rtx array.
14798 (rs6000_expand_extract_even): Update call accordingly.
14799 (rs6000_expand_interleave): Likewise.
14800 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14801 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14802 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14803 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14804 (rs6000_expand_vec_perm_const): Delete.
14805 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14806 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14807 (altivec_expand_vec_perm_const_le): Take each operand individually.
14808 Operate on constant selectors rather than rtxes.
14809 (altivec_expand_vec_perm_const): Likewise. Update call to
14810 altivec_expand_vec_perm_const_le.
14811 (rs6000_expand_vec_perm_const): Delete.
14812 (rs6000_vectorize_vec_perm_const_ok): Delete.
14813 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14814 reference to the SPE evmerge intructions.
14815 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14816 an element count and rtx array.
14817 (rs6000_expand_extract_even): Update call accordingly.
14818 (rs6000_expand_interleave): Likewise.
14819 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14820 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14822 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14826 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14827 vector mode and that that mode matches the mode of the data
14829 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14830 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14831 directly using expand_vec_perm_1 when forcing selectors into
14833 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14835 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14837 * optabs-query.h (can_vec_perm_p): Delete.
14838 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14839 * optabs-query.c (can_vec_perm_p): Split into...
14840 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14841 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14842 particular selector is valid.
14843 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14844 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14845 (vect_grouped_load_supported): Likewise.
14846 (vect_shift_permute_load_chain): Likewise.
14847 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14848 (vect_transform_slp_perm_load): Likewise.
14849 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14850 (vectorizable_bswap): Likewise.
14851 (vect_gen_perm_mask_checked): Likewise.
14852 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14853 implementations of variable permutation vectors into account
14854 when deciding which selector to use.
14855 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14856 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14857 with a false third argument.
14858 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14859 to test whether the constant selector is valid and can_vec_perm_var_p
14860 to test whether a variable selector is valid.
14862 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14864 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14865 * optabs-query.c (can_vec_perm_p): Likewise.
14866 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14867 instead of vec_perm_indices.
14868 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14869 (vect_gen_perm_mask_checked): Likewise,
14870 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14871 (vect_gen_perm_mask_checked): Likewise,
14873 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14875 * optabs-query.h (qimode_for_vec_perm): Declare.
14876 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14877 (qimode_for_vec_perm): ...this new function.
14878 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14880 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14882 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14883 does not have a conditional at the top.
14885 2018-01-02 Richard Biener <rguenther@suse.de>
14887 * ipa-inline.c (big_speedup_p): Fix expression.
14889 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14892 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14895 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14899 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14900 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14901 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14902 cond_taken_branch_cost 3->4.
14904 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14906 PR tree-optimization/83581
14907 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14908 TODO_cleanup_cfg if any changes have been made.
14910 PR middle-end/83608
14911 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14912 convert_modes if target mode has the right side, but different mode
14915 PR middle-end/83609
14916 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14917 last argument when extracting from CONCAT. If either from_real or
14918 from_imag is NULL, use expansion through memory. If result is not
14919 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14920 the parts directly to inner mode, if even that fails, use expansion
14923 PR middle-end/83623
14924 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14925 check for bswap in mode rather than HImode and use that in expand_unop
14928 Copyright (C) 2018 Free Software Foundation, Inc.
14930 Copying and distribution of this file, with or without modification,
14931 are permitted in any medium without royalty provided the copyright
14932 notice and this notice are preserved.