1 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2 Cesar Philippidis <cesar@codesourcery.com>
5 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6 when emitting error on private/firstprivate reductions.
7 * omp-low.c (lower_omp_target): Avoid reference-type processing
8 on pointers for firstprivate clause.
10 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15 * config/aarch64/aarch64-simd.md
16 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17 (aarch64_ld1_x3_<mode>): Likewise
18 (aarch64_st1x2<VALLDIF:mode>): Likewise
19 (aarch64_st1_x2_<mode>): Likewise
20 (aarch64_st1x3<VALLDIF:mode>): Likewise
21 (aarch64_st1_x3_<mode>): Likewise
22 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
23 (vld1_s8_x3): Likewise.
24 (vld1_u16_x3): Likewise.
25 (vld1_s16_x3): Likewise.
26 (vld1_u32_x3): Likewise.
27 (vld1_s32_x3): Likewise.
28 (vld1_u64_x3): Likewise.
29 (vld1_s64_x3): Likewise.
30 (vld1_f16_x3): Likewise.
31 (vld1_f32_x3): Likewise.
32 (vld1_f64_x3): Likewise.
33 (vld1_p8_x3): Likewise.
34 (vld1_p16_x3): Likewise.
35 (vld1_p64_x3): Likewise.
36 (vld1q_u8_x3): Likewise.
37 (vld1q_s8_x3): Likewise.
38 (vld1q_u16_x3): Likewise.
39 (vld1q_s16_x3): Likewise.
40 (vld1q_u32_x3): Likewise.
41 (vld1q_s32_x3): Likewise.
42 (vld1q_u64_x3): Likewise.
43 (vld1q_s64_x3): Likewise.
44 (vld1q_f16_x3): Likewise.
45 (vld1q_f32_x3): Likewise.
46 (vld1q_f64_x3): Likewise.
47 (vld1q_p8_x3): Likewise.
48 (vld1q_p16_x3): Likewise.
49 (vld1q_p64_x3): Likewise.
50 (vst1_s64_x2): Likewise.
51 (vst1_u64_x2): Likewise.
52 (vst1_f64_x2): Likewise.
53 (vst1_s8_x2): Likewise.
54 (vst1_p8_x2): Likewise.
55 (vst1_s16_x2): Likewise.
56 (vst1_p16_x2): Likewise.
57 (vst1_s32_x2): Likewise.
58 (vst1_u8_x2): Likewise.
59 (vst1_u16_x2): Likewise.
60 (vst1_u32_x2): Likewise.
61 (vst1_f16_x2): Likewise.
62 (vst1_f32_x2): Likewise.
63 (vst1_p64_x2): Likewise.
64 (vst1q_s8_x2): Likewise.
65 (vst1q_p8_x2): Likewise.
66 (vst1q_s16_x2): Likewise.
67 (vst1q_p16_x2): Likewise.
68 (vst1q_s32_x2): Likewise.
69 (vst1q_s64_x2): Likewise.
70 (vst1q_u8_x2): Likewise.
71 (vst1q_u16_x2): Likewise.
72 (vst1q_u32_x2): Likewise.
73 (vst1q_u64_x2): Likewise.
74 (vst1q_f16_x2): Likewise.
75 (vst1q_f32_x2): Likewise.
76 (vst1q_f64_x2): Likewise.
77 (vst1q_p64_x2): Likewise.
78 (vst1_s64_x3): Likewise.
79 (vst1_u64_x3): Likewise.
80 (vst1_f64_x3): Likewise.
81 (vst1_s8_x3): Likewise.
82 (vst1_p8_x3): Likewise.
83 (vst1_s16_x3): Likewise.
84 (vst1_p16_x3): Likewise.
85 (vst1_s32_x3): Likewise.
86 (vst1_u8_x3): Likewise.
87 (vst1_u16_x3): Likewise.
88 (vst1_u32_x3): Likewise.
89 (vst1_f16_x3): Likewise.
90 (vst1_f32_x3): Likewise.
91 (vst1_p64_x3): Likewise.
92 (vst1q_s8_x3): Likewise.
93 (vst1q_p8_x3): Likewise.
94 (vst1q_s16_x3): Likewise.
95 (vst1q_p16_x3): Likewise.
96 (vst1q_s32_x3): Likewise.
97 (vst1q_s64_x3): Likewise.
98 (vst1q_u8_x3): Likewise.
99 (vst1q_u16_x3): Likewise.
100 (vst1q_u32_x3): Likewise.
101 (vst1q_u64_x3): Likewise.
102 (vst1q_f16_x3): Likewise.
103 (vst1q_f32_x3): Likewise.
104 (vst1q_f64_x3): Likewise.
105 (vst1q_p64_x3): Likewise.
107 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
109 * config/msp430/msp430.c (msp430_output_labelref): Prepend
110 user_label_prefix to name.
112 * tree-core.h: Update comment about the format of NAME string
113 passed to handler in attribute_spec.
115 * config/msp430/msp430.md: Remove erroneous subreg expression from
116 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
117 zero_extend{q,h}isi2.
119 2018-05-30 Borislav Petkov <bp@suse.de>
121 * doc/extend.texi: Document some architecture specific
122 constraints and sort entries.
124 2018-05-30 Martin Sebor <msebor@redhat.com>
127 * builtins.c (expand_builtin_stpcpy_1): New function.
128 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
129 only if the former succeeds.
131 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
133 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
136 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
138 * doc/invoke.texi (-flinker-output): Document
140 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
142 * passes.c (ipa_write_summaries): Only modify statements if body
144 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
145 incrementally linking.
146 (ipa_passes): Likewise.
147 * lto-cgraph.c (lto_output_node): When incrementally linking do not
148 pass down resolution info.
149 * common.opt (flag_incremental_link): Update info.
150 * gcc.c (plugin specs): Turn flinker-output=* to
151 -plugin-opt=-linker-output-known
152 * toplev.c (compile_file): Also cut compilation when doing incremental
154 * flag-types. (enum lto_partition_model): Add
155 LTO_LINKER_OUTPUT_NOLTOREL.
156 (invoke.texi): Add -flinker-output docs.
157 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
158 link same way as WPA; do not stream in dead initializers.
160 * dwarf2out.c (dwarf2out_die_ref_for_decl,
161 darf2out_register_external_decl): Support incremental link.
163 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
165 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
167 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
169 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
170 it down to simple_object_copy_lto_debug_sections.
171 (run_gcc): Determine incremental LTO link time and configure
172 lto1 into non-wpa mode, disable renaming of debug sections.
174 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
176 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
177 descriptions of various incorrectly documented functions.
179 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
182 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
183 address check not strict.
185 2018-05-30 Richard Biener <rguenther@suse.de>
187 PR tree-optimization/85964
188 * tracer.c (better_p): Drop initialized count check, we only
189 call the function with initialized counts now.
190 (find_best_successor): Do find a best edge if one
191 has uninitialized count.
192 (find_best_predecessor): Likewise. Do BB frequency check only
193 if count is initialized.
195 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
197 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
198 (aarch64_ldrstr_offset_compare): New.
199 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
200 load/store orderings.
201 (aarch64_gen_adjusted_ldpstp): Likewise.
203 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
205 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
206 Check for subset of GENERAL_REGS and FP_REGS.
207 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
210 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
212 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
213 and wi::to_poly_offset. Add the current offset and then check
214 whether the sum fits, rather than using an unchecked addition of
215 a checked term. Check for a shwi rather than a uhwi.
216 * expr.c (get_bit_range): Use tree_to_poly_uint64.
217 (store_constructor): Use poly_int_tree_p.
218 (expand_expr_real_1): Likewise.
219 * function.c (assign_temp): Likewise.
220 * fold-const.c (const_binop): Use poly_int_tree_p and
222 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
224 * ipa-icf-gimple.c (func_checker::compare_operand): Use
225 to_poly_offset for MEM offsets.
226 * ipa-icf.c (sem_variable::equals): Likewise.
227 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
228 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
229 wi::to_poly_offset for BIT_FIELD_REF offsets.
230 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
232 * var-tracking.c (emit_note_insn_var_location): Use
235 2018-05-29 Jim Wilson <jimw@sifive.com>
237 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
239 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
242 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
243 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
245 (sse4_1_round<mode>2): Use nonimmediate_operand
246 for operand 1 predicate.
248 2018-05-29 Martin Sebor <msebor@redhat.com>
249 Richard Biener <rguenther@suse.de>
252 * calls.c (get_size_range): Call determine_value_range instead
254 * tree-vrp.h (determine_value_range): Declared new function.
255 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
257 2018-05-29 Richard Biener <rguenther@suse.de>
259 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
260 sure to use non-pattern stmts for get_earlier_stmt arguments.
261 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
262 called on pattern stmts.
263 (get_later_stmt): Likewise.
265 2018-05-29 Martin Liska <mliska@suse.cz>
267 PR gcov-profile/85759
268 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
271 2018-05-29 Jakub Jelinek <jakub@redhat.com>
273 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
275 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
279 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
280 VEC_PACK_FLOAT_EXPR): New tree codes.
281 * tree-pretty-print.c (op_code_prio): Handle
282 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
283 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
284 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
285 * tree-inline.c (estimate_operator_cost): Likewise.
286 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
287 * fold-const.c (const_binop): Likewise.
288 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
289 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
290 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
291 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
292 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
293 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
294 * expr.c (expand_expr_real_2): Likewise.
295 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
296 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
297 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
299 * optabs.c (expand_widen_pattern_expr): For
300 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
301 sign from result type rather than operand's type.
302 (expand_binop_directly): For vec_packu_float_optab and
303 vec_packs_float_optab allow result type to be different from operand's
305 * optabs-tree.c (optab_for_tree_code): Handle
306 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
307 VEC_PACK_FLOAT_EXPR. Formatting fixes.
308 * tree-vect-generic.c (expand_vector_operations_1): Handle
309 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
311 * tree-vect-stmts.c (supportable_widening_operation): Handle
313 (supportable_narrowing_operation): Handle FLOAT_EXPR.
314 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
315 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
316 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
317 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
319 (vec_pack<floatprefix>_float_<mode>): New expander.
320 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
322 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
323 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
324 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
325 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
326 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
328 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
329 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
330 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
331 VEC_PACK_FLOAT_EXPR): Document.
333 2018-05-29 Richard Biener <rguenther@suse.de>
335 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
337 (stmt_vec_info_vec): Make pointer.
338 (init_stmt_vec_info_vec): Remove.
339 (free_stmt_vec_info_vec): Likewise.
340 (set_stmt_vec_info_vec): New function.
341 (free_stmt_vec_infos): Likewise.
342 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
343 (set_vinfo_for_stmt): Likewise.
344 (get_earlier_stmt): Likewise.
345 (get_later_stmt): Likewise.
346 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
347 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
348 (vec_info::~vec_info): Free stmt_vec_infos.
349 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
350 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
351 (pass_slp_vectorize::execute): Likewise.
352 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
353 (free_stmt_vec_info_vec): Likewise.
354 (set_stmt_vec_info_vec): New function.
355 (free_stmt_vec_infos): Likewise.
356 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
357 the global stmt_vec_info_vec.
358 * tree-parloops.c (gather_scalar_reductions): Use
359 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
362 2018-05-29 Richard Biener <rguenther@suse.de>
364 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
366 2018-05-29 Martin Liska <mliska@suse.cz>
367 David Malcolm <dmalcolm@redhat.com>
369 * vec.c (test_reverse): New.
370 (vec_c_tests): Add new test.
371 * vec.h (vl_ptr>::reverse): New function.
373 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
375 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
377 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
380 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
382 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
384 2018-05-28 Richard Biener <rguenther@suse.de>
386 PR tree-optimization/85933
387 * tree-vect-data-refs.c (vect_record_base_alignments): Only
388 look at stmts marked as vectorizable.
390 2018-05-28 Richard Biener <rguenther@suse.de>
392 PR tree-optimization/85934
393 * tree-vect-generic.c (expand_vector_operations_1): Hoist
394 vector boolean check before scalar optimization.
396 2018-05-28 Jakub Jelinek <jakub@redhat.com>
398 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
401 2018-05-28 Mark Wielaard <mark@klomp.org>
403 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
404 if it is an expression containing a minus sign.
406 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
408 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
410 2018-05-27 Paul Koning <ni1d@arrl.net>
412 * config/pdp11/pdp11.md (truncsihi2): Remove.
414 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
415 Chung-Ju Wu <jasonwucj@gmail.com>
417 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
419 (unaligned_store_dw): Ditto.
420 * config/nds32/nds32-memory-manipulation.c
421 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
422 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
423 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
424 (emit_setmem_word_loop): Rename to ...
425 (emit_setmem_doubleword_loop): ... this.
426 (nds32_gen_dup_4_byte_to_word_value): New function.
427 (nds32_gen_dup_8_byte_to_double_word_value): New function.
428 (nds32_expand_setmem_loop): Refine implementation.
429 (nds32_expand_setmem_loop_v3m): Ditto.
430 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
433 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
435 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
437 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
439 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
440 (nds32_init_machine_status): Initialize machine->attr_naked_p and
441 machine->attr_no_prologue_p.
442 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
443 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
444 (nds32_expand_epilogue): Consider attr_naked_p.
445 (nds32_expand_epilogue_v3pop): Likewise.
446 (nds32_can_use_return_insn): Likewise.
447 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
448 attr_no_prologue_p fields.
449 * config/nds32/nds32.opt (mret-in-naked-func): New option.
451 2018-05-27 Jakub Jelinek <jakub@redhat.com>
454 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
457 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
459 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
461 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
463 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
465 (*<floatsuffix>floatv2div2sf2): Rename to ...
466 (*float<floatunssuffix>v2div2sf2): ... this.
467 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
468 (float<floatunssuffix>v2div2sf2_mask): ... this.
469 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
470 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
471 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
473 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
475 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
477 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
479 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
481 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
483 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
484 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
485 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
486 gen_ufix_truncv8dfv8si2.
487 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
488 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
489 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
490 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
491 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
492 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
493 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
494 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
496 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
500 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
502 2018-05-25 Jim Wilson <jimw@sifive.com>
504 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
505 * config/riscv/riscv.c (struct machine_function): Add
506 interrupt_handler_p and attribute_checked_p fields.
507 (riscv_attribute_table): Add interrupt.
508 (riscv_interrupt_type_p): New.
509 (riscv_save_reg_p): Save extra regs for interrupt handler.
510 (riscv_use_save_libcall): Return false for interrupt handler.
511 (riscv_first_stack_step): Add forward declaration.
512 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
513 for interrupt handler with large frame. Use it for saved reg list.
514 (riscv_expand_prologue): Move flag_stack_usage_info support to
515 eliminate duplication.
516 (riscv_expand_epilogue): Generate mret for interrupt handler.
517 (riscv_epilogue_uses): New.
518 (riscv_can_use_return_insn): Return false for interrupt handler.
519 (riscv_function_ok_for_sibcall): Likewise.
520 (riscv_set_current_function): Add interrupt handler support.
521 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
522 * config/riscv/riscv.md (UNSPECV_MRET): New.
524 (riscv_frflags, riscv_fsflags): Use tab after opcode.
526 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
528 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
530 PR tree-optimization/85712
531 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
532 this candidate has already been replaced in-situ by a copy.
534 2018-05-25 Jason Merrill <jason@redhat.com>
536 PR c++/80485 - inline function non-zero address.
537 * symtab.c (nonzero_address): Check DECL_COMDAT.
539 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
542 * config/alpha/alpha.md (ashlsi3): New insn pattern.
543 (*ashlsi_se): Rename from *ashldi_se. Define as sign
544 extension of SImode operation. Use const123_operand predicate.
546 (*saddl_se_1): Ditto.
548 (*ssubl_se_1): Ditto.
549 * config/alpha/predicates.md (const123_operand): New predicate.
550 * config/alpha/constraints.md (P): Use IN_RANGE.
552 2018-05-25 Richard Biener <rguenther@suse.de>
554 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
556 (ref_maybe_used_by_stmt_p): Likewise.
557 (stmt_may_clobber_ref_p): Likewise.
558 (stmt_may_clobber_ref_p_1): Likewise.
559 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
561 (ref_maybe_used_by_stmt_p): Likewise.
562 (stmt_may_clobber_ref_p): Likewise.
563 (stmt_may_clobber_ref_p_1): Likewise.
564 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
565 the alias oracle to disambiguate DRs with stmts DR analysis
567 (vect_analyze_data_refs): Do not give up on not analyzable
568 DRs for BB vectorization. Remove code truncating the dataref
571 2018-05-25 Jakub Jelinek <jakub@redhat.com>
574 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
575 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
576 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
578 2018-05-25 Richard Biener <rguenther@suse.de>
580 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
581 function, combining stmt data ref gathering and fatal analysis
583 (vect_analyze_data_refs): Remove now redudnant code and simplify.
584 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
585 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
586 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
587 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
589 2018-05-25 Bin Cheng <bin.cheng@arm.com>
591 PR tree-optimization/85720
592 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
593 SCC if all partitions are builtins.
594 (version_loop_by_alias_check): New parameter. Generate cancelable
595 runtime alias check if all partitions are builtins.
596 (distribute_loop): Update call to above function.
598 2018-05-25 Bin Cheng <bin.cheng@arm.com>
600 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
601 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
602 (parm_default_def_partition_arg): Ditto.
603 (set_parm_default_def_partition): Ditto.
604 (get_parm_default_def_partitions): Ditto and make it static.
605 (get_undefined_value_partitions): Ditto and make it static.
606 (remove_ssa_form): Refactor call to init_var_map here.
607 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
608 computation for loop region.
609 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
610 (register_default_def): Delete.
611 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
612 (parm_default_def_partition_arg): Ditto.
613 (set_parm_default_def_partition): Ditto.
614 (get_parm_default_def_partitions): Ditto and make it static.
615 (get_undefined_value_partitions): Ditto and make it static.
616 (coalesce_with_default, coalesce_with_default): Update comment.
617 (create_coalesce_list_for_region): New func factored out from
618 create_outofssa_var_map.
619 (populate_coalesce_list_for_outofssa): New func factored out from
620 create_outofssa_var_map and coalesce_ssa_name.
621 (create_outofssa_var_map): Delete.
622 (coalesce_ssa_name): Refactor to support live range computation.
623 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
624 (get_parm_default_def_partitions): Delete.
625 (get_undefined_value_partitions): Ditto.
626 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
627 computation for loop region.
628 (new_tree_live_info, loe_visit_block): Ditto.
629 (live_worklist, set_var_live_on_entry): Ditto.
630 (calculate_live_on_exit, verify_live_on_entry): Ditto.
631 * tree-ssa-live.h (struct _var_map): New fields.
632 (init_var_map): Change decl.
633 (region_contains_p): New.
635 2018-05-25 Bin Cheng <bin.cheng@arm.com>
637 * tree-ssa-live.h (live_merge_and_clear): Delete.
639 2018-05-25 Richard Biener <rguenther@suse.de>
642 * tree-dump.c (dequeue_and_dump): Remove access to removed
643 operand 2 of a SWITCH_EXPR.
645 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
647 * doc/sourcebuild.texi (vect_double_cond_arith): Include
648 multiplication and division.
649 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
650 (cond_udiv@var{m}, cond_umod@var{m}): Document.
651 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
652 (cond_udiv_optab, cond_umod_optab): New optabs.
653 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
654 (IFN_COND_RDIV): New internal functions.
655 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
656 TRUNC_MOD_EXPR and RDIV_EXPR.
657 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
658 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
660 (SVE_INT_BINARY): Include mult.
661 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
662 (optab, sve_int_op): Handle mult.
663 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
665 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
666 for SVE_INT_BINARY_SD.
668 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
670 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
671 (optab, sve_int_op): Handle div and udiv.
672 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
673 for SVE_INT_BINARY_SD.
674 (*<optab><mode>3): New insn for the same.
676 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
678 * tree-vect-patterns.c: Include predict.h.
679 (vect_recog_divmod_pattern): Restrict check for division support
680 to when optimizing for size.
682 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
684 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
685 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
686 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
687 (gimple_match_op::set_op): Likewise.
688 (gimple_resimplify4): Declare.
689 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
690 (expr::gen_transform): Likewise.
691 (decision_tree::gen): Generate a simplification routine for 4 operands.
692 * gimple-match-head.c (gimple_simplify): Add an overload for
693 4 operands. In the top-level function, handle up to 4 call
694 arguments and call gimple_resimplify4.
695 (gimple_resimplify4): New function.
696 (build_call_internal): Pass a fourth operand.
697 (maybe_push_to_seq): Likewise.
698 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
699 Fold VEC_COND_EXPRs of an operation and a default value into
700 an IFN_COND_* function if possible.
701 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
703 (SVE_COND_FP_BINARY): Include them.
704 (optab, sve_fp_op): Handle them.
705 (SVE_INT_BINARY_REV): New code iterator.
706 (SVE_COND_FP_BINARY_REV): New int iterator.
707 (commutative): New int attribute.
708 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
710 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
712 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
713 (*cond_<optab><mode>): New patterns for reversed operands.
715 2018-05-25 Richard Biener <rguenther@suse.de>
717 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
718 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
719 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
720 (STMT_VINFO_GROUPED_ACCESS): Adjust.
721 * tree-vect-data-refs.c (everywhere): Adjust users.
722 * tree-vect-loop.c (everywhere): Likewise.
723 * tree-vect-slp.c (everywhere): Likewise.
724 * tree-vect-stmts.c (everywhere): Likewise.
725 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
727 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
729 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
731 (gcc_cv_as_section_exclude): ... this.
732 Try Solaris as #exclude syntax.
733 * configure: Regenerate.
734 * config.in: Regenerate.
735 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
737 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
738 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
740 * varasm.c (default_elf_asm_named_section): Don't check if
741 HAVE_GAS_SECTION_EXCLUDE is defined.
743 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
745 * doc/md.texi: Update the documentation of the cond_* optabs
746 to mention the new final operand. Fix GET_MODE_NUNITS call.
747 Describe the scalar case too.
748 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
749 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
751 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
752 (get_conditional_internal_fn): Update comment.
753 * tree-vect-loop.c (vectorizable_reduction): Pass the original
754 accumulator value as a final argument to conditional functions.
755 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
756 a define_expand and add an "else" operand. Assert for now that
757 the else operand is equal to operand 2. Use SVE_INT_BINARY and
758 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
759 (*cond_<optab><mode>): New patterns.
760 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
761 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
762 (UNSPEC_COND_EOR): Delete.
763 (optab): Remove associated mappings.
764 (SVE_INT_BINARY): New code iterator.
765 (sve_int_op): Remove int attribute and add "minus" to the code
767 (SVE_COND_INT_OP): Delete.
768 (SVE_COND_FP_OP): Rename to...
769 (SVE_COND_FP_BINARY): ...this.
771 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
773 * optabs.c (can_reuse_operands_p): New function.
774 (maybe_legitimize_operands): Try to reuse the results for
777 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
779 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
780 Add {q} suffix to insn mnemonic.
782 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
784 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
785 (msp430_warn_func_return): New.
787 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
789 * fold-const.c (tree_nonzero_bits): New function.
790 * fold-const.h (tree_nonzero_bits): Likewise.
791 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
792 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
794 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
798 * varasm.c (assemble_alias): Check ifunc_resolver only on
801 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
804 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
805 when memory input operand is handled.
807 2018-05-24 Luis Machado <luis.machado@linaro.org>
809 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
811 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
813 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
815 * match.pd: Delay FMA folds until after vectorization.
817 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
820 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
821 address check not strict.
823 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
825 * gimple-match.h (gimple_match_op): New class.
826 (mprts_hook): Replace parameters with a gimple_match_op *.
827 (maybe_build_generic_op): Likewise.
828 (gimple_simplified_result_is_gimple_val): Replace parameters with
829 a const gimple_match_op *.
830 (gimple_simplify): Replace code_helper * and tree * parameters with
831 a gimple_match_op * parameter.
832 (gimple_resimplify1): Replace code_helper *, tree and tree *
833 parameters with a gimple_match_op * parameter.
834 (gimple_resimplify2): Likewise.
835 (gimple_resimplify3): Likewise.
836 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
837 parameters with a gimple_match_op * parameter.
838 * gimple-match-head.c (gimple_simplify): Change prototypes of
839 auto-generated functions to take a gimple_match_op * instead of
840 separate code_helper * and tree * parameters. Make the same
841 change in the top-level overload and update calls to the
842 gimple_resimplify routines. Update calls to the auto-generated
843 functions and to maybe_push_res_to_seq in the publicly-facing
844 operation-specific gimple_simplify overloads.
845 (gimple_match_op::MAX_NUM_OPS): Define.
846 (gimple_resimplify1): Replace rcode and ops with a single res_op
847 parameter. Update call to gimple_simplify.
848 (gimple_resimplify2): Likewise.
849 (gimple_resimplify3): Likewise.
850 (mprts_hook): Replace parameters with a gimple_match_op *.
851 (maybe_build_generic_op): Likewise.
852 (build_call_internal): Replace type, nargs and ops with
854 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
855 with a single gimple_match_op *. Update calls to mprts_hook,
856 build_call_internal and gimple_simplified_result_is_gimple_val.
857 Factor out code that is common to the tree_code and combined_fn cases.
858 * genmatch.c (expr::gen_transform): Replace tem_code and
859 tem_ops with a gimple_match_op called tem_op. Update calls
860 to the gimple_resimplify functions and maybe_push_res_to_seq.
861 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
862 res_ops. Update call to the gimple_resimplify functions.
863 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
864 (decision_tree::gen): Make the functions take a gimple_match_op *
865 called res_op instead of separate res_code and res_ops parameters.
866 Update call accordingly.
867 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
868 and ops with a single res_op parameter. Update calls to
869 maybe_build_generic_op and maybe_push_res_to_seq.
870 (fold_stmt_1): Update calls to gimple_simplify and
871 replace_stmt_with_simplification.
872 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
873 and gimple_simplified_result_is_gimple_val.
874 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
876 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
877 with a gimple_match_op *.
878 (vn_nary_build_or_lookup): Likewise. Update call to
879 vn_nary_build_or_lookup_1.
880 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
881 gimple_match_op *. Update calls to the gimple_resimplify routines
882 and to gimple_simplified_result_is_gimple_val.
883 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
884 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
885 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
886 (visit_nary_op): Likewise.
887 (visit_reference_op_load): Likewise.
889 2018-05-23 Luis Machado <luis.machado@linaro.org>
891 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
892 modifier for printing the step amount.
894 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
897 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
900 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
902 * doc/sourcebuild.texi (Endianness): New subsubsection.
904 2018-05-23 Luis Machado <luis.machado@linaro.org>
906 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
907 <prefetch_dynamic_strides>: New const bool field.
908 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
909 prefetch_dynamic_strides.
910 (exynosm1_prefetch_tune): Likewise.
911 (thunderxt88_prefetch_tune): Likewise.
912 (thunderx_prefetch_tune): Likewise.
913 (thunderx2t99_prefetch_tune): Likewise.
914 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
916 (aarch64_override_options_internal): Update to set
917 PARAM_PREFETCH_DYNAMIC_STRIDES.
918 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
919 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
920 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
921 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
922 prefetch-dynamic-strides setting.
924 2018-05-23 Luis Machado <luis.machado@linaro.org>
926 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
927 <minimum_stride>: New const int field.
928 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
929 minimum_stride field defaulting to -1.
930 (exynosm1_prefetch_tune): Likewise.
931 (thunderxt88_prefetch_tune): Likewise.
932 (thunderx_prefetch_tune): Likewise.
933 (thunderx2t99_prefetch_tune): Likewise.
934 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
935 <default_opt_level>: Set to 3.
936 (aarch64_override_options_internal): Update to set
937 PARAM_PREFETCH_MINIMUM_STRIDE.
938 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
939 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
940 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
941 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
942 stride is constant and is below the minimum stride threshold.
944 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
946 * config/arm/arm-cpus.in (mode26): Delete.
947 (armv4): Delete mode26 reference.
948 * config/arm/arm.c (arm_configure_build_target): Delete use of
951 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
953 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
955 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
956 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
958 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
959 (floatunsdidf2): Ditto.
961 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
963 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
964 (fixuns_trunc<mode>si2_avx512f): Ditto.
965 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
966 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
967 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
969 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
971 PR rtl-optimization/79985
972 * df-scan.c (df_insn_refs_collect): Remove special case for
973 global registers and asm statements.
975 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
977 * extend.texi (Global Register Variables): Rewrite the bullet list.
978 Note that the register is available for allocation. Note that access
979 via inline asm must use constraints. Add note about async-signal
980 handlers. Remove paragraph about automagic register selection.
982 2018-05-23 Richard Biener <rguenther@suse.de>
984 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
985 of fixed offset from memset VN.
987 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
989 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
991 (alloc_cand_and_find_basis): Initialize first_interp field.
992 (slsr_process_mul): Modify first_interp field.
993 (slsr_process_add): Likewise.
994 (slsr_process_cast): Modify first_interp field for each new
996 (slsr_process_copy): Likewise.
997 (dump_candidate): Dump first_interp field.
998 (replace_mult_candidate): Process all interpretations, not just
1000 (replace_rhs_if_not_dup): Likewise.
1001 (replace_one_candidate): Likewise.
1003 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1005 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1007 (aarch64_needs_frame_chain): New function.
1008 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1010 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1013 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1014 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1015 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1016 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1017 as true for strict-align.
1018 (aarch64_can_inline_p): Perform checks even when callee has no
1019 attributes to check for strict alignment.
1020 * doc/extend.texi (AArch64 Function Attributes): Document
1022 * doc/invoke.texi: (AArch64 Options): Likewise.
1024 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1026 PR tree-optimization/85853
1027 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1028 the handling of the root of the node to...
1029 (vect_slp_analyze_node_operations_1): ...this new function,
1030 and run the whole thing with the child nodes' def types
1031 set according to their SLP node's def type.
1033 2018-05-23 Richard Biener <rguenther@suse.de>
1036 * tree-data-ref.c (create_runtime_alias_checks): Defer
1037 and ignore overflow warnings.
1039 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1041 PR tree-optimization/85822
1042 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1045 2018-05-23 Richard Biener <rguenther@suse.de>
1047 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1048 memset constants via native_interpret_expr.
1050 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1053 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1055 (cgraph_node::create_alias): Likewise.
1056 (cgraph_node::get_availability): Check ifunc_resolver instead
1057 of looking up ifunc attribute.
1058 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1059 * varasm.c (do_assemble_alias): Likewise.
1060 (assemble_alias): Likewise.
1061 (default_binds_local_p_3): Likewise.
1062 * cgraph.h (cgraph_node): Add ifunc_resolver.
1063 (cgraph_node::only_called_directly_or_aliased_p): Return false
1065 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1067 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1068 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1069 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1070 instead of looking up ifunc attribute.
1072 2018-05-22 Luis Machado <luis.machado@linaro.org>
1074 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1076 2018-05-22 Martin Sebor <msebor@redhat.com>
1079 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1080 only when expasion succeeds.
1081 (expand_builtin_strcmp): Same.
1082 (expand_builtin_strncmp): Same.
1084 2018-05-22 Martin Sebor <msebor@redhat.com>
1086 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1088 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1089 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1091 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1092 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1094 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1095 Add check for legitimate_address.
1096 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1097 (aarch64_swap_ldrstr_operands): New.
1098 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1101 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1102 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1104 * config/aarch64/aarch64.md: New patterns to generate stp
1106 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1107 single words and double words.
1108 (load_pair_sw, load_pair_dw): Likewise.
1109 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1111 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1113 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1114 for different mode ldpstp and add peephole for merged zero stores.
1116 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1118 (aarch64_gen_store_pair): Rename calls to match new patterns.
1119 (aarch64_gen_load_pair): Rename calls to match new patterns.
1120 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1121 (load_pair<DREG:mode><DREG2:mode>): ... This.
1122 (store_pair<mode>): Rename to...
1123 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1124 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1126 (V_INT_EQUIV): Handle SImode.
1127 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1130 2018-05-22 Martin Sebor <msebor@redhat.com>
1133 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1134 or ajust the presumed bound on an operation to avoid unnecessary
1137 2018-05-22 Martin Sebor <msebor@redhat.com>
1139 PR tree-optimization/85826
1140 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1141 assuming that a DECL necesarily has a constant size.
1143 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1146 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1148 2018-05-22 Richard Biener <rguenther@suse.de>
1150 PR tree-optimization/85834
1151 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1152 non-constant and non-zero memset arguments.
1154 2018-05-22 Martin Liska <mliska@suse.cz>
1157 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1159 2018-05-22 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/85863
1162 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1163 comparisons when vectype is specified.
1164 (vectorizable_condition): Do not specify vectype for
1165 vect_is_simple_cond when SLP vectorizing.
1167 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1170 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1171 define __ibm128 as long double.
1172 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1173 as a distinct type when IEEE 128-bit support is enabled.
1174 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1175 128-bit types to use the correct functions.
1176 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1177 convert between 128-bit floating point types that have different
1178 modes but the same representation, instead of using gen_lowpart to
1180 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1182 (IFKF_reg): New attributes to give the register constraints for
1184 (extend<mode>tf2_internal): New insns to mark an explicit
1185 conversion between 128-bit floating point types that have a
1186 different mode but share the same representation.
1188 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1190 PR tree-optimization/85814
1191 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1192 a null return from get_strinfo when unsharing the next
1193 strinfo in the chain.
1195 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1198 * varasm.c (weak_finish): Clean up weak_decls.
1200 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1202 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1203 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1204 UNSPEC_UADALP values.
1205 * config/aarch64/iterators.md (ABAL): New int iterator.
1208 (sur): Add mappings for the above.
1209 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1211 (aarch64_<sur>abal<mode>_4): Likewise.
1212 (aarch64_<sur>adalp<mode>_3): Likewise.
1213 (<sur>sadv16qi): New define_expand.
1215 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1217 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1218 (*movdf_internal): Ditto.
1219 (*rcpsf2_sse): Ditto.
1220 (*rsqrtsf2_sse): Ditto.
1221 (*sqrt<mode>2_sse): Ditto.
1223 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1225 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1227 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1228 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1229 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1230 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1232 * config/aarch64/arm_neon.h: Likewise.
1233 * config/aarch64/iterators.md (VQ_I): New.
1235 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1237 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1239 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1242 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1244 * config/nds32/constraints.md (S): New constraint.
1245 * config/nds32/nds32.md (call_internal): Use constraint S.
1246 (call_value_internal): Likewise.
1247 (sibcall_internal): Likewise.
1248 (sibcall_value_internal): Likewise.
1250 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1251 Chung-Ju Wu <jasonwucj@gmail.com>
1253 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1256 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1257 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1260 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1261 (nds32_rtx_costs_impl): Simplify.
1262 (nds32_address_cost_impl): Simplify.
1263 (nds32_init_rtx_costs): New function.
1264 (nds32_rtx_costs_speed_prefer): Likewise.
1265 (nds32_rtx_costs_size_prefer): Likewise.
1266 (nds32_address_cost_speed_prefer): Likewise.
1267 (nds32_address_cost_speed_fwprop): Likewise.
1268 (nds32_address_cost_size_prefer): Likewise.
1269 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1270 * config/nds32/nds32.c (nds32_option_override): Use
1271 nds32_init_rtx_costs function.
1273 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1275 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1276 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1277 (TARGET_PIPELINE_N8): Likewise.
1278 (TARGET_PIPELINE_N10): Likewise.
1279 (TARGET_PIPELINE_N13): Likewise.
1280 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1282 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1284 * config/nds32/nds32-fpu.md: Update copyright year.
1286 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1290 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1292 * config/nds32/nds32.c
1293 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1294 * config/nds32/nds32.opt (minline-asm-r15): New option.
1296 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1298 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1300 * config/nds32/nds32.md (abssi2): New pattern.
1302 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1304 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1305 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1306 Merge insn pattern from sse_cvtsi2ss<round_name> and
1307 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1308 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1309 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1310 using SWI48 mode iterator.
1311 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1312 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1313 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1314 pattern from sse_cvttss2si<round_saeonly_name>
1315 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1316 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1317 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1318 using SWI48 mode iterator.
1319 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1320 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1321 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1322 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1323 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1324 using SWI48 mode iterator.
1325 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1326 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1327 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1328 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1329 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1330 SWI48 mode iterator.
1331 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1332 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1333 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1334 pattern from sse_cvttsd2si<round_saeonly_name>
1335 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1337 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1339 * config/nds32/nds32-md-auxiliary.c
1340 (nds32_valid_smw_lwm_base_p): Refine.
1341 (nds32_output_smw_single_word): Refine.
1342 (nds32_output_smw_double_word): New.
1343 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1345 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1347 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1348 (nds32_output_stack_pop): Refine.
1349 (nds32_expand_unaligned_load): Refine.
1350 (nds32_expand_unaligned_store): Refine.
1352 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1353 Chung-Ju Wu <jasonwucj@gmail.com>
1355 * config/nds32/constants.md: Add TP_REGNUM constant.
1356 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1357 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1359 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1360 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1361 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1362 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1363 and PIC code generation.
1364 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1366 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1368 * config/nds32/nds32.md: Support TLS and PIC.
1369 * config/nds32/nds32.c: Support TLS and PIC.
1370 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1371 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1374 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1376 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1377 mode with E_ prefix.
1379 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1380 Chung-Ju Wu <jasonwucj@gmail.com>
1382 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1383 * config/nds32/nds32-md-auxiliary.c
1384 (symbolic_reference_mentioned_p): New.
1385 (nds32_legitimize_ict_address): New.
1386 (nds32_expand_ict_move): New.
1387 (nds32_indirect_call_referenced_p): New.
1388 (nds32_symbol_binds_local_p): Delete.
1389 (nds32_long_call_p): Modify.
1390 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1391 * config/nds32/nds32-protos.h
1392 (symbolic_reference_mentioned_p): Declare.
1393 (nds32_legitimize_ict_address): Declare.
1394 (nds32_expand_ict_move): Declare.
1395 (nds32_indirect_call_referenced_p): Declare.
1396 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1397 (nds32_relax_group): Use nds32_ict_const_p as condition.
1398 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1399 (nds32_asm_file_start): Output ict_model directive in asm code.
1400 (nds32_legitimate_address_p): Consider indirect call.
1401 (nds32_print_operand): Consider indirect call.
1402 (nds32_print_operand_address): Consider indirect call.
1403 (nds32_insert_attributes): Handle "indirect_call" attribute.
1404 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1405 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1406 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1407 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1408 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1409 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1410 (TARGET_ICT_MODEL_SMALL): Define.
1411 (TARGET_ICT_MODEL_LARGE): Define.
1412 * config/nds32/nds32.md (movsi): Consider ict model.
1413 (call, call_value): Consider ict model.
1414 (sibcall, sibcall_value): Consider ict model.
1415 * config/nds32/nds32.opt (mict-model): New option.
1416 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1419 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1420 Monk Chiang <sh.chiang04@gmail.com>
1421 Jim Wilson <jimw@sifive.com>
1423 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1424 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1425 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1426 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1427 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1428 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1429 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1430 compute save_libcall_adjustment properly.
1431 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1432 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1433 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1434 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1435 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1436 (ABI_SPEC): Handle mabi=ilp32e.
1437 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1438 (RVE): Add RVE mask.
1439 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1440 <-march>: Add rv32e as an example.
1442 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1445 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1446 (intra_create_variable_infos): Handle C++ constructors.
1448 2018-05-18 Martin Liska <mliska@suse.cz>
1450 * passes.def: Remove a redundant pass.
1452 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1455 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1457 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1459 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1461 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1463 (armv2, armv2a, armv3, armv3m): Delete architectures.
1464 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1465 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1466 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1468 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1469 (*mulsidi3adddi): Likewise.
1470 (mulsidi3): Likewise.
1471 (*mulsidi3_nov6): Likewise.
1472 (umulsidi3): Likewise.
1473 (umulsidi3_nov6): Likewise.
1474 (umaddsidi4): Likewise.
1475 (*umulsidi3adddi): Likewise.
1476 (smulsi3_highpart): Likewise.
1477 (*smulsi3_highpart_nov6): Likewise.
1478 (umulsi3_highpart): Likewise.
1479 (*umulsi3_highpart_nov6): Likewise.
1480 * config/arm/arm.h (arm_arch3m): Delete.
1481 * config/arm/arm.c (arm_arch3m): Delete.
1482 (arm_option_override_internal): Update armv3-related comment.
1483 (arm_configure_build_target): Delete use of isa_bit_mode32.
1484 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1485 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1486 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1488 (mulusa3): Likewise.
1489 * config/arm/arm-protos.h (arm_arch3m): Delete.
1490 * config/arm/arm-tables.opt: Regenerate.
1491 * config/arm/arm-tune.md: Likewise.
1492 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1493 deleted architectures.
1495 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1497 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1498 (armv5t, armv5te): New features.
1499 (ARMv5, ARMv5e): Delete fgroups.
1500 (ARMv5t, ARMv5te): Adjust for above changes.
1502 (armv5, armv5e): Delete arches.
1503 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1505 (*call_reg_arm): Likewise.
1506 (*call_value_reg_armv5): Likewise.
1507 (*call_value_reg_arm): Likewise.
1508 (*call_symbol): Likewise.
1509 (*call_value_symbol): Likewise.
1510 (*sibcall_insn): Likewise.
1511 (*sibcall_value_insn): Likewise.
1513 (prefetch): Likewise.
1514 (define_split and define_peephole2 dependent on arm_arch5):
1516 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1518 (TARGET_ARM_QBIT): Likewise.
1519 (TARGET_DSP_MULTIPLY): Likewise.
1520 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1521 (arm_arch5, arm_arch5e): Delete.
1522 (arm_arch5t, arm_arch5te): Declare.
1523 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1524 (arm_arch5t): Declare.
1525 (arm_option_reconfigure_globals): Update for the above.
1526 (arm_options_perform_arch_sanity_checks): Update comment, replace
1527 use of arm_arch5 with arm_arch5t.
1528 (use_return_insn): Likewise.
1529 (arm_emit_call_insn): Likewise.
1530 (output_return_instruction): Likewise.
1531 (arm_final_prescan_insn): Likewise.
1532 (arm_coproc_builtin_available): Likewise.
1533 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1534 arm_arch5e with arm_arch5t and arm_arch5te.
1535 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1536 (arm_arch5t, arm_arch5te): Declare.
1537 * config/arm/arm-tables.opt: Regenerate.
1538 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1539 * config/arm/t-multilib: Likewise.
1540 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1541 instead of arm_arch5.
1542 (*call_reg_thumb1): Likewise.
1543 (*call_value_reg_thumb1_v5): Likewise.
1544 (*call_value_reg_thumb1): Likewise.
1545 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1547 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1549 2018-05-18 Martin Liska <mliska@suse.cz>
1551 PR gcov-profile/84846
1552 * doc/gcov.texi: Document -t option of gcov tool.
1554 2018-05-18 Martin Liska <mliska@suse.cz>
1556 PR gcov-profile/84846
1557 * gcov.c (print_usage): Add new -t option.
1558 (process_args): Handle the option.
1559 (generate_results): Use stdout as output when requested by
1562 2018-05-18 Martin Liska <mliska@suse.cz>
1564 PR gcov-profile/84846
1565 * coverage.c (coverage_init): Write PWD to .gcno file.
1566 * doc/gcov.texi: Document how working directory is printed.
1567 * gcov-dump.c (dump_gcov_file): Print PWD.
1568 * gcov.c (output_intermediate_file): Likewise.
1569 (read_graph_file): Read PWD string.
1570 (output_lines): Print PWD.
1572 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1575 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1576 for retval and return false if all args to phi are zero.
1578 2018-05-18 Richard Biener <rguenther@suse.de>
1580 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1582 (evrp_dom_walker::before_dom_children): Call it.
1584 2018-05-18 Richard Biener <rguenther@suse.de>
1586 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1587 results when processing array refs with variable index.
1589 2018-05-18 Toon Moene <toon@moene.org>
1591 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1592 directly after that of -floop-interchange. Indicate that both
1593 options are enabled by default when specifying -O3.
1595 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1597 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1598 iterator. Delete separate integer-mode vec_set<mode> expander.
1599 (aarch64_simd_vec_setv2di): Delete.
1600 (vec_setv2di): Delete.
1601 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1602 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1603 the "w, r" alternative.
1605 2018-05-18 Martin Liska <mliska@suse.cz>
1607 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1608 * tree-pass.h (make_pass_lower_switch_O0): New function.
1609 * tree-switch-conversion.c (node_has_low_bound): Remove.
1610 (node_has_high_bound): Likewise.
1611 (node_is_bounded): Likewise.
1612 (class pass_lower_switch): Make it a template type and create
1614 (pass_lower_switch::execute): Add template argument.
1615 (make_pass_lower_switch): New function.
1616 (make_pass_lower_switch_O0): New function.
1617 (do_jump_if_equal): Remove.
1618 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1619 all the hard work to tree optimization passes.
1621 2018-05-18 Martin Liska <mliska@suse.cz>
1623 * dbgcnt.c (limit_low): Renamed from limit.
1624 (limit_high): New variable.
1625 (dbg_cnt_is_enabled): Check for upper limit.
1626 (dbg_cnt): Adjust dumping.
1627 (dbg_cnt_set_limit_by_index): Add new argument for high
1629 (dbg_cnt_set_limit_by_name): Likewise.
1630 (dbg_cnt_process_single_pair): Parse new format.
1631 (dbg_cnt_process_opt): Use strtok.
1632 (dbg_cnt_list_all_counters): Remove 'value' and add
1634 * doc/invoke.texi: Document changes.
1636 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1638 * doc/sourcebuild.texi (scalar_all_fma): Document.
1639 * tree.def (FMA_EXPR): Delete.
1640 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1641 * internal-fn.c (ternary_direct): New macro.
1642 (expand_ternary_optab_fn): Likewise.
1643 (direct_ternary_optab_supported_p): Likewise.
1644 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1645 * builtins.c (fold_builtin_fma): Delete.
1646 (fold_builtin_3): Don't call it.
1647 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1648 * expr.c (expand_expr_real_2): Likewise.
1649 * fold-const.c (operand_equal_p): Likewise.
1650 (fold_ternary_loc): Likewise.
1651 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1652 * gimple.c (DEFTREECODE): Likewise.
1653 * gimplify.c (gimplify_expr): Likewise.
1654 * optabs-tree.c (optab_for_tree_code): Likewise.
1655 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1656 * tree-eh.c (operation_could_trap_p): Likewise.
1657 (stmt_could_throw_1_p): Likewise.
1658 * tree-inline.c (estimate_operator_cost): Likewise.
1659 * tree-pretty-print.c (dump_generic_node): Likewise.
1660 (op_code_prio): Likewise.
1661 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1662 * tree-ssa-operands.c (get_expr_operands): Likewise.
1663 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1664 * fold-const-call.h (fold_fma): Delete.
1665 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1666 CFN_FNMA and CFN_FNMS.
1668 * genmatch.c (combined_fn): New enum.
1669 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1670 (commutative_op): New function.
1671 (commutate): Use it. Handle more than 2 operands.
1672 (dt_operand::gen_gimple_expr): Use commutative_op.
1673 (parser::parse_expr): Allow :c to be used with non-binary
1674 operators if the commutative operand is known.
1675 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1676 CFN_FMS, CFN_FNMA and CFN_FNMS.
1677 (backprop::process_assign_use): Remove FMA_EXPR handling.
1678 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1679 (gen_hsa_fma): New function.
1680 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1681 IFN_FNMA and IFN_FNMS.
1682 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1683 * gimple-fold.h (follow_all_ssa_edges): Declare.
1684 * gimple-fold.c (follow_all_ssa_edges): New function.
1685 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1686 gimple_build interface and use follow_all_ssa_edges to fold the result.
1687 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1688 instead of checking for optabs directly.
1689 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1690 rather than FMA_EXPRs.
1691 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1692 call to IFN_FMA instead of an FMA_EXPR.
1694 2018-05-17 Jim Wilson <jimw@sifive.com>
1696 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1697 sign extended promoted subreg, and the range does not have the sign bit
1698 set, then do a sign extend.
1700 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1701 test, check for sign extended subreg and/or constant operands, and
1702 do a sign extend in that case.
1704 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1706 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1707 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1709 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1710 Change logics_shift_reg to logics_shift_imm.
1711 (thunderx2t99_fp_loadpair_basic): Delete.
1712 (thunderx2t99_fp_storepair_basic): Delete.
1713 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1714 (thunderx2t99_asimd_polynomial): Delete.
1715 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1716 and neon_fp_mul_d_scalar_q.
1717 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1718 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1719 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1720 (thunderx2t99_asimd_lut): Add missing tbl types.
1721 (thunderx2t99_asimd_ext): Delete.
1722 (thunderx2t99_asimd_load1_1_mult): Delete.
1723 (thunderx2t99_asimd_load1_2_mult): Delete.
1724 (thunderx2t99_asimd_load1_ldp): New.
1725 (thunderx2t99_asimd_load1): New.
1726 (thunderx2t99_asimd_load2): Add missing *load2* types.
1727 (thunderx2t99_asimd_load3): New.
1728 (thunderx2t99_asimd_load4): New.
1729 (thunderx2t99_asimd_store1_1_mult): Delete.
1730 (thunderx2t99_asimd_store1_2_mult): Delete.
1731 (thunderx2t99_asimd_store2_mult): Delete.
1732 (thunderx2t99_asimd_store2_onelane): Delete.
1733 (thunderx2t99_asimd_store_stp): New.
1734 (thunderx2t99_asimd_store1): New.
1735 (thunderx2t99_asimd_store2): New.
1736 (thunderx2t99_asimd_store3): New.
1737 (thunderx2t99_asimd_store4): New.
1739 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1741 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1742 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1744 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1745 Segher Boessenkool <segher@kernel.crashing.org>
1748 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1751 2018-05-17 Richard Biener <rguenther@suse.de>
1753 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1754 for pruning loop and prune defs feeding only already visited PHIs.
1756 2018-05-17 Richard Biener <rguenther@suse.de>
1758 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1760 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1761 Richard Biener <rguenther@suse.de>
1763 PR tree-optimization/85793
1764 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1765 for VMAT_ELEMENTWISE.
1767 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1769 * internal-fn.h (lookup_internal_fn): Declare
1770 * internal-fn.c (lookup_internal_fn): New function.
1771 * gimple.c (gimple_build_call_from_tree): Handle calls to
1773 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1774 internal function names.
1775 * tree-pretty-print.c (dump_generic_node): Likewise.
1776 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1778 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1780 * gimple-fold.h (gimple_build): Make the function forms take
1781 combined_fn rather than built_in_function.
1782 (gimple_simplify): Likewise.
1783 * gimple-match-head.c (gimple_simplify): Likewise.
1784 * gimple-fold.c (gimple_build): Likewise.
1785 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1786 rather than gimple_build_call_internal.
1787 (get_initial_defs_for_reduction): Likewise.
1788 (vect_create_epilog_for_reduction): Likewise.
1789 (vectorizable_live_operation): Likewise.
1791 2018-05-17 Martin Liska <mliska@suse.cz>
1793 * gimple-ssa-sprintf.c (format_directive): Do not use
1794 space in between 'G_' and '('.
1796 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1799 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1800 even if the mask is not all ones.
1803 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1805 (ix86_gimple_fold_builtin): Likewise.
1808 * config/i386/i386.c: Include tree-vector-builder.h.
1809 (ix86_vector_shift_count): New function.
1810 (ix86_fold_builtin): Fold shift builtins by scalar count.
1811 (ix86_gimple_fold_builtin): Likewise.
1813 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1814 _mm512_setzero): New intrinsics.
1816 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1817 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1819 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1820 code generation for cases where splatting a value is not useful.
1821 * simplify-rtx.c (simplify_ternary_operation): Simplify
1822 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1823 mode to a vec_concat.
1825 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1827 * config.gcc: Support "goldmont-plus".
1828 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1830 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1831 PROCESSOR_GOLDMONT_PLUS.
1832 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1833 (processor_target_table): Add "goldmont-plus".
1834 (PTA_GOLDMONT_PLUS): Define.
1835 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1836 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1837 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1838 (fold_builtin_cpu): Add "goldmont-plus".
1839 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1840 (ix86_option_override_internal): Add "goldmont-plus".
1841 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1842 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1843 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1844 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1846 2018-05-17 Richard Biener <rguenther@suse.de>
1848 PR tree-optimization/85757
1849 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1850 remove defs that only feed that PHI from further processing.
1852 2018-05-16 Jim Wilson <jimw@sifive.com>
1854 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1856 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1858 2018-05-16 Mark Wielaard <mark@klomp.org>
1860 * dwarf2out.c (count_index_strings): New function.
1861 (output_indirect_strings): Call count_index_strings and generate
1862 header for dwarf_version >= 5.
1864 2018-05-16 Mark Wielaard <mark@klomp.org>
1866 * dwarf2out.c (dwarf_FORM): New function.
1867 (set_indirect_string): Use dwarf_FORM.
1868 (reset_indirect_string): Likewise.
1869 (size_of_die): Likewise.
1870 (value_format): Likewise.
1871 (output_die): Likewise.
1872 (add_skeleton_AT_string): Likewise.
1873 (output_macinfo_op): Likewise.
1874 (index_string): Likewise.
1875 (output_index_string_offset): Likewise.
1876 (output_index_string): Likewise.
1877 (count_index_strings): Likewise.
1879 2018-05-16 Carl Love <cel@us.ibm.com>
1881 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1882 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1884 2018-05-16 Martin Jambor <mjambor@suse.cz>
1886 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1887 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1889 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1891 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1892 (fnma<mode>4): Likewise.
1893 (fms<mode>4): Likewise.
1894 (fnms<mode>4): Likewise.
1895 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1896 (aarch64_fnma<mode>4): Likewise.
1897 (aarch64_fms<mode>4): Likewise.
1898 (aarch64_fnms<mode>4): Likewise.
1899 (aarch64_fnmadd<mode>4): Likewise.
1901 2018-05-16 Jason Merrill <jason@redhat.com>
1903 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1905 2018-05-16 Richard Biener <rguenther@suse.de>
1907 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1908 (dump_stmt_cost): Declare.
1909 (add_stmt_cost): Dump cost we add.
1910 (add_stmt_costs): New function.
1911 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1913 (vect_analyze_stmt): Adjust prototype.
1914 (vectorizable_condition): Likewise.
1915 (vectorizable_live_operation): Likewise.
1916 (vectorizable_reduction): Likewise.
1917 (vectorizable_induction): Likewise.
1918 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1919 cost vector to pass to vectorizable_ and record afterwards.
1920 (vect_model_reduction_cost): Take cost vector argument and adjust.
1921 (vect_model_induction_cost): Likewise.
1922 (vectorizable_reduction): Likewise.
1923 (vectorizable_induction): Likewise.
1924 (vectorizable_live_operation): Likewise.
1925 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1926 SLP_TREE_NUMBER_OF_VEC_STMTS.
1927 (vect_analyze_slp_cost_1): Remove.
1928 (vect_analyze_slp_cost): Likewise.
1929 (vect_slp_analyze_node_operations): Take visited args and
1930 a target cost vector. Avoid processing already visited stmt sets.
1931 (vect_slp_analyze_operations): Use a local cost vector to gather
1932 costs and register those of non-discarded instances.
1933 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1934 (vect_schedule_slp_instance): Remove copying of
1935 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1937 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1938 adding cost. Record cost entry location.
1939 (vect_prologue_cost_for_slp_op): Function to compute cost of
1940 a constant or invariant generated for SLP vect in the prologue,
1941 split out from vect_analyze_slp_cost_1.
1942 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1943 (vect_model_promotion_demotion_cost): Likewise.
1944 (vect_model_store_cost): Likewise, make static.
1945 (vect_model_load_cost): Likewise.
1946 (vectorizable_bswap): Add cost vector arg and adjust.
1947 (vectorizable_call): Likewise.
1948 (vectorizable_simd_clone_call): Likewise.
1949 (vectorizable_conversion): Likewise.
1950 (vectorizable_assignment): Likewise.
1951 (vectorizable_shift): Likewise.
1952 (vectorizable_operation): Likewise.
1953 (vectorizable_store): Likewise.
1954 (vectorizable_load): Likewise.
1955 (vectorizable_condition): Likewise.
1956 (vectorizable_comparison): Likewise.
1957 (can_vectorize_live_stmts): Likewise.
1958 (vect_analyze_stmt): Likewise.
1959 (vect_transform_stmt): Adjust calls to vectorizable_*.
1960 * tree-vectorizer.c: Include gimple-pretty-print.h.
1961 (dump_stmt_cost): New function.
1963 2018-05-16 Richard Biener <rguenther@suse.de>
1965 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1966 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1967 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1968 (check_name): New callback.
1969 (dse_classify_store): Track cycles via a visited bitmap of PHI
1970 defs and simplify handling of in-loop and across loop dead stores
1971 and properly fail for loop-variant refs. Handle byte-tracking with
1972 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1975 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1977 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1978 (vect_get_mask_type_for_stmt): Likewise.
1979 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1981 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1982 to determine the statement's vector type and the vector type that
1983 should be used for calculating nunits. Deal with cases in which
1984 the type has to be deferred.
1985 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1986 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1987 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1988 (vect_determine_vf_for_stmt): New functions, split out from...
1989 (vect_determine_vectorization_factor): ...here.
1990 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1991 (vect_get_mask_type_for_stmt): New functions, split out from
1992 vect_determine_vectorization_factor.
1994 2018-05-16 Richard Biener <rguenther@suse.de>
1996 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1997 verify the [VEC_]COND_EXPR embedded comparison.
1999 2018-05-15 Martin Sebor <msebor@redhat.com>
2001 PR tree-optimization/85753
2002 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2003 RECORD_TYPE in addition to ARRAY_TYPE.
2005 2018-05-15 Martin Sebor <msebor@redhat.com>
2008 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2010 2018-05-15 Richard Biener <rguenther@suse.de>
2012 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2013 add by_clobber_p one. Change algorithm to collect all defs
2014 representing uses we need to walk and try reducing them to
2015 a single one before failing.
2016 (dse_dom_walker::dse_optimize_stmt): Adjust.
2018 2018-05-13 Mark Wielaard <mark@klomp.org>
2020 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2021 (size_of_loc_descr): Likewise.
2022 (output_loc_operands): Likewise.
2023 (output_loc_operands_raw): Likewise.
2024 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2025 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2026 (hash_loc_operands): Likewise.
2027 (compare_loc_operands): Likewise.
2029 2018-05-14 Mark Wielaard <mark@klomp.org>
2031 * dwarf2out.c (count_index_addrs): New function.
2032 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2034 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2036 PR tree-optimization/83648
2037 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2038 return value as malloc candidate.
2040 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2043 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2044 as true in call to suggest_attribute.
2046 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2048 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2049 -mreadonly-in-sdata.
2051 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2053 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2055 (aarch64_crypto_aesd_fused): Likewise.
2057 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2059 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2060 (movsi_aarch64): Likewise.
2061 (load_pairsi): Likewise.
2062 (load_pairdi): Likewise.
2063 (store_pairsi): Likewise.
2064 (store_pairdi): Likewise.
2065 (load_pairsf): Likewise.
2066 (load_pairdf): Likewise.
2067 (store_pairsf): Likewise.
2068 (store_pairdf): Likewise.
2069 (zero_extend): Likewise.
2070 (trunc): Swap alternatives.
2071 (fcvt_target): Add '?' to prefer w over r.
2073 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2076 * config/i386/i386.md: Disallow non-commutative arithmetics in
2077 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2078 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2079 in the peephole2 before it.
2081 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2083 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2084 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2085 (ix86_handle_option): Handle -mcldemote.
2086 * config.gcc: New header.
2087 * config/i386/cldemoteintrin.h: New file.
2088 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2089 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2092 OPTION_MASK_ISA_CLDEMOTE.
2093 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2094 (ix86_valid_target_attribute_inner_p): Ditto.
2095 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2096 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2097 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2098 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2099 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2101 * config/i386/i386.opt: Add -mcldemote.
2102 * config/i386/x86intrin.h: New header.
2103 * doc/invoke.texi: Add -mcldemote.
2105 2018-05-14 Richard Biener <rguenther@suse.de>
2107 * doc/match-and-simplify.texi: Adjust :s documentation.
2109 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2111 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2112 intended memcpy size.
2113 (REORDER_45): Likewise.
2115 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2117 * sort.cc: New file.
2118 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2119 * vec.c (qsort_chk): Use gcc_qsort.
2120 * Makefile.in (OBJS-libcommon): Add sort.o.
2121 (build/sort.o): New target. Use it...
2122 (BUILD_RTL): ... here, and...
2123 (build/gencfn-macros): ... here, and...
2124 (build/genmatch): ... here.
2126 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2127 Chung-Ju Wu <jasonwucj@gmail.com>
2129 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2130 * config/nds32/nds32-graywolf.md: New file.
2131 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2132 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2134 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2135 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2136 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2137 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2138 * config/nds32/pipelines.md: Include n15 settings.
2140 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2141 Chung-Ju Wu <jasonwucj@gmail.com>
2143 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2144 * config/nds32/nds32-n13.md: New file.
2145 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2146 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2148 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2149 * config/nds32/nds32.md (pipeline_model): Add n13.
2150 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2151 * config/nds32/pipelines.md: Include n13 settings.
2153 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2154 Chung-Ju Wu <jasonwucj@gmail.com>
2156 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2157 * config/nds32/nds32-n10.md: New file.
2158 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2159 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2161 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2162 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2163 * config/nds32/nds32.md (pipeline_model): Add n10.
2164 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2165 * config/nds32/pipelines.md: Include n10 settings.
2167 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2168 Kito Cheng <kito.cheng@gmail.com>
2169 Chung-Ju Wu <jasonwucj@gmail.com>
2171 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2172 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2173 Add enum values for DSP extension instructions.
2174 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2176 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2177 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2179 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2180 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2181 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2182 * config/nds32/nds32-intrinsic.md: Likewise.
2183 * config/nds32/nds32_intrinsic.h: Likewise.
2184 * config/nds32/nds32-md-auxiliary.c: Likewise.
2185 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2186 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2187 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2188 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2189 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2190 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2191 TYPE_DMAC in switch statement.
2192 * config/nds32/nds32.c: New checking and implementation for DSP
2193 extension instructions.
2194 * config/nds32/nds32.h: Likewise.
2195 * config/nds32/nds32.md: Likewise.
2196 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2197 * config/nds32/predicates.md: Implement new predicates for DSP
2200 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2202 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2203 Reformat alternatives and attributes so it is easier to identify
2204 which constraints/attributes go with which instruction.
2205 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2206 (mov<mode>_softfloat32, FMOVE64): Likewise.
2207 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2208 (mov<mode>_softfloat64, FMOVE64): Likewise.
2210 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2212 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2214 (Basic PowerPC Built-in Functions): The new name of the
2215 subsection previously known as "PowerPC Built-in Functions".
2216 (Basic PowerPC Built-in Functions Available on all Configurations):
2218 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2219 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2220 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2221 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2223 2018-05-11 Martin Jambor <mjambor@suse.cz>
2226 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2229 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2232 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2234 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2236 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2237 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2238 (ix86_handle_option): Handle -mwaitpkg.
2239 * config.gcc: New header.
2240 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2241 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2242 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2244 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2245 OPTION_MASK_ISA_WAITPKG.
2246 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2247 (ix86_option_override_internal): Add PTA_WAITPKG.
2248 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2249 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2250 IX86_BUILTIN_TPAUSE.
2251 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2252 __builtin_ia32_umwait and __builtin_ia32_tpause.
2253 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2254 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2255 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2256 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2257 UNSPECV_TPAUSE): New.
2258 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2259 * config/i386/i386.opt: Add -mwaitpkg.
2260 * config/i386/waitpkgintrin.h: New file.
2261 * config/i386/x86intrin.h: New header.
2262 * doc/invoke.texi: Add -mwaitpkg.
2264 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2267 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2269 (cortex-m0): Use armv6s-m isa.
2270 (cortex-m0plus): Likewise.
2271 (cortex-m1): Likewise.
2272 (cortex-m0.small-multiply): Likewise.
2273 (cortex-m0plus.small-multiply): Likewise.
2274 (cortex-m1.small-multiply): Likewise.
2276 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2277 Jakub Jelinek <jakub@redhat.com>
2279 PR tree-optimization/85692
2280 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2281 source permute as well.
2283 2018-05-11 Martin Liska <mliska@suse.cz>
2286 * doc/extend.texi: Document LLVM style format for no_sanitize
2289 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2291 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2292 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2293 (mode_supports_vsx_dform_quad): Likewise.
2294 (mode_supports_vmx_dform): Move these functions to be next to the
2295 other mode_supports functions.
2296 (mode_supports_dq_form): Likewise.
2297 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2298 mode_supports_dq_form.
2299 (reg_offset_addressing_ok_p): Likewise.
2300 (offsettable_ok_by_alignment): Likewise.
2301 (rs6000_legitimate_offset_address_p): Likewise.
2302 (legitimate_lo_sum_address_p): Likewise.
2303 (rs6000_legitimize_address): Likewise.
2304 (rs6000_legitimize_reload_address): Likewise.
2305 (rs6000_secondary_reload_inner): Likewise.
2306 (rs6000_preferred_reload_class): Likewise.
2307 (rs6000_output_move_128bit): Likewise.
2309 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2311 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2312 Generate SImode target register for null target.
2313 <case IX86_BUILTIN_XGETBV>: Ditto.
2314 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2315 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2317 2018-05-10 Carl Love <cel@us.ibm.com>
2319 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2320 dcbtt and dcbtstt if operands[2] is 0.
2322 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2325 * config/i386/sse.md (usadv64qi): New expander.
2327 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2329 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2330 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2331 -maltivec=be support.
2332 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2333 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2334 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2335 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2336 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2337 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2338 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2339 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2340 altivec_vsumsws): Adjust.
2341 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2342 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2343 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2345 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2346 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2347 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2348 (altivec_lve<VI_char>x): Delete expand.
2349 (*altivec_lve<VI_char>x_internal): Rename to...
2350 (altivec_lve<VI_char>x): ... this.
2351 (altivec_lvxl_<mode>): Delete expand.
2352 (*altivec_lvxl_<mode>_internal): Rename to ...
2353 (altivec_lvxl_<mode>): ... this.
2354 (altivec_stvxl_<mode>): Delete expand.
2355 (*altivec_stvxl_<mode>_internal): Rename to ...
2356 (altivec_stvxl_<mode>): ... this.
2357 (altivec_stve<VI_char>x): Delete expand.
2358 (*altivec_stve<VI_char>x_internal): Rename to ...
2359 (altivec_stve<VI_char>x): ... this.
2360 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2361 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2362 reduc_plus_scal_<mode>): Adjust.
2363 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2365 (rs6000_cpu_cpp_builtins): Adjust.
2366 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2367 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2368 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2369 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2370 -maltivec=be support.
2371 (rs6000_split_vec_extract_var): Adjust.
2372 (rs6000_split_v4si_init): Adjust.
2373 (swap_selector_for_mode): Delete.
2374 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2375 altivec_expand_stvex_be): Delete.
2376 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2377 -maltivec=be support.
2378 (rs6000_gimple_fold_builtin): Ditto.
2379 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2381 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2382 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2383 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2384 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2385 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2386 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2387 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2388 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2389 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2390 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2391 anonymous split): Adjust.
2392 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2393 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2395 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2397 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2398 when --with-gxx-include-dir is also specified.
2399 * configure: Regenerate.
2401 2018-05-09 Jim Wilson <jimw@sifive.com>
2404 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2405 * config/riscv/t-withmultilib: New.
2406 * config/riscv/withmultilib.h: New.
2407 * doc/install.texi: Document RISC-V --with-multilib-list support.
2409 2018-05-09 Richard Biener <rguenther@suse.de>
2411 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2413 (vect_bb_vectorization_profitable_p): Adjust. Compute
2414 actual scalar cost using the cost vector and the add_stmt_cost
2417 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2419 PR rtl-optimization/85645
2420 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2421 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2423 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2425 PR rtl-optimization/85645
2426 * shrink-wrap.c (spread_components): Return a boolean saying if
2427 anything was changed.
2428 (try_shrink_wrapping_separate): Iterate spread_components until
2429 nothing changes anymore.
2431 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2433 PR rtl-optimization/85645
2434 * regrename.c (build_def_use): Also kill the chains that include the
2435 destination of a REG_CFA_REGISTER note.
2437 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2439 PR rtl-optimization/85645
2440 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2441 insn that has a REG_CFA_REGISTER note.
2443 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2445 * cfgexpand.c (expand_clobber): New function.
2446 (expand_gimple_stmt_1): Use it.
2447 * tree-vect-stmts.c (vect_clobber_variable): New function,
2449 (vectorizable_simd_clone_call): ...here.
2450 (vectorizable_store): Emit a clobber either side of an
2451 IFN_STORE_LANES sequence.
2452 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2454 2018-05-09 Tom de Vries <tom@codesourcery.com>
2457 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2458 (define_insn "trap_if_false"): Add exit after trap.
2460 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2462 PR rtl-optimization/85638
2463 * bb-reorder.c: Include common/common-target.h.
2464 (create_forwarder_block): New function extracted from...
2465 (fix_up_crossing_landing_pad): ...here. Rename into...
2466 (dw2_fix_up_crossing_landing_pad): ...this.
2467 (sjlj_fix_up_crossing_landing_pad): New function.
2468 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2469 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2470 from both partitions and exit the loop after one iteration.
2472 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2475 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2477 (Basic PowerPC Built-in Functions): The new name of the
2478 subsection previously known as "PowerPC Built-in Functions".
2479 (Basic PowerPC Built-in Functions Available on all Configurations):
2481 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2483 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2485 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2487 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2490 2018-05-08 Jim Wilson <jimw@sifive.com>
2492 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2493 (LD_EMUL_SUFFIX): New.
2494 (LINK_SPEC): Use it.
2496 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2498 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2500 (Basic PowerPC Built-in Functions): The new name of the
2501 subsection previously known as "PowerPC Built-in Functions".
2502 (Basic PowerPC Built-in Functions Available on all Configurations):
2504 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2506 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2508 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2510 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2513 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2516 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2517 after cmpelim optimization.
2519 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2521 * config.gcc: Support "goldmont".
2522 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2523 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2525 * config/i386/i386.c (m_GOLDMONT): Define.
2526 (processor_target_table): Add "goldmont".
2527 (PTA_GOLDMONT): Define.
2528 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2529 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2530 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2531 (fold_builtin_cpu): Add "goldmont".
2532 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2533 (ix86_option_override_internal): Add "goldmont".
2534 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2535 (processor_type): Add PROCESSOR_GOLDMONT.
2536 * config/i386/i386.md: Add CPU "glm".
2537 * config/i386/glm.md: New file.
2538 * config/i386/x86-tune.def: Add m_GOLDMONT.
2539 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2541 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2544 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2546 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2547 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2548 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2551 * config/i386/i386.c (ix86_fold_builtin): Handle
2552 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2555 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2556 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2558 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2561 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2562 (check_arch): Likewise.
2563 (check_fpu): Return the result rather than printing it.
2564 (end arch): Fix operator precedence.
2565 (end cpu): Likewise.
2566 (END): Print the result from check_fpu.
2568 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2569 Alan Hayward <alan.hayward@arm.com>
2570 David Sherwood <david.sherwood@arm.com>
2572 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2573 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2574 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2575 (*fcmuo<mode>_and): New patterns.
2577 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2579 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2580 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2581 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2582 (cmp_op, sve_imm_con): New code attributes.
2583 (SVE_COND_INT_CMP, imm_con): Delete.
2584 (cmp_op): Remove above unspecs from int attribute.
2585 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2587 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2588 comparison-specific unspecs.
2589 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2590 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2591 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2592 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2593 (*vec_fcm<cmp_op><mode>): Rename to...
2594 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2595 (*vec_fcmuo<mode>): Rename to...
2596 (*fcmuo<mode>): ...this and adjust likewise.
2597 (*pred_fcm<cmp_op><mode>): New pattern.
2598 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2599 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2601 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2603 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2604 (aarch64_emit_sve_predicated_cond): New function.
2605 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2606 (aarch64_emit_unspec_cond_or): Replace with...
2607 (aarch64_emit_sve_or_conds): ...this new function. Use
2608 aarch64_emit_sve_ptrue_op for the individual comparisons and
2609 aarch64_emit_binop to OR them together.
2610 (aarch64_emit_inverted_unspec_cond): Replace with...
2611 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2612 aarch64_emit_sve_ptrue_op for the comparison and
2613 aarch64_emit_unop to invert the result.
2614 (aarch64_expand_sve_vec_cmp_float): Update after the above
2615 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2617 2018-05-07 Nathan Sidwell <nathan@acm.org>
2619 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2620 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2621 (Backwards Compatibility): Likewise.
2623 2018-05-07 Luis Machado <luis.machado@linaro.org>
2627 2018-05-07 Luis Machado <luis.machado@linaro.org>
2629 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2630 <prefetch_dynamic_strides>: New const bool field.
2631 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2632 prefetch_dynamic_strides.
2633 (exynosm1_prefetch_tune): Likewise.
2634 (thunderxt88_prefetch_tune): Likewise.
2635 (thunderx_prefetch_tune): Likewise.
2636 (thunderx2t99_prefetch_tune): Likewise.
2637 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2638 (aarch64_override_options_internal): Update to set
2639 PARAM_PREFETCH_DYNAMIC_STRIDES.
2640 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2641 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2642 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2643 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2644 prefetch-dynamic-strides setting.
2646 2018-05-07 Luis Machado <luis.machado@linaro.org>
2648 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2649 <minimum_stride>: New const int field.
2650 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2651 minimum_stride field.
2652 (exynosm1_prefetch_tune): Likewise.
2653 (thunderxt88_prefetch_tune): Likewise.
2654 (thunderx_prefetch_tune): Likewise.
2655 (thunderx2t99_prefetch_tune): Likewise.
2656 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2657 (aarch64_override_options_internal): Update to set
2658 PARAM_PREFETCH_MINIMUM_STRIDE.
2659 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2660 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2661 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2662 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2663 stride is constant and is below the minimum stride threshold.
2665 2018-05-07 Luis Machado <luis.machado@linaro.org>
2667 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2670 2018-05-07 Luis Machado <luis.machado@linaro.org>
2672 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2673 <prefetch_dynamic_strides>: New const bool field.
2674 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2675 prefetch_dynamic_strides.
2676 (exynosm1_prefetch_tune): Likewise.
2677 (thunderxt88_prefetch_tune): Likewise.
2678 (thunderx_prefetch_tune): Likewise.
2679 (thunderx2t99_prefetch_tune): Likewise.
2680 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2681 (aarch64_override_options_internal): Update to set
2682 PARAM_PREFETCH_DYNAMIC_STRIDES.
2683 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2684 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2685 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2686 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2687 prefetch-dynamic-strides setting.
2689 2018-05-07 Luis Machado <luis.machado@linaro.org>
2691 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2692 <minimum_stride>: New const int field.
2693 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2694 minimum_stride field.
2695 (exynosm1_prefetch_tune): Likewise.
2696 (thunderxt88_prefetch_tune): Likewise.
2697 (thunderx_prefetch_tune): Likewise.
2698 (thunderx2t99_prefetch_tune): Likewise.
2699 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2700 (aarch64_override_options_internal): Update to set
2701 PARAM_PREFETCH_MINIMUM_STRIDE.
2702 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2703 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2704 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2705 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2706 stride is constant and is below the minimum stride threshold.
2708 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2711 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2712 the type is addressable. Don't force op into register if it has
2715 2018-05-05 Roland McGrath <mcgrathr@google.com>
2718 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2719 any section for which we don't know a specific type it should have,
2720 regardless of name. Previously this was done only for the exact
2721 names ".init_array", ".fini_array", and ".preinit_array".
2722 (default_elf_asm_named_section): Add comment about
2723 relationship with default_section_type_flags and SECTION_NOTYPE.
2724 (get_section): Don't consider it a type conflict if one side has
2725 SECTION_NOTYPE and the other doesn't, as long as neither has the
2726 SECTION_BSS et al used in the default_section_type_flags logic.
2728 2018-05-05 Tom de Vries <tom@codesourcery.com>
2731 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2732 (workaround_barsyncs): New function.
2733 (nvptx_reorg): Use workaround_barsyncs.
2734 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2735 (define_expand "nvptx_membar_cta"): New define_expand.
2736 (define_insn "*nvptx_membar_cta"): New insn.
2738 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2740 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2741 To improve optimization opportunities.
2742 * builtin-types.def: The new needed builtin types for the above.
2744 2018-05-04 Richard Biener <rguenther@suse.de>
2746 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2747 * gimple-ssa-store-merging.c
2748 (imm_store_chain_info::output_merged_store): Remove redundant create,
2749 release split_store vector contents on failure.
2750 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2751 scalar stmt vector on cache hit.
2753 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2755 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2757 * config.gcc (powerpc-xilinx-eabi*): Remove.
2758 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2760 (fusion_addis_mem_combo_load): Ditto.
2761 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2763 (rs6000_cpu_cpp_builtins): Ditto.
2764 * config/rs6000/rs6000-linux.c
2765 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2766 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2767 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2769 (rs6000_setup_reg_addr_masks): Ditto.
2770 (rs6000_init_hard_regno_mode_ok): Ditto.
2771 (rs6000_option_override_internal): Ditto.
2772 (legitimate_lo_sum_address_p): Ditto.
2773 (rs6000_legitimize_address): Ditto.
2774 (rs6000_legitimize_reload_address): Ditto.
2775 (rs6000_legitimate_address_p): Ditto.
2776 (abi_v4_pass_in_fpr): Ditto.
2777 (setup_incoming_varargs): Ditto.
2778 (rs6000_gimplify_va_arg): Ditto.
2779 (rs6000_split_multireg_move): Ditto.
2780 (rs6000_savres_strategy): Ditto.
2781 (rs6000_emit_prologue_components): Ditto.
2782 (rs6000_emit_epilogue_components): Ditto.
2783 (rs6000_emit_prologue): Ditto.
2784 (rs6000_emit_epilogue): Ditto.
2785 (rs6000_elf_file_end): Ditto.
2786 (rs6000_function_value): Ditto.
2787 (rs6000_libcall_value): Ditto.
2788 * config/rs6000/rs6000.h: Ditto.
2789 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2790 (TARGET_MINMAX): ... this. New.
2791 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2792 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2793 (*movsi_internal1_single): Delete.
2794 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2795 mfpu=, mxilinx-fpu): Delete.
2796 * config/rs6000/singlefp.h: Delete.
2797 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2798 * config/rs6000/t-rs6000: Ditto.
2799 * config/rs6000/t-xilinx: Delete.
2800 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2801 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2802 (VStype_simple): Delete.
2803 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2804 * config/rs6000/xfpu.h: Delete.
2805 * config/rs6000/xfpu.md: Delete.
2806 * config/rs6000/xilinx.h: Delete.
2807 * config/rs6000/xilinx.opt: Delete.
2808 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2809 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2811 2018-05-04 Tom de Vries <tom@codesourcery.com>
2814 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2817 2018-05-04 Richard Biener <rguenther@suse.de>
2820 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2821 (expand_complex_div_wide): Likewise.
2822 (expand_complex_operations_1): Likewise.
2823 (expand_complex_libcall): Preserve EH info of the original stmt.
2824 (tree_lower_complex): Handle removed blocks.
2825 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2826 on complex multiplication and division libcall builtins.
2828 2018-05-04 Richard Biener <rguenther@suse.de>
2831 * fold-const.c (negate_expr_p): Restrict negation of operand
2832 zero of a division to when we know that can happen without
2834 (fold_negate_expr_1): Likewise.
2836 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2839 * real.h (real_nextafter): Declare.
2840 * real.c (real_nextafter): New function.
2841 * fold-const-call.c (fold_const_nextafter): New function.
2842 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2843 CASE_CFN_NEXTTOWARD.
2844 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2845 even when arg1_mode is different from arg0_mode.
2847 2018-05-03 Nathan Sidwell <nathan@acm.org>
2849 * doc/extend.texi (Deprecated Features): Remove
2851 (Backwards Compatibility): Likewise.
2852 * doc/invoke.texi (C++ Language Options): Likewise.
2853 (C++ Dialect Options): Likewise.
2855 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2858 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2859 _mm512_mask_mullox_epi64): New intrinsics.
2861 2018-05-03 Tom de Vries <tom@codesourcery.com>
2864 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2865 dump files): Add offload-tree.
2867 2018-05-03 Richard Biener <rguenther@suse.de>
2869 PR tree-optimization/85615
2870 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2871 to loops not nested in BBs loop father to avoid creating multi-entry
2874 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2876 PR tree-optimization/70291
2877 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2878 arguments. Change return type to tree. Emit libcall as a new
2879 statement rather than replacing existing one when inplace_p is true.
2880 (expand_complex_multiplication_components): New function.
2881 (expand_complex_multiplication): Expand floating-point complex
2882 multiplication using the above.
2883 (expand_complex_division): Rename inner_type parameter to type.
2884 Update expand_complex_libcall call-site.
2885 (expand_complex_operations_1): Update expand_complex_multiplication
2886 and expand_complex_division call-sites.
2888 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2891 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2892 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2893 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2894 the highest significant bit of the shift count mask is clear. In
2895 check whether and[sq]i3 is needed verify that all significant bits
2896 of the shift count other than the highest are set.
2898 2018-05-02 Tom de Vries <tom@codesourcery.com>
2901 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2902 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2903 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2904 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2905 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2906 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2907 __builtin_goacc_parlevel_size.
2909 2018-05-02 Richard Biener <rguenther@suse.de>
2911 PR tree-optimization/85597
2912 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2913 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2916 2018-05-02 Tom de Vries <tom@codesourcery.com>
2919 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2920 dump files): Add ltrans-tree.
2922 2018-05-02 Tom de Vries <tom@codesourcery.com>
2925 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2926 dump files): Add wpa-ipa.
2928 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2930 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2931 powerpc*-*-linux*paired* target.
2932 * config/rs6000/750cl.h: Delete.
2933 * config/rs6000/paired.h: Delete.
2934 * config/rs6000/paired.md: Delete.
2935 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2937 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2938 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2939 comment. Remove paired float support.
2940 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2941 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2943 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2944 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2946 * config/rs6000/rs6000.c: Remove paired float support.
2947 (paired_expand_vector_init, paired_expand_vector_move,
2948 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2949 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2950 paired_expand_builtin, paired_expand_predicate_builtin,
2951 paired_init_builtins): Delete.
2952 * config/rs6000/rs6000.h: Remove paired float support.
2953 * config/rs6000/rs6000.md: Remove paired float support.
2954 (move_from_CR_ov_bit): Delete.
2955 * config/rs6000/rs6000.opt (mpaired): Delete.
2956 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2957 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2959 2018-05-02 Richard Biener <rguenther@suse.de>
2962 * gimplify.c (gimplify_save_expr): When in SSA form allow
2963 SAVE_EXPRs to compute to SSA vars.
2965 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2968 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2969 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2970 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2971 clobber operands[2], instead use a new pseudo. Formatting fixes.
2973 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2975 PR tree-optimization/85586
2976 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2977 exit early for statements in the same group if the accesses are
2980 2018-05-02 Tom de Vries <tom@codesourcery.com>
2983 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2986 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2988 PR tree-optimization/85143
2989 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2991 2018-05-01 Tom de Vries <tom@codesourcery.com>
2994 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2995 not found" error message.
2997 2018-05-01 Tom de Vries <tom@codesourcery.com>
3000 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3001 * vec.c (test_ordered_remove_if): New function.
3002 (vec_c_tests): Call test_ordered_remove_if.
3003 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3004 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3005 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3006 VEC_ORDERED_REMOVE_IF.
3008 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3010 PR tree-optimization/82665
3011 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3012 pointer subtraction where arguments come from a memchr call.
3014 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3016 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3017 --push-state --as-needed and --pop-state instead of --as-needed and
3018 --no-as-needed if ld supports it.
3019 * configure: Regenerated.
3022 * doc/install.texi2html: Replace _002d with - and _002a with * in
3023 generated html files using sed.
3025 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3028 * gcc-rich-location.c (blank_line_before_p): New function.
3029 (use_new_line): New function.
3030 (gcc_rich_location::add_fixit_insert_formatted): New function.
3031 * gcc-rich-location.h
3032 (gcc_rich_location::add_fixit_insert_formatted): New function.
3034 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3036 * selftest.c (assert_streq): Rename "expected" and "actual" to
3037 "val1" and "val2". Extend NULL-handling to cover both inputs
3038 symmetrically, while still requiring both to be non-NULL for a pass.
3039 * selftest.h (assert_streq): Rename "expected" and "actual" to
3041 (ASSERT_EQ): Likewise.
3042 (ASSERT_EQ_AT): Likewise.
3043 (ASSERT_KNOWN_EQ): Likewise.
3044 (ASSERT_KNOWN_EQ_AT): Likewise.
3045 (ASSERT_NE): Likewise.
3046 (ASSERT_MAYBE_NE): Likewise.
3047 (ASSERT_MAYBE_NE_AT): Likewise.
3048 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3049 the assertion to pass.
3050 (ASSERT_STREQ_AT): Likewise.
3052 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3054 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3055 interaction with -pie.
3057 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3059 * selftest.h: Fix alphabetization of per-source-file selftest
3062 2018-04-30 Jason Merrill <jason@redhat.com>
3064 PR c++/61982 - dead stores to destroyed objects.
3065 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3068 2018-04-30 Jason Merrill <jason@redhat.com>
3070 * tree.c (build_clobber): New.
3071 * tree.h: Declare it.
3072 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3074 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3076 * diagnostic-show-locus.c (layout::layout): Update for
3077 location_get_source_line returning a char_span.
3078 (struct char_span): Move to input.h.
3079 (struct correction): Update for fields in char_span becoming
3081 (struct source_line): Update for location_get_source_line
3082 returning a char_span.
3083 (layout::print_line): Likewise.
3084 * edit-context.c (edited_file::print_content): Likewise.
3085 (edited_file::print_diff_hunk): Likewise.
3086 (edited_file::print_run_of_changed_lines): Likewise.
3087 (edited_file::get_num_lines): Likewise.
3088 (edited_line::edited_line): Likewise.
3089 * final.c (asm_show_source): Likewise.
3090 * input.c (location_get_source_line): Convert return type
3091 from const char * to char_span, losing the final "line_len"
3093 (dump_location_info): Update for the above.
3094 (get_substring_ranges_for_loc): Likewise. Use a char_span
3095 when handling the literal within the line.
3096 (test_reading_source_line): Update for location_get_source_line
3097 returning a char_span.
3098 * input.h (class char_span): Move here from
3099 diagnostic-show-locus.c, converting from a struct to a class.
3100 Make data members private.
3101 (char_span::operator bool): New.
3102 (char_span::length): New.
3103 (char_span::get_buffer): New.
3104 (char_span::operator[]): New.
3105 (char_span::subspan): Make const.
3106 (char_span::xstrdup): New.
3107 (location_get_source_line): Convert return type from const char *
3108 to char_span, losing the final "line_size" param.
3110 2018-04-30 Jan Hubicka <jh@suse.cz>
3112 * lto-wrapper.c (ltrans_priorities): New static var.
3113 (cmp_priority): New.
3114 (run_gcc): Read priorities and if doing parallel build order
3115 the Makefile by them.
3117 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3119 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3121 2018-04-30 Richard Biener <rguenther@suse.de>
3123 * tree-cfg.c (verify_address): Remove base argument, add
3124 flag whether to check TREE_ADDRESSABLE and do that.
3125 (verify_expr): Remove.
3126 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3127 (verify_gimple_assign_single): Likewise.
3128 (verify_gimple_switch): Likewise.
3129 (verify_expr_location_1): Dereference tp once. Add (disabled)
3130 piece from verify_expr.
3131 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3133 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3135 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3137 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3139 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3140 (small_data_pattern): Likewise.
3141 (arc_rewrite_small_data): Likewise.
3142 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3143 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3144 (get_symbol_alignment): New function.
3145 (legitimate_small_data_address_p): Likewise.
3146 (legitimate_scaled_address): Update, call
3147 legitimate_small_data_address_p.
3148 (output_sdata): New static variable.
3149 (arc_print_operand): Update how we handle small data operands.
3150 (arc_print_operand_address): Likewise.
3151 (arc_legitimate_address_p): Update, use
3152 legitimate_small_data_address_p.
3153 (arc_rewrite_small_data_p): Remove.
3154 (arc_rewrite_small_data_1): Likewise.
3155 (arc_rewrite_small_data): Likewise.
3156 (small_data_pattern): Likewise.
3157 (compact_sda_memory_operand): Update to use
3158 legitimate_small_data_address_p and get_symbol_alignment.
3159 (prepare_move_operands): Don't rewite sdata pattern.
3160 (prepare_extend_operands): Remove.
3161 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3163 (zero_extendqisi2): Likewise.
3164 (zero_extendhisi2): Likewise.
3165 (extendqihi2): Likewise.
3166 (extendqisi2): Likewise.
3167 (extendhisi2): Likewise.
3171 * config/arc/constraints.md (Usd): Change it to memory constraint.
3173 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3175 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3176 as source of std instructions.
3177 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3178 allow 6-bit constants as source for store instructions.
3179 (movdi_insn): Update instruction pattern to allow 6-bit constants
3180 as source for store instructions.
3182 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3184 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3186 2018-04-30 Nathan Sidwell <nathan@acm.org>
3187 Sandra Loosemore <sandra@codesourcery.com>
3189 * dumpfile.c (dump_open): Allow '-' for stdout.
3190 * doc/invoke.texi (Developer Options): Document dump filename
3191 determination early. Document stdin/stdout selection.
3193 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3195 Microblaze Target: PIC data text relative
3197 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3198 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3200 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3201 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3202 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3203 New addressing mode for data-text relative position indepenedent code.
3204 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3205 'ADDRESS_SYMBOLIC_TXT_REL'.
3206 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3207 (microblaze_legitimate_pic_operand): Exclude function calls from
3208 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3209 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3211 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3212 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3213 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3214 for 'address + offset'.
3215 (microblaze_expand_prologue): Add new function prologue call for
3217 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3218 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3219 table in case of TARGET_PIC_DATA_TEXT_REL.
3220 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3221 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3222 Add new macros 'UNSPEC_TEXT',
3223 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3224 + exclude function calls from 'UNSPEC_PLT' in case of data text
3226 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3227 new target hook for generating address diff vector tables in case of
3229 * doc/tm.texi : Regenerate.
3230 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3231 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3232 of addr diff vector generation.
3233 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3234 target hook definition.
3235 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3236 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3237 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3238 option for data text relative.
3240 2018-04-30 Richard Biener <rguenther@suse.de>
3242 * tree-chrec.h (evolution_function_is_constant_p): Remove
3244 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3246 2018-04-30 Richard Biener <rguenther@suse.de>
3249 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3251 2018-04-30 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/28364
3254 PR tree-optimization/85275
3255 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3256 copying first exit test.
3258 2018-04-28 Mark Wielaard <mark@klomp.org>
3260 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3262 (dwarf_AT): Handle DW_AT_addr_base.
3263 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3265 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3268 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3269 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3270 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3271 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3273 2018-04-28 Richard Biener <rguenther@suse.de>
3275 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3276 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3277 to reflect use. Only add interesting stmts.
3279 2018-04-27 Martin Jambor <mjambor@suse.cz>
3282 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3283 the jump function allows for passing through aggregate values.
3285 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3287 * input.h (in_system_header_at): Convert from macro to inline
3289 (from_macro_expansion_at): Likewise.
3290 (from_macro_definition_at): Likewise.
3292 2018-04-27 Jeff Law <law@redhat.com>
3294 * config.gcc: Mark tile* targets as deprecated/obsolete.
3296 2018-04-27 Richard Biener <rguenther@suse.de>
3298 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3301 2018-04-27 Richard Biener <rguenther@suse.de>
3303 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3305 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3307 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3308 with Yd constraint. Set "preferred_for_speed" attribute from
3309 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3311 (*movdi_internal): Ditto.
3312 (movti_interunit splitters): Remove
3313 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3314 (movdi_interunit splitters): Ditto.
3315 * config/i386/constraints.md (Ye): Remove.
3316 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3318 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3321 * config/aarch64/constraints.md (Usg): Limit to 31.
3324 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3326 PR tree-optimization/85529
3327 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3328 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3329 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3330 zero extension or masking of the MSB bit.
3331 (optimize_range_tests): Add FIRST_BB argument, pass it through
3332 to optimize_range_tests_var_bound.
3333 (maybe_optimize_range_tests, reassociate_bb): Adjust
3334 optimize_range_tests callers.
3336 2018-04-26 Richard Biener <rguenther@suse.de>
3337 Jakub Jelinek <jakub@redhat.com>
3339 * cgraph.h (symbol_table): Just declare debug method here.
3340 * symtab.c (symbol_table::debug): Define.
3342 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3344 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3346 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3348 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3349 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3350 (*movdi_internal): Substitute Yi and Yj constraint with x
3351 and Ym and Yn constraint with y constraint. Update "isa"
3352 attribute and set "preferred_for_speed" attribute from
3353 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3354 (*movsi_internal): Ditto.
3355 (*movdf_internal): Ditto.
3356 (*movsf_internal): Ditto.
3357 (*zero_extendsidi2): Ditto.
3358 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3359 (sse2_loadld): Ditto.
3360 (*vec_extract<ssevecmodelower>_0): Ditto.
3361 (*vec_extractv4si_0_zext_sse4): Ditto.
3362 (vec_concatv2di): Ditto.
3363 (*vec_dup<mode>): Ditto.
3364 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3365 * config/i386/constraints.md (Yi): Remove.
3370 2018-04-26 Nathan Sidwell <nathan@acm.org>
3372 * dumpfile.c (dump_open): New.
3373 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3374 (dump_finish): Detect stdio/stderr by value not name.
3376 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3378 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3380 2018-04-26 Tom de Vries <tom@codesourcery.com>
3383 * config/nvptx/nvptx.c (verify_neutering_jumps)
3384 (verify_neutering_labels): New function
3385 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3387 2018-04-26 Tom de Vries <tom@codesourcery.com>
3390 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3391 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3394 2018-04-26 Richard Biener <rguenther@suse.de>
3395 Tom de Vries <tom@codesourcery.com>
3398 * lto-streamer-out.c (output_function): Fixup loops if required to match
3399 discovery done in the reader.
3401 2018-04-26 Richard Biener <rguenther@suse.de>
3403 PR tree-optimization/85116
3404 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3405 have a loop exit from the single latch predecessor. Remove
3406 case of header with just condition.
3407 (ch_base::copy_headers): Exclude infinite loops from any
3409 (pass_ch::execute): Record exits.
3411 2018-04-26 Richard Biener <rguenther@suse.de>
3413 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3414 prologue cost vector and pass it to vect_get_load_cost.
3415 (vect_get_peeling_costs_all_drs): Likewise.
3416 (vect_peeling_hash_get_lowest_cost): Likewise.
3417 (vect_enhance_data_refs_alignment): Likewise.
3419 2018-04-26 Richard Biener <rguenther@suse.de>
3422 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3423 checking of integer<->pointer conversions.
3424 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3425 sign-/zero-extending pointer types.
3426 (expand_omp_for_static_chunk): Likewise.
3428 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3429 Jean Lee <xiaoyur347@gmail.com>
3431 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3432 (TARGET_ASAN_SHADOW_OFFSET): Define.
3433 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3434 true for -fsanitize=address.
3436 2018-04-25 Mark Wielaard <mark@klomp.org>
3438 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3441 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3443 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3444 than "alu", remove explicit "memory" and "imm_disp" attributes.
3445 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3448 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3449 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3450 gen_lowpart_no_emit.
3452 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3455 * config/i386/i386.c (ix86_expand_builtin): Change memory
3456 operand to XI, extend p0 to Pmode.
3457 * config/i386/i386.md: Change unspec volatile and operand
3458 1 mode to XI, change operand 0 mode to P.
3460 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3462 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3463 GET_MODE_MASK before any checking.
3464 (nds32_can_use_bset_p): Likewise.
3465 (nds32_can_use_btgl_p): Likewise.
3467 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3469 * config/nds32/nds32-doubleword.md: New define_split pattern for
3470 illegal register number.
3472 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3474 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3476 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3478 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3480 2018-04-25 Richard Biener <rguenther@suse.de>
3482 * lto-streamer.h (LTO_major_version): Bump to 8.
3484 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3486 * BASE-VER: Set to 9.0.0.
3488 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3490 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3491 in __abskf2 and __powikf2.
3493 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3496 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3497 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3498 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3499 Use the above on operand 2. Reindent.
3500 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3502 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3505 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3507 * config/i386/i386.opt (mcet): Removed.
3508 * doc/install.texi: Remove -mcet documentation.
3509 * doc/invoke.texi: Likewise.
3511 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3514 * doc/install.texi: Remove -mcet from bootstrap-cet.
3516 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3519 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3520 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3524 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3525 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3526 containing a CONST_VECTOR.
3528 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3530 * doc/install.texi: Update newlib dependency for nvptx.
3532 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3535 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3536 instead of INTVAL when shifting x left.
3538 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3540 PR tree-optimization/85478
3541 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3542 vect_grouped_store_supported for single element vectors.
3544 2018-04-24 Richard Biener <rguenther@suse.de>
3547 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3548 load cost increase to the case of non-constant step.
3550 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3553 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3554 destination if any_malformed_asm.
3556 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3559 * expr.c (store_field): In the bitfield case, if the value comes from
3560 a function call and is returned in registers by means of a PARALLEL,
3561 do not change the mode of the temporary unless BLKmode and VOIDmode.
3563 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3565 PR rtl-optimization/85423
3566 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3567 dependencies to debug insns when the previous insn is non-debug.
3569 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3571 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3572 enums into a single definition.
3573 (fls): Fix predicates and printing.
3576 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3578 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3579 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3580 and short u6 immediate.
3581 (check_if_valid_sleep_operand): Remove.
3582 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3584 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3586 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3587 flag_always_save_lp condition.
3588 * config/nds32/nds32.opt (malways-save-lp): New option.
3590 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3592 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3593 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3594 * config/nds32/nds32.h
3595 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3596 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3598 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3600 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3601 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3603 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3604 Chung-Ju Wu <jasonwucj@gmail.com>
3606 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3608 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3609 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3611 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3613 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3615 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3617 * config/nds32/nds32-protos.h (nds32_data_alignment,
3618 nds32_local_alignment): Declare.
3619 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3620 nds32_local_alignment): New functions.
3621 (TARGET_CONSTANT_ALIGNMENT): Define.
3622 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3624 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3626 * config/nds32/nds32.c
3627 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3628 (TARGET_MODES_TIEABLE_P): Likewise.
3630 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3632 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3635 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3636 Chung-Ju Wu <jasonwucj@gmail.com>
3638 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3639 for unaligned access.
3640 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3642 * config/nds32/nds32-intrinsic.md: Likewise.
3643 * config/nds32/nds32_intrinsic.h: Likewise.
3644 * config/nds32/nds32.h (nds32_builtins): Likewise.
3645 * config/nds32/nds32.opt (munaligned-access): New option.
3646 * config/nds32/nds32.c (nds32_asm_file_start): Display
3647 flag_unaligned_access status.
3649 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3651 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3652 -mno-relax is present.
3653 * config/riscv/linux.h (LINK_SPEC): Ditto.
3655 2018-04-20 Martin Sebor <msebor@redhat.com>
3658 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3660 (gimple_fold_builtin_stxcpy_chk): Same.
3661 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3663 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3666 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3667 __powikf2 when long double is IEEE 128-bit.
3669 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3671 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3672 step to make sure stack always aligned.
3674 2018-04-20 Carl Love <cel@us.ibm.com>
3677 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3678 size check for arg0.
3680 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3681 Tom de Vries <tom@codesourcery.com>
3684 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3685 Emit insns for calls too.
3686 (nvptx_find_par): Always look for worker-level predecessor insn.
3687 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3689 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3690 (nvptx_process_pars): Propagate frames for calls.
3692 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3695 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3697 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3698 (ix86_handle_option): Don't handle OPT_mibt.
3699 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3701 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3703 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3704 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3705 (ix86_target_macros): Define __CET__ with flag_cf_protection
3706 for -fcf-protection.
3707 * config/i386/i386.c (isa2_opts): Remove -mibt.
3708 * config/i386/i386.h (TARGET_IBT): Removed.
3709 (TARGET_IBT_P): Likewise.
3710 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3711 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3712 * config/i386/i386.opt (mcet): Update help message.
3715 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3716 -mcet as an alias for -mshstk.
3718 2018-04-20 Richard Biener <rguenther@suse.de>
3721 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3722 complexity by forcing a single use of the multiply operand.
3724 2018-04-20 Martin Jambor <mjambor@suse.cz>
3727 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3728 recursion dependency to only apply to non-clones.
3730 2018-04-20 Martin Jambor <mjambor@suse.cz>
3733 * ipa-cp.c (create_specialized_node): Check that clones of
3734 self-recursive edges exist during IPA-CP.
3736 2018-04-19 Toon Moene <toon@moene.org>
3738 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3741 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3743 PR tree-optimization/85467
3744 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3745 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3746 VECTOR_CST element to type.
3748 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3751 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3752 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3753 (builtin_longjmp): Likewise.
3754 (save_stack_nonlocal): New pattern.
3755 (restore_stack_nonlocal): Likewise.
3757 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3760 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3761 Replace ASM_OUTPUT_LABEL with fprintf.
3763 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3766 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3767 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3768 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3769 define __IBT__ and __SHSTK__ for -fcf-protection.
3770 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3772 (ix86_trampoline_init): Likewise.
3773 (x86_output_mi_thunk): Likewise.
3774 (ix86_notrack_prefixed_insn_p): Likewise.
3775 (ix86_option_override_internal): Don't disallow -fcf-protection.
3776 * config/i386/i386.md (rdssp<mode>): Also enable for
3778 (incssp<mode>): Likewise.
3779 (nop_endbr): Likewise.
3780 * config/i386/i386.opt (mcet): Change help message to built-in
3784 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3785 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3786 enable CET built-in functions.
3788 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3790 * common/config/i386/i386-common.c
3791 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3792 OPTION_MASK_ISA_MOVDIRI_UNSET,
3793 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3794 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3795 * config.gcc (movdirintrin.h): New header.
3796 * config/i386/cpuid.h (bit_MOVDIRI,
3797 bit_MOVDIR64B): New bits.
3798 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3800 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3801 (VOID, PVOID, PCVOID)): New function types.
3802 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3803 __builtin_ia32_directstoreu_u64,
3804 __builtin_ia32_movdir64b): New builtins.
3805 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3806 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3808 (ix86_valid_target_attribute_inner_p): Ditto.
3809 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3810 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3811 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3812 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3813 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3814 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3815 (movdiri<mode>, movdir64b_<mode>): New.
3816 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3817 * config/i386/immintrin.h: Include movdirintrin.h.
3818 * config/i386/movdirintrin.h: New file.
3819 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3821 2018-04-19 Richard Biener <rguenther@suse.de>
3824 * cfg.c (clear_bb_flags): When loop state says we have
3825 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3827 2018-04-19 Richard Biener <rguenther@suse.de>
3829 PR tree-optimization/84737
3830 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3831 copying restrict info.
3832 (vect_setup_realignment): Use it.
3833 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3834 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3835 the first DR to all generated stores.
3836 (vectorizable_load): Likewise for loads.
3838 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3840 PR tree-optimization/85446
3841 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3842 the integral and pointer types to have the same precision.
3844 * doc/install.texi: Document --disable-cet being the default and
3847 2018-04-18 Martin Liska <mliska@suse.cz>
3849 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3852 2018-04-18 Martin Liska <mliska@suse.cz>
3855 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3858 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3859 arguments if they are comparable.
3861 2018-04-18 Martin Liska <mliska@suse.cz>
3864 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3867 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3870 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3873 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3874 ENDBR after calling __morestack.
3876 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3879 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3880 by using gcc_base_ver to generate a gcc_driver_version, and use
3881 it when generating GCC_DRIVER_NAME.
3882 * configure: Regenerate.
3884 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3887 * config.gcc: Obsolete powerpc*-*-*spe*.
3889 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3892 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3893 (stabstr_D): Change type of unum from unsigned int to
3894 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3897 2018-04-17 Jim Wilson <jimw@sifive.com>
3900 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3901 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3902 Set arg_pointer_offset after using pretend_args_size.
3904 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3906 PR rtl-optimization/85431
3907 * dse.c (record_store): Ignore zero width stores.
3910 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3911 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3912 __builtin_stack_restore rather than after it.
3913 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3914 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3915 argument instead of virtual_dynamic_stack_rtx.
3917 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3919 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3921 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3922 Add note to error message to explain internal mapping of overloaded
3923 built-in function name to non-overloaded built-in function name.
3924 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3927 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3930 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3931 where the inputs overlap with the output.
3933 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3935 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3936 (=v, v) alternative and explicit "memory" attribute.
3937 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3938 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3940 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3941 "sselog1" type instead of "sselog".
3942 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3943 "sselog". Remove explicit "memory" attribute.
3944 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3945 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3947 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3948 "sselog1" type instead of "sselog". Remove explicit "memory"
3950 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3951 use "sselog1" type instead of "sselog". Remove explicit "memory"
3953 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3954 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3956 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3957 "sselog1" type instead of "sselog". Remove explicit "memory"
3959 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3960 use "sselog1" type instead of "sselog". Remove explicit "memory"
3964 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3967 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3970 2018-04-17 Martin Jambor <mjambor@suse.cz>
3973 * ipa-cp.c (create_specialized_node): Call
3974 expand_all_artificial_thunks if necessary.
3976 2018-04-17 Martin Liska <mliska@suse.cz>
3979 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3980 in message, remote space in between '_G' and '('.
3982 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3985 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3986 avx512f_vmcmp<mode>3<round_saeonly_name>,
3987 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3988 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3989 avx512f_rndscale<mode><round_saeonly_name>,
3990 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3991 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3992 Use %<iptr>2 instead of %2 for -masm=intel.
3993 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3994 avx512f_vcvttss2usi<round_saeonly_name>,
3995 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3997 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3998 avx512f_vcvttsd2usi<round_saeonly_name>,
3999 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4000 Use %q1 instead of %1 for -masm=intel.
4001 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4002 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4003 of %3 for -masm=intel.
4004 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4006 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4008 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4010 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4011 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4013 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4015 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4016 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4017 %g1 and one with %0 and %1.
4018 (avx512er_vmrcp28<mode><round_saeonly_name>,
4019 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4021 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4022 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4023 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4024 of %0 and %{%4%} for -masm=intel.
4025 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4026 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4027 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4028 order of %0 and %{%5%}%{z%} for -masm=intel.
4030 2018-04-17 Jan Hubicka <jh@suse.cz>
4033 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4035 2018-04-17 Martin Liska <mliska@suse.cz>
4038 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4039 for target_clone error message. Make default implementation
4040 clone to be a local declaration.
4041 (separate_attrs): Add new argument and check for an empty
4043 (expand_target_clones): Handle it.
4044 (ipa_target_clone): Make redirection just for target_clones
4047 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4048 Tom de Vries <tom@codesourcery.com>
4051 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4052 tiled basic blocks without omp continue statements.
4054 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4057 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4058 vec_extract expression as having side effects to make sure it gets
4061 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4064 * config/i386/i386.c (get_builtin_code_for_version): Check
4067 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4070 * gcc/config.gcc: Support "skylake".
4071 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4073 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4074 (processor_target_table): Add "skylake".
4075 (ix86_option_override_internal): Add "skylake".
4076 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4077 PROCESSOR_CANNONLAKE.
4078 (get_builtin_code_for_version): Fix priority for
4079 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4080 PROCESSOR_SKYLAKE-AVX512.
4081 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4082 (processor_type): Add PROCESSOR_SKYLAKE.
4084 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4085 Jason Merrill <jason@redhat.com>
4088 * convert.c (convert_to_integer_1): Use direct recursion for
4089 enumeral types and types with a precision less than the number
4090 of bits in their mode.
4092 2018-04-16 Julia Koval <julia.koval@intel.com>
4095 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4096 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4098 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4101 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4102 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4103 and -mno-direct-move.
4105 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4108 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4109 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4110 Ensure negative shifts result in {0}.
4112 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4114 PR rtl-optimization/79916
4115 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4116 regs (if any) to define how to gnerate SD moves when LRA is in
4119 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4121 PR rtl-optimization/85393
4122 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4123 * except.c (expand_dw2_landing_pad_for_region): Make static.
4124 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4125 a label and unconditional jump to old_bb, rather than
4126 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4129 PR rtl-optimization/85376
4130 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4131 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4132 instead of a specific value.
4134 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4135 Bin Cheng <bin.cheng@arm.com>
4137 PR tree-optimization/82965
4138 PR tree-optimization/83991
4139 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4140 by_profile_only parameter.
4141 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4142 information if the loop was predicted to iterate too many times.
4143 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4145 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4148 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4151 2018-04-13 Martin Liska <mliska@suse.cz>
4152 Jakub Jelinek <jakub@redhat.com>
4155 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4156 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4157 * builtins.c (expand_builtin_memory_copy_args): Use
4158 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4159 handle dest_addr == pc_rtx.
4161 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4164 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4165 asked to not generate direct moves.
4166 (fix_trunc<mode>si2_stfiwx): Similar.
4167 (fix_trunc<mode>si2_internal): Similar.
4169 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4172 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4173 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4174 lookup if dest in some wider mode is known to be const0_rtx and
4175 if so, record permanent equivalence for it to be ZERO_EXTEND of
4176 the narrower mode destination.
4178 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4180 * lto-streamer-out.c (output_function): Revert 259346.
4181 * omp-expand.c (expand_oacc_for): Likewise.
4183 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4185 PR rtl-optimization/85354
4186 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4187 * sel-sched.c (sel_global_init): ... here.
4189 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4192 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4193 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4194 mode for PE-COFF targets.
4195 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4196 (i386_pe_asm_lto_end): Likewise.
4197 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4198 (TARGET_ASM_LTO_END): Likewise.
4199 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4200 (i386_pe_asm_lto_start): New function.
4201 (i386_pe_asm_lto_end): Likewise.
4203 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4204 Richard Biener <rguenther@suse.de>
4207 * lto-streamer-out.c (output_function): Fix CFG loop state before
4209 * omp-expand.c (expand_oacc_for): Handle calls to internal
4210 functions like regular functions.
4212 2018-04-12 Richard Biener <rguenther@suse.de>
4215 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4216 for the early LTO debug to properly generate references to it
4217 during DIE emission. Do not re-use that for the skeleton for
4219 (dwarf2out_early_finish): Likewise.
4221 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4224 * config/i386/sse.md
4225 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4226 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4227 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4228 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4229 and output is a reg, avoid creating invalid lowpart subreg, but
4230 instead split into a 512-bit move. Don't split if not AVX512VL,
4231 input is xmm16+ reg and output is a mem.
4232 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4233 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4234 xmm16+ reg and output is a mem.
4236 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4238 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4239 also for flag_dwarf2_cfi_asm.
4241 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4243 PR rtl-optimization/85342
4244 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4245 a bool scalar var inside of the loop instead. Don't try to update
4246 recog_data.operand after failed apply_change_group.
4248 2018-04-12 Tom de Vries <tom@codesourcery.com>
4251 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4252 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4253 array with flexible array member as array without given dimension.
4254 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4255 argument for undefined param to true.
4257 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4260 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4261 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4262 from PowerPC section.
4263 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4264 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4265 help text that is too long.
4266 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4267 help text that is too long.
4268 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4269 help text that is too long.
4271 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4273 * config/alpha/alpha.md (stack_probe_internal): Rename
4274 from "probe_stack". Update all callers.
4276 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4278 PR rtl-optimization/84566
4279 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4280 sched_macro_fuse_insns.
4282 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4285 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4286 (compute_block_dependences): ... from here.
4288 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4290 PR tree-optimization/85331
4291 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4292 from int to HOST_WIDE_INT.
4294 2018-04-11 Martin Jambor <mjambor@suse.cz>
4297 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4298 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4299 not the same as the source val.
4300 (cgraph_edge_brings_value_p): New parameter.
4301 (gather_edges_for_value): Pass destination value to
4302 cgraph_edge_brings_value_p.
4303 (perhaps_add_new_callers): Likewise.
4304 (get_info_about_necessary_edges): Likewise and exclude values brought
4305 only by self-recursive edges.
4306 (create_specialized_node): Redirect only clones of self-calling edges.
4307 (+self_recursive_pass_through_p): New function.
4308 (find_more_scalar_values_for_callers_subset): Use it.
4309 (find_aggregate_values_for_callers_subset): Likewise.
4310 (known_aggs_to_agg_replacement_list): Removed.
4311 (decide_whether_version_node): Re-calculate known constants for all
4312 remaining context clones.
4314 2018-04-11 Richard Biener <rguenther@suse.de>
4317 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4318 from early DWARF output.
4319 (dwarf2out_early_finish): Output line info unconditionally into
4320 early DWARF and add reference to it.
4322 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4325 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4326 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4327 other than V2DFmode using iptr mode attribute.
4328 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4330 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4332 PR rtl-optimization/84659
4333 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4335 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4338 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4340 (output_loc_list): Pass address of a dummy size variable even in the
4341 locview handling loop.
4342 (index_location_lists): Add comment on why skip_loc_list_entry can't
4345 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4348 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4351 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4354 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4355 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4356 and -mstring-compare-inline-limit.
4358 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4361 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4362 for stack clash protection in a register whenever we need it to be in
4365 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4367 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4368 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4370 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4373 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4375 (mlong-double-): Ditto.
4376 * config/rs6000/sysv4.opt (msdata=): Ditto.
4377 (mtls-size=): Ditto.
4379 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4381 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4382 erroneous entries for
4383 "vector int vec_ldl (int, long int *)", and
4384 "vector unsigned int vec_ldl (int, unsigned long int *)".
4385 Add comments and entries for
4386 "vector bool char vec_ldl (int, bool char *)",
4387 "vector bool short vec_ldl (int, bool short *)",
4388 "vector bool int vec_ldl (int, bool int *)",
4389 "vector bool long long vec_ldl (int, bool long long *)",
4390 "vector pixel vec_ldl (int, pixel *)",
4391 "vector long long vec_ldl (int, long long *)",
4392 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4393 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4394 type tree bool_long_long_type_node and correct definition of
4395 bool_V2DI_type_node to make reference to this new type tree.
4396 (rs6000_mangle_type): Replace erroneous reference to
4397 bool_long_type_node with bool_long_long_type_node.
4398 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4399 comments to emphasize sign distinctions for char and int types and
4400 replace RS6000_BTI_bool_long constant with
4401 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4402 use of RS6000_BTI_pixel.
4403 (bool_long_type_node): Remove this macro definition.
4404 (bool_long_long_type_node): New macro definition
4406 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4408 PR rtl-optimization/85300
4409 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4410 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4411 simplify_unary_operation fails.
4413 2018-04-10 Martin Liska <mliska@suse.cz>
4415 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4416 cgraph_edge and ipa_ref.
4418 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4422 * config/i386/sse.md
4423 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4424 computation of the VEC_MERGE selector from mask.
4425 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4426 Fix decoding of the VEC_MERGE selector into mask.
4428 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4430 PR tree-optimization/85286
4431 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4433 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4435 * final.c (final_1): Set insn_last_address as well as
4436 insn_current_address.
4438 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4441 * explow.c (emit_stack_probe): Call validize_mem on memory location
4442 before passing it to gen_probe_stack. Create address operand and
4443 legitimize it for the probe_stack_address case.
4445 2018-04-09 Jan Hubicka <jh@suse.cz>
4448 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4449 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4450 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4452 (free_lang_data): Rebuild type inheritance graph.
4454 2018-04-09 Martin Sebor <msebor@redhat.com>
4456 * invoke.texi (-finline-small-functions): Mention other optimization
4458 (-findirect-inlining, -fpartial-inlining): Same.
4459 (-finline-functions-called-once): Same.
4460 (-freorder-blocks-and-partition): Same.
4462 2018-04-09 Jan Hubicka <jh@suse.cz>
4465 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4466 jumps; choose last target that matches the criteria (i.e.
4467 no partition changes for non-crossing jumps).
4468 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4469 support for redirecting crossing jumps to non-crossing.
4471 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4473 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4474 also for naked functions.
4476 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4478 * config/arc/arc.md (add_shift): New pattern.
4479 (add_shift2): Likewise.
4480 (sub_shift): Likewise.
4481 (sub_shift_cmp0_noout): Likewise.
4482 (compare_si_ashiftsi): Likewise.
4483 (xbfu_cmp0_noout): New combine pattern.
4484 (xbfu_cmp0"): Likewise.
4485 (movsi_set_cc_insn): Place the predicable variant first.
4486 (commutative_binary_cmp0_noout): Remove clobber.
4487 (commutative_binary_cmp0): New pattern.
4488 (noncommutative_binary_cmp0): Likewise.
4489 (noncommutative_binary_cmp0_noout): Likewise.
4490 (noncommutative_binary_comparison_result_used): Removed.
4491 (rsub_cmp0): New pattern.
4492 (rsub_cmp0_noout): Likewise.
4493 (extzvsi): Changed, keep only meaningful variants.
4494 (SQH, SEZ): New iterators.
4495 (SQH_postfix): New mode attribute.
4496 (SEZ_prefix): New code attribute.
4497 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4498 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4499 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4501 (noncommutative_operator): Check the availability of barrel
4504 2018-04-09 Richard Biener <rguenther@suse.de>
4506 PR tree-optimization/85284
4507 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4508 Only use the niter constraining form of simple_iv when the exit
4511 2018-04-09 Tom de Vries <tom@codesourcery.com>
4514 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4515 (define_expand "*memory_barrier"): New define_expand.
4516 (define_insn "memory_barrier"): New insn.
4518 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4520 PR rtl-optimization/80463
4521 PR rtl-optimization/83972
4522 PR rtl-optimization/83480
4524 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4525 correct producer for the insn.
4526 (tidy_control_flow): Fixup seqnos in case of debug insns.
4528 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4530 PR rtl-optimization/83913
4532 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4533 different sched-times when merging exprs.
4535 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4537 PR rtl-optimization/83962
4539 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4540 tidy_fallthru_edge and tidy_control_flow.
4542 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4544 PR rtl-optimization/83530
4546 * sel-sched.c (force_next_insn): New global variable.
4547 (remove_insn_for_debug): When force_next_insn is true, also leave only
4548 next insn in the ready list.
4549 (sel_sched_region): When the region wasn't scheduled, make another pass
4550 over it with force_next_insn set to 1.
4552 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4554 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4556 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4557 for interrupt control.
4558 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4559 functions for interrupt control.
4560 * config/nds32/nds32-intrinsic.md: Likewise.
4561 * config/nds32/nds32_intrinsic.h: Likewise.
4562 * config/nds32/nds32.h (nds32_builtins): Likewise.
4564 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4566 * config/nds32/nds32.c (nds32_init_machine_status,
4567 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4568 strict_aligned_p field.
4569 (nds32_expand_to_rtl_hook): New function.
4570 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4571 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4573 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4574 Chung-Ju Wu <jasonwucj@gmail.com>
4576 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4577 * config/nds32/nds32-n7.md: New file.
4578 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4579 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4581 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4582 * config/nds32/nds32.md (pipeline_model): Add n7.
4583 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4584 * config/nds32/pipelines.md: Include n7 settings.
4586 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4587 Chung-Ju Wu <jasonwucj@gmail.com>
4589 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4590 * config/nds32/nds32-e8.md: New file.
4591 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4594 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4595 * config/nds32/nds32.md (pipeline_model): Add e8.
4596 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4597 * config/nds32/pipelines.md: Include e8 settings.
4599 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4600 Chung-Ju Wu <jasonwucj@gmail.com>
4602 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4603 * config/nds32/nds32-n8.md: New file.
4604 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4605 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4607 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4608 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4609 * config/nds32/nds32.md (pipeline_model): Add n8.
4610 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4611 * config/nds32/pipelines.md: Include n8 settings.
4613 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4614 Chung-Ju Wu <jasonwucj@gmail.com>
4616 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4617 * config/nds32/nds32-n9-2r1w.md: New file.
4618 * config/nds32/nds32-n9-3r2w.md: New file.
4619 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4620 nds32_register_ports): New or modify for cpu n9.
4621 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4623 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4624 * config/nds32/nds32-utils.c: New file.
4625 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4626 TARGET_MUL_SLOW): Define.
4627 * config/nds32/nds32.md (pipeline_model): New attribute.
4628 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4629 New options that support cpu n9.
4630 * config/nds32/pipelines.md: Include n9 settings.
4631 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4633 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4636 information if necessary.
4637 (output_cond_branch_compare_zero): Likewise.
4638 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4639 (nds32_target_alignment): Refine for alignment.
4640 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4641 (FUNCTION_BOUNDARY): Modify.
4642 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4644 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4646 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4648 * config/nds32/constants.md (unspec_volatile_element): Add values for
4649 TLB operation and data prefetch.
4650 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4651 functions for TLB operation and data prefetch.
4652 * config/nds32/nds32-intrinsic.md: Likewise.
4653 * config/nds32/nds32_intrinsic.h: Likewise.
4654 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4655 (nds32_print_operand): Likewise.
4656 * config/nds32/nds32.h (nds32_builtins): Likewise.
4658 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4659 Andrew Pinski <pinsika@gcc.gnu.org>
4662 * match.pd: Use constant_boolean_node of correct type instead of
4663 boolean_true_node or boolean_false_node for simplifying
4664 pointer comparisons to zero.
4666 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4668 PR tree-optimization/80021
4669 * tree.c (verify_type_variant): Make error call in verify_variant_match
4670 translatable and remove final full stop.
4672 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4674 * config/nds32/constants.md (unspec_volatile_element): Add
4675 UNSPEC_VOLATILE_EH_RETURN.
4676 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4677 nds32_output_stack_pop): Support dwarf exception handling process.
4678 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4679 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4680 exception handling process.
4681 (nds32_compute_stack_frame): Likewise.
4682 (nds32_return_addr_rtx): Likewise.
4683 (nds32_initial_elimination_offset): Likewise.
4684 (nds32_expand_prologue): Likewise.
4685 (nds32_expand_epilogue): Likewise.
4686 (nds32_dynamic_chain_address): New function.
4687 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4689 (DYNAMIC_CHAIN_ADDRESS): Define.
4690 (EH_RETURN_DATA_REGNO): Define.
4691 (EH_RETURN_STACKADJ_RTX): Define.
4692 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4693 patterns for dwarf exception handling.
4695 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4697 * config/nds32/nds32.h: Clean up obsolete macros.
4699 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4701 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4702 Add enum values for particular instructions.
4703 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4704 particular intrinsic functions.
4705 * config/nds32/nds32-intrinsic.md: Likewise.
4706 * config/nds32/nds32_intrinsic.h: Likewise.
4707 * config/nds32/nds32.h (nds32_builtins): Likewise.
4708 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4709 (btst, ave): New patterns for particular instructions.
4711 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4713 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4714 Add enum values for atomic load/store and memory sync.
4715 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4717 * config/nds32/nds32-intrinsic.md: Likewise.
4718 * config/nds32/nds32_intrinsic.h: Likewise.
4719 * config/nds32/nds32.h (nds32_builtins): Likewise.
4721 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4723 PR tree-optimization/85257
4724 * fold-const.c (native_encode_vector): If not all elts could fit
4725 and off is -1, return 0 rather than offset.
4726 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4727 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4728 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4729 adjust buffer in native_interpret_expr call.
4731 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4733 * config/nds32/constants.md (unspec_volatile_element): Add cache
4734 control enum values.
4735 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4736 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4737 * config/nds32/nds32.c (nds32_cctl_names): New.
4738 (nds32_print_operand): Handle cache control register names.
4739 * config/nds32/nds32.h (nds32_builtins): New enum values.
4740 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4742 * config/nds32/nds32.md (type): Add mmu.
4743 * config/nds32/pipelines.md (simple_insn): Add mmu.
4745 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4747 * config/nds32/nds32.md (type): Remove call.
4748 * config/nds32/pipelines.md (simple_insn): Likewise.
4750 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4752 * config/nds32/constants.md (unspec_volatile_element): Add
4753 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4754 UNSPEC_VOLATILE_FMFCFG.
4755 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4756 description for fmfcfg and fmfcsr.
4757 (bdesc_1arg): Add fmtcsr.
4758 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4759 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4760 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4761 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4762 unspec_fmfcfg): New patterns.
4763 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4764 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4765 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4766 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4767 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4768 __nds32__fmfcfg): Define.
4770 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4772 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4773 intrinsic register names.
4774 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4775 intrinsic register enum values and macros.
4777 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4779 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4780 for load/store addressing form.
4781 (nds32_print_operand_address): Likewise.
4783 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4786 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4787 based on LABEL_REF. Remove useless assertion.
4788 (pic_address_needs_scratch): Fix formatting.
4789 (sparc_legitimize_pic_address): Minor tweaks.
4790 (sparc_delegitimize_address): Adjust assertion accordingly.
4791 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4792 into symbolic_operand.
4793 (movsi_high_pic_label_ref): Likewise.
4794 (movsi_lo_sum_pic_label_ref): Likewise.
4795 (movdi_pic_label_ref): Likewise.
4796 (movdi_high_pic_label_ref): Likewise.
4797 (movdi_lo_sum_pic_label_ref): Likewise.
4799 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4801 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4802 custom LIB_SPEC setup.
4804 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4805 Kito Cheng <kito.cheng@gmail.com>
4807 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4808 * config/riscv/freebsd.h: New.
4810 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4812 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4813 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4816 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4817 Kito Cheng <kito.cheng@gmail.com>
4819 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4820 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4821 * config/nds32/nds32-protos.h (nds32_output_call,
4822 nds32_output_return): Declare.
4823 * config/nds32/nds32.md: Refine all the call and return patterns.
4825 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4828 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4829 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4831 PR rtl-optimization/84872
4832 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4833 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4836 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4838 * expr.c (copy_blkmode_to_reg): Revert 254862.
4839 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4841 2018-04-06 Richard Biener <rguenther@suse.de>
4844 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4845 after seeing a component reference with an adjacent field. Treat
4846 refs to arrays at struct end of external decls similar to
4847 refs to unconstrained commons.
4849 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4852 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4853 look through SAVE_EXPRs with non-side-effects argument. Adjust
4855 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4858 2018-04-06 Richard Biener <rguenther@suse.de>
4861 * alias.c (find_base_term): New wrapper around find_base_term
4862 unwinding CSELIB_VAL_PTR changes.
4863 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4866 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4868 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4870 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4871 constant definitions.
4872 ("nop"): lr 0,0 -> nopr r0
4873 ("nop_lr0", "nop_lr1"): New insn definitions.
4875 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4877 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4878 NDS32_V3PUSH_AVAILABLE_P macro.
4880 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4881 Chung-Ju Wu <jasonwucj@gmail.com>
4883 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4884 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4885 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4886 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4887 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4888 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4889 * config/nds32/constraints.md: New constraints and checking for hard
4890 float configuration.
4891 * config/nds32/iterators.md: New mode iterator and attribute for hard
4892 float configuration.
4893 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4895 * config/nds32/nds32-fpu.md: New file.
4896 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4897 deal with hard float code generation.
4898 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4900 (abi_type, float_reg_number): New enum type.
4901 * config/nds32/nds32-predicates.c: New predicates for hard float.
4902 * config/nds32/nds32-protos.h: Declare functions for hard float.
4903 * config/nds32/nds32.c: Implementation for hard float configuration.
4904 * config/nds32/nds32.h: Definitions for hard float configuration.
4905 * config/nds32/nds32.md: Include hard float machine description and
4906 modify patterns for hard float configuration.
4907 * config/nds32/nds32.opt: New options for hard float configuration.
4908 * config/nds32/predicates.md: New predicates for hard float
4911 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4913 * common/config/nds32/nds32-common.c
4914 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4916 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4919 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4920 CONSTRUCTOR_ELT (ctor, ...)->value.
4922 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4925 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4927 2018-04-05 Tom de Vries <tom@codesourcery.com>
4930 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4933 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4934 Kito Cheng <kito.cheng@gmail.com>
4936 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4937 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4938 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4939 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4941 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4942 Kito Cheng <kito.cheng@gmail.com>
4944 * config/nds32/constraints.md (Ufe): New memory constraint.
4945 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4946 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4947 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4949 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4950 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4952 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4954 * config/nds32/nds32.md: Use optimize_size in the condition for
4955 alu-shift instructions.
4957 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4959 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4961 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4963 * config/nds32/nds32.md (negsi2): Refine pattern.
4965 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4966 Chung-Ju Wu <jasonwucj@gmail.com>
4968 * config/nds32/iterators.md (shift_rotate): New code iterator.
4969 (shift): New code attribute.
4970 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4971 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4972 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4973 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4974 bit-wise operations.
4975 (andsi3, *andsi3): Ditto.
4976 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4977 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4978 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4979 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4980 nds32_ior_operand, nds32_xor_operand): New predicates.
4982 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4984 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4985 (addsi3, subsi3): ... this.
4987 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4991 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4993 * config/nds32/nds32.md: Adjust indention.
4995 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4997 * config/nds32/nds32.md (feature): New attribute.
4999 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5001 * config/nds32/nds32.md (subtype): New attribute.
5003 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5006 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5007 expansion to perform a bitwise AND of the argument followed by a
5008 boolean negation of the result.
5010 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5012 PR rtl-optimization/84878
5013 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5014 the basic block. Assert the use reference is not artificial and that
5015 it has an associated insn.
5017 2018-04-04 Michael Matz <matz@suse.de>
5019 * builtins.c (compute_objsize): Pass correct operand
5020 to array_at_struct_end_p.
5022 2018-04-04 Richard Biener <rguenther@suse.de>
5025 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5026 from contexts for DINFO_LEVEL_TERSE and below.
5028 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5030 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5031 resiter_operand condition.
5032 * config/nds32/nds32.md (*move<mode>): Ditto.
5034 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5035 Monk Chiang <sh.chiang04@gmail.com>
5037 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5039 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5041 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5043 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5044 Kito Cheng <kito.cheng@gmail.com>
5046 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5047 nds32_cond_code_str, output_cond_branch,
5048 output_cond_branch_compare_zero, nds32_expand_cbranch,
5049 nds32_expand_cstore, nds32_expand_movcc,
5050 nds32_output_cbranchsi4_equality_zero,
5051 nds32_output_cbranchsi4_equality_reg,
5052 nds32_output_cbranchsi4_equality_reg_or_const_int,
5053 nds32_output_cbranchsi4_greater_less_zero: New functions.
5054 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5055 nds32_expand_cstore, nds32_expand_movcc,
5056 nds32_output_cbranchsi4_equality_zero,
5057 nds32_output_cbranchsi4_equality_reg,
5058 nds32_output_cbranchsi4_equality_reg_or_const_int,
5059 nds32_output_cbranchsi4_greater_less_zero): Declare.
5060 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5061 nds32_rimm11s_operand): New predicates.
5062 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5063 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5066 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5068 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5069 * config/nds32/nds32.md: Ditto.
5070 * config/nds32/pipelines.md: Ditto.
5072 2018-04-04 Richard Biener <rguenther@suse.de>
5074 PR tree-optimization/85168
5075 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5076 propagating abnormals.
5078 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5080 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5082 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5083 Kito Cheng <kito.cheng@gmail.com>
5085 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5086 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5087 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5088 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5089 * config/nds32/nds32.md (sibcall_internal): New.
5090 (sibcall_register): Remove.
5091 (sibcall_immediate): Remove.
5092 (sibcall_value_internal): New.
5093 (sibcall_value_register): Remove.
5094 (sibcall_value_immediate): Remove.
5095 * config/nds32/predicates.md (nds32_general_register_operand): New.
5096 (nds32_call_address_operand): New.
5098 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5100 PR rtl-optimization/85167
5101 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5102 bb_defs if *split_p, instead preinitialize it to NULL.
5104 PR tree-optimization/85156
5105 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5106 evaluating the argument multiple times.
5108 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5110 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5112 (_mm_cvtpd_ps): Likewise.
5113 (_mm_cvttpd_epi32): Likewise.
5114 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5115 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5116 vector, pixel, and bool following altivec.h include.
5118 2018-04-03 Martin Sebor <msebor@redhat.com>
5120 * doc/extend.texi (Common Function Attributes): Clarify.
5121 (const attribute): Likewise.
5122 (pure attribute): Likewise.
5124 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5127 * config/i386/i386.c (ix86_expand_vector_set): Use
5128 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5130 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5132 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5133 instructions when changing rounding bits to preserve precision bits
5134 in the x87 control word.
5136 2018-04-03 Martin Liska <mliska@suse.cz>
5138 PR tree-optimization/82491
5139 * rtl.h (strip_offset_and_add): Replace += suboffset with
5140 poly_uint64 () + suboffset.
5142 2018-03-29 Martin Liska <mliska@suse.cz>
5143 Martin Jambor <mjambor@suse.cz>
5146 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5147 param_type is not an integral or pointer type.
5149 2018-04-03 Richard Biener <rguenther@suse.de>
5151 * sese.h (recompute_all_dominators): Remove.
5153 2018-04-02 Martin Sebor <msebor@redhat.com>
5155 * doc/invoke.texi (-Wrestrict): Fix typos.
5157 2018-04-02 Jim Wilson <jimw@sifive.com>
5159 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5160 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5161 (<optab>di3, <optab>si3_extend): Likewise.
5162 (<optab>si3_mask, <optab>si3_mask_1): New.
5163 (<optab>di3_mask, <optab>di3_mask_1): New.
5164 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5165 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5166 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5168 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5170 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5173 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5175 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5176 (nds32_canonicalize_comparison): New function.
5178 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5179 Kito Cheng <kito.cheng@gmail.com>
5180 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5182 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5183 * config/nds32/constants.md (unspec_volatile_element): Add
5184 UNSPEC_VOLATILE_RELAX_GROUP.
5185 * config/nds32/nds32-relax-opt.c: New file.
5186 * config/nds32/nds32-predicates.c
5187 (nds32_symbol_load_store_p): New function.
5188 * config/nds32/nds32-protos.h
5189 (nds32_symbol_load_store_p): Declare function.
5190 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5191 * config/nds32/nds32.c
5192 (nds32_register_pass): New function to register pass.
5193 (nds32_register_passes): New function to register passes.
5194 * config/nds32/nds32.md (relax_group): New pattern.
5195 * config/nds32/nds32.opt (mrelax-hint): New option.
5196 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5198 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5200 * config/nds32/t-nds32: Modify files dependency.
5202 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5204 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5205 (PROFILE_HOOK): Define its implementation.
5207 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5209 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5210 type and 32-bit size.
5212 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5215 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5216 (V_128_256): New mode iterator.
5217 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5218 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5219 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5221 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5222 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5224 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5227 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5228 NaN inputs correctly.
5230 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5233 * config/rs6000/vsx.md (??r): New mode attribute.
5234 (*vsx_mov<mode>_64bit): Use it.
5235 (*vsx_mov<mode>_32bit): Likewise.
5237 2018-03-30 Martin Sebor <msebor@redhat.com>
5239 PR tree-optimization/84818
5240 * builtins.c (check_access): Use warning_n.
5242 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5245 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5247 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5250 2018-03-30 Julia Koval <julia.koval@intel.com>
5253 * x86-tune.def (movx, partial_reg_dependency): Enable for
5256 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5259 * lra-constraints.c (process_alt_operands): Move setting
5260 this_alternative_matches below.
5262 2018-03-29 Martin Liska <mliska@suse.cz>
5265 * doc/invoke.texi: Document how LTO works with debug info.
5266 Describe auto-load support of binutils. Mention 'x86-64'
5267 as valid option value of -march option.
5269 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5271 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5274 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5275 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5276 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5279 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5282 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5283 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5284 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5285 for RS6000_BTM_POWERPC64.
5286 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5287 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5288 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5293 2018-03-28 Carl Love <cel@us.ibm.com>
5296 2017-09-27 Carl Love <cel@us.ibm.com>
5298 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5299 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5300 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5303 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5305 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5306 instead of __vector bool.
5307 (_mm_max_pu8): Likewise.
5308 (_mm_min_pi16): Likewise.
5310 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5313 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5314 (DIVWEUO): Likewise.
5316 (DIVDEUO): Likewise.
5317 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5318 DIVWEUO and DIVDEUO.
5319 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5320 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5321 (div_extend): Likewise.
5322 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5324 (__builtin_divweuo): Likewise.
5325 (__builtin_divdeo): Likewise.
5326 (__builtin_divdeuo): Likewise.
5328 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5331 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5332 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5334 PR tree-optimization/82004
5335 * gimple-match-head.c (optimize_pow_to_exp): New function.
5336 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5337 Don't fold to exp if optimize_pow_to_exp is false.
5339 2018-03-28 Martin Liska <mliska@suse.cz>
5342 * calls.c (initialize_argument_information): Fix trailing space.
5343 * common.opt: Fix typo and provide better explanation for
5344 -fsanitize-coverage option.
5345 * config/i386/i386.opt: Fix typo.
5347 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5348 Martin Liska <mliska@suse.cz>
5351 * gimplify.c (asan_poison_variable): Don't do the check for
5352 gimplify_omp_ctxp here.
5353 (gimplify_decl_expr): Do it here.
5354 (gimplify_target_expr): Likewise.
5356 2018-03-28 Martin Liska <mliska@suse.cz>
5359 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5360 chkp_type_bounds_count if MPX is not enabled.
5362 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5364 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5366 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5369 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5370 function to create the function decl for complex long double
5371 multiply and divide for -mabi=ieeelongdouble.
5372 (init_float128_ieee): Call it.
5374 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5377 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5378 -fcf-protection=branch -mibt.
5379 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5381 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5384 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5386 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5389 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5390 extern array declarations.
5392 2018-03-27 Richard Biener <rguenther@suse.de>
5395 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5396 explicit single_use checks.
5398 2018-03-27 Richard Biener <rguenther@suse.de>
5400 PR tree-optimization/85082
5401 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5404 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5406 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5407 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5408 Turn on fasynchronous-unwind-tables and funwind-tables.
5410 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5413 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5414 (*bmi_blsr_<mode>_ccz): Ditto.
5416 2018-03-26 Tom de Vries <tom@codesourcery.com>
5418 PR tree-optimization/85063
5419 * omp-general.c (offloading_function_p): New function. Factor out
5421 * omp-offload.c (pass_omp_target_link::gate): ... here.
5422 * omp-general.h (offloading_function_p): Declare.
5423 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5424 with attribute omp declare target for offloading functions.
5426 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5428 PR tree-optimization/84005
5429 * tree-data-ref.h (get_base_for_alignment): Declare.
5430 * tree-data-ref.c (get_base_for_alignment_1): New function.
5431 (get_base_for_alignment): Likewise.
5432 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5433 get_base_for_alignment to find a suitable base object, instead
5434 of always using drb->base_address.
5436 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5439 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5440 known size by default.
5442 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5445 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5446 and non BLKmode operands.
5448 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5451 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5452 Clean up attributes.
5454 2018-03-23 Richard Biener <rguenther@suse.de>
5457 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5458 we are going to emit early debug for LTO.
5460 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5463 * function.c (match_asm_constraints_1): Don't optimize if input
5464 doesn't satisfy general_operand predicate for output's mode.
5467 * alias.c (write_dependence_p): Don't require for x_canonicalized
5468 non-VOIDmode if x has VOIDmode.
5471 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5472 just don't try to optimize it rather than assert it never happens.
5474 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5476 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5477 macro expansions for definition of ST_INTERNAL_<mode> and
5478 LD_INTERNAL_<mode> builtins.
5479 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5481 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5483 (altivec_expand_st_builtin): Likewise.
5484 (altivec_expand_builtin): Remove calls to deleted functions.
5485 (rs6000_address_for_altivec): Delete this function.
5486 * config/rs6000/vector.md: Remove expands for
5487 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5489 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5492 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5493 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5494 re-computing once computed.
5495 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5496 (arm_init_machine_status): Initialize
5497 machine->static_chain_stack_bytes.
5499 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5502 * doc/extend.texi: Add four new prototypes for vec_ld.
5503 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5504 definitions for more logical presentation.
5505 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5506 entries for V1TI variants of __builtin_altivec_ld builtin.
5507 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5508 handling of V1TI variant of LVX icode pattern.
5509 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5510 (rs6000_gimple_fold_builtin): Likewise.
5511 (altivec_init_builtins): Add code to define
5512 __builtin_altivec_lvx_v1ti function.
5514 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5517 * function.c (match_asm_constraints_1): Don't do the optimization
5518 if input isn't a REG, SUBREG, MEM or constant.
5520 2018-03-22 Tom de Vries <tom@codesourcery.com>
5522 PR tree-optimization/84956
5523 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5524 bb_has_abnormal_pred.
5526 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5529 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5530 DECL_INITIAL (decl) to decl at the end.
5531 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5534 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5536 * doc/extend.texi (__builtin_tgmath): Document when complex
5537 integer types are treated as _Complex _Float64.
5539 2018-03-21 Tom de Vries <tom@codesourcery.com>
5541 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5543 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5545 PR tree-optimization/84960
5546 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5547 if it is ENTRY block, move them into single succ of ENTRY in that case.
5549 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5551 PR tree-optimization/84811
5552 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5553 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5554 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5555 (known_subrange_p): Update accordingly. Cast each value involved
5556 in the size comparison, rather than casting the result of the
5559 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5561 PR tree-optimization/84982
5562 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5563 by flipping the least significant bit rather than all bits from
5564 bitpos to bitpos + bitsize - 1.
5566 2018-03-21 Nathan Sidwell <nathan@acm.org>
5568 * doc/extend.texi (Deprecated Features): Remove mention of
5569 long-deleted deprecations.
5571 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5574 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5575 * configure: Regenerate.
5577 2018-03-21 Tom de Vries <tom@codesourcery.com>
5579 PR tree-optimization/83126
5580 * tree-parloops.c (num_phis): New function.
5581 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5583 2018-03-21 Nathan Sidwell <nathan@acm.org>
5585 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5586 mention anon-struct/union members and trailing attributes.
5588 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5590 PR tree-optimization/84969
5591 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5592 builtin memset partitions if they set different rhs values.
5594 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5596 PR rtl-optimization/84989
5597 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5598 VEC_DUPLICATE with scalar result mode.
5600 2018-03-21 Martin Liska <mliska@suse.cz>
5603 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5604 not intended return statement.
5606 2018-03-21 Martin Liska <mliska@suse.cz>
5609 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5610 (chkp_find_bound_slots_1): Limit number of iterations.
5612 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5615 * Minor grammar fixes for x86 options.
5617 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5620 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5621 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5623 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5626 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5627 (altivec_lvx_<mode>_1op): Likewise.
5628 (altivec_stvx_<mode>_2op): Likewise.
5629 (altivec_stvx_<mode>_1op): Likewise.
5630 (altivec_lvx_<VM2:mode>): New define_expand.
5631 (altivec_stvx_<VM2:mode>): Likewise.
5632 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5633 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5634 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5635 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5636 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5637 (rs6000_gen_lvx): Likewise.
5638 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5639 (altivec_expand_stv_builtin): Likewise.
5640 (altivec_expand_builtin): Likewise.
5641 * config/rs6000/vector.md: Likewise.
5643 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5646 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5649 2018-03-20 Richard Biener <rguenther@suse.de>
5652 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5653 sign-conversions as zero, fall back to standard scalar_stmt
5656 2018-03-20 Martin Liska <mliska@suse.cz>
5659 * predict.c (rebuild_frequencies): Handle case when we have
5660 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5662 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5665 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5666 flag_section_anchors.
5667 * varasm.c (use_blocks_for_decl_p): Remove hack for
5668 dw2_force_const_mem.
5671 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5673 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5674 be created, use lowpart_subreg of operands[0] rather than operands[0]
5676 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5677 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5678 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5679 and n constraint instead of aarch64_shift_imm_di and Usd.
5680 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5681 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5683 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5686 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5687 to favor GPR over NEON registers.
5688 (<shift>di3_neon): Likewise.
5690 2018-03-20 Tom de Vries <tom@codesourcery.com>
5693 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5694 (nvptx_process_pars): Emit bar.sync asap and alap.
5696 2018-03-20 Tom de Vries <tom@codesourcery.com>
5699 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5700 seen_label if seen_label is already set.
5702 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5705 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5706 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5707 Use 1U instead of 1. Formatting fixes.
5710 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5711 instead of TREE_TYPE (s1) for the return value.
5713 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5715 PR tree-optimization/84946
5716 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5717 bitsize + bitsize in poly_uint64 rather than poly_int64.
5720 * dwarf2asm.c: Include fold-const.c.
5721 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5722 of decl rather than decl itself.
5724 PR rtl-optimization/84643
5725 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5727 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5730 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5731 calling assemble_variable.
5733 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5736 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5737 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5739 2018-03-19 Jim Wilson <jimw@sifive.com>
5742 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5743 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5744 (riscv_first_stack_step): Likewise.
5745 (riscv_option_override): Use STACK_BOUNDARY instead of
5747 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5749 (BIGGEST_ALIGNMENT): Set to 128.
5750 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5751 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5754 2018-03-19 Richard Biener <rguenther@suse.de>
5756 PR tree-optimization/84933
5757 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5758 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5760 2018-03-19 Richard Biener <rguenther@suse.de>
5762 PR tree-optimization/84859
5763 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5764 (cond_if_else_store_replacement): Perform sinking operation on
5765 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5766 Generalize what a BB with a single eligible store is.
5768 2018-03-19 Richard Biener <rguenther@suse.de>
5770 PR tree-optimization/84929
5771 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5772 chrec_is_positive against non-chrec arg.
5774 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5777 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5779 2018-03-18 Martin Liska <mliska@suse.cz>
5781 PR rtl-optimization/84635
5782 * regrename.c (build_def_use): Use matches_mode only when
5785 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5787 PR tree-optimization/84913
5788 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5789 vectorize chains of COND_EXPRs.
5791 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5793 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5795 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5797 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5799 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5801 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5803 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5804 Kito Cheng <kito.cheng@gmail.com>
5806 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5807 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5808 (nds32_adjust_reg_alloc_order): New function.
5809 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5811 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5813 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5814 nds32_print_operand, nds32_print_operand_address): Use
5815 HOST_WIDE_INT_PRINT_DEC instead.
5817 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5819 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5821 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5824 * config/i386/i386.c (initial_ix86_tune_features,
5825 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5827 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5828 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5829 rather than 1u << ix86_tune. Formatting fix.
5830 (ix86_option_override_internal): Change ix86_arch_mask from
5831 unsigned int to unsigned HOST_WIDE_INT, initialize to
5832 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5833 (ix86_function_specific_restore): Likewise.
5835 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5838 * postreload.c (reload_combine_recognize_pattern): Perform
5839 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5840 truncate_int_for_mode the result for the destination's mode.
5843 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5844 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5846 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5849 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5852 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5853 regno_allocno_class_array and sorted_pseudos.
5854 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5855 insns where regno is used.
5857 2018-03-16 Martin Liska <mliska@suse.cz>
5860 * multiple_target.c (create_dispatcher_calls): Redirect
5861 reference in the symbol table.
5863 2018-03-16 Martin Liska <mliska@suse.cz>
5866 * multiple_target.c (create_dispatcher_calls): Redirect also
5869 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5873 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5874 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5875 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5877 2018-03-16 Julia Koval <julia.koval@intel.com>
5879 * doc/invoke.texi (Skylake Server): Add CLWB.
5880 Cannonlake): Remove CLWB.
5882 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5884 PR tree-optimization/84841
5885 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5887 (FLOAT_ONE_CONST_TYPE): Define.
5888 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5889 (sort_by_operand_rank): Put entries with higher constant_type last
5890 rather than first to match comments.
5892 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5894 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5897 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5900 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5901 check and use address of target always.
5903 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5906 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5907 (indirect_thunk_bnd_needed): Likewise.
5908 (indirect_thunks_used): Likewise.
5909 (indirect_thunks_bnd_used): Likewise.
5910 (indirect_return_needed): New.
5911 (indirect_return_bnd_needed): Likewise.
5912 (output_indirect_thunk_function): Add a bool argument for
5914 (output_indirect_thunk_function): Don't generate alias for
5915 function return thunk.
5916 (ix86_code_end): Call output_indirect_thunk_function to generate
5917 function return thunks.
5918 (ix86_output_function_return): Set indirect_return_bnd_needed
5919 and indirect_return_needed instead of indirect_thunk_bnd_needed
5920 and indirect_thunk_needed.
5922 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5924 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5925 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5926 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5928 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5929 Paul Hua <paul.hua.gm@gmail.com>
5932 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5934 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5936 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5937 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5940 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5943 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5944 instead of GET_MODE_SIZE when comparing Units.
5946 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5949 * varasm.c (hash_section): Return an unchangeble hash value
5950 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5951 Return !aarch64_can_use_per_function_literal_pools_p ().
5953 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5956 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5957 rather than address of cmode as last argument to prepare_cmp_insn.
5959 2018-03-15 Julia Koval <julia.koval@intel.com>
5961 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5962 F_AVX512VNNI, F_AVX512BITALG): New.
5964 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5967 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5968 insn for floating-point loads and stores.
5970 2018-03-14 Carl Love <cel@us.ibm.com>
5972 * config/rs6000/rs6000-c.c: Add macro definitions for
5973 ALTIVEC_BUILTIN_VEC_PERMXOR.
5974 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5975 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5976 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5977 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5979 * config/doc/extend.texi: Add prototypes for vec_permxor.
5981 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5984 * diagnostic-show-locus.c (class layout_point): Convert m_line
5985 from int to linenum_type.
5986 (line_span::comparator): Use linenum "compare" function when
5987 comparing line numbers.
5988 (test_line_span): New function.
5989 (layout_range::contains_point): Convert param "row" from int to
5991 (layout_range::intersects_line_p): Likewise.
5992 (layout::will_show_line_p): Likewise.
5993 (layout::print_source_line): Likewise.
5994 (layout::should_print_annotation_line_p): Likewise.
5995 (layout::print_annotation_line): Likewise.
5996 (layout::print_leading_fixits): Likewise.
5997 (layout::annotation_line_showed_range_p): Likewise.
5998 (struct line_corrections): Likewise for field m_row.
5999 (line_corrections::line_corrections): Likewise for param "row".
6000 (layout::print_trailing_fixits): Likewise.
6001 (layout::get_state_at_point): Likewise.
6002 (layout::get_x_bound_for_row): Likewise.
6003 (layout::print_line): Likewise.
6004 (diagnostic_show_locus): Likewise for locals "last_line" and
6006 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6007 * input.c (selftest::test_linenum_comparisons): New function.
6008 (selftest::input_c_tests): Call it.
6009 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6010 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6011 * selftest.h (ASSERT_GT): New macro.
6012 (ASSERT_GT_AT): New macro.
6013 (ASSERT_LT): New macro.
6014 (ASSERT_LT_AT): New macro.
6016 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6018 PR rtl-optimization/84780
6019 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6021 2018-03-14 Martin Liska <mliska@suse.cz>
6023 * tree.c (record_node_allocation_statistics): Use
6024 get_stats_node_kind.
6025 (get_stats_node_kind): New function extracted from
6026 record_node_allocation_statistics.
6027 (free_node): Use get_stats_node_kind.
6029 2018-03-14 Richard Biener <rguenther@suse.de>
6031 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6032 that the value-set of ANTIC_IN doesn't grow.
6035 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6037 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6038 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6040 2018-03-14 Julia Koval <julia.koval@intel.com>
6042 * config.gcc (icelake-client, icelake-server): New.
6044 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6045 (initial_ix86_arch_features): Ditto.
6046 (PTA_SKYLAKE): Add SGX.
6047 (PTA_ICELAKE): Remove.
6048 (PTA_ICELAKE_CLIENT): New.
6049 (PTA_ICELAKE_SERVER): New.
6050 (ix86_option_override_internal): Split up icelake on icelake client and
6052 (get_builtin_code_for_version): Ditto.
6053 (fold_builtin_cpu): Ditto.
6054 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6055 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6056 * config/i386/i386.h (processor_type): Ditto.
6057 * doc/invoke.texi: Ditto.
6059 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6062 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6063 INTEGER_CST offset, add it together with bitpos / 8 and
6064 sign extend based on POINTER_SIZE.
6068 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6071 * config/i386/constraints.md (Yc): New register constraint.
6072 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6073 Use Yc constraint for alternative 2 of operand 0. Remove
6074 preferred_for_speed attribute.
6076 2018-03-14 Richard Biener <rguenther@suse.de>
6078 PR tree-optimization/84830
6079 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6080 with the old one to avoid oscillations.
6082 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6085 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6087 (assign_by_spills): Return a flag of reload assignment failure.
6088 Do not process the reload assignment failures. Do not spill other
6089 reload pseudos if they has the same reg class. Update n if
6091 (lra_assign): Add a return arg. Set up from the result of
6092 assign_by_spills call.
6093 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6094 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6095 usage_insns if it is not NULL.
6096 (spill_hard_reg_in_range): New function.
6097 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6098 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6099 function prototypes.
6100 (lra_assign): Change prototype.
6101 * lra.c (lra): Add code to deal with fails by splitting hard reg
6104 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6106 * config/riscv/riscv.opt (mrelax): New option.
6107 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6108 "norelax" when riscv_mrelax is disabled.
6109 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6111 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6114 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6115 reassociation for int modes.
6117 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6119 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6120 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6122 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6123 * config/aarch64/aarch64-sve.md
6124 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6125 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6126 (*extend<mode><Vwide>2): Rename to...
6127 (aarch64_sve_extend<mode><Vwide>2): ...this.
6128 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6129 renaming the old pattern to...
6130 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6132 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6133 define_expand, renaming the old pattern to...
6134 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6135 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6136 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6137 account when deciding which SVE instruction the optab should use.
6138 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6140 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6142 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6143 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6144 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6145 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6146 (tlsdesc_small_<mode>): Turn a define_expand and use
6147 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6148 (tlsdesc_small_advsimd_<mode>): ...this.
6149 (tlsdesc_small_sve_<mode>): New pattern.
6151 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6153 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6154 (UNSPEC_UMUL_HIGHPART): New constants.
6155 (MUL_HIGHPART): New int iteraor.
6156 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6157 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6159 (*<su>mul<mode>3_highpart): New define_insn.
6161 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6164 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6167 2018-03-13 Martin Liska <mliska@suse.cz>
6170 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6172 (sem_item_optimizer::~sem_item_optimizer): Release it.
6173 (sem_item_optimizer::merge_classes): Register variable aliases.
6174 (sem_item_optimizer::fixup_pt_set): New function.
6175 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6176 * ipa-icf.h: Declare new variables and functions.
6178 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6181 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6182 integer_pow2p@2 and test integer_pow2p in condition.
6183 (A < 0 ? C : 0): Similarly for @1.
6186 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6187 characters starting at p contain '\0' character, don't look beyond
6191 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6192 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6195 * reg-stack.c (change_stack): Change update_end var from int to
6196 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6197 also call set_block_for_insn on the newly added insns and rescan.
6200 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6201 on the last operand.
6204 * tree.c (stabilize_reference_1): Return save_expr (e) for
6205 STATEMENT_LIST even if it doesn't have side-effects.
6207 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6209 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6211 2018-03-12 Renlin Li <renlin.li@arm.com>
6213 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6214 aarch64_output_scalar_simd_mov_immediate.
6216 2018-03-12 Martin Sebor <msebor@redhat.com>
6218 PR tree-optimization/83456
6219 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6220 for perfectly overlapping calls to memcpy.
6221 (gimple_fold_builtin_memory_chk): Same.
6222 (gimple_fold_builtin_strcpy): Handle no-warning.
6223 (gimple_fold_builtin_stxcpy_chk): Same.
6224 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6226 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6228 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6229 parameter. Use it for SFmode.
6230 (rs6000_function_arg_advance_1): Adjust.
6231 (rs6000_function_arg): Adjust.
6232 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6234 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6236 PR rtl-optimization/84169
6237 PR rtl-optimization/84780
6238 * combine.c (can_combine_p): Check for a 2-insn combination whether
6239 the destination register is used between the two insns, too.
6241 2018-03-12 Richard Biener <rguenther@suse.de>
6243 PR tree-optimization/84803
6244 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6245 for refs DR analysis didn't process.
6247 2018-03-12 Richard Biener <rguenther@suse.de>
6249 PR tree-optimization/84777
6250 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6251 force-vectorize loops ignore whether we are optimizing for size.
6253 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6255 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6256 (TARGET_MD_ASM_ADJUST): Define.
6258 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6259 Kito Cheng <kito.cheng@gmail.com>
6260 Chung-Ju Wu <jasonwucj@gmail.com>
6262 * config/nds32/nds32.c (nds32_compute_stack_frame,
6263 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6264 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6265 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6266 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6267 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6268 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6269 * config/nds32/nds32.md (prologue, epilogue): Use macro
6270 NDS32_V3PUSH_AVAILABLE_P to do checking.
6272 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6275 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6276 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6277 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6278 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6279 addition of most attributes on !orig_type_die or the attribute not
6280 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6282 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6283 Chung-Ju Wu <jasonwucj@gmail.com>
6285 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6287 * config/nds32/nds32.opt (mvh): New option.
6289 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6290 Chung-Ju Wu <jasonwucj@gmail.com>
6292 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6294 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6295 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6298 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6299 Chung-Ju Wu <jasonwucj@gmail.com>
6301 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6303 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6304 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6306 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6307 Kito Cheng <kito.cheng@gmail.com>
6308 Chung-Ju Wu <jasonwucj@gmail.com>
6310 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6311 UNSPEC_FFMISM and UNSPEC_FLMISM.
6312 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6313 for ffb, ffmism and flmism.
6314 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6315 (unspec_ffmism): Ditto.
6316 (unspec_flmism): Ditto.
6317 (nds32_expand_builtin_impl): Check if string extension is available.
6318 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6319 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6321 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6324 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6327 * lra-assigns.c (assign_by_spills): Return a flag of reload
6328 assignment failure. Do not process the reload assignment
6329 failures. Do not spill other reload pseudos if they has the same
6331 (lra_assign): Add a return arg. Set up from the result of
6332 assign_by_spills call.
6333 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6334 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6335 usage_insns if it is not NULL.
6336 (spill_hard_reg_in_range): New function.
6337 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6338 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6339 function prototypes.
6340 (lra_assign): Change prototype.
6341 * lra.c (lra): Add code to deal with fails by splitting hard reg
6344 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6347 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6349 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6352 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6353 (dw_val_node): Add val_symbolic_view.
6354 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6355 (symview_upper_bound): New.
6356 (new_line_info_table): Initialize symviews_since_reset.
6357 (dwarf2out_source_line): Count symviews_since_reset and set
6358 symview_upper_bound.
6359 (dw_val_equal_p): Handle symview.
6360 (add_AT_symview): New.
6361 (print_dw_val): Handle symview.
6362 (attr_checksum, attr_checksum_ordered): Likewise.
6363 (same_dw_val_p, size_of_die): Likewise.
6364 (value_format, output_die): Likewise.
6365 (add_high_low_attributes): Use add_AT_symview for entry_view.
6366 (dwarf2out_finish): Reset symview_upper_bound, clear
6369 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6372 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6373 Add strict argument and use it.
6374 (rs6000_split_multireg_move): Update for new strict argument.
6375 (mem_operand_gpr): Disallow all non-offsettable addresses.
6376 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6378 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6381 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6382 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6383 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6386 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6387 decl, use remap_type if we want to use the type.
6389 2018-03-09 Martin Sebor <msebor@redhat.com>
6391 PR tree-optimization/84526
6392 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6394 (builtin_access::generic_overlap): Be prepared to handle non-array
6397 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6399 PR rtl-optimization/84682
6400 * lra-constraints.c (process_address_1): Check is_address flag
6401 for address constraints.
6402 (process_alt_operands): Likewise.
6403 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6404 preprocess_constraints.
6405 * recog.h (preprocess_constraints): Add oploc parameter.
6407 * recog.c (preprocess_constraints): Test address_operand for
6408 CT_ADDRESS constraints.
6410 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6413 * lra-assigns.c (assign_by_spills): Return a flag of reload
6414 assignment failure. Do not process the reload assignment
6415 failures. Do not spill other reload pseudos if they has the same
6417 (lra_assign): Add a return arg. Set up from the result of
6418 assign_by_spills call.
6419 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6420 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6421 usage_insns if it is not NULL.
6422 (spill_hard_reg_in_range): New function.
6423 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6424 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6425 function prototypes.
6426 (lra_assign): Change prototype.
6427 * lra.c (lra): Add code to deal with fails by splitting hard reg
6430 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6433 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6434 Accept complain bool parameter. Only emit errors if it is true.
6435 (arm_parse_cpu_option_name): Likewise.
6436 (arm_target_thumb_only): Adjust callers of the above.
6437 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6438 prototype to take a default true bool parameter.
6439 (arm_parse_arch_option_name): Likewise.
6441 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6442 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6446 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6447 * configure: Regenerate.
6448 * configure.ac ("linker --version-script option"): New.
6449 ("linker soname option"): New.
6451 2018-03-09 Richard Biener <rguenther@suse.de>
6453 PR tree-optimization/84775
6454 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6455 immediate uses of predicate stmts and mark them modified.
6458 PR tree-optimization/84178
6459 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6461 (version_loop_for_if_conversion): Delay update_ssa call.
6462 (tree_if_conversion): Delay update_ssa until after predicate
6465 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6468 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6469 when the function accesses prior frames.
6471 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6474 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6475 gen_llsym, otherwise call maybe_gen_llsym.
6478 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6479 has ',' character inside of it.
6481 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6484 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6485 as clobbering CC_REGNUM.
6487 2018-03-08 Richard Biener <rguenther@suse.de>
6490 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6491 (follow_copies_to_constant): Do not follow SSA names registered
6494 2018-03-08 Richard Biener <rguenther@suse.de>
6496 PR tree-optimization/84178
6497 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6499 (version_loop_for_if_conversion): Delay update_ssa call.
6500 (tree_if_conversion): Delay update_ssa until after predicate
6503 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6505 PR tree-optimization/84178
6506 * tree-if-conv.c (release_bb_predicate): Remove the
6507 the assertion that the stmts have NULL use_ops.
6508 Discard the statements, asserting that they haven't
6509 yet been added to a BB.
6511 2018-03-08 Richard Biener <rguenther@suse.de>
6513 PR tree-optimization/84746
6514 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6515 (phi_translate): Pass in destination ANTIC_OUT set.
6516 (phi_translate_1): Likewise. For a simplified result lookup
6517 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6518 (phi_translate_set): Adjust.
6519 (do_pre_regular_insertion): Likewise.
6520 (do_pre_partial_partial_insertion): Likewise.
6522 2018-03-08 Martin Liska <mliska@suse.cz>
6524 PR gcov-profile/84735
6525 * doc/gcov.texi: Document usage of profile files.
6526 * gcov-io.h: Document changes in the format.
6528 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6532 * dwarf2out.c (struct dw_line_info_table): Update comments for
6534 (FORCE_RESET_NEXT_VIEW): New.
6535 (FORCE_RESETTING_VIEW_P): New.
6536 (RESETTING_VIEW_P): Check for -1 too.
6537 (ZERO_VIEW_P): Likewise.
6538 (new_line_info_table): Force-reset next view.
6539 (dwarf2out_begin_function): Likewise.
6540 (dwarf2out_source_line): Simplify zero_view_p initialization.
6541 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6542 view directly. Omit view when omitting .loc at line 0.
6544 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6546 PR tree-optimization/84740
6547 * tree-switch-conversion.c (process_switch): Call build_constructors
6548 only if info.phi_count is non-zero.
6550 PR tree-optimization/84739
6551 * tree-tailcall.c (find_tail_calls): Check call arguments against
6552 DECL_ARGUMENTS (current_function_decl) rather than
6553 DECL_ARGUMENTS (func) when checking for tail recursion.
6555 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6557 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6558 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6559 Volker Reichelt's entry and add entries for people that perform
6560 GCC fuzzy testing and report numerous bugs.
6562 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6565 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6566 readonly data in sdata, if that is disabled.
6567 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6568 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6569 -mreadonly-in-sdata option.
6571 2018-03-07 Martin Sebor <msebor@redhat.com>
6573 PR tree-optimization/84468
6574 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6575 basic block when looking for nul assignment.
6577 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6580 * except.h (output_function_exception_table): Adjust prototype.
6581 * except.c (output_function_exception_table): Remove FNNAME parameter
6582 and add SECTION parameter. Ouput one part of the table at a time.
6583 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6584 the first part of the exception table and emit unwind directives.
6585 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6586 (i386_pe_seh_cold_init): Likewise.
6587 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6588 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6589 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6590 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6591 * config/i386/winnt.c: Include except.h.
6592 (struct seh_frame_state): Add reg_offset, after_prologue and
6593 in_cold_section fields.
6594 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6595 (i386_pe_seh_cold_init): New function.
6596 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6597 to seh->in_cold_section.
6598 (seh_emit_push): Record the offset of the push.
6599 (seh_emit_save): Record the offet of the save.
6600 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6601 Test seh->after_prologue to disregard the epilogue.
6602 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6603 (i386_pe_end_cold_function): New function.
6605 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6608 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6609 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6612 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6614 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6615 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6618 * multiple_target.c: Include tree-inline.h and intl.h.
6619 (expand_target_clones): Diagnose and fail if node->definition and
6620 !tree_versionable_function_p (node->decl).
6622 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6624 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6626 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6627 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6628 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6630 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6633 * combine.c (try_combine): Use reg_or_subregno instead of handling
6634 just paradoxical SUBREGs and REGs.
6636 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6638 * config/arc/arc.c (arc_finalize_pic): Remove function.
6639 (arc_must_save_register): We use single base PIC register, remove
6640 checks to save/restore the PIC register.
6641 (arc_expand_prologue): Likewise.
6642 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6644 (arc_verify_short): Likewise.
6645 (arc_attr_type): Likewise.
6646 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6647 (walk_stores): Likewise.
6648 (arc_address_cost): Make it static.
6649 (arc_verify_short): Likewise.
6650 (branch_dest): Likewise.
6651 (arc_attr_type): Likewise.
6652 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6653 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6654 (arc_final_prescan_insn): Remove inserting the nops due to
6655 hardware hazards. It is done in reorg step.
6656 (insn_length_variant_t): Remove.
6657 (insn_length_parameters_t): Likewise.
6658 (arc_insn_length_parameters): Likewise.
6659 (arc_get_insn_variants): Likewise.
6660 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6662 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6665 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6668 PR tree-optimization/84687
6669 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6671 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6673 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6675 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6676 Rename to ppc_speculation_barrier.
6677 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6678 __builtin_ppc_speculation_barrier.
6680 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6683 * combine.c (combine_simplify_rtx): Don't try to simplify if
6684 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6687 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6689 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6690 to 32 bytes when compiling for POWER9.
6692 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6695 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6696 regparm >= 3 with no arg reg available also for calls with
6697 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6700 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6702 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6704 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6707 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6709 2018-03-05 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/84486
6712 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6713 When inserting a __builtin_assume_aligned call set the LHS
6714 SSA name alignment info accordingly.
6716 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6718 PR tree-optimization/84114
6719 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6720 Avoid reassociation of FLOAT_MODE addition.
6722 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6724 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6725 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6726 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6727 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6728 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6729 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6730 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6732 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6733 __builtin_ia32_wbinvd): New builtins.
6734 (SPECIAL_ARGS2): New.
6735 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6736 (SPECIAL_ARGS2): New.
6737 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6738 (ix86_valid_target_attribute_inner_p): Ditto.
6739 (ix86_init_mmx_sse_builtins): Add special_args2.
6740 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6741 TARGET_WBNOINVD_P): New.
6742 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6743 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6744 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6745 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6746 * config/i386/pconfigintrin.h: New file.
6747 * config/i386/wbnoinvdintrin.h: Ditto.
6748 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6749 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6751 2018-03-05 Richard Biener <rguenther@suse.de>
6753 PR tree-optimization/84670
6754 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6756 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6757 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6758 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6759 doesn't grow if all successors (recursively) were visited at least
6762 2018-03-05 Richard Biener <rguenther@suse.de>
6764 PR tree-optimization/84650
6765 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6766 if executed in the loop pipeline.
6768 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6770 * doc/configfiles.texi (Configuration Files): Move info about
6771 conditionalizing $target-protos.h to...
6772 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6773 differs from $target-protos.h.
6775 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6776 Chung-Ju Wu <jasonwucj@gmail.com>
6778 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6779 * config/nds32/nds32-multiple.md (setmemsi): Define.
6780 * config/nds32/nds32-memory-manipulation.c
6781 (nds32_gen_dup_4_byte_to_word_value): New.
6782 (emit_setmem_word_loop): New.
6783 (emit_setmem_byte_loop): New.
6784 (nds32_expand_setmem_loop): New.
6785 (nds32_expand_setmem_loop_v3m): New.
6786 (nds32_expand_setmem_unroll): New.
6787 (nds32_expand_setmem): New.
6789 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6790 Chung-Ju Wu <jasonwucj@gmail.com>
6792 * config/nds32/nds32-memory-manipulation.c
6793 (nds32_emit_load_store): New.
6794 (nds32_emit_post_inc_load_store): New.
6795 (nds32_emit_mem_move): New.
6796 (nds32_emit_mem_move_block): New.
6797 (nds32_expand_movmemsi_loop_unknown_size): New.
6798 (nds32_expand_movmemsi_loop_known_size): New.
6799 (nds32_expand_movmemsi_loop): New.
6800 (nds32_expand_movmemsi_unroll): New.
6801 (nds32_expand_movmemqi): Rename ...
6802 (nds32_expand_movmemsi): ... to this.
6803 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6804 (movmemsi): ... to this.
6805 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6806 (nds32_expand_movmemsi): ... to this.
6808 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6809 Monk Chiang <sh.chiang04@gmail.com>
6810 Chung-Ju Wu <jasonwucj@gmail.com>
6812 * config/nds32/nds32-protos.h
6813 (nds32_expand_load_multiple): New arguments.
6814 (nds32_expand_store_multiple): Ditto.
6815 (nds32_valid_multiple_load_store): Rename ...
6816 (nds32_valid_multiple_load_store_p): ... to this.
6817 * config/nds32/nds32-memory-manipulation.c
6818 (nds32_expand_load_multiple): Refine implementation.
6819 (nds32_expand_store_multiple): Ditto.
6820 * config/nds32/nds32-multiple.md
6821 (load_multiple): Update nds32_expand_load_multiple interface.
6822 (store_multiple): Update nds32_expand_store_multiple interface.
6823 * config/nds32/nds32-predicates.c
6824 (nds32_valid_multiple_load_store): Rename ...
6825 (nds32_valid_multiple_load_store_p): ... to this and refine
6827 * config/nds32/predicates.md
6828 (nds32_load_multiple_and_update_address_operation): New predicate.
6829 (nds32_store_multiple_and_update_address_operation): New predicate.
6831 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6832 Chung-Ju Wu <jasonwucj@gmail.com>
6834 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6835 (combo): New attribute.
6836 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6838 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6840 * config/nds32/nds32.opt: Change -mcmodel= default value.
6842 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6843 Monk Chiang <sh.chiang04@gmail.com>
6844 Chung-Ju Wu <jasonwucj@gmail.com>
6846 * config/nds32/constants.md (unspec_element): New enum.
6847 * config/nds32/constraints.md (Umw): New constraint.
6848 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6849 * config/nds32/nds32-intrinsic.md: Likewise.
6850 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6851 (nds32_valid_smw_lwm_base_p): New.
6852 (nds32_output_smw_single_word): New.
6853 (nds32_output_lmw_single_word): New.
6854 (nds32_expand_unaligned_load): New.
6855 (nds32_expand_unaligned_store): New.
6856 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6857 (nds32_output_smw_single_word): Declare.
6858 (nds32_output_lmw_single_word): Declare.
6859 (nds32_expand_unaligned_load): Declare.
6860 (nds32_expand_unaligned_store): Declare.
6861 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6862 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6863 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6864 NDS32_BUILTIN_UASTORE_DW.
6865 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6868 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6869 Kito Cheng <kito.cheng@gmail.com>
6870 Chung-Ju Wu <jasonwucj@gmail.com>
6872 * config/nds32/nds32-intrinsic.c
6873 (nds32_expand_builtin_null_ftype_reg): Delete.
6874 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6875 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6876 (nds32_read_argument): New.
6877 (nds32_legitimize_target): Ditto.
6878 (nds32_legitimize_argument): Ditto.
6879 (nds32_check_constant_argument): Ditto.
6880 (nds32_expand_unop_builtin): Ditto.
6881 (nds32_expand_unopimm_builtin): Ditto.
6882 (nds32_expand_binop_builtin): Ditto.
6883 (nds32_builtin_decl_impl): Ditto.
6884 (builtin_description): Ditto.
6885 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6886 (nds32_init_builtins_impl): Ditto.
6887 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6888 (nds32_builtin_decl): New.
6889 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6890 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6892 2018-03-02 Jeff Law <law@redhat.com>
6894 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6895 (redundant_insn, fill_simple_delay_slots): Likewise.
6896 (fill_slots_from_thread): Likewise.
6897 * resource.c (mark_referenced_resources): Likewise.
6898 (mark_set_resources, find_dead_or_set_registers): Likewise.
6900 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6902 * substring-locations.h (format_warning_va): Formatting fix for
6904 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6906 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6907 * substring-locations.c: Include intl.h.
6908 (format_warning_va): Turned into small wrapper around
6909 format_warning_n_va, renamed to ...
6910 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6911 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6913 (format_warning_at_substring_n): New function.
6914 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6915 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6916 format_warning_at_substring with just a shorter name instead of
6917 const function pointer.
6918 (fmtwarn_n): New function.
6919 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6920 appropriate, get rid of all the fmtstr temporaries, move conditionals
6921 with G_() wrapped string literals directly into fmtwarn arguments,
6922 cast dir.len to (int), formatting fixes.
6924 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6926 * doc/invoke.texi: Remove "Cilk Plus" references.
6928 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6929 Richard Biener <rguenther@suse.de>
6932 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6933 for error or warning attributes if CALL_FROM_THUNK_P is set.
6936 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6939 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6940 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6943 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6944 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6946 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6947 instead of -1U in last predictors element's probability member.
6949 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6952 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6953 arguments if they are comparable.
6955 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6957 PR tree-optimization/84634
6958 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6959 masks and masked_loop_p with a single loop_masks, making sure it's
6960 null for bb vectorization.
6962 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6964 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6965 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6966 loop->force_vectorize to check whether there is no alias.
6968 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6971 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6973 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6975 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6976 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6977 instead of a loop around prev_real_insn.
6978 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6982 * config/i386/i386.c (ix86_print_operand): Use conditional
6983 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6986 2018-03-02 Richard Biener <rguenther@suse.de>
6988 PR tree-optimization/84427
6989 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6990 (bitmap_set_subtract_values): Rewrite to handle multiple
6993 (prune_clobbered_mems): Likewise.
6994 (phi_translate): Take edge instead of pred/phiblock.
6995 (phi_translate_1): Likewise.
6996 (phi_translate_set): Likewise. Insert all translated
6997 exprs for a value into the set, keeping possibly multiple
6998 expressions per value.
6999 (compute_antic_aux): Adjust for phi_translate changes.
7000 When intersecting union the expressions and prune those
7001 not in the final value set, keeping possibly multiple
7002 expressions per value. Do not use value-insertion
7003 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7004 all expressions. Add verification that the value-sets
7005 only shrink during iteration.
7006 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7007 (do_pre_regular_insertion): Likewise.
7008 (do_pre_partial_partial_insertion): Likewise.
7010 2018-03-02 Richard Biener <rguenther@suse.de>
7013 * config/darwin.c (saved_debug_info_level): New static global.
7014 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7015 (darwin_asm_lto_end): Restore debug info generation settings.
7017 2018-03-01 Martin Liska <mliska@suse.cz>
7020 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7023 2018-03-01 Richard Biener <rguenther@suse.de>
7026 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7027 type annotation with LTO when debug was disabled at compile-time.
7029 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7031 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7033 (mips_final_postscan_insn): Likewise.
7035 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7037 PR rtl-optimization/84528
7038 * alias.c (init_alias_target): Add commentary.
7039 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7040 a unique base value if the frame pointer is not eliminated
7041 to the stack pointer.
7043 2018-03-01 Tom de Vries <tom@codesourcery.com>
7045 PR rtl-optimization/83327
7046 * lra-int.h (hard_regs_spilled_into): Declare.
7047 * lra.c (hard_regs_spilled_into): Define.
7048 (init_reg_info): Init hard_regs_spilled_into.
7049 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7050 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7051 (process_bb_lives): Handle hard_regs_spilled_into.
7052 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7053 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7055 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7057 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7058 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7059 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7060 * config/rs6000/aix72.h: New file.
7062 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7064 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7065 instead of warning_at with conditional singular and plural messages
7069 * stor-layout.c (update_alignment_for_field): For
7070 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7071 && !DECL_PACKED (field), do the alignment update, just use
7072 only desired_align instead of MAX (type_align, desired_align)
7074 (place_field): Don't do known_align < desired_align handling
7075 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7076 is non-NULL, instead do it after rli->prev_field handling and
7077 only if not within a bitfield word. For DECL_PACKED (field)
7078 use type_align of BITS_PER_UNIT.
7080 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7082 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7083 superfluous parentheses and trailing spaces.
7085 2018-02-28 Richard Biener <rguenther@suse.de>
7087 PR tree-optimization/84584
7088 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7089 SCoPs with fake exit edge.
7091 2018-02-28 Martin Liska <mliska@suse.cz>
7094 * timevar.c (timer::print): Fix format to properly print 100%
7097 2018-02-28 Richard Biener <rguenther@suse.de>
7100 * genmatch.c (capture_info::walk_match): Do not mark
7101 captured expressions without operands as expr_p given
7102 they act more like predicates and should be subject to
7103 "lost tail" side-effect preserving.
7105 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7107 PR rtl-optimization/81611
7108 * auto-inc-dec.c (attempt_change): Move dead note from
7109 mem_insn if it's the next use of regno
7110 (find_address): Take address use of reg holding
7111 non-incremented value. Add parm to limit search to the named
7113 (merge_in_block): Attempt to use a mem insn that is the next
7114 use of the original regno.
7116 2018-02-27 Martin Sebor <msebor@redhat.com>
7119 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7120 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7122 2018-02-27 Martin Sebor <msebor@redhat.com>
7124 PR translation/84207
7125 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7126 n argument to unsigned HOST_WIDE_INT.
7127 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7128 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7129 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7130 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7132 2018-02-27 Richard Biener <rguenther@suse.de>
7134 PR tree-optimization/84512
7135 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7136 Do not use the estimate returned from record_stmt_cost for
7137 the scalar iteration cost but sum properly using add_stmt_cost.
7139 2018-02-27 Richard Biener <rguenther@suse.de>
7141 PR tree-optimization/84466
7142 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7143 Adjust last change to less strictly validate use operands.
7145 2018-02-27 Martin Liska <mliska@suse.cz>
7147 PR gcov-profile/84548
7148 * gcov.c (process_file): Allow partial overlap and consider it
7149 also as group functions.
7150 (output_lines): Properly calculate range of lines for a group.
7152 2018-02-27 Martin Liska <mliska@suse.cz>
7154 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7155 'ggc' suffixes. Change first column width.
7156 (timer::print): Fix formatting of the column.
7158 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7160 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7161 preserve inline entry blocks for the sake of debug inline
7162 entry point markers alone.
7163 (remove_unused_locals): Suggest in comments a better place to
7164 force the preservation of inline entry blocks that are
7165 otherwise unused, but do not preserve them.
7167 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7169 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7171 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7174 * config/i386/constraints.md (Bs): Replace
7175 ix86_indirect_branch_register with
7176 TARGET_INDIRECT_BRANCH_REGISTER.
7178 * config/i386/i386.md (indirect_jump): Likewise.
7179 (tablejump): Likewise.
7180 (*sibcall_memory): Likewise.
7181 (*sibcall_value_memory): Likewise.
7182 Peepholes of indirect call and jump via memory: Likewise.
7183 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7184 (*sibcall_value_GOT_32): Likewise.
7185 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7186 (GOT_memory_operand): Likewise.
7187 (call_insn_operand): Likewise.
7188 (sibcall_insn_operand): Likewise.
7189 (GOT32_symbol_operand): Likewise.
7190 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7192 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7194 PR rtl-optimization/83496
7195 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7196 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7197 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7198 redundant insn, if any.
7199 (relax_delay_slots): Likewise.
7200 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7202 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7204 PR tree-optimization/83965
7205 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7206 that grouped statements are part of a reduction chain. Return
7207 true if the statement is not marked as a reduction itself but
7209 (vect_recog_dot_prod_pattern): Don't check whether the statement
7210 is part of a group here.
7211 (vect_recog_sad_pattern): Likewise.
7212 (vect_recog_widen_sum_pattern): Likewise.
7214 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7217 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7219 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7222 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7224 (ix86_output_indirect_function_return): New prototype.
7225 (ix86_split_simple_return_pop_internal): Likewise.
7226 * config/i386/i386.c (indirect_return_via_cx): New.
7227 (indirect_return_via_cx_bnd): Likewise.
7228 (indirect_thunk_name): Handle return va CX_REG.
7229 (output_indirect_thunk_function): Create alias for
7230 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7231 (ix86_output_indirect_jmp): Remove the bool argument.
7232 (ix86_output_indirect_function_return): New function.
7233 (ix86_split_simple_return_pop_internal): Likewise.
7234 * config/i386/i386.md (*indirect_jump): Don't pass false
7235 to ix86_output_indirect_jmp.
7236 (*tablejump_1): Likewise.
7237 (simple_return_pop_internal): Change it to define_insn_and_split.
7238 Call ix86_split_simple_return_pop_internal to split it for
7240 (simple_return_indirect_internal): Call
7241 ix86_output_indirect_function_return instead of
7242 ix86_output_indirect_jmp.
7244 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7247 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7248 memset and value initialization afterwards.
7250 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7252 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7254 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7257 * common/config/aarch64/aarch64-common.c
7258 (aarch_option_optimization_table[]): Switch
7259 off fomit-frame-pointer
7261 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7262 Chung-Ju Wu <jasonwucj@gmail.com>
7264 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7266 (store_multiple): Ditto.
7268 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7270 * config.gcc: Add --with-cpu support for nds32 target.
7271 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7272 * config/nds32/nds32.opt: Add -mcpu= option.
7274 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7276 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7277 isel=yes): Warn for these deprecated options.
7279 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7281 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7282 ISA_2_5_MASKS_EMBEDDED.
7284 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7286 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7287 p->max as pointers rather than using iterative_hash_expr.
7289 2018-02-23 Carl Love <cel@us.ibm.com>
7291 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7292 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7294 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7295 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7296 P8V_BUILTIN_VEC_VUNSIGNED2.
7298 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7301 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7302 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7304 * lra-constraints.c (curr_insn_transform): Set up
7305 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7307 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7308 * lra-eliminations.c (spill_pseudos): Ditto.
7309 (process_insn_for_elimination): Ditto.
7310 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7311 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7312 LRA_NON_CLOBBERED_ALT.
7314 2018-02-22 Martin Sebor <msebor@redhat.com>
7316 PR tree-optimization/84480
7317 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7318 to maybe_diag_stxncpy_trunc. Call it.
7319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7320 from gimple_fold_builtin_strcpy. Print inlining stack.
7321 (handle_builtin_stxncpy): Print inlining stack.
7322 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7324 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7327 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7328 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7329 and -fcheck-pointer-bounds are used together.
7330 (indirect_thunk_prefix): New enum.
7331 (indirect_thunk_need_prefix): New function.
7332 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7333 "_nt" instead of "_bnd" for NOTRACK prefix.
7334 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7335 (output_indirect_thunk_function): Likewise.
7337 (ix86_code_end): Update output_indirect_thunk_function calls.
7338 (ix86_output_indirect_branch_via_reg): Replace
7339 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7340 (ix86_output_indirect_branch_via_push): Likewise.
7341 (ix86_output_function_return): Likewise.
7342 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7343 incompatible with -fcf-protection=branch and
7344 -fcheck-pointer-bounds.
7346 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7349 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7350 Change gcc_assert call to output_operand_lossage.
7352 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7354 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7356 2018-02-22 DJ Delorie <dj@redhat.com>
7357 Sebastian Perta <sebastian.perta@renesas.com>
7358 Oleg Endo <olegendo@gcc.gnu.org>
7360 * config/rx/rx.c (rx_rtx_costs): New function.
7361 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7363 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7365 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7367 2018-02-22 Martin Liska <mliska@suse.cz>
7370 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7371 Add "native" as a possible value.
7373 2018-02-22 Martin Liska <mliska@suse.cz>
7376 * config/i386/i386.c (ix86_option_override_internal):
7377 Add "native" as a possible value for -march and -mtune.
7379 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7382 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7383 to all type variants.
7385 PR tree-optimization/84503
7386 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7387 width as info->bitpos + info->bitsize - start.
7388 (merged_store_group::merge_overlapping): Simplify width computation.
7389 (check_no_overlap): New function.
7390 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7391 start + width and last_order of the group, fail if check_no_overlap
7393 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7394 to group if check_no_overlap fails.
7396 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7398 * config/rs6000/altivec.md: Delete contraint arguments to
7399 define_expand, define_split, and define_peephole2, and in
7400 define_insn_and_split if always unused.
7401 * config/rs6000/darwin.md: Ditto.
7402 * config/rs6000/dfp.md: Ditto.
7403 * config/rs6000/rs6000.md: Ditto.
7404 * config/rs6000/sync.md: Ditto.
7405 * config/rs6000/vector.md: Ditto.
7406 * config/rs6000/vsx.md: Ditto.
7408 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7410 * config/rs6000/altivec.md: Write output control strings as braced
7411 blocks instead of double-quoted strings.
7412 * config/rs6000/darwin.md: Ditto.
7413 * config/rs6000/rs6000.md: Ditto.
7414 * config/rs6000/vector.md: Ditto.
7415 * config/rs6000/vsx.md: Ditto.
7417 2018-02-21 Jason Merrill <jason@redhat.com>
7419 PR c++/84314 - ICE with templates and fastcall attribute.
7420 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7422 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7424 * ipa-cp.c (determine_versionability): Fix comment typos.
7426 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7429 * ipa-cp.c (determine_versionability): Do not version functions caling
7432 2018-02-21 Martin Liska <mliska@suse.cz>
7435 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7436 Add "native" as a possible value.
7437 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7438 the macro when native cpu detection is available.
7440 2018-02-21 Martin Liska <mliska@suse.cz>
7443 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7444 Add "native" as a possible value.
7445 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7446 when native cpu detection is available.
7448 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7449 Martin Sebor <msebor@redhat.com>
7451 PR tree-optimization/84478
7452 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7454 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7455 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7456 support which is conservatively correct, for 2 only stay conservative
7457 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7458 argument to the 2 argument get_range_strlen, adjust 6 arg
7459 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7461 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7462 (gimple_fold_builtin_strlen): Pass true as last argument to
7465 2018-02-20 Martin Sebor <msebor@redhat.com>
7468 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7469 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7470 (builtin_memref::builtin_memref): Factor out parts into
7471 set_base_and_offset and call it.
7473 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7476 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7477 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7478 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7479 search at the associated MODE_INT.
7481 2018-02-20 Jeff Law <law@redhat.com>
7484 PR tree-optimization/81592
7486 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7487 for range data rather than using global data.
7488 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7489 range data rather than using global data.
7490 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7491 pass it to children as needed.
7492 (struct directive::fmtresult): Similarly.
7493 (struct directive::set_width): Similarly.
7494 (struct directive::set_precision): Similarly.
7495 (format_integer, format_directive, parse_directive): Similarly.
7496 (format_none): Accept unnamed vr_values parameter.
7497 (format_percent, format_floating, format_character): Similarly.
7498 (format_string, format_plain): Similarly.
7499 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7500 the EVRP range analyzer for range data rather than using global data.
7501 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7502 gimple-ssa-evrp-analyze.h
7503 (class sprintf_dom_walker): Add after_dom_children member function.
7504 Add evrp_range_analyzer member.
7505 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7506 range analyzer as needed.
7507 (sprintf_dom_walker::after_dom_children): New member function.
7508 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7510 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7511 (evrp_range_analyzer::pop_to_marker): Likewise.
7513 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7515 PR tree-optimization/84419
7516 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7517 with the required type if its current type is compatible but
7520 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7523 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7524 after vectorization.
7526 2018-02-20 Martin Liska <mliska@suse.cz>
7529 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7530 possible values if we don't have a hint.
7532 2018-02-20 Martin Liska <mliska@suse.cz>
7536 * final.c (shorten_branches): Build align_tab array with one
7538 * opts.c (finish_options): Add alignment option limit check.
7539 (MAX_CODE_ALIGN): Likewise.
7540 (MAX_CODE_ALIGN_VALUE): Likewise.
7541 * doc/invoke.texi: Document maximum allowed option value for
7542 all -falign-* options.
7544 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7547 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7548 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7549 * var-tracking.c (emit_note_insn_var_location): Remove all references
7550 to NOTE_INSN_CALL_ARG_LOCATION.
7551 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7552 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7553 Use copy_rtx_if_shared.
7554 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7555 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7556 (dwarf2out_var_location): Remove handling of
7557 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7559 * final.c (final_scan_insn): Remove all references to
7560 NOTE_INSN_CALL_ARG_LOCATION.
7561 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7562 before dumping final insns.
7563 * except.c (emit_note_eh_region_end): Remove all references to
7564 NOTE_INSN_CALL_ARG_LOCATION.
7565 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7566 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7567 * config/arc/arc.c (hwloop_optimize): Likewise.
7568 * config/arm/arm.c (create_fix_barrier): Likewise.
7569 * config/s390/s390.c (s390_chunkify_start): Likewise.
7570 * config/sh/sh.c (find_barrier): Likewise.
7571 * config/i386/i386.c (rest_of_insert_endbranch,
7572 ix86_seh_fixup_eh_fallthru): Likewise.
7573 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7574 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7575 * config/frv/frv.c (frv_function_prologue): Likewise.
7576 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7578 (note_outside_basic_block_p): Remove all references to
7579 NOTE_INSN_CALL_ARG_LOCATION.
7580 * gengtype.c (adjust_field_rtx_def): Likewise.
7581 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7583 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7584 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7587 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7590 PR tree-optimization/84452
7591 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7592 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7595 2018-02-19 Martin Liska <mliska@suse.cz>
7598 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7600 2018-02-19 Martin Liska <mliska@suse.cz>
7601 Richard Sandiford <richard.sandiford@linaro.org>
7603 PR tree-optimization/82491
7604 * gimple-fold.c (get_base_constructor): Make earlier bail out
7607 2018-02-19 Carl Love <cel@us.ibm.com>
7609 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7610 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7612 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7613 P8V_BUILTIN_VEC_NEG.
7615 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7617 * config/rl78/rl78.md (movdf): New define expand.
7619 2018-02-19 Martin Liska <mliska@suse.cz>
7622 * doc/invoke.texi: Fix typo.
7623 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7625 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7627 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7628 handle rs6000_single_float and rs6000_double_float specially for
7631 2018-02-16 Jeff Law <law@redhat.com>
7633 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7636 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7639 * ipa-inline.c (inline_small_functions): Fix a typo.
7641 2018-02-16 Nathan Sidwell <nathan@acm.org>
7643 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7645 2018-02-16 Carl Love <cel@us.ibm.com>
7647 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7648 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7649 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7650 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7651 expansion to P8V_BUILTIN_VEC_FLOAT2.
7653 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7655 PR rtl-optimization/70023
7656 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7657 src_regno into account.
7659 2018-02-16 Carl Love <cel@us.ibm.com>
7661 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7662 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7663 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7664 * config/rs6000/rs6000.c: Remove case statements for
7665 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7666 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7667 and P9V_BUILTIN_VEC_VINSERT4B.
7668 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7669 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7670 * config/rs6000/vsx.md:
7671 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7674 2018-02-16 Carl Love <cel@us.ibm.com>
7676 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7678 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7680 * config/rs6000/rs6000-c.c: Add the definitions for
7681 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7682 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7683 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7684 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7685 definition for insert4b and define insn *insert3b_internal.
7686 * doc/extend.texi: Add documentation for vec_extract4b.
7688 2018-02-16 Nathan Sidwell <nathan@acm.org>
7690 * doc/extend.texi (Backwards Compatibility): Mention friend
7691 injection. Note for-scope is deprecated.
7692 * doc/invoke.texi (-ffriend-injection): Deprecate.
7694 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7696 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7697 that moved to I2, also allow destinations that are a paradoxical
7698 subreg (instead of a normal reg).
7700 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7703 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7706 2018-02-16 Richard Biener <rguenther@suse.de>
7708 PR tree-optimization/84037
7709 PR tree-optimization/84016
7711 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7712 Adjust vec_construct for the fact we need additional higher latency
7713 128bit inserts for AVX256 and AVX512 vector builds.
7714 (ix86_add_stmt_cost): Scale vector construction cost for
7717 2018-02-16 Richard Biener <rguenther@suse.de>
7719 PR tree-optimization/84417
7720 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7721 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7722 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7724 2018-02-16 Martin Liska <mliska@suse.cz>
7727 * internal-fn.def (ASAN_CHECK): Set proper flags.
7728 (ASAN_MARK): Likewise.
7730 2018-02-16 Julia Koval <julia.koval@intel.com>
7732 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7733 from PTA_CANNONLAKE.
7735 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7738 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7739 Use ++iter rather than iter++ for std::list iterators.
7740 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7741 defer deleting them until all nodes in the forest are processed. Do
7742 free even leaf nodes. Change to_process into auto_vec.
7745 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7746 * vec.h (vec_default_construct): Use memset instead of placement new
7747 if BROKEN_VALUE_INITIALIZATION is defined.
7748 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7749 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7752 PR rtl-optimization/83723
7753 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7754 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7755 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7757 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7759 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7761 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7763 PR rtl-optimization/81443
7764 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7765 from inner REGs to paradoxical SUBREGs.
7767 2018-02-16 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/84399
7770 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7771 For operands we can analyze at their definition make sure we can
7772 analyze them at each use as well.
7774 2018-02-16 Richard Biener <rguenther@suse.de>
7776 PR tree-optimization/84190
7777 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7778 volatile accesses if the decl isn't volatile.
7780 2018-02-15 Jason Merrill <jason@redhat.com>
7782 PR c++/84314 - ICE with templates and fastcall attribute.
7783 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7784 TYPE_CANONICAL on an existing type.
7786 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7788 PR tree-optimization/84383
7789 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7790 dstoff nor call operand_equal_p if dstbase is NULL.
7792 PR tree-optimization/84334
7793 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7794 also a CONSTANT_CLASS_P, punt.
7796 2018-02-14 Jim Wilson <jimw@sifive.com>
7798 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7799 first SMALL_OPERAND check. New local min_second_step. Move assert
7800 to where locals are set. Add TARGET_RVC support.
7801 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7803 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7805 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7807 2018-02-14 Martin Sebor <msebor@redhat.com>
7809 PR tree-optimization/83698
7810 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7811 arrays constrain the offset range to their bounds.
7812 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7813 (builtin_access::overlap): Avoid setting the size of overlap if it's
7815 (maybe_diag_overlap): Also consider arrays when deciding what values
7816 of offsets to include in diagnostics.
7818 2018-02-14 Martin Sebor <msebor@redhat.com>
7821 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7822 that correspond to the kind of a declaration.
7824 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7827 * config/pa/pa.md: Load address of PIC label using the linkage table
7828 if the label is nonlocal.
7830 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7832 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7833 warning message if user requests -maltivec=be.
7834 * doc/invoke.texi: Document deprecation of -maltivec=be.
7836 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7839 * config/rs6000/rs6000-c.c: Update definitions for
7840 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7841 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7843 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7846 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7847 add _get_ssp intrinsics. Remove argument from
7848 __builtin_ia32_rdssp[d|q].
7849 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7850 * config/i386/i386-builtin.def: Remove argument from
7851 __builtin_ia32_rdssp[d|q].
7852 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7853 ix86_expand_special_args_builtin for _rdssp[d|q].
7854 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7855 Clear register before usage.
7856 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7857 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7859 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7861 PR tree-optimization/84357
7862 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7863 operand 1 of an ARRAY_REF too.
7865 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7868 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7869 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7871 (set_of_reg): New struct.
7872 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7873 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7874 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7876 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7877 Split into bitclr, bitset, bitinvert patterns if appropriate.
7878 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7879 use rx_fuse_in_memory_bitop.
7880 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7881 to named insn, correct maximum insn length.
7883 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7886 * machmode.def: Define a complex mode for PARTIAL_INT.
7887 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7889 * doc/rtl.texi: Document CSPImode.
7890 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7892 (msp430_hard_regno_nregs_with_padding): Likewise.
7894 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7897 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7899 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7901 PR rtl-optimization/84169
7902 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7903 we generated a parallel as new i3 and we split that to new i2 and i3
7904 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7905 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7906 those to i2, not i1. Partially rewrite this scan code.
7908 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7911 * stor-layout.c (place_field): For variable length fields, adjust
7912 offset_align afterwards not just based on the field's alignment,
7913 but also on the size.
7916 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7917 of exps and logs in the use_exp2 case.
7919 2018-02-13 Jeff Law <law@redhat.com>
7921 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7924 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7927 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7931 * common.opt (gas-loc-support, gas-locview-support): New.
7932 (ginline-points, ginternal-reset-location-views): New.
7933 * doc/invoke.texi: Document them. Use @itemx where intended.
7934 (gvariable-location-views): Adjust.
7935 * target.def (reset_location_view): New.
7936 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7937 (TARGET_RESET_LOCATION_VIEW): New.
7938 * doc/tm.texi: Rebuilt.
7939 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7940 (dwarf2out_default_as_locview_support): New.
7941 (output_asm_line_debug_info): Use option variables.
7942 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7943 (output_loc_list): Likewise.
7944 (add_high_low_attributes): Check option variables.
7945 Don't output entry view attribute in strict mode.
7946 (gen_inlined_subroutine_die): Check option variables.
7947 (dwarf2out_inline_entry): Likewise.
7948 (init_sections_and_labels): Likewise.
7949 (dwarf2out_early_finish): Likewise.
7950 (maybe_reset_location_view): New, from...
7951 (dwarf2out_var_location): ... here. Call it.
7952 * debug.h (dwarf2out_default_as_loc_support): Declare.
7953 (dwarf2out_default_as_locview_support): Declare.
7954 * hooks.c (hook_int_rtx_insn_0): New.
7955 * hooks.h (hook_int_rtx_insn_0): Declare.
7956 * toplev.c (process_options): Take -gas-loc-support and
7957 -gas-locview-support from dwarf2out. Enable
7958 -gvariable-location-views by default only with locview
7959 assembler support. Enable -ginternal-reset-location-views by
7960 default only if the target defines the corresponding hook.
7961 Enable -ginline-points by default if location views are
7962 enabled; force it disabled if statement frontiers are
7964 * tree-inline.c (expand_call_inline): Check option variables.
7965 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7967 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7969 PR tree-optimization/84321
7970 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7971 handling. Also check whether the anti-range contains any values
7972 that satisfy the mask; switch to a VR_RANGE if not.
7974 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7977 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7979 2018-02-13 Martin Jambor <mjambor@suse.cz>
7982 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7983 of call statements, also set location of a load to a temporary.
7985 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7987 * config/rl78/rl78.c (add_vector_labels): New function.
7988 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7989 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7990 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7991 which checks that no arguments are passed.
7992 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7993 * doc/extend.texi: Documentation for the new attribute.
7995 2018-02-13 Andreas Schwab <schwab@suse.de>
7997 * config/riscv/linux.h (CPP_SPEC): Define.
7999 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8002 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8003 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8004 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8005 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8006 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8007 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8008 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8009 temporarily for AES and PCLMUL builtins.
8011 PR tree-optimization/84339
8012 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8013 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8017 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8018 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8019 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8021 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8023 * omp-simd-clone.h: New file.
8024 * omp-simd-clone.c: Include omp-simd-clone.h.
8025 (expand_simd_clones): No longer static.
8026 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8027 cgraph.h and omp-simd-clone.h.
8028 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8029 (vect_recog_widen_shift_pattern): Formatting fix.
8030 (vect_pattern_recog_1): Don't check optab for calls.
8033 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8034 operands[2] into a REG before using gen_lowpart on it.
8036 2018-02-12 Jeff Law <law@redhat.com>
8039 * config/sh/sh.c (find_barrier): Consider a sibling call
8042 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8043 successfully back substituting a reg.
8045 2018-02-12 Richard Biener <rguenther@suse.de>
8047 PR tree-optimization/84037
8048 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8049 parameter, move visited init to caller.
8050 (vect_slp_analyze_operations): Separate cost from validity
8051 check, initialize visited once for all instances.
8052 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8054 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8056 (vect_model_promotion_demotion_cost): Likewise.
8057 (vectorizable_bswap): Guard cost modeling with !slp_node
8058 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8060 (vectorizable_call): Likewise.
8061 (vectorizable_conversion): Likewise.
8062 (vectorizable_assignment): Likewise.
8063 (vectorizable_shift): Likewise.
8064 (vectorizable_operation): Likewise.
8065 (vectorizable_store): Likewise.
8066 (vectorizable_load): Likewise.
8067 (vectorizable_condition): Likewise.
8068 (vectorizable_comparison): Likewise.
8070 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8073 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8074 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8075 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8077 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8080 * params.def (inline-min-speedup): Increase from 8 to 15.
8081 (max-inline-insns-auto): Decrease from 40 to 30.
8082 * ipa-split.c (consider_split): Add some buffer for function to
8083 be considered inlining candidate.
8084 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8087 2018-02-12 Richard Biener <rguenther@suse.de>
8089 PR tree-optimization/84037
8090 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8091 matched stmts if we cannot swap the non-matched ones.
8093 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8095 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8096 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8097 _mm_maskz_scalef_round_ss): New intrinsics.
8098 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8099 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8100 __builtin_ia32_scalefss_round): Remove.
8101 (__builtin_ia32_scalefsd_mask_round,
8102 __builtin_ia32_scalefss_mask_round): New intrinsics.
8103 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8104 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8105 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8106 "<round_constraint>")): Changed to ...
8107 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8108 "<round_scalar_constraint>")): ... this.
8109 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8110 %0, %1, %2<round_op3>}"): Changed to ...
8111 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8112 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8113 %2<round_scalar_mask_op3>}"): ... this.
8114 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8116 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8118 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8119 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8120 (_mm_maskz_sqrt_round_ss): New intrinsics.
8121 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8122 (__builtin_ia32_sqrtsd_mask_round)
8123 (__builtin_ia32_sqrtss_mask_round): New builtins.
8124 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8125 (__builtin_ia32_sqrtss_round): Remove.
8126 (__builtin_ia32_sqrtsd_mask_round)
8127 (__builtin_ia32_sqrtss_mask_round): New builtins.
8128 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8129 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8130 ((match_operand:VF_128 1 "vector_operand"
8131 "xBm,<round_constraint>")): Changed to ...
8132 ((match_operand:VF_128 1 "vector_operand"
8133 "xBm,<round_scalar_constraint>")): ... this.
8134 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8135 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8136 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8137 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8138 %<iptr>1<round_scalar_mask_op3>}): ... this.
8139 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8140 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8142 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8145 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8146 Cast vec_cmpeq result to correct type.
8147 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8148 Cast vec_cmpgt result to correct type.
8150 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8152 * final.c (final_scan_insn_1): Renamed from...
8153 (final_scan_insn): ... this. New wrapper, to recover
8154 seen from the outermost call in recursive ones.
8155 * config/sparc/sparc.c (output_return): Drop seen from call.
8156 (output_sibcall): Likewise.
8157 * config/visium/visium.c (output_branch): Likewise.
8159 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8161 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8164 2018-02-10 Alan Modra <amodra@gmail.com>
8167 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8168 Specify LR as an input.
8170 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8173 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8174 remove_member_access_dummy_vars): New functions.
8175 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8176 lower_omp_1, execute_lower_omp): Use them.
8178 PR rtl-optimization/84308
8179 * shrink-wrap.c (spread_components): Release todo vector.
8181 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8183 PR rtl-optimization/57193
8184 * ira-color.c (struct allocno_color_data): Add member
8185 conflict_allocno_hard_prefs.
8186 (update_conflict_allocno_hard_prefs): New.
8187 (bucket_allocno_compare_func): Add a preference based on
8188 conflict_allocno_hard_prefs.
8189 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8190 (color_allocnos): Remove a dead code. Initiate
8191 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8193 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8196 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8197 constraint from =wa to wa. Avoid a subreg on the output operand,
8198 instead use a pseudo and subreg it in a move.
8199 (p9_xxbrd_<mode>): Changed to ...
8200 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8201 (p9_xxbrd_v2df): New expander.
8202 (p9_xxbrw_<mode>): Changed to ...
8203 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8204 (p9_xxbrw_v4sf): New expander.
8206 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8208 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8210 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8213 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8214 multiply in 32-bit mode.
8215 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8216 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8219 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8221 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8222 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8223 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8224 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8226 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8229 * dwarf2out.c (is_trivial_indirect_ref): New function.
8230 (dwarf2out_late_global_decl): Do not generate a location
8231 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8232 and that are not defined in the current unit.
8234 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8236 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8237 instead of a libcall for UNORDERED.
8239 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8242 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8243 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8245 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8248 * config/s390/s390.c (s390_set_current_function): Invoke
8249 s390_indirect_branch_settings also if fndecl didn't change.
8251 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8253 * config/rs6000/rs6000.md (blockage): Set length to zero.
8255 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8257 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8259 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8262 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8263 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8267 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8268 PARALLEL incoming that failed vt_get_decl_and_offset check.
8271 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8272 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8274 (get_variable_section): For decls in named .bss* sections pass true as
8275 second argument to bss_initializer_p.
8277 2018-02-09 Marek Polacek <polacek@redhat.com>
8278 Jakub Jelinek <jakub@redhat.com>
8281 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8282 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8283 Sync some changes from cxx_fold_indirect_ref.
8285 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8287 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8289 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8290 (BLOCK_INLINE_ENTRY_LABEL): New.
8291 (dwarf2out_var_location): Disregard inline entry markers.
8292 (inline_entry_data): New struct.
8293 (inline_entry_data_hasher): New hashtable type.
8294 (inline_entry_data_hasher::hash): New.
8295 (inline_entry_data_hasher::equal): New.
8296 (inline_entry_data_table): New variable.
8297 (add_high_low_attributes): Add DW_AT_entry_pc and
8298 DW_AT_GNU_entry_view attributes if a pending entry is found
8299 in inline_entry_data_table. Add old entry_pc attribute only
8300 if debug nonbinding markers are disabled.
8301 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8302 markers are enabled.
8303 (block_within_block_p, dwarf2out_inline_entry): New.
8304 (dwarf2out_finish): Check that no entries remained in
8305 inline_entry_data_table.
8306 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8307 (final_scan_insn, notice_source_line): Likewise.
8308 (rest_of_clean_state): Skip inline entry markers.
8309 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8311 * gimple.c (gimple_build_debug_inline_entry): New.
8312 * gimple.h (enum gimple_debug_subcode): Add
8313 GIMPLE_DEBUG_INLINE_ENTRY.
8314 (gimple_build_debug_inline_entry): Declare.
8315 (gimple_debug_inline_entry_p): New.
8316 (gimple_debug_nonbind_marker_p): Adjust.
8317 * insn-notes.def (INLINE_ENTRY): New.
8318 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8319 inline entry marker notes.
8320 (print_insn): Likewise.
8321 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8322 (INSN_DEBUG_MARKER_KIND): Likewise.
8323 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8324 * tree-inline.c (expand_call_inline): Build and insert
8325 debug_inline_entry stmt.
8326 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8327 inline entry blocks early, if nonbind markers are enabled.
8328 (dump_scope_block): Dump fragment info.
8329 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8330 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8331 (gimple_build_debug_inline_entry): New.
8332 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8333 Enable/disable inline entry points too.
8334 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8335 (DEBUG_INSN): Describe inline entry markers.
8337 * common.opt (gvariable-location-views): New.
8338 (gvariable-location-views=incompat5): New.
8339 * config.in: Rebuilt.
8340 * configure: Rebuilt.
8341 * configure.ac: Test assembler for view support.
8342 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8343 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8344 * dwarf2out.c (var_loc_view): New typedef.
8345 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8346 (dwarf2out_locviews_in_attribute): New.
8347 (dwarf2out_locviews_in_loclist): New.
8348 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8349 (enum dw_line_info_opcode): Add LI_adv_address.
8350 (struct dw_line_info_table): Add view.
8351 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8352 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8353 (zero_view_p): New variable.
8354 (ZERO_VIEW_P): New macro.
8355 (output_asm_line_debug_info): New.
8356 (struct var_loc_node): Add view.
8357 (add_AT_view_list, AT_loc_list): New.
8358 (add_var_loc_to_decl): Add view param. Test it against last.
8359 (new_loc_list): Add view params. Record them.
8360 (AT_loc_list_ptr): Handle loc and view lists.
8361 (view_list_to_loc_list_val_node): New.
8362 (print_dw_val): Handle dw_val_class_view_list.
8363 (size_of_die): Likewise.
8364 (value_format): Likewise.
8365 (loc_list_has_views): New.
8366 (gen_llsym): Set vl_symbol too.
8367 (maybe_gen_llsym, skip_loc_list_entry): New.
8368 (dwarf2out_maybe_output_loclist_view_pair): New.
8369 (output_loc_list): Output view list or entries too.
8370 (output_view_list_offset): New.
8371 (output_die): Handle dw_val_class_view_list.
8372 (output_dwarf_version): New.
8373 (output_compilation_unit_header): Use it.
8374 (output_skeleton_debug_sections): Likewise.
8375 (output_rnglists, output_line_info): Likewise.
8376 (output_pubnames, output_aranges): Update version comments.
8377 (output_one_line_info_table): Output view numbers in asm comments.
8378 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8379 Call maybe_gen_llsym.
8380 (loc_list_from_tree_1): Adjust.
8381 (add_AT_location_description): Create view list attribute if
8382 needed, check it's absent otherwise.
8383 (convert_cfa_to_fb_loc_list): Adjust.
8384 (maybe_emit_file): Call output_asm_line_debug_info for test.
8385 (dwarf2out_var_location): Reset views as needed. Precompute
8386 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8387 attribute. Set view.
8388 (new_line_info_table): Reset next view.
8389 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8390 (dwarf2out_source_line): Likewise. Output view resets and labels to
8391 the assembler, or select appropriate line info opcodes.
8392 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8393 (optimize_string_length): Catch it. Adjust.
8394 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8395 dw_val_class_view_list, and remove it if no longer needed.
8396 (hash_loc_list): Hash view numbers.
8397 (loc_list_hasher::equal): Compare them.
8398 (optimize_location_lists): Check whether a view list symbol is
8399 needed, and whether the locview attribute is present, and
8400 whether they match. Remove the locview attribute if no longer
8402 (index_location_lists): Call skip_loc_list_entry for test.
8403 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8404 Use output_dwarf_version.
8405 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8406 (struct dw_val_node): Add val_view_list.
8407 * final.c (SEEN_NEXT_VIEW): New.
8408 (set_next_view_needed): New.
8409 (clear_next_view_needed): New.
8410 (maybe_output_next_view): New.
8411 (final_start_function): Rename to...
8412 (final_start_function_1): ... this. Take pointer to FIRST,
8413 add SEEN parameter. Emit param bindings in the initial view.
8414 (final_start_function): Reintroduce SEEN-less interface.
8415 (final): Rename to...
8416 (final_1): ... this. Take SEEN parameter. Output final pending
8417 next view at the end.
8418 (final): Reintroduce seen-less interface.
8419 (final_scan_insn): Output pending next view before switching
8420 sections or ending a block. Mark the next view as needed when
8421 outputting variable locations. Notify debug backend of section
8422 changes, and of location view changes.
8423 (rest_of_handle_final): Adjust.
8424 * toplev.c (process_options): Autodetect value for debug variable
8425 location views option. Warn on incompat5 without -gdwarf-5.
8426 * doc/invoke.texi (gvariable-location-views): New.
8427 (gvariable-location-views=incompat5): New.
8428 (gno-variable-location-views): New.
8430 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8432 PR tree-optimization/84136
8433 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8434 that the result of find_edge is non-NULL.
8436 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8439 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8440 storing integer register in SImode. Fix cost of 256 and 512
8441 byte aligned SSE register store.
8443 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8445 * config/i386/i386.c (ix86_multiplication_cost): Fix
8446 multiplication cost for TARGET_AVX512DQ.
8448 2018-02-08 Marek Polacek <polacek@redhat.com>
8450 PR tree-optimization/84238
8451 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8454 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8456 PR tree-optimization/84265
8457 * tree-vect-stmts.c (vectorizable_store): Don't treat
8458 VMAT_CONTIGUOUS accesses as grouped.
8459 (vectorizable_load): Likewise.
8461 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8463 PR tree-optimization/81635
8464 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8465 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8466 (test_round_for_mask): New functions.
8467 (wide_int_cc_tests): Call test_round_for_mask.
8468 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8469 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8470 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8471 range returned by get_range_info.
8473 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8476 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8477 * symtab.c: Include builtins.h
8478 (symtab_node::output_to_lto_symbol_table_p): Move here
8479 from lto-streamer-out.c:output_symbol_p.
8480 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8481 (output_symbol_p): Move all logic to symtab.c
8482 (produce_symtab): Update.
8484 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8486 * config/s390/s390-opts.h (enum indirect_branch): Define.
8487 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8488 (s390_indirect_branch_via_thunk)
8489 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8490 (enum s390_indirect_branch_type): Define.
8491 * config/s390/s390.c (struct s390_frame_layout, struct
8492 machine_function): Remove.
8493 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8494 (indirect_branch_table_label_no, indirect_branch_table_name):
8496 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8497 (enum s390_indirect_branch_option): Define.
8498 (s390_return_addr_from_memory): New function.
8499 (s390_handle_string_attribute): New function.
8500 (s390_attribute_table): Add new attribute handler.
8501 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8502 (s390_indirect_branch_via_thunk): New function.
8503 (s390_indirect_branch_via_inline_thunk): New function.
8504 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8505 sibling call optimization for non z10 compiles.
8506 (s390_emit_call): Force indirect branch target to be a single
8507 register. Add r1 clobber for non-z10 compiles.
8508 (s390_emit_epilogue): Emit return jump via return_use expander.
8509 (s390_reorg): Handle JUMP_INSNs as execute targets.
8510 (s390_option_override_internal): Perform validity checks for the
8511 new command line options.
8512 (s390_indirect_branch_attrvalue): New function.
8513 (s390_indirect_branch_settings): New function.
8514 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8515 (s390_output_indirect_thunk_function): New function.
8516 (s390_code_end): Implement target hook.
8517 (s390_case_values_threshold): Implement target hook.
8518 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8520 * config/s390/s390.h (struct s390_frame_layout)
8521 (struct machine_function): Move here from s390.c.
8522 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8523 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8524 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8525 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8526 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8527 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8528 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8529 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8530 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8531 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8532 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8533 (mnemonic attribute): Add values which aren't recognized
8535 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8536 pattern for branch conversion. Fix mnemonic attribute.
8537 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8538 indirect branch via thunk if requested.
8539 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8540 ("*indirect_jump"): Disable for branch conversion using out of
8542 ("indirect_jump_via_thunk<mode>_z10")
8543 ("indirect_jump_via_thunk<mode>")
8544 ("indirect_jump_via_inlinethunk<mode>_z10")
8545 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8546 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8547 ("casesi_jump_via_inlinethunk<mode>_z10")
8548 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8549 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8550 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8551 ("*indirect2_jump"): Disable for branch conversion.
8552 ("casesi_jump"): Turn into expander and expand patterns for branch
8554 ("return_use"): New expander.
8555 ("*return"): Emit return via thunk and rename it to ...
8556 ("*return<mode>"): ... this one.
8557 * config/s390/s390.opt: Add new options and and enum for the
8560 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8562 * lra-constraints.c (match_reload): Unconditionally use
8563 gen_lowpart_SUBREG, rather than selecting between that
8564 and equivalent gen_rtx_SUBREG code.
8566 2018-02-08 Richard Biener <rguenther@suse.de>
8568 PR tree-optimization/84233
8569 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8570 changed flag instead of boguously re-using phi_inserted.
8572 2018-02-08 Martin Jambor <mjambor@suse.cz>
8574 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8575 static local variables.
8577 2018-02-08 Richard Biener <rguenther@suse.de>
8579 PR tree-optimization/84278
8580 * tree-vect-stmts.c (vectorizable_store): When looking for
8581 smaller vector types to perform grouped strided loads/stores
8582 make sure the mode is supported by the target.
8583 (vectorizable_load): Likewise.
8585 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8587 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8588 Increase LDP/STP opportunities by adding adjacent callee-saves.
8590 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8592 PR rtl-optimization/84068
8593 PR rtl-optimization/83459
8594 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8596 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8598 PR tree-optimization/84224
8599 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8600 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8603 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8606 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8607 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8608 count, remove one USE.
8610 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8612 * doc/install.texi (Configuration): Document the
8613 --with-long-double-format={ibm,ieee} PowerPC configuration
8617 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8618 Convert from define_expand to be define_insn_and_split. Rework
8619 float/double/_Float128 conversions to QI/HI/SImode to work with
8620 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8621 conversions to QI/HImode types did a store and then a load to
8622 truncate the value. For conversions to VSX registers, don't split
8623 the insn, instead emit the code directly. Use the code iterator
8624 any_fix to combine signed and unsigned conversions.
8625 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8626 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8627 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8628 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8629 (fix_<mode>di2_hw): Likewise.
8630 (fixuns_<mode>di2_hw): Likewise.
8631 (fix_<mode>si2_hw): Likewise.
8632 (fixuns_<mode>si2_hw): Likewise.
8633 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8634 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8635 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8636 fix<uns>_trunc<SFDF:mode>si2_p8.
8637 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8639 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8640 (fix<uns>_<mode>_mem): Likewise.
8641 (fctiw<u>z_<mode>_mem): Likewise.
8642 (fix<uns>_<mode>_mem): Likewise.
8643 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8644 the register allocator from doing a direct move to the GPRs to do
8645 a store, and instead use the ISA 3.0 store byte/half-word from
8646 vector register instruction. For IEEE 128-bit floating point,
8647 also optimize stores of 32-bit ints.
8648 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8650 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8652 * genextract.c (push_pathstr_operand): New function to support
8654 (walk_rtx): Call push_pathstr_operand.
8655 (print_path): Support [a-zA-Z].
8657 2018-02-07 Richard Biener <rguenther@suse.de>
8659 PR tree-optimization/84037
8660 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8661 (cse_and_gimplify_to_preheader): Declare.
8662 (vect_get_place_in_interleaving_chain): Likewise.
8663 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8665 (_loop_vec_info::~_loop_vec_info): Delete it.
8666 (cse_and_gimplify_to_preheader): New function.
8667 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8668 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8669 (vectorizable_load): Likewise. For grouped stores always base
8670 the IV on the first element.
8671 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8672 condition before gimplifying.
8674 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8676 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8677 *DIV_EXPR and *MOD_EXPR.
8679 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8682 * config/i386/i386.c (ix86_option_override_internal): Mask out
8683 the CF_SET bit when checking -fcf-protection.
8685 2018-02-07 Tom de Vries <tom@codesourcery.com>
8688 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8691 2018-02-07 Richard Biener <rguenther@suse.de>
8693 PR tree-optimization/84204
8694 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8697 PR tree-optimization/84205
8698 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8699 special-case isl_ast_op_zdiv_r.
8701 PR tree-optimization/84223
8702 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8703 Only add conditions from within the region.
8704 (gather_bbs::after_dom_children): Adjust.
8706 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8709 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8710 * config/avr/avr.md: Only post-reload split REG-REG moves if
8711 either register is GENERAL_REG_P.
8713 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8715 PR tree-optimization/84235
8716 * tree-ssa-scopedtables.c
8717 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8718 if the subtraction is performed in floating point type where NaNs are
8719 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8720 build 1. Formatting fix.
8722 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8725 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8726 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8727 and skip it regardless of bb boundaries. Use CALL_P macro,
8728 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8729 unnecessarily, formatting fix.
8731 2018-02-06 Michael Collison <michael.collison@arm.com>
8733 * config/arm/thumb2.md:
8734 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8735 (*thumb_mov_notscc): Ditto.
8737 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8740 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8741 unsigned_fix, not "s".
8743 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8745 * configure.ac (gcc_fn_eh_frame_ro): New function.
8746 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8747 correct .eh_frame permissions.
8748 * configure: Regenerate.
8750 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8752 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8755 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8757 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8758 Display warning message for -mno-speculate-indirect-jumps.
8760 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8762 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8764 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8766 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8768 PR tree-optimization/84225
8769 * tree-eh.c (find_trapping_overflow): Only call
8770 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8772 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8775 * config/i386/i386.c: Reimplement the check of possible options
8776 -mibt/-mshstk conbination. Change error messages.
8777 * doc/invoke.texi: Fix a typo: remove extra '='.
8779 2018-02-06 Marek Polacek <polacek@redhat.com>
8781 PR tree-optimization/84228
8782 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8784 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8787 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8788 emitted arch directives.
8789 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8790 __ARM_FEATURE_COPROC before changing architectures.
8792 2018-02-06 Richard Biener <rguenther@suse.de>
8794 * config/i386/i386.c (print_reg): Fix typo.
8795 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8797 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8799 * configure: Regenerate.
8801 2018-02-05 Martin Sebor <msebor@redhat.com>
8803 PR tree-optimization/83369
8804 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8807 2018-02-05 Martin Liska <mliska@suse.cz>
8809 * doc/invoke.texi: Cherry-pick upstream r323995.
8811 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8813 * ira.c (ira_init_register_move_cost): Adjust comment.
8815 2018-02-05 Martin Liska <mliska@suse.cz>
8817 PR gcov-profile/84137
8818 * doc/gcov.texi: Fix typo in documentation.
8820 2018-02-05 Martin Liska <mliska@suse.cz>
8822 PR gcov-profile/83879
8823 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8824 using dlopen functionality.
8826 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8828 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8829 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8830 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8831 _mm_maskz_range_round_ss): New intrinsics.
8832 (__builtin_ia32_rangesd128_round)
8833 (__builtin_ia32_rangess128_round): Remove.
8834 (__builtin_ia32_rangesd128_mask_round,
8835 __builtin_ia32_rangess128_mask_round): New builtins.
8836 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8837 __builtin_ia32_rangess128_round): Remove.
8838 (__builtin_ia32_rangesd128_mask_round,
8839 __builtin_ia32_rangess128_mask_round): New builtins.
8840 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8841 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8842 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8843 "<round_saeonly_constraint>")): Changed to ...
8844 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8845 "<round_saeonly_scalar_constraint>")): ... this.
8846 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8847 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8848 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8849 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8850 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8852 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8854 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8856 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8857 Remove all values except native, 8540 and 8548.
8859 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8861 * config/i386/i386.c (ix86_output_function_return): Pass
8862 INVALID_REGNUM, instead of -1, as invalid register number to
8863 indirect_thunk_name and output_indirect_thunk.
8865 2018-02-02 Julia Koval <julia.koval@intel.com>
8867 * config.gcc: Add -march=icelake.
8868 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8869 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8870 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8871 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8872 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8873 (processor_target_table): Add icelake.
8874 (ix86_option_override_internal): Handle new PTAs.
8875 (get_builtin_code_for_version): Handle icelake.
8876 (M_INTEL_COREI7_ICELAKE): New.
8877 (fold_builtin_cpu): Handle icelake.
8878 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8879 * doc/invoke.texi: Add -march=icelake.
8881 2018-02-02 Julia Koval <julia.koval@intel.com>
8883 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8884 to wide_int_bitmask.
8885 * wide-int-bitmask.h: New.
8887 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8890 * config/i386/i386.md: Replace Pmode with word_mode in
8891 builtin_setjmp_setup and builtin_longjmp to support x32.
8893 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8897 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8899 (rs6000_supported_cpu_names): New static variable.
8900 (linux_cpu_translation_table): Likewise.
8901 (elf_platform) <cpu>: Define new static variable and use it.
8902 Translate kernel AT_PLATFORM name to canonical name if needed.
8903 Error if platform name is unknown.
8905 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8908 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8910 2018-02-01 Jeff Law <law@redhat.com>
8913 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8914 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8915 the scratch if RELEASE_VIA_POP is false.
8916 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8917 If we have to save a temporary register, decrement SIZE appropriately.
8918 Pass new arguments to release_scratch_register_on_entry.
8919 (ix86_adjust_stack_and_probe): Likewise.
8920 (ix86_emit_probe_stack_range): Pass new arguments to
8921 release_scratch_register_on_entry.
8923 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8925 PR rtl-optimization/84157
8926 * combine.c (change_zero_ext): Use REG_P predicate in
8927 front of HARD_REGISTER_P predicate.
8929 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8931 * config/avr/avr.c (avr_option_override): Move disabling of
8932 -fdelete-null-pointer-checks to...
8933 * common/config/avr/avr-common.c (avr_option_optimization_table):
8936 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8938 PR tree-optimization/81635
8939 * tree-data-ref.c (split_constant_offset_1): For types that
8940 wrap on overflow, try to use range info to prove that wrapping
8943 2018-02-01 Renlin Li <renlin.li@arm.com>
8946 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8948 (aarch64_register_move_cost): Likewise.
8949 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8951 (REG_CLASS_NAMES): Likewise.
8952 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8953 TAILCALL_ADDR_REGS. Remove IP registers.
8954 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8956 2018-02-01 Richard Biener <rguenther@suse.de>
8958 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8959 for specifying RPO order and allow NULL for that.
8960 * domwalk.c (dom_walker::dom_walker): Likewise.
8961 (dom_walker::walk): Handle NULL RPO order.
8962 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8964 (rewrite_update_dom_walker): Likewise.
8965 (mark_def_dom_walker): Likewise.
8967 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8969 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8970 (aarch64_maybe_expand_sve_subreg_move): Declare.
8971 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8972 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8974 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8975 that are semantically a reverse operation.
8976 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8977 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8978 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8980 (aarch64_can_change_mode_class): For big-endian, forbid changes
8981 between two SVE modes if they have different element sizes.
8983 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8985 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8986 the TImode handling for big-endian targets.
8988 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8990 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8991 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8993 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8994 Remove BSWAP handing for big-endian targets and use the form of
8995 LD1RQ appropariate for the mode.
8997 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8999 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9000 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9003 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9006 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9007 check for operands that need to go through aarch64_sve_reload_be.
9009 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9011 PR tree-optimization/81661
9012 PR tree-optimization/84117
9013 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9014 * tree-eh.c: Include gimplify.h.
9015 (find_trapping_overflow, replace_trapping_overflow,
9016 rewrite_to_non_trapping_overflow): New functions.
9017 * tree-vect-loop.c: Include tree-eh.h.
9018 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9019 * tree-data-ref.c: Include tree-eh.h.
9020 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9022 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9024 PR rtl-optimization/84123
9025 * combine.c (change_zero_ext): Check if hard register satisfies
9026 can_change_dest_mode before calling gen_lowpart_SUBREG.
9028 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9031 * ira.c (ira_init_register_move_cost): Remove assert.
9033 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9035 PR rtl-optimization/84071
9036 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9037 * doc/tm.texi: Regenerate.
9039 2018-01-31 Richard Biener <rguenther@suse.de>
9041 PR tree-optimization/84132
9042 * tree-data-ref.c (analyze_miv_subscript): Properly
9043 check whether evolution_function_is_affine_multivariate_p
9044 before calling gcd_of_steps_may_divide_p.
9046 2018-01-31 Julia Koval <julia.koval@intel.com>
9049 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9050 * config/i386/i386.md (rdpid_rex64) New.
9051 (rdpid): Make 32bit only.
9053 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9056 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9057 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9059 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9062 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9064 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9066 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9068 PR rtl-optimization/84071
9069 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9070 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9072 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9074 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9075 (arc_attribute_table): Add 'aux' attribute.
9076 (arc_in_small_data_p): Consider aux like variables.
9077 (arc_is_aux_reg_p): New function.
9078 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9079 (arc_get_aux_arg): New function.
9080 (prepare_move_operands): Handle aux-register access.
9081 (arc_handle_aux_attribute): New function.
9082 * doc/extend.texi (ARC Variable attributes): Add subsection.
9084 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9086 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9087 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9088 (arc_attribute_table): Add 'uncached' attribute.
9089 (arc_print_operand): Print '.di' flag for uncached memory
9091 (arc_in_small_data_p): Do not consider for small data the uncached
9093 (arc_is_uncached_mem_p): New function.
9094 * config/arc/predicates.md (compact_store_memory_operand): Check
9095 for uncached memory accesses.
9096 (nonvol_nonimm_operand): Likewise.
9097 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9099 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9102 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9103 falign-loops=): Add Optimization flag.
9105 2018-01-30 Jeff Law <law@redhat.com>
9108 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9109 INT_REGISTERS_SAVED. Check it prior to calling
9110 get_scratch_register_on_entry.
9111 (ix86_adjust_stack_and_probe): Similarly.
9112 (ix86_emit_probe_stack_range): Similarly.
9113 (ix86_expand_prologue): Corresponding changes.
9115 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9118 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9119 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9121 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9124 * lra-constraints.c (curr_insn_transform): Process AND in the
9127 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9129 PR rtl-optimization/83986
9130 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9131 dependence against last_pending_memory_flush in addition to
9134 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9136 PR tree-optimization/81611
9137 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9140 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9143 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9146 2018-01-30 Richard Biener <rguenther@suse.de>
9147 Jakub Jelinek <jakub@redhat.com>
9149 PR tree-optimization/84111
9150 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9151 inner loops added during recursion, as they don't have up-to-date
9154 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9157 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9158 (can_inline_edge_by_limits_p): ... here.
9159 (can_early_inline_edge_p, check_callers,
9160 update_caller_keys, update_callee_keys, recursive_inlining,
9161 add_new_edges_to_heap, speculation_useful_p,
9162 inline_small_functions,
9163 inline_small_functions, flatten_function,
9164 inline_to_all_callers_1): Update.
9166 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9168 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9171 2018-01-30 Richard Biener <rguenther@suse.de>
9173 PR tree-optimization/83008
9174 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9175 invariant and constant vector uses in stmts when they need
9178 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9181 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9182 * configure: Regenerate.
9184 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9186 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9188 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9189 Use gen_rtx_REG rather than gen_lowpart.
9191 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9193 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9194 rather than 0 when creating partial subregs.
9196 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9198 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9201 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9204 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9205 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9206 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9207 flags. This restores the settings used before the 2017-07-24.
9208 Turning off pre increment/decrement/modify allows IVOPTS to
9209 optimize DF/SF loops where the index is an int.
9211 2018-01-29 Richard Biener <rguenther@suse.de>
9212 Kelvin Nilsen <kelvin@gcc.gnu.org>
9215 * tree-vect-stmts.c (vectorizable_call): Don't call
9216 targetm.vectorize_builtin_md_vectorized_function if callee is
9219 2018-01-22 Carl Love <cel@us.ibm.com>
9221 * doc/extend.tex: Fix typo in second arg in
9222 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9224 2018-01-29 Richard Biener <rguenther@suse.de>
9226 PR tree-optimization/84086
9227 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9228 (flush_ssaname_freelist): When SSA names were released reset
9229 the SCEV hash table.
9231 2018-01-29 Richard Biener <rguenther@suse.de>
9233 PR tree-optimization/84057
9234 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9235 removed paths when removing edges.
9237 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9239 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9240 -mfunction-return=@var{choice}.
9242 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9245 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9247 (layout::print_source_line): Likewise.
9248 (test_get_line_width_without_trailing_whitespace): Add test cases.
9250 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9253 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9256 2018-01-26 Jim Wilson <jimw@sifive.com>
9258 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9260 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9263 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9265 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9266 and CMP + SUB-immediate -> SUBS.
9268 2018-01-26 Martin Sebor <msebor@redhat.com>
9270 PR tree-optimization/83896
9271 * tree-ssa-strlen.c (get_string_len): Rename...
9272 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9273 Avoid assuming length is constant.
9274 (handle_char_store): Use HOST_WIDE_INT for string length.
9276 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9279 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9280 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9282 2018-01-26 Richard Biener <rguenther@suse.de>
9284 PR rtl-optimization/84003
9285 * dse.c (record_store): Only record redundant stores when
9286 the earlier store aliases at least all accesses the later one does.
9288 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9290 PR rtl-optimization/83985
9291 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9292 REG_CFA_RESTORE insns.
9293 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9294 REG_CFA_RESTORE insns here.
9297 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9298 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9300 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9302 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9303 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9304 (arc_init): Likewise.
9305 (arc_override_options): Likewise.
9306 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9308 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9310 * config/arc/arc.h (TARGET_DBNZ): Define.
9311 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9312 properly set the tune attribute.
9313 (dbnz): Use TARGET_DBNZ guard.
9314 * config/arc/arc.opt (mtune): Add core3 option.
9316 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9318 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9319 recognize new pic like addresses.
9320 (arc_delegitimize_address): Clean up.
9322 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9324 * config/arc/arc-arches.def: Option mrf16 valid for all
9326 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9327 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9328 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9329 * config/arc/arc-tables.opt: Regenerate.
9330 * config/arc/arc.c (arc_conditional_register_usage): Handle
9331 reduced register file case.
9332 (arc_file_start): Set must have build attributes.
9333 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9335 * config/arc/arc.opt (mrf16): Add new option.
9336 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9337 * config/arc/genmultilib.awk: Handle new mrf16 option.
9338 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9339 * config/arc/t-multilib: Regenerate.
9340 * doc/invoke.texi (ARC Options): Document mrf16 option.
9342 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9344 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9345 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9346 (arc_attribute_table): Add 'secure_call' attribute.
9347 (arc_print_operand): Print secure call operand.
9348 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9350 (arc_is_secure_call_p): New function. * config/arc/arc.md
9351 (call_i): Add support for sjli instruction.
9352 (call_value_i): Likewise.
9353 * config/arc/constraints.md (Csc): New constraint.
9355 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9356 John Eric Martin <John.Martin@emmicro-us.com>
9358 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9359 * config/arc/arc.c (_arc_jli_section): New struct.
9360 (arc_jli_section): New type.
9361 (rc_jli_sections): New static variable.
9362 (arc_handle_jli_attribute): New function.
9363 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9364 (arc_file_end): New function.
9365 (TARGET_ASM_FILE_END): Define.
9366 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9367 (arc_add_jli_section): New function.
9368 (jli_call_scan): Likewise.
9369 (arc_reorg): Call jli_call_scan.
9370 (arc_output_addsi): Remove 'S' from printing asm operand.
9371 (arc_is_jli_call_p): New function.
9372 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9374 (movhi_insn): Likewise.
9375 (movsi_insn): Likewise.
9376 (movsi_set_cc_insn): Likewise.
9377 (loadqi_update): Likewise.
9378 (load_zeroextendqisi_update): Likewise.
9379 (load_signextendqisi_update): Likewise.
9380 (loadhi_update): Likewise.
9381 (load_zeroextendhisi_update): Likewise.
9382 (load_signextendhisi_update): Likewise.
9383 (loadsi_update): Likewise.
9384 (loadsf_update): Likewise.
9385 (movsicc_insn): Likewise.
9386 (bset_insn): Likewise.
9387 (bxor_insn): Likewise.
9388 (bclr_insn): Likewise.
9389 (bmsk_insn): Likewise.
9390 (bicsi3_insn): Likewise.
9391 (cmpsi_cc_c_insn): Likewise.
9392 (movsi_ne): Likewise.
9393 (movsi_cond_exec): Likewise.
9394 (clrsbsi2): Likewise.
9405 (call_i): Remove 'S' asm letter, add jli instruction.
9406 (call_value_i): Likewise.
9407 * config/arc/arc.op (mjli-always): New option.
9408 * config/arc/constraints.md (Cji): New constraint.
9409 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9411 (subsf3_fpx): Likewise.
9412 (mulsf3_fpx): Likewise.
9413 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9415 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9417 * doc/invoke.texi (ARC): Document mjli-always option.
9419 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9421 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9422 avoid addition with 0 and use incw and decw where possible.
9424 2018-01-26 Richard Biener <rguenther@suse.de>
9426 PR tree-optimization/81082
9427 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9428 association if it requires casting to unsigned.
9429 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9430 from fold_plusminus_mult_expr to catch important cases late when
9431 range info is available.
9433 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9435 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9436 * configure.ac (hidden_linkonce): New test.
9437 * configure: Regenerate.
9438 * config.in: Regenerate.
9440 2018-01-26 Julia Koval <julia.koval@intel.com>
9442 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9443 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9444 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9445 _mm_mask_bitshuffle_epi64_mask): Fix type.
9446 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9447 USI_FTYPE_V4DI_V4DI_USI): Remove.
9448 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9449 __builtin_ia32_vpshufbitqmb256_mask,
9450 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9451 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9452 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9454 2018-01-26 Alan Modra <amodra@gmail.com>
9457 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9458 UNSPEC_VBPERMQ. Sort other unspecs.
9460 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9462 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9464 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9467 * predict.c (drop_profile): Do not push/pop cfun; update also
9469 (handle_missing_profiles): Fix logic looking for zero profiles.
9471 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9474 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9475 on functions with #pragma omp declare simd or functions with simd
9477 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9478 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9479 Remove trailing \n from warning_at calls.
9481 2018-01-25 Tom de Vries <tom@codesourcery.com>
9484 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9485 for neutered workers.
9487 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9490 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9491 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9493 2018-01-24 Jeff Law <law@redhat.com>
9496 * i386.c (get_probe_interval): Move to earlier point.
9497 (ix86_compute_frame_layout): If -fstack-clash-protection and
9498 the frame is larger than the probe interval, then use pushes
9499 to save registers rather than reg->mem moves.
9500 (ix86_expand_prologue): Remove conditional for int_registers_saved
9503 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9506 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9507 min/max for never referenced object.
9509 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9512 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9514 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9515 attributes from DECL_ATTRIBUTES (decl) without affecting
9516 DECL_ATTRIBUTES (current_function_decl).
9517 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9518 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9520 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9522 PR tree-optimization/83979
9523 * fold-const.c (fold_comparison): Use constant_boolean_node
9524 instead of boolean_{true,false}_node.
9526 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9528 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9531 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9533 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9534 Simplify the clause that sets the length attribute.
9535 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9536 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9537 clause that sets the length attribute.
9538 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9540 2018-01-24 Tom de Vries <tom@codesourcery.com>
9543 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9544 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9545 Add strict parameter.
9546 (prevent_branch_around_nothing): Insert dummy insn between branch to
9547 label and label with no ptx insn inbetween.
9548 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9550 2018-01-24 Tom de Vries <tom@codesourcery.com>
9553 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9554 for neutered threads in warp.
9555 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9557 2018-01-24 Richard Biener <rguenther@suse.de>
9559 PR tree-optimization/83176
9560 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9563 2018-01-24 Richard Biener <rguenther@suse.de>
9565 PR tree-optimization/82819
9566 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9567 code generating pluses that are no-ops in the target precision.
9569 2018-01-24 Richard Biener <rguenther@suse.de>
9572 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9574 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9576 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9577 to merge probabilities.
9578 * predict.c (probably_never_executed): Also mark as cold functions
9579 with global 0 profile and guessed local profile.
9580 * profile-count.c (profile_probability::combine_with_count): New
9582 * profile-count.h (profile_probability::operator*,
9583 profile_probability::operator*=, profile_probability::operator/,
9584 profile_probability::operator/=): Reduce precision to adjusted
9585 and set value to guessed on contradictory divisions.
9586 (profile_probability::combine_with_freq): Remove.
9587 (profile_probability::combine_wiht_count): Declare.
9588 (profile_count::force_nonzero):: Set to adjusted.
9589 (profile_count::probability_in):: Set quality to adjusted.
9590 * tree-ssa-tail-merge.c (replace_block_by): Use
9593 2018-01-23 Andrew Waterman <andrew@sifive.com>
9594 Jim Wilson <jimw@sifive.com>
9596 * config/riscv/riscv.c (riscv_stack_boundary): New.
9597 (riscv_option_override): Set riscv_stack_boundary. Handle
9598 riscv_preferred_stack_boundary_arg.
9599 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9600 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9601 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9602 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9603 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9604 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9606 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9609 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9610 of struct ix86_frame.
9611 (ix86_expand_epilogue): Likewise. Add a local variable for
9612 the reg_save_offset field in struct ix86_frame.
9614 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9616 PR tree-optimization/82604
9617 * tree-loop-distribution.c (enum partition_kind): New enum item
9618 PKIND_PARTIAL_MEMSET.
9619 (partition_builtin_p): Support above new enum item.
9620 (generate_code_for_partition): Ditto.
9621 (compute_access_range): Differentiate cases that equality can be
9622 proven at all loops, the innermost loops or no loops.
9623 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9624 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9625 (finalize_partitions, distribute_loop): Don't fuse partition of
9626 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9627 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9630 2018-01-23 Martin Liska <mliska@suse.cz>
9632 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9633 order to ignore the predictor.
9634 (PRED_POLYMORPHIC_CALL): Likewise.
9635 (PRED_RECURSIVE_CALL): Likewise.
9637 2018-01-23 Martin Liska <mliska@suse.cz>
9639 * tree-profile.c (tree_profiling): Print function header to
9640 aware reader which function we are working on.
9641 * value-prof.c (gimple_find_values_to_profile): Do not print
9642 not interesting value histograms.
9644 2018-01-23 Martin Liska <mliska@suse.cz>
9646 * profile-count.h (enum profile_quality): Add
9647 profile_uninitialized as the first value. Do not number values
9648 as they are zero based.
9649 (profile_count::verify): Update sanity check.
9650 (profile_probability::verify): Likewise.
9652 2018-01-23 Nathan Sidwell <nathan@acm.org>
9654 * doc/invoke.texi (ffor-scope): Deprecate.
9656 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9658 PR tree-optimization/83510
9659 * domwalk.c (set_all_edges_as_executable): New function.
9660 (dom_walker::dom_walker): Convert bool param
9661 "skip_unreachable_blocks" to enum reachability. Move setup of
9662 edge flags to set_all_edges_as_executable and only do it when
9663 reachability is REACHABLE_BLOCKS.
9664 * domwalk.h (enum dom_walker::reachability): New enum.
9665 (dom_walker::dom_walker): Convert bool param
9666 "skip_unreachable_blocks" to enum reachability.
9667 (set_all_edges_as_executable): New decl.
9668 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9669 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9671 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9672 but converting true to REACHABLE_BLOCKS.
9673 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9675 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9676 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9677 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9679 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9680 if check_all_array_refs will be called.
9682 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9684 * tree.c (selftest::test_location_wrappers): Add more test
9687 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9689 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9690 (selftest::test_bit_in_range): Likewise.
9692 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9695 * doc/sourcebuild.texi (vect_float): Say that the selector
9696 only describes the situation when -funsafe-math-optimizations is on.
9697 (vect_float_strict): Document.
9699 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9701 PR tree-optimization/83965
9702 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9703 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9704 instead of checking only for a reduction.
9705 (vect_recog_widen_sum_pattern): Likewise.
9707 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9709 * predict.c (probably_never_executed): Only use precise profile info.
9710 (compute_function_frequency): Skip after inlining hack since we now
9711 have quality checking.
9713 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9715 * profile-count.h (profile_probability::very_unlikely,
9716 profile_probability::unlikely, profile_probability::even): Set
9717 precision to guessed.
9719 2018-01-23 Richard Biener <rguenther@suse.de>
9721 PR tree-optimization/83963
9722 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9723 Properly terminate dominator walk when crossing the exit edge not
9724 when visiting its source block.
9726 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9729 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9730 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9732 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9734 PR tree-optimization/83957
9735 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9736 semicolon after for body surrounded by braces.
9738 PR tree-optimization/83081
9739 * profile-count.h (profile_probability::split): New method.
9740 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9741 Use profile_probability::split.
9742 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9743 when splitting a single conditional jump into 2.
9745 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9747 PR tree-optimization/69452
9748 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9751 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9753 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9754 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9755 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9757 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9759 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9760 * config/rl78/rl78.md (movdi): New define_expand.
9761 * config/rl78/rl78.c (rl78_split_movdi): New function.
9763 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9766 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9768 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9769 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9770 128-bit to produce an UNSPEC move to get the double word with the
9771 signbit and then a shift directly to do signbit.
9772 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9773 implementation with a new version that just does either a direct
9774 move or a regular move. Move memory interface to separate insns.
9775 Move insns so they are next to the expander.
9776 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9777 with signbit move. Split big and little endian case.
9778 (signbit<mode>2_dm_mem_le): Likewise.
9779 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9780 (signbit<mode>2_dm2): Likewise.
9782 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9784 * config/rl78/rl78.md (anddi3): New define_expand.
9786 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9788 * config/rl78/rl78.md (umindi3): New define_expand.
9790 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9792 * config/rl78/rl78.md (smindi3): New define_expand.
9794 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9796 * config/rl78/rl78.md (smaxdi3): New define_expand.
9798 2018-01-22 Carl Love <cel@us.ibm.com>
9800 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9801 LVX_V1TI): Add macro expansion.
9802 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9803 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9804 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9805 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9806 Change check to determine if the instruction is a byte reversing
9807 entry. Fix typo in comment.
9808 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9809 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9810 Add def_builtin calls for new builtins.
9811 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9812 Add define_insn expansion.
9814 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9816 * config/rl78/rl78.md (umaxdi3): New define_expand.
9818 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9820 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9821 for non-QImode registers.
9823 2018-01-22 Richard Biener <rguenther@suse.de>
9825 PR tree-optimization/83963
9826 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9827 including the loop exit block.
9828 (scop_detection::merge_sese): Likewise.
9829 (scop_detection::add_scop): Do it here instead.
9831 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9833 * doc/sourcebuild.texi (arm_softfloat): Document.
9835 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9838 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9839 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9840 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9842 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9843 David Edelsohn <dje.gcc@gmail.com>
9846 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9847 Change "crset eq" to "crset 2".
9848 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9849 (*call_indirect_aix<mode>_nospec): Likewise.
9850 (*call_value_indirect_aix<mode>_nospec): Likewise.
9851 (*call_indirect_elfv2<mode>_nospec): Likewise.
9852 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9853 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9854 change assembly output from . to $.
9855 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9856 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9857 (*tablejump<mode>_internal1_nospec): Likewise.
9859 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9862 * config/sh/sh_optimize_sett_clrt.cc:
9863 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9865 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9867 PR tree-optimization/83940
9868 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9869 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9870 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9871 use it to pass back the definition type.
9872 (vect_check_store_rhs): Likewise rhs_dt_out.
9873 (vect_build_gather_load_calls): Add a mask_dt argument and use
9874 it instead of a call to vect_is_simple_use.
9875 (vectorizable_store): Update calls to vect_check_load_store_mask
9876 and vect_check_store_rhs. Use the dt returned by the latter instead
9877 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9878 instead of calls to vect_is_simple_use. Pass the scalar rather
9879 than the vector operand to vect_is_simple_use when handling
9880 second and subsequent copies of an rhs value.
9881 (vectorizable_load): Update calls to vect_check_load_store_mask
9882 and vect_build_gather_load_calls. Use the cached mask_dt and
9883 gs_info.offset_dt instead of calls to vect_is_simple_use.
9885 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9888 * tree-emutls.c: Include gimplify.h.
9889 (lower_emutls_2): New function.
9890 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9891 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9892 it before further processing.
9895 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9896 UINTVAL (trueop1) instead of INTVAL (op1).
9898 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9902 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9903 INCOMING_FRAME_SP_OFFSET if not defined.
9904 (scan_trace): Add ENTRY argument. If true and
9905 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9906 emit a note to adjust the CFA offset.
9907 (create_cfi_notes): Adjust scan_trace callers.
9908 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9909 INCOMING_FRAME_SP_OFFSET in the CIE.
9910 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9911 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9913 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9914 * doc/tm.texi: Regenerated.
9916 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9918 PR rtl-optimization/83147
9919 * lra-constraints.c (remove_inheritance_pseudos): Use
9920 lra_substitute_pseudo_within_insn.
9922 2018-01-19 Tom de Vries <tom@codesourcery.com>
9923 Cesar Philippidis <cesar@codesourcery.com>
9926 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9928 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9931 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9932 spaces for function labels.
9934 2018-01-19 Martin Liska <mliska@suse.cz>
9936 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9937 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9938 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9939 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9940 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9941 (PRED_CONST_RETURN): Change from 69 to 65.
9942 (PRED_NULL_RETURN): Change from 91 to 71.
9943 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9944 (PRED_LOOP_GUARD): Change from 66 to 73.
9946 2018-01-19 Martin Liska <mliska@suse.cz>
9948 * predict.c (predict_insn_def): Add new assert.
9949 (struct branch_predictor): Change type to signed integer.
9950 (test_prediction_value_range): Amend test to cover
9952 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9953 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9954 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9955 (PRED_LOOP_IV_COMPARE): Likewise.
9956 * predict.h (PROB_UNINITIALIZED): Define new constant.
9958 2018-01-19 Martin Liska <mliska@suse.cz>
9960 * predict.c (dump_prediction): Add new format for
9961 analyze_brprob.py script which is enabled with -details
9963 * profile-count.h (precise_p): New function.
9965 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9967 PR tree-optimization/83922
9968 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9969 there are no statements that need masking.
9970 (vect_active_double_reduction_p): New function.
9971 (vect_analyze_loop_operations): Use it when handling phis that
9972 are not in the loop header.
9974 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9976 PR tree-optimization/83914
9977 * tree-vect-loop.c (vectorizable_induction): Don't convert
9978 init_expr or apply the peeling adjustment for inductions
9979 that are nested within the vectorized loop.
9981 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9983 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9986 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9990 * function.h (gimplify_parameters): Add gimple_seq * argument.
9991 * function.c: Include gimple.h and options.h.
9992 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9993 for the added local temporaries if needed.
9994 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9995 if there are any parameter cleanups, wrap whole body into a
9996 try/finally with the cleanups.
9998 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10001 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10002 Use GET_MODE_CLASS for scalar floating point.
10004 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10008 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10009 Fix call of call_cgraph_insertion_hooks.
10011 2018-01-18 Martin Sebor <msebor@redhat.com>
10013 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10015 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10018 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10021 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10024 * common.opt: (-ffile-prefix-map): New option.
10025 * opts.c (common_handle_option): Defer it.
10026 * opts-global.c (handle_common_deferred_options): Handle it.
10027 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10028 * file-prefix-map.h: New file.
10029 (remap_debug_filename, add_debug_prefix_map): ...here.
10030 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10031 * final.c (debug_prefix_map, add_debug_prefix_map
10032 remap_debug_filename): Move to...
10033 * file-prefix-map.c: New file.
10034 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10035 generalize, get rid of alloca(), use strrchr() instead of strchr().
10036 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10037 Implement in terms of add_prefix_map().
10038 (remap_macro_filename, remap_debug_filename): Implement in term of
10040 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10041 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10042 * dbxout.c: Include file-prefix-map.h.
10043 * varasm.c: Likewise.
10044 * vmsdbgout.c: Likewise.
10045 * xcoffout.c: Likewise.
10046 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10047 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10048 * doc/invoke.texi (-ffile-prefix-map): Document.
10049 (-fdebug-prefix-map): Update description.
10051 2018-01-18 Martin Liska <mliska@suse.cz>
10053 * config/i386/i386.c (indirect_thunk_name): Document that also
10055 (output_indirect_thunk): Document why both instructions
10056 (pause and lfence) are generated.
10058 2018-01-18 Richard Biener <rguenther@suse.de>
10060 PR tree-optimization/83887
10061 * graphite-scop-detection.c
10062 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10063 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10064 (scop_detection::merge_sese): Re-implement with a flood-fill
10065 algorithm that properly finds a SESE region if it exists.
10067 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10070 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10071 pointer_diff optimizations use view_convert instead of convert.
10073 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10075 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10076 Generate different code for -mno-speculate-indirect-jumps.
10077 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10078 (*call_indirect_aix<mode>): Disable for
10079 -mno-speculate-indirect-jumps.
10080 (*call_indirect_aix<mode>_nospec): New define_insn.
10081 (*call_value_indirect_aix<mode>): Disable for
10082 -mno-speculate-indirect-jumps.
10083 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10084 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10085 -mno-speculate-indirect-jumps.
10086 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10088 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10090 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10091 long double type, set the flags for noting the default long double
10092 type, even if we don't pass or return a long double type.
10094 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10097 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10100 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10102 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10103 support for merge[hl].
10104 (fold_mergehl_helper): New helper function.
10105 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10106 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10107 (altivec_vmrglw_direct): Add xxmrglw insn.
10109 2018-01-17 Andrew Waterman <andrew@sifive.com>
10111 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10112 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10114 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10117 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10118 call the lto_location_cache before reading the
10119 DECL_SOURCE_LOCATION of the types.
10121 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10122 Richard Sandiford <richard.sandiford@linaro.org>
10124 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10125 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10126 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10127 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10128 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10130 * config/aarch64/constraints.md (aarch64_movti_operand):
10132 * config/aarch64/predicates.md (Uti): Add new constraint.
10134 2018-01-17 Carl Love <cel@us.ibm.com>
10136 * config/rs6000/vsx.md (define_expand xl_len_r,
10137 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10138 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10140 (define_expand, define_insn): Move the shift left from the
10141 define_insn to the define_expand for lxvl and stxvl instructions.
10142 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10143 and XL_LEN_R definitions to PURE.
10145 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10147 * config/i386/i386.c (indirect_thunk_name): Declare regno
10148 as unsigned int. Compare regno with INVALID_REGNUM.
10149 (output_indirect_thunk): Ditto.
10150 (output_indirect_thunk_function): Ditto.
10151 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10152 in the call to output_indirect_thunk_function.
10154 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10156 PR middle-end/83884
10157 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10158 rather than the size of inner_type to determine the stack slot size
10159 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10161 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10164 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10167 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10169 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10170 endian Linux systems to optionally enable multilibs for selecting
10171 the long double type if the user configured an explicit type.
10172 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10173 have no long double multilibs if not defined.
10174 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10175 warn if the user used -mabi={ieee,ibm}longdouble and we built
10176 multilibs for long double.
10177 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10178 appropriate multilib option.
10179 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10181 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10182 for building long double multilibs.
10183 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10185 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10187 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10190 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10192 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10195 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10198 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10201 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10203 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10204 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10206 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10208 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10209 different rtl trees depending on TARGET_64BIT.
10210 (rs6000_gen_lvx): Likewise.
10212 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10214 * config/visium/visium.md (nop): Tweak comment.
10215 (hazard_nop): Likewise.
10217 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10219 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10220 -mspeculate-indirect-jumps.
10221 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10222 for -mno-speculate-indirect-jumps.
10223 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10224 (*call_value_indirect_elfv2<mode>): Disable for
10225 -mno-speculate-indirect-jumps.
10226 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10227 (indirect_jump): Emit different RTL for
10228 -mno-speculate-indirect-jumps.
10229 (*indirect_jump<mode>): Disable for
10230 -mno-speculate-indirect-jumps.
10231 (*indirect_jump<mode>_nospec): New define_insn.
10232 (tablejump): Emit different RTL for
10233 -mno-speculate-indirect-jumps.
10234 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10235 (tablejumpsi_nospec): New define_expand.
10236 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10237 (tablejumpdi_nospec): New define_expand.
10238 (*tablejump<mode>_internal1): Disable for
10239 -mno-speculate-indirect-jumps.
10240 (*tablejump<mode>_internal1_nospec): New define_insn.
10241 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10244 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10246 * caller-save.c (insert_save): Drop unnecessary parameter. All
10249 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10250 Richard Biener <rguenth@suse.de>
10253 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10254 return early, inline manually is_gimple_sizepos. Make sure if we
10255 call gimplify_expr we don't end up with a gimple constant.
10256 * tree.c (variably_modified_type_p): Don't return true for
10257 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10258 * gimplify.h (is_gimple_sizepos): Remove.
10260 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10262 PR tree-optimization/83857
10263 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10264 vectorizable_live_operation for pure SLP statements.
10265 (vectorizable_live_operation): Handle PHIs.
10267 2018-01-16 Richard Biener <rguenther@suse.de>
10269 PR tree-optimization/83867
10270 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10271 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10273 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10276 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10277 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10278 If off is not INTEGER_CST, issue a may not be aligned warning
10279 rather than isn't aligned. Use isn%'t rather than isn't.
10280 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10282 <case MULT_EXPR>: Improve the case when bottom and one of the
10283 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10284 operand, in that case check if the other operand is multiple of
10285 bottom divided by the INTEGER_CST operand.
10287 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10290 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10291 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10292 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10293 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10294 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10295 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10296 (pa_function_arg_size): New function.
10298 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10300 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10301 in a separate statement.
10303 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10305 PR tree-optimization/83847
10306 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10307 group gathers and scatters.
10309 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10311 PR rtl-optimization/86620
10312 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10314 PR rtl-optimization/83213
10315 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10316 to last if both are JUMP_INSNs.
10318 PR tree-optimization/83843
10319 * gimple-ssa-store-merging.c
10320 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10321 store_immediate_info for bswap/nop orig_stores.
10323 2018-01-15 Andrew Waterman <andrew@sifive.com>
10325 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10327 <UDIV>: Increase cost if !TARGET_DIV.
10329 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10331 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10332 (define_attr "cr_logical_3op"): New.
10333 (cceq_ior_compare): Adjust.
10334 (cceq_ior_compare_complement): Adjust.
10335 (*cceq_rev_compare): Adjust.
10336 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10337 (is_cracked_insn): Adjust.
10338 (insn_must_be_first_in_group): Adjust.
10339 * config/rs6000/40x.md: Adjust.
10340 * config/rs6000/440.md: Adjust.
10341 * config/rs6000/476.md: Adjust.
10342 * config/rs6000/601.md: Adjust.
10343 * config/rs6000/603.md: Adjust.
10344 * config/rs6000/6xx.md: Adjust.
10345 * config/rs6000/7450.md: Adjust.
10346 * config/rs6000/7xx.md: Adjust.
10347 * config/rs6000/8540.md: Adjust.
10348 * config/rs6000/cell.md: Adjust.
10349 * config/rs6000/e300c2c3.md: Adjust.
10350 * config/rs6000/e500mc.md: Adjust.
10351 * config/rs6000/e500mc64.md: Adjust.
10352 * config/rs6000/e5500.md: Adjust.
10353 * config/rs6000/e6500.md: Adjust.
10354 * config/rs6000/mpc.md: Adjust.
10355 * config/rs6000/power4.md: Adjust.
10356 * config/rs6000/power5.md: Adjust.
10357 * config/rs6000/power6.md: Adjust.
10358 * config/rs6000/power7.md: Adjust.
10359 * config/rs6000/power8.md: Adjust.
10360 * config/rs6000/power9.md: Adjust.
10361 * config/rs6000/rs64.md: Adjust.
10362 * config/rs6000/titan.md: Adjust.
10364 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10366 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10367 ix86_indirect_branch_register logic.
10369 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10371 * config/i386/constraints.md (Bs): Update
10372 ix86_indirect_branch_register check. Don't check
10373 ix86_indirect_branch_register with GOT_memory_operand.
10375 * config/i386/predicates.md (GOT_memory_operand): Don't check
10376 ix86_indirect_branch_register here.
10377 (GOT32_symbol_operand): Likewise.
10379 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10381 * config/i386/predicates.md (constant_call_address_operand):
10382 Rewrite ix86_indirect_branch_register logic.
10383 (sibcall_insn_operand): Likewise.
10385 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10387 * config/i386/constraints.md (Bs): Replace
10388 ix86_indirect_branch_thunk_register with
10389 ix86_indirect_branch_register.
10391 * config/i386/i386.md (indirect_jump): Likewise.
10392 (tablejump): Likewise.
10393 (*sibcall_memory): Likewise.
10394 (*sibcall_value_memory): Likewise.
10395 Peepholes of indirect call and jump via memory: Likewise.
10396 * config/i386/i386.opt: Likewise.
10397 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10398 (GOT_memory_operand): Likewise.
10399 (call_insn_operand): Likewise.
10400 (sibcall_insn_operand): Likewise.
10401 (GOT32_symbol_operand): Likewise.
10403 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10405 PR middle-end/83837
10406 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10407 type rather than type addr's type points to.
10408 (expand_omp_atomic_mutex): Likewise.
10409 (expand_omp_atomic): Likewise.
10411 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10414 * config/i386/i386.c (output_indirect_thunk_function): Use
10415 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10416 for __x86_return_thunk.
10418 2018-01-15 Richard Biener <rguenther@suse.de>
10420 PR middle-end/83850
10421 * expmed.c (extract_bit_field_1): Fix typo.
10423 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10426 * config/arm/iterators.md (VF): New mode iterator.
10427 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10428 Remove integer-related logic from pattern.
10429 (neon_vabd<mode>_3): Likewise.
10431 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10433 PR middle-end/82694
10434 * common.opt (fstrict-overflow): No longer an alias.
10435 (fwrapv-pointer): New option.
10436 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10437 also for pointer types based on flag_wrapv_pointer.
10438 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10439 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10440 opts->x_flag_wrapv got set.
10441 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10442 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10443 POINTER_TYPE_OVERFLOW_UNDEFINED.
10444 * match.pd: Likewise in address comparison pattern.
10445 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10447 2018-01-15 Richard Biener <rguenther@suse.de>
10450 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10451 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10452 Reset type names to their identifier if their TYPE_DECL doesn't
10453 have linkage (and thus is used for ODR and devirt).
10454 (save_debug_info_for_decl): Remove.
10455 (save_debug_info_for_type): Likewise.
10456 (add_tree_to_fld_list): Adjust.
10457 * tree-pretty-print.c (dump_generic_node): Make dumping of
10458 type names more robust.
10460 2018-01-15 Richard Biener <rguenther@suse.de>
10462 * BASE-VER: Bump to 8.0.1.
10464 2018-01-14 Martin Sebor <msebor@redhat.com>
10467 * builtins.c (check_access): Avoid warning when the no-warning bit
10470 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10472 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10473 * ira-color (allocno_hard_regs_compare): Likewise.
10475 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10478 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10479 Use .pushsection/.popsection.
10481 2018-01-14 Martin Sebor <msebor@redhat.com>
10484 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10486 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10488 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10489 entry from extra_headers.
10490 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10491 extra_headers, make the list bitwise identical to the i?86-*-* one.
10493 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10495 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10496 -mcmodel=large with -mindirect-branch=thunk,
10497 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10498 -mfunction-return=thunk-extern.
10499 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10500 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10501 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10503 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10505 * config/i386/i386.c (print_reg): Print the name of the full
10506 integer register without '%'.
10507 (ix86_print_operand): Handle 'V'.
10508 * doc/extend.texi: Document 'V' modifier.
10510 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10512 * config/i386/constraints.md (Bs): Disallow memory operand for
10513 -mindirect-branch-register.
10515 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10516 (GOT_memory_operand): Likewise.
10517 (call_insn_operand): Likewise.
10518 (sibcall_insn_operand): Likewise.
10519 (GOT32_symbol_operand): Likewise.
10520 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10521 for -mindirect-branch-register.
10522 (tablejump): Likewise.
10523 (*sibcall_memory): Likewise.
10524 (*sibcall_value_memory): Likewise.
10525 Disallow peepholes of indirect call and jump via memory for
10526 -mindirect-branch-register.
10527 (*call_pop): Replace m with Bw.
10528 (*call_value_pop): Likewise.
10529 (*sibcall_pop_memory): Replace m with Bs.
10530 * config/i386/i386.opt (mindirect-branch-register): New option.
10531 * doc/invoke.texi: Document -mindirect-branch-register option.
10533 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10535 * config/i386/i386-protos.h (ix86_output_function_return): New.
10536 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10537 set function_return_type.
10538 (indirect_thunk_name): Add ret_p to indicate thunk for function
10540 (output_indirect_thunk_function): Pass false to
10541 indirect_thunk_name.
10542 (ix86_output_indirect_branch_via_reg): Likewise.
10543 (ix86_output_indirect_branch_via_push): Likewise.
10544 (output_indirect_thunk_function): Create alias for function
10545 return thunk if regno < 0.
10546 (ix86_output_function_return): New function.
10547 (ix86_handle_fndecl_attribute): Handle function_return.
10548 (ix86_attribute_table): Add function_return.
10549 * config/i386/i386.h (machine_function): Add
10550 function_return_type.
10551 * config/i386/i386.md (simple_return_internal): Use
10552 ix86_output_function_return.
10553 (simple_return_internal_long): Likewise.
10554 * config/i386/i386.opt (mfunction-return=): New option.
10555 (indirect_branch): Mention -mfunction-return=.
10556 * doc/extend.texi: Document function_return function attribute.
10557 * doc/invoke.texi: Document -mfunction-return= option.
10559 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10561 * config/i386/i386-opts.h (indirect_branch): New.
10562 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10563 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10564 with local indirect jump when converting indirect call and jump.
10565 (ix86_set_indirect_branch_type): New.
10566 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10567 (indirectlabelno): New.
10568 (indirect_thunk_needed): Likewise.
10569 (indirect_thunk_bnd_needed): Likewise.
10570 (indirect_thunks_used): Likewise.
10571 (indirect_thunks_bnd_used): Likewise.
10572 (INDIRECT_LABEL): Likewise.
10573 (indirect_thunk_name): Likewise.
10574 (output_indirect_thunk): Likewise.
10575 (output_indirect_thunk_function): Likewise.
10576 (ix86_output_indirect_branch_via_reg): Likewise.
10577 (ix86_output_indirect_branch_via_push): Likewise.
10578 (ix86_output_indirect_branch): Likewise.
10579 (ix86_output_indirect_jmp): Likewise.
10580 (ix86_code_end): Call output_indirect_thunk_function if needed.
10581 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10583 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10584 (ix86_attribute_table): Add indirect_branch.
10585 * config/i386/i386.h (machine_function): Add indirect_branch_type
10586 and has_local_indirect_jump.
10587 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10589 (tablejump): Likewise.
10590 (*indirect_jump): Use ix86_output_indirect_jmp.
10591 (*tablejump_1): Likewise.
10592 (simple_return_indirect_internal): Likewise.
10593 * config/i386/i386.opt (mindirect-branch=): New option.
10594 (indirect_branch): New.
10597 (thunk-inline): Likewise.
10598 (thunk-extern): Likewise.
10599 * doc/extend.texi: Document indirect_branch function attribute.
10600 * doc/invoke.texi: Document -mindirect-branch= option.
10602 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10605 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10607 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10609 * ipa-inline.c (want_inline_small_function_p): Return false if
10610 inlining has already failed with CIF_FINAL_ERROR.
10611 (update_caller_keys): Call want_inline_small_function_p before
10613 (update_callee_keys): Likewise.
10615 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10617 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10619 (rs6000_quadword_masked_address_p): Likewise.
10620 (quad_aligned_load_p): Likewise.
10621 (quad_aligned_store_p): Likewise.
10622 (const_load_sequence_p): Add comment to describe the outer-most loop.
10623 (mimic_memory_attributes_and_flags): New function.
10624 (rs6000_gen_stvx): Likewise.
10625 (replace_swapped_aligned_store): Likewise.
10626 (rs6000_gen_lvx): Likewise.
10627 (replace_swapped_aligned_load): Likewise.
10628 (replace_swapped_load_constant): Capitalize argument name in
10629 comment describing this function.
10630 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10631 and stores that access quad-word aligned addresses and replace
10632 with stvx or lvx instructions when appropriate.
10633 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10634 New function prototype.
10635 (rs6000_quadword_masked_address_p): Likewise.
10636 (rs6000_gen_lvx): Likewise.
10637 (rs6000_gen_stvx): Likewise.
10638 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10639 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10640 when memory address is aligned.
10641 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10642 this split to select lvx instruction when memory address is aligned.
10643 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10644 instruction when memory address is aligned.
10645 (*vsx_le_perm_load_v16qi): Likewise.
10646 (four unnamed splitters): Modify to select the stvx instruction
10647 when memory is aligned.
10649 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10651 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10652 which appears in the queue multiple times.
10654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10655 Alan Hayward <alan.hayward@arm.com>
10656 David Sherwood <david.sherwood@arm.com>
10658 * tree-vectorizer.h (vec_lower_bound): New structure.
10659 (_loop_vec_info): Add check_nonzero and lower_bounds.
10660 (LOOP_VINFO_CHECK_NONZERO): New macro.
10661 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10662 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10663 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10664 fields. Make seg_len the distance travelled, not including the
10666 (dr_direction_indicator): Declare.
10667 (dr_zero_step_indicator): Likewise.
10668 (dr_known_forward_stride_p): Likewise.
10669 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10671 (runtime_alias_check_p): Allow runtime alias checks with
10673 (operator ==): Compare access_size and align.
10674 (prune_runtime_alias_test_list): Rework for new distinction between
10675 the access_size and seg_len.
10676 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10678 (get_segment_min_max): New function.
10679 (create_intersect_range_checks): Use it.
10680 (dr_step_indicator): New function.
10681 (dr_direction_indicator): Likewise.
10682 (dr_zero_step_indicator): Likewise.
10683 (dr_known_forward_stride_p): Likewise.
10684 * tree-loop-distribution.c (data_ref_segment_size): Return
10685 DR_STEP * (niters - 1).
10686 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10688 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10689 (vect_preserves_scalar_order_p): New function, split out from...
10690 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10691 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10692 (vect_vfa_access_size): New function.
10693 (vect_vfa_align): Likewise.
10694 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10695 (dump_lower_bound): New function.
10696 (vect_check_lower_bound): Likewise.
10697 (vect_small_gap_p): Likewise.
10698 (vectorizable_with_step_bound_p): Likewise.
10699 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10700 depencies if the vectorization factor is 1. Convert the checks
10701 for nonzero steps into checks on the bounds of DR_STEP. Try using
10702 a bunds check for variable steps if the minimum required step is
10703 relatively small. Update calls to the dr_with_seg_len
10704 constructor and to vect_compile_time_alias.
10705 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10707 (vect_loop_versioning): Call it.
10708 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10710 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10713 Alan Hayward <alan.hayward@arm.com>
10714 David Sherwood <david.sherwood@arm.com>
10716 * doc/sourcebuild.texi (vect_scatter_store): Document.
10717 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10719 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10721 * genopinit.c (main): Add supports_vec_scatter_store and
10722 supports_vec_scatter_store_cached to target_optabs.
10723 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10724 IFN_MASK_SCATTER_STORE.
10725 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10727 * internal-fn.h (internal_store_fn_p): Declare.
10728 (internal_fn_stored_value_index): Likewise.
10729 * internal-fn.c (scatter_store_direct): New macro.
10730 (expand_scatter_store_optab_fn): New function.
10731 (direct_scatter_store_optab_supported_p): New macro.
10732 (internal_store_fn_p): New function.
10733 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10734 IFN_MASK_SCATTER_STORE.
10735 (internal_fn_mask_index): Likewise.
10736 (internal_fn_stored_value_index): New function.
10737 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10738 for scatter stores.
10739 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10740 * optabs-query.c (supports_vec_scatter_store_p): New function.
10741 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10742 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10743 true for scatter stores.
10744 (vect_gather_scatter_fn_p): Handle scatter stores too.
10745 (vect_check_gather_scatter): Consider using scatter stores if
10746 supports_vec_scatter_store_p.
10747 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10748 scatter stores too.
10749 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10750 internal_fn_stored_value_index.
10751 (check_load_store_masking): Handle scatter stores too.
10752 (vect_get_store_rhs): Make public.
10753 (vectorizable_call): Use internal_store_fn_p.
10754 (vectorizable_store): Handle scatter store internal functions.
10755 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10756 when deciding whether the end of the group has been reached.
10757 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10758 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10759 (mask_scatter_store<mode>): New insns.
10761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10762 Alan Hayward <alan.hayward@arm.com>
10763 David Sherwood <david.sherwood@arm.com>
10765 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10766 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10767 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10769 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10770 Use vect_truncate_gather_scatter_offset if we can't treat the
10771 operation as a normal gather load or scatter store.
10772 (get_group_load_store_type): Take the gather_scatter_info
10773 as argument. Try using a gather load or scatter store for
10774 single-element groups.
10775 (get_load_store_type): Update calls to get_group_load_store_type
10776 and vect_use_strided_gather_scatters_p.
10778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10779 Alan Hayward <alan.hayward@arm.com>
10780 David Sherwood <david.sherwood@arm.com>
10782 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10783 optional tree argument.
10784 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10786 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10787 but continue to use the current value as a fallback.
10788 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10789 to compare the updates.
10790 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10791 (get_load_store_type): Use it when handling a strided access.
10792 (vect_get_strided_load_store_ops): New function.
10793 (vect_get_data_ptr_increment): Likewise.
10794 (vectorizable_load): Handle strided gather loads. Always pass
10795 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10798 Alan Hayward <alan.hayward@arm.com>
10799 David Sherwood <david.sherwood@arm.com>
10801 * doc/md.texi (gather_load@var{m}): Document.
10802 (mask_gather_load@var{m}): Likewise.
10803 * genopinit.c (main): Add supports_vec_gather_load and
10804 supports_vec_gather_load_cached to target_optabs.
10805 * optabs-tree.c (init_tree_optimization_optabs): Use
10806 ggc_cleared_alloc to allocate target_optabs.
10807 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10808 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10810 * internal-fn.h (internal_load_fn_p): Declare.
10811 (internal_gather_scatter_fn_p): Likewise.
10812 (internal_fn_mask_index): Likewise.
10813 (internal_gather_scatter_fn_supported_p): Likewise.
10814 * internal-fn.c (gather_load_direct): New macro.
10815 (expand_gather_load_optab_fn): New function.
10816 (direct_gather_load_optab_supported_p): New macro.
10817 (direct_internal_fn_optab): New function.
10818 (internal_load_fn_p): Likewise.
10819 (internal_gather_scatter_fn_p): Likewise.
10820 (internal_fn_mask_index): Likewise.
10821 (internal_gather_scatter_fn_supported_p): Likewise.
10822 * optabs-query.c (supports_at_least_one_mode_p): New function.
10823 (supports_vec_gather_load_p): Likewise.
10824 * optabs-query.h (supports_vec_gather_load_p): Declare.
10825 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10826 and memory_type field.
10827 (NUM_PATTERNS): Bump to 15.
10828 * tree-vect-data-refs.c: Include internal-fn.h.
10829 (vect_gather_scatter_fn_p): New function.
10830 (vect_describe_gather_scatter_call): Likewise.
10831 (vect_check_gather_scatter): Try using internal functions for
10832 gather loads. Recognize existing calls to a gather load function.
10833 (vect_analyze_data_refs): Consider using gather loads if
10834 supports_vec_gather_load_p.
10835 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10836 (vect_get_gather_scatter_offset_type): Likewise.
10837 (vect_convert_mask_for_vectype): Likewise.
10838 (vect_add_conversion_to_patterm): Likewise.
10839 (vect_try_gather_scatter_pattern): Likewise.
10840 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10841 (vect_vect_recog_func_ptrs): Add it.
10842 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10843 internal_fn_mask_index and internal_gather_scatter_fn_p.
10844 (check_load_store_masking): Take the gather_scatter_info as an
10845 argument and handle gather loads.
10846 (vect_get_gather_scatter_ops): New function.
10847 (vectorizable_call): Check internal_load_fn_p.
10848 (vectorizable_load): Likewise. Handle gather load internal
10850 (vectorizable_store): Update call to check_load_store_masking.
10851 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10852 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10853 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10854 (aarch64_gather_scale_operand_d): New predicates.
10855 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10856 (mask_gather_load<mode>): New insns.
10858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10859 Alan Hayward <alan.hayward@arm.com>
10860 David Sherwood <david.sherwood@arm.com>
10862 * optabs.def (fold_left_plus_optab): New optab.
10863 * doc/md.texi (fold_left_plus_@var{m}): Document.
10864 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10865 * internal-fn.c (fold_left_direct): Define.
10866 (expand_fold_left_optab_fn): Likewise.
10867 (direct_fold_left_optab_supported_p): Likewise.
10868 * fold-const-call.c (fold_const_fold_left): New function.
10869 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10870 * tree-parloops.c (valid_reduction_p): New function.
10871 (gather_scalar_reductions): Use it.
10872 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10873 (vect_finish_replace_stmt): Declare.
10874 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10875 (needs_fold_left_reduction_p): New function, split out from...
10876 (vect_is_simple_reduction): ...here. Accept reductions that
10877 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10878 (vect_force_simple_reduction): Also store the reduction type in
10879 the assignment's STMT_VINFO_REDUC_TYPE.
10880 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10881 (merge_with_identity): New function.
10882 (vect_expand_fold_left): Likewise.
10883 (vectorize_fold_left_reduction): Likewise.
10884 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10885 scalar phi in place for it. Check for target support and reject
10886 cases that would reassociate the operation. Defer the transform
10887 phase to vectorize_fold_left_reduction.
10888 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10889 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10890 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10894 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10895 call to ifc_temp_var.
10897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10898 Alan Hayward <alan.hayward@arm.com>
10899 David Sherwood <david.sherwood@arm.com>
10901 * target.def (legitimize_address_displacement): Take the original
10902 offset as a poly_int.
10903 * targhooks.h (default_legitimize_address_displacement): Update
10905 * targhooks.c (default_legitimize_address_displacement): Likewise.
10906 * doc/tm.texi: Regenerate.
10907 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10908 as an argument, moving assert of ad->disp == ad->disp_term to...
10909 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10910 Try calling targetm.legitimize_address_displacement before expanding
10911 the address rather than afterwards, and adjust for the new interface.
10912 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10913 Match the new hook interface. Handle SVE addresses.
10914 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10915 new hook interface.
10917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10919 * Makefile.in (OBJS): Add early-remat.o.
10920 * target.def (select_early_remat_modes): New hook.
10921 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10922 * doc/tm.texi: Regenerate.
10923 * targhooks.h (default_select_early_remat_modes): Declare.
10924 * targhooks.c (default_select_early_remat_modes): New function.
10925 * timevar.def (TV_EARLY_REMAT): New timevar.
10926 * passes.def (pass_early_remat): New pass.
10927 * tree-pass.h (make_pass_early_remat): Declare.
10928 * early-remat.c: New file.
10929 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10931 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10938 vfm1 with a bound_epilog parameter.
10939 (vect_do_peeling): Update calls accordingly, and move the prologue
10940 call earlier in the function. Treat the base bound_epilog as 0 for
10941 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10942 this base when peeling for gaps.
10943 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10944 with fully-masked loops.
10945 (vect_estimate_min_profitable_iters): Handle the single peeled
10946 iteration in that case.
10948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10949 Alan Hayward <alan.hayward@arm.com>
10950 David Sherwood <david.sherwood@arm.com>
10952 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10953 single-element interleaving even if the size is not a power of 2.
10954 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10955 accesses for single-element interleaving if the group size is
10958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10959 Alan Hayward <alan.hayward@arm.com>
10960 David Sherwood <david.sherwood@arm.com>
10962 * doc/md.texi (fold_extract_last_@var{m}): Document.
10963 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10964 * optabs.def (fold_extract_last_optab): New optab.
10965 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10966 * internal-fn.c (fold_extract_direct): New macro.
10967 (expand_fold_extract_optab_fn): Likewise.
10968 (direct_fold_extract_optab_supported_p): Likewise.
10969 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10970 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10971 EXTRACT_LAST_REDUCTION.
10972 (get_initial_def_for_reduction): Do not create an initial vector
10973 for EXTRACT_LAST_REDUCTION reductions.
10974 (vectorizable_reduction): Leave the scalar phi in place for
10975 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10976 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10977 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10978 transform phase to vectorizable_condition.
10979 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10981 (vect_finish_stmt_generation): ...here.
10982 (vect_finish_replace_stmt): New function.
10983 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10984 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10986 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10989 Alan Hayward <alan.hayward@arm.com>
10990 David Sherwood <david.sherwood@arm.com>
10992 * doc/md.texi (extract_last_@var{m}): Document.
10993 * optabs.def (extract_last_optab): New optab.
10994 * internal-fn.def (EXTRACT_LAST): New internal function.
10995 * internal-fn.c (cond_unary_direct): New macro.
10996 (expand_cond_unary_optab_fn): Likewise.
10997 (direct_cond_unary_optab_supported_p): Likewise.
10998 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10999 loops using EXTRACT_LAST.
11000 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11001 (extract_last_<mode>): ...this optab.
11002 (vec_extract<mode><Vel>): Update accordingly.
11004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11005 Alan Hayward <alan.hayward@arm.com>
11006 David Sherwood <david.sherwood@arm.com>
11008 * target.def (empty_mask_is_expensive): New hook.
11009 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11010 * doc/tm.texi: Regenerate.
11011 * targhooks.h (default_empty_mask_is_expensive): Declare.
11012 * targhooks.c (default_empty_mask_is_expensive): New function.
11013 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11014 if the target says that empty masks are expensive.
11015 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11017 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11020 Alan Hayward <alan.hayward@arm.com>
11021 David Sherwood <david.sherwood@arm.com>
11023 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11024 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11025 (vect_use_loop_mask_for_alignment_p): New function.
11026 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11027 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11028 niters_skip argument. Make sure that the first niters_skip elements
11029 of the first iteration are inactive.
11030 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11031 Update call to vect_set_loop_masks_directly.
11032 (get_misalign_in_elems): New function, split out from...
11033 (vect_gen_prolog_loop_niters): ...here.
11034 (vect_update_init_of_dr): Take a code argument that specifies whether
11035 the adjustment should be added or subtracted.
11036 (vect_update_init_of_drs): Likewise.
11037 (vect_prepare_for_masked_peels): New function.
11038 (vect_do_peeling): Skip prologue peeling if we're using a mask
11039 instead. Update call to vect_update_inits_of_drs.
11040 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11042 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11043 alignment. Do not include the number of peeled iterations in
11044 the minimum threshold in that case.
11045 (vectorizable_induction): Adjust the start value down by
11046 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11047 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11048 Take the number of skipped iterations into account when calculating
11050 * tree-vect-stmts.c (vect_gen_while_not): New function.
11052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11053 Alan Hayward <alan.hayward@arm.com>
11054 David Sherwood <david.sherwood@arm.com>
11056 * doc/sourcebuild.texi (vect_fully_masked): Document.
11057 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11058 default value to 0.
11059 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11061 (vect_analyze_loop_2): ...here. Don't check the vectorization
11062 factor against the number of loop iterations if the loop is
11065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11066 Alan Hayward <alan.hayward@arm.com>
11067 David Sherwood <david.sherwood@arm.com>
11069 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11070 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11071 (dump_groups): Update accordingly.
11072 (iv_use::mem_type): New member variable.
11073 (address_p): New function.
11074 (record_use): Add a mem_type argument and initialize the new
11076 (record_group_use): Add a mem_type argument. Use address_p.
11077 Remove obsolete null checks of base_object. Update call to record_use.
11078 (find_interesting_uses_op): Update call to record_group_use.
11079 (find_interesting_uses_cond): Likewise.
11080 (find_interesting_uses_address): Likewise.
11081 (get_mem_type_for_internal_fn): New function.
11082 (find_address_like_use): Likewise.
11083 (find_interesting_uses_stmt): Try find_address_like_use before
11084 calling find_interesting_uses_op.
11085 (addr_offset_valid_p): Use the iv mem_type field as the type
11086 of the addressed memory.
11087 (add_autoinc_candidates): Likewise.
11088 (get_address_cost): Likewise.
11089 (split_small_address_groups_p): Use address_p.
11090 (split_address_groups): Likewise.
11091 (add_iv_candidate_for_use): Likewise.
11092 (autoinc_possible_for_pair): Likewise.
11093 (rewrite_groups): Likewise.
11094 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11095 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11096 (get_alias_ptr_type_for_ptr_address): New function.
11097 (rewrite_use_address): Rewrite address uses in calls that were
11098 identified by find_address_like_use.
11100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11101 Alan Hayward <alan.hayward@arm.com>
11102 David Sherwood <david.sherwood@arm.com>
11104 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11106 * gimple-expr.h (is_gimple_addressable: Likewise.
11107 * gimple-expr.c (is_gimple_address): Likewise.
11108 * internal-fn.c (expand_call_mem_ref): New function.
11109 (expand_mask_load_optab_fn): Use it.
11110 (expand_mask_store_optab_fn): Likewise.
11112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11113 Alan Hayward <alan.hayward@arm.com>
11114 David Sherwood <david.sherwood@arm.com>
11116 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11117 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11118 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11119 (cond_umax@var{mode}): Document.
11120 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11121 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11122 (cond_umin_optab, cond_umax_optab): New optabs.
11123 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11124 (COND_IOR, COND_XOR): New internal functions.
11125 * internal-fn.h (get_conditional_internal_fn): Declare.
11126 * internal-fn.c (cond_binary_direct): New macro.
11127 (expand_cond_binary_optab_fn): Likewise.
11128 (direct_cond_binary_optab_supported_p): Likewise.
11129 (get_conditional_internal_fn): New function.
11130 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11131 Cope with reduction statements that are vectorized as calls rather
11133 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11134 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11135 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11136 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11137 (UNSPEC_COND_EOR): New unspecs.
11138 (optab): Add mappings for them.
11139 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11140 (sve_int_op, sve_fp_op): New int attributes.
11142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11143 Alan Hayward <alan.hayward@arm.com>
11144 David Sherwood <david.sherwood@arm.com>
11146 * optabs.def (while_ult_optab): New optab.
11147 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11148 * internal-fn.def (WHILE_ULT): New internal function.
11149 * internal-fn.h (direct_internal_fn_supported_p): New override
11150 that takes two types as argument.
11151 * internal-fn.c (while_direct): New macro.
11152 (expand_while_optab_fn): New function.
11153 (convert_optab_supported_p): Likewise.
11154 (direct_while_optab_supported_p): New macro.
11155 * wide-int.h (wi::udiv_ceil): New function.
11156 * tree-vectorizer.h (rgroup_masks): New structure.
11157 (vec_loop_masks): New typedef.
11158 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11159 and fully_masked_p.
11160 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11161 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11162 (vect_max_vf): New function.
11163 (slpeel_make_loop_iterate_ntimes): Delete.
11164 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11165 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11166 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11167 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11168 internal-fn.h, stor-layout.h and optabs-query.h.
11169 (vect_set_loop_mask): New function.
11170 (add_preheader_seq): Likewise.
11171 (add_header_seq): Likewise.
11172 (interleave_supported_p): Likewise.
11173 (vect_maybe_permute_loop_masks): Likewise.
11174 (vect_set_loop_masks_directly): Likewise.
11175 (vect_set_loop_condition_masked): Likewise.
11176 (vect_set_loop_condition_unmasked): New function, split out from
11177 slpeel_make_loop_iterate_ntimes.
11178 (slpeel_make_loop_iterate_ntimes): Rename to..
11179 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11180 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11181 (vect_do_peeling): Update call accordingly.
11182 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11184 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11185 mask_compare_type, can_fully_mask_p and fully_masked_p.
11186 (release_vec_loop_masks): New function.
11187 (_loop_vec_info): Use it to free the loop masks.
11188 (can_produce_all_loop_masks_p): New function.
11189 (vect_get_max_nscalars_per_iter): Likewise.
11190 (vect_verify_full_masking): Likewise.
11191 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11192 retries, and free the mask rgroups before retrying. Check loop-wide
11193 reasons for disallowing fully-masked loops. Make the final decision
11194 about whether use a fully-masked loop or not.
11195 (vect_estimate_min_profitable_iters): Do not assume that peeling
11196 for the number of iterations will be needed for fully-masked loops.
11197 (vectorizable_reduction): Disable fully-masked loops.
11198 (vectorizable_live_operation): Likewise.
11199 (vect_halve_mask_nunits): New function.
11200 (vect_double_mask_nunits): Likewise.
11201 (vect_record_loop_mask): Likewise.
11202 (vect_get_loop_mask): Likewise.
11203 (vect_transform_loop): Handle the case in which the final loop
11204 iteration might handle a partial vector. Call vect_set_loop_condition
11205 instead of slpeel_make_loop_iterate_ntimes.
11206 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11207 (check_load_store_masking): New function.
11208 (prepare_load_store_mask): Likewise.
11209 (vectorizable_store): Handle fully-masked loops.
11210 (vectorizable_load): Likewise.
11211 (supportable_widening_operation): Use vect_halve_mask_nunits for
11213 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11214 (vect_gen_while): New function.
11215 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11216 (aarch64_uqdec<mode>): New insn.
11218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11219 Alan Hayward <alan.hayward@arm.com>
11220 David Sherwood <david.sherwood@arm.com>
11222 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11223 (reduc_xor_scal_optab): New optabs.
11224 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11225 (reduc_xor_scal_@var{m}): Document.
11226 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11227 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11228 internal functions.
11229 * fold-const-call.c (fold_const_call): Handle them.
11230 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11231 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11232 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11233 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11234 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11235 (UNSPEC_XORV): New unspecs.
11236 (optab): Add entries for them.
11237 (BITWISEV): New int iterator.
11238 (bit_reduc_op): New int attributes.
11240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11241 Alan Hayward <alan.hayward@arm.com>
11242 David Sherwood <david.sherwood@arm.com>
11244 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11245 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11246 * optabs.def (vec_shl_insert_optab): New optab.
11247 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11248 (duplicate_and_interleave): Likewise.
11249 * tree-vect-loop.c: Include internal-fn.h.
11250 (neutral_op_for_slp_reduction): New function, split out from
11251 get_initial_defs_for_reduction.
11252 (get_initial_def_for_reduction): Handle option 2 for variable-length
11253 vectors by loading the neutral value into a vector and then shifting
11254 the initial value into element 0.
11255 (get_initial_defs_for_reduction): Replace the code argument with
11256 the neutral value calculated by neutral_op_for_slp_reduction.
11257 Use gimple_build_vector for constant-length vectors.
11258 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11259 but the first group_size elements have a neutral value.
11260 Use duplicate_and_interleave otherwise.
11261 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11262 Update call to get_initial_defs_for_reduction. Handle SLP
11263 reductions for variable-length vectors by creating one vector
11264 result for each scalar result, with the elements associated
11265 with other scalar results stubbed out with the neutral value.
11266 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11267 Require IFN_VEC_SHL_INSERT for double reductions on
11268 variable-length vectors, or SLP reductions that have
11269 a neutral value. Require can_duplicate_and_interleave_p
11270 support for variable-length unchained SLP reductions if there
11271 is no neutral value, such as for MIN/MAX reductions. Also require
11272 the number of vector elements to be a multiple of the number of
11273 SLP statements when doing variable-length unchained SLP reductions.
11274 Update call to vect_create_epilog_for_reduction.
11275 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11276 and remove initial values.
11277 (duplicate_and_interleave): Make public.
11278 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11279 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11282 Alan Hayward <alan.hayward@arm.com>
11283 David Sherwood <david.sherwood@arm.com>
11285 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11286 (can_duplicate_and_interleave_p): New function.
11287 (vect_get_and_check_slp_defs): Take the vector of statements
11288 rather than just the current one. Remove excess parentheses.
11289 Restriction rejectinon of vect_constant_def and vect_external_def
11290 for variable-length vectors to boolean types, or types for which
11291 can_duplicate_and_interleave_p is false.
11292 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11293 (duplicate_and_interleave): New function.
11294 (vect_get_constant_vectors): Use gimple_build_vector for
11295 constant-length vectors and suitable variable-length constant
11296 vectors. Use duplicate_and_interleave for other variable-length
11297 vectors. Don't defer the update when inserting new statements.
11299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11300 Alan Hayward <alan.hayward@arm.com>
11301 David Sherwood <david.sherwood@arm.com>
11303 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11304 min_profitable_iters doesn't go negative.
11306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11307 Alan Hayward <alan.hayward@arm.com>
11308 David Sherwood <david.sherwood@arm.com>
11310 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11311 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11312 * optabs.def (vec_mask_load_lanes_optab): New optab.
11313 (vec_mask_store_lanes_optab): Likewise.
11314 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11315 (MASK_STORE_LANES): Likewise.
11316 * internal-fn.c (mask_load_lanes_direct): New macro.
11317 (mask_store_lanes_direct): Likewise.
11318 (expand_mask_load_optab_fn): Handle masked operations.
11319 (expand_mask_load_lanes_optab_fn): New macro.
11320 (expand_mask_store_optab_fn): Handle masked operations.
11321 (expand_mask_store_lanes_optab_fn): New macro.
11322 (direct_mask_load_lanes_optab_supported_p): Likewise.
11323 (direct_mask_store_lanes_optab_supported_p): Likewise.
11324 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11326 (vect_load_lanes_supported): Likewise.
11327 * tree-vect-data-refs.c (strip_conversion): New function.
11328 (can_group_stmts_p): Likewise.
11329 (vect_analyze_data_ref_accesses): Use it instead of checking
11330 for a pair of assignments.
11331 (vect_store_lanes_supported): Take a masked_p parameter.
11332 (vect_load_lanes_supported): Likewise.
11333 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11334 vect_store_lanes_supported and vect_load_lanes_supported.
11335 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11336 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11337 parameter. Don't allow gaps for masked accesses.
11338 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11339 and vect_load_lanes_supported.
11340 (get_load_store_type): Take a masked_p parameter and update
11341 call to get_group_load_store_type.
11342 (vectorizable_store): Update call to get_load_store_type.
11343 Handle IFN_MASK_STORE_LANES.
11344 (vectorizable_load): Update call to get_load_store_type.
11345 Handle IFN_MASK_LOAD_LANES.
11347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11348 Alan Hayward <alan.hayward@arm.com>
11349 David Sherwood <david.sherwood@arm.com>
11351 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11353 * config/aarch64/aarch64-protos.h
11354 (aarch64_sve_struct_memory_operand_p): Declare.
11355 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11356 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11357 (VPRED, vpred): Handle SVE structure modes.
11358 * config/aarch64/constraints.md (Utx): New constraint.
11359 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11360 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11361 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11362 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11363 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11364 structure modes. Split into pieces after RA.
11365 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11366 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11368 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11369 SVE structure modes.
11370 (aarch64_classify_address): Likewise.
11371 (sizetochar): Move earlier in file.
11372 (aarch64_print_operand): Handle SVE register lists.
11373 (aarch64_array_mode): New function.
11374 (aarch64_sve_struct_memory_operand_p): Likewise.
11375 (TARGET_ARRAY_MODE): Redefine.
11377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11378 Alan Hayward <alan.hayward@arm.com>
11379 David Sherwood <david.sherwood@arm.com>
11381 * target.def (array_mode): New target hook.
11382 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11383 * doc/tm.texi: Regenerate.
11384 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11385 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11386 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11387 targetm.array_mode.
11388 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11392 Alan Hayward <alan.hayward@arm.com>
11393 David Sherwood <david.sherwood@arm.com>
11395 * fold-const.c (fold_binary_loc): Check the argument types
11396 rather than the result type when testing for a vector operation.
11398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11400 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11401 * doc/tm.texi: Regenerate.
11403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11404 Alan Hayward <alan.hayward@arm.com>
11405 David Sherwood <david.sherwood@arm.com>
11407 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11408 (sve): Document new AArch64 extension.
11409 * doc/md.texi (w): Extend the description of the AArch64
11410 constraint to include SVE vectors.
11411 (Upl, Upa): Document new AArch64 predicate constraints.
11412 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11414 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11415 (msve-vector-bits=): New option.
11416 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11417 SVE when these are disabled.
11418 (sve): New extension.
11419 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11420 modes. Adjust their number of units based on aarch64_sve_vg.
11421 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11422 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11423 aarch64_addr_query_type.
11424 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11425 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11426 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11427 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11428 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11429 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11430 (aarch64_simd_imm_zero_p): Delete.
11431 (aarch64_check_zero_based_sve_index_immediate): Declare.
11432 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11433 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11434 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11435 (aarch64_sve_float_mul_immediate_p): Likewise.
11436 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11437 rather than an rtx.
11438 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11439 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11440 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11441 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11442 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11443 (aarch64_regmode_natural_size): Likewise.
11444 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11445 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11447 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11448 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11449 for VG and the SVE predicate registers.
11450 (V_ALIASES): Add a "z"-prefixed alias.
11451 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11452 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11453 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11454 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11455 (REG_CLASS_NAMES): Add entries for them.
11456 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11457 and the predicate registers.
11458 (aarch64_sve_vg): Declare.
11459 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11460 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11461 (REGMODE_NATURAL_SIZE): Define.
11462 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11464 * config/aarch64/aarch64.c: Include cfgrtl.h.
11465 (simd_immediate_info): Add a constructor for series vectors,
11466 and an associated step field.
11467 (aarch64_sve_vg): New variable.
11468 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11469 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11470 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11471 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11472 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11473 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11474 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11475 (aarch64_get_mask_mode): New functions.
11476 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11477 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11478 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11479 predicate modes and predicate registers. Explicitly restrict
11480 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11481 to store a vector mode if it is recognized by
11482 aarch64_classify_vector_mode.
11483 (aarch64_regmode_natural_size): New function.
11484 (aarch64_hard_regno_caller_save_mode): Return the original mode
11486 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11487 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11488 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11489 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11491 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11492 does not overlap dest if the function is frame-related. Handle
11494 (aarch64_split_add_offset): New function.
11495 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11496 them aarch64_add_offset.
11497 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11498 and update call to aarch64_sub_sp.
11499 (aarch64_add_cfa_expression): New function.
11500 (aarch64_expand_prologue): Pass extra temporary registers to the
11501 functions above. Handle the case in which we need to emit new
11502 DW_CFA_expressions for registers that were originally saved
11503 relative to the stack pointer, but now have to be expressed
11504 relative to the frame pointer.
11505 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11507 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11508 IP0 and IP1 values for SVE frames.
11509 (aarch64_expand_vec_series): New function.
11510 (aarch64_expand_sve_widened_duplicate): Likewise.
11511 (aarch64_expand_sve_const_vector): Likewise.
11512 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11513 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11514 into the register, rather than emitting a SET directly.
11515 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11516 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11517 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11518 (offset_9bit_signed_scaled_p): New functions.
11519 (aarch64_replicate_bitmask_imm): New function.
11520 (aarch64_bitmask_imm): Use it.
11521 (aarch64_cannot_force_const_mem): Reject expressions involving
11522 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11523 (aarch64_classify_index): Handle SVE indices, by requiring
11524 a plain register index with a scale that matches the element size.
11525 (aarch64_classify_address): Handle SVE addresses. Assert that
11526 the mode of the address is VOIDmode or an integer mode.
11527 Update call to aarch64_classify_symbol.
11528 (aarch64_classify_symbolic_expression): Update call to
11529 aarch64_classify_symbol.
11530 (aarch64_const_vec_all_in_range_p): New function.
11531 (aarch64_print_vector_float_operand): Likewise.
11532 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11533 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11534 and the FP immediates 1.0 and 0.5.
11535 (aarch64_print_address_internal): Handle SVE addresses.
11536 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11537 (aarch64_regno_regclass): Handle predicate registers.
11538 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11540 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11541 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11542 (aarch64_convert_sve_vector_bits): New function.
11543 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11544 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11545 rather than an rtx.
11546 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11547 Handle SVE vector and predicate modes. Accept VL-based constants
11548 that need only one temporary register, and VL offsets that require
11549 no temporary registers.
11550 (aarch64_conditional_register_usage): Mark the predicate registers
11551 as fixed if SVE isn't available.
11552 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11553 Return true for SVE vector and predicate modes.
11554 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11555 rather than an unsigned int. Handle SVE modes.
11556 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11558 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11560 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11561 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11562 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11563 (aarch64_sve_float_mul_immediate_p): New functions.
11564 (aarch64_sve_valid_immediate): New function.
11565 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11566 Explicitly reject structure modes. Check for INDEX constants.
11567 Handle PTRUE and PFALSE constants.
11568 (aarch64_check_zero_based_sve_index_immediate): New function.
11569 (aarch64_simd_imm_zero_p): Delete.
11570 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11571 vector modes. Accept constants in the range of CNT[BHWD].
11572 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11573 ask for an Advanced SIMD mode.
11574 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11575 (aarch64_simd_vector_alignment): Handle SVE predicates.
11576 (aarch64_vectorize_preferred_vector_alignment): New function.
11577 (aarch64_simd_vector_alignment_reachable): Use it instead of
11579 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11580 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11582 (MAX_VECT_LEN): Delete.
11583 (expand_vec_perm_d): Add a vec_flags field.
11584 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11585 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11586 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11588 (aarch64_evpc_rev): Rename to...
11589 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11590 (aarch64_evpc_rev_global): New function.
11591 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11592 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11594 (aarch64_evpc_sve_tbl): New function.
11595 (aarch64_expand_vec_perm_const_1): Update after rename of
11596 aarch64_evpc_rev. Handle SVE permutes too, trying
11597 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11598 than aarch64_evpc_tbl.
11599 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11600 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11601 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11602 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11603 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11604 (aarch64_expand_sve_vcond): New functions.
11605 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11606 of aarch64_vector_mode_p.
11607 (aarch64_dwarf_poly_indeterminate_value): New function.
11608 (aarch64_compute_pressure_classes): Likewise.
11609 (aarch64_can_change_mode_class): Likewise.
11610 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11611 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11612 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11613 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11614 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11615 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11616 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11617 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11619 (Dn, Dl, Dr): Accept const as well as const_vector.
11620 (Dz): Likewise. Compare against CONST0_RTX.
11621 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11622 of "vector" where appropriate.
11623 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11624 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11625 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11626 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11627 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11628 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11629 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11630 (v_int_equiv): Extend to SVE modes.
11631 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11633 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11634 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11635 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11636 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11637 (SVE_COND_FP_CMP): New int iterators.
11638 (perm_hilo): Handle the new unpack unspecs.
11639 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11641 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11642 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11643 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11644 (aarch64_equality_operator, aarch64_constant_vector_operand)
11645 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11646 (aarch64_sve_nonimmediate_operand): Likewise.
11647 (aarch64_sve_general_operand): Likewise.
11648 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11649 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11650 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11651 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11652 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11653 (aarch64_sve_float_arith_immediate): Likewise.
11654 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11655 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11656 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11657 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11658 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11659 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11660 (aarch64_sve_float_arith_operand): Likewise.
11661 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11662 (aarch64_sve_float_mul_operand): Likewise.
11663 (aarch64_sve_vec_perm_operand): Likewise.
11664 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11665 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11666 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11667 as well as const_vector.
11668 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11669 in file. Use CONST0_RTX and CONSTM1_RTX.
11670 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11671 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11672 Use aarch64_simd_imm_zero.
11673 * config/aarch64/aarch64-sve.md: New file.
11674 * config/aarch64/aarch64.md: Include it.
11675 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11676 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11677 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11678 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11679 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11680 (sve): New attribute.
11681 (enabled): Disable instructions with the sve attribute unless
11683 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11684 aarch64_expand_mov_immediate.
11685 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11686 CNT[BHSD] immediates.
11687 (movti): Split CONST_POLY_INT moves into two halves.
11688 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11689 Split additions that need a temporary here if the destination
11690 is the stack pointer.
11691 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11692 (*add<mode>3_poly_1): New instruction.
11693 (set_clobber_cc): New expander.
11695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11697 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11698 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11699 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11700 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11701 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11702 Change innermode from fixed_mode_size to machine_mode.
11703 (simplify_subreg): Update call accordingly. Handle a constant-sized
11704 subreg of a variable-length CONST_VECTOR.
11706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11707 Alan Hayward <alan.hayward@arm.com>
11708 David Sherwood <david.sherwood@arm.com>
11710 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11711 (add_offset_to_base): New function, split out from...
11712 (create_mem_ref): ...here. When handling a scale other than 1,
11713 check first whether the address is valid without the offset.
11714 Add it into the base if so, leaving the index and scale as-is.
11716 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11719 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11720 fold_for_warn before checking if arg2 is INTEGER_CST.
11722 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11724 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11725 (store_multiple_operation): Delete.
11726 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11727 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11728 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11729 guarded by TARGET_STRING.
11730 (rs6000_output_load_multiple): Delete.
11731 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11732 OPTION_MASK_STRING / TARGET_STRING handling.
11733 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11734 (const rs6000_opt_masks) <"string">: Change mask to 0.
11735 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11736 (MASK_STRING): Delete.
11737 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11739 (load_multiple): Delete.
11746 (store_multiple): Delete.
11753 (movmemsi_8reg): Delete.
11754 (corresponding unnamed define_insn): Delete.
11755 (movmemsi_6reg): Delete.
11756 (corresponding unnamed define_insn): Delete.
11757 (movmemsi_4reg): Delete.
11758 (corresponding unnamed define_insn): Delete.
11759 (movmemsi_2reg): Delete.
11760 (corresponding unnamed define_insn): Delete.
11761 (movmemsi_1reg): Delete.
11762 (corresponding unnamed define_insn): Delete.
11763 * config/rs6000/rs6000.opt (mno-string): New.
11764 (mstring): Replace by deprecation warning stub.
11765 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11767 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11769 * regrename.c (regrename_do_replace): If replacing the same
11770 reg multiple times, try to reuse last created gen_raw_REG.
11773 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11774 main to workaround a bug in GDB.
11776 2018-01-12 Tom de Vries <tom@codesourcery.com>
11779 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11781 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11783 PR rtl-optimization/80481
11784 * ira-color.c (get_cap_member): New function.
11785 (allocnos_conflict_by_live_ranges_p): Use it.
11786 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11787 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11789 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11792 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11793 (*saddl_se_1): Ditto.
11794 (*ssubsi_1): Ditto.
11795 (*ssubl_se_1): Ditto.
11797 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11799 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11800 rather than wi::to_widest for DR_INITs.
11801 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11802 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11803 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11805 (vect_analyze_group_access_1): Note that here.
11807 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11809 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11810 polynomial type sizes.
11812 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11814 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11815 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11816 (gimple_add_tmp_var): Likewise.
11818 2018-01-12 Martin Liska <mliska@suse.cz>
11820 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11821 (gimple_alloc_sizes): Likewise.
11822 (dump_gimple_statistics): Use PRIu64 in printf format.
11823 * gimple.h: Change uint64_t to int.
11825 2018-01-12 Martin Liska <mliska@suse.cz>
11827 * tree-core.h: Use uint64_t instead of int.
11828 * tree.c (tree_node_counts): Likewise.
11829 (tree_node_sizes): Likewise.
11830 (dump_tree_statistics): Use PRIu64 in printf format.
11832 2018-01-12 Martin Liska <mliska@suse.cz>
11834 * Makefile.in: As qsort_chk is implemented in vec.c, add
11835 vec.o to linkage of gencfn-macros.
11836 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11837 passing the info to record_node_allocation_statistics.
11838 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11840 * ggc-common.c (struct ggc_usage): Add operator== and use
11841 it in operator< and compare function.
11842 * mem-stats.h (struct mem_usage): Likewise.
11843 * vec.c (struct vec_usage): Remove operator< and compare
11844 function. Can be simply inherited.
11846 2018-01-12 Martin Jambor <mjambor@suse.cz>
11849 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11850 * tree-ssa-math-opts.c: Include domwalk.h.
11851 (convert_mult_to_fma_1): New function.
11852 (fma_transformation_info): New type.
11853 (fma_deferring_state): Likewise.
11854 (cancel_fma_deferring): New function.
11855 (result_of_phi): Likewise.
11856 (last_fma_candidate_feeds_initial_phi): Likewise.
11857 (convert_mult_to_fma): Added deferring logic, split actual
11858 transformation to convert_mult_to_fma_1.
11859 (math_opts_dom_walker): New type.
11860 (math_opts_dom_walker::after_dom_children): New method, body moved
11861 here from pass_optimize_widening_mul::execute, added deferring logic
11863 (pass_optimize_widening_mul::execute): Moved most of code to
11864 math_opts_dom_walker::after_dom_children.
11865 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11866 * config/i386/i386.c (ix86_option_override_internal): Added
11867 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11869 2018-01-12 Richard Biener <rguenther@suse.de>
11872 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11873 inline instance vars.
11875 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11878 * config/rx/rx.c (rx_is_restricted_memory_address):
11879 Handle SUBREG case.
11881 2018-01-12 Richard Biener <rguenther@suse.de>
11883 PR tree-optimization/80846
11884 * target.def (split_reduction): New target hook.
11885 * targhooks.c (default_split_reduction): New function.
11886 * targhooks.h (default_split_reduction): Declare.
11887 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11888 target requests first reduce vectors by combining low and high
11890 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11891 (get_vectype_for_scalar_type_and_size): Export.
11892 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11893 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11894 * doc/tm.texi: Regenerate.
11895 * config/i386/i386.c (ix86_split_reduction): Implement
11896 TARGET_VECTORIZE_SPLIT_REDUCTION.
11898 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11901 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11902 in PIC mode except for TARGET_VXWORKS_RTP.
11903 * config/sparc/sparc.c: Include cfgrtl.h.
11904 (TARGET_INIT_PIC_REG): Define.
11905 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11906 (sparc_pic_register_p): New predicate.
11907 (sparc_legitimate_address_p): Use it.
11908 (sparc_legitimize_pic_address): Likewise.
11909 (sparc_delegitimize_address): Likewise.
11910 (sparc_mode_dependent_address_p): Likewise.
11911 (gen_load_pcrel_sym): Remove 4th parameter.
11912 (load_got_register): Adjust call to above. Remove obsolete stuff.
11913 (sparc_expand_prologue): Do not call load_got_register here.
11914 (sparc_flat_expand_prologue): Likewise.
11915 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11916 (sparc_use_pseudo_pic_reg): New function.
11917 (sparc_init_pic_reg): Likewise.
11918 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11919 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11921 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11923 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11924 Add item for branch_cost.
11926 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11928 PR rtl-optimization/83565
11929 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11930 not extend the result to a larger mode for rotate operations.
11931 (num_sign_bit_copies1): Likewise.
11933 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11936 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11938 Use values-Xc.o for -pedantic.
11939 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11941 2018-01-12 Martin Liska <mliska@suse.cz>
11944 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11946 (possible_polymorphic_call_targets): Use it.
11947 (ipa_devirt): Likewise.
11949 2018-01-12 Martin Liska <mliska@suse.cz>
11951 * profile-count.h (enum profile_quality): Use 0 as invalid
11952 enum value of profile_quality.
11954 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11956 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11957 -mext-string options.
11959 2018-01-12 Richard Biener <rguenther@suse.de>
11961 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11962 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11963 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11965 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11967 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11969 * configure.ac (--with-long-double-format): Add support for the
11970 configuration option to change the default long double format on
11972 * config.gcc (powerpc*-linux*-*): Likewise.
11973 * configure: Regenerate.
11974 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11975 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11976 used without modification.
11978 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11980 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11981 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11982 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11983 MISC_BUILTIN_SPEC_BARRIER.
11984 (rs6000_init_builtins): Likewise.
11985 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11987 (speculation_barrier): New define_insn.
11988 * doc/extend.texi: Document __builtin_speculation_barrier.
11990 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11993 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11994 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11995 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11997 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11998 integral modes instead of "ss" and "sd".
11999 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12000 vectors with 32-bit and 64-bit elements.
12001 (vecdupssescalarmodesuffix): New mode attribute.
12002 (vec_dup<mode>): Use it.
12004 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12007 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12008 frame if argument is passed on stack.
12010 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12013 * ree.c (combine_reaching_defs): Optimize also
12014 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12015 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12017 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12019 PR middle-end/83189
12020 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12022 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12024 PR middle-end/83718
12025 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12026 after they are computed.
12028 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12030 PR tree-optimization/83695
12031 * gimple-loop-linterchange.cc
12032 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12033 reset cached scev information after interchange.
12034 (pass_linterchange::execute): Remove call to scev_reset_htab.
12036 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12038 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12039 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12040 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12041 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12042 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12043 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12044 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12045 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12046 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12047 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12048 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12049 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12050 (V_lane_reg): Likewise.
12051 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12053 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12054 (vfmal_lane_low<mode>_intrinsic,
12055 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12056 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12057 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12058 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12059 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12060 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12062 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12064 * config/arm/arm-cpus.in (fp16fml): New feature.
12065 (ALL_SIMD): Add fp16fml.
12066 (armv8.2-a): Add fp16fml as an option.
12067 (armv8.3-a): Likewise.
12068 (armv8.4-a): Add fp16fml as part of fp16.
12069 * config/arm/arm.h (TARGET_FP16FML): Define.
12070 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12072 * config/arm/arm-modes.def (V2HF): Define.
12073 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12074 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12075 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12076 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12077 vfmsl_low, vfmsl_high): New set of builtins.
12078 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12079 (vfml_op): New code attribute.
12080 (VFMLHALVES): New int iterator.
12081 (VFML, VFMLSEL): New mode attributes.
12082 (V_reg): Define mapping for V2HF.
12083 (V_hi, V_lo): New mode attributes.
12084 (VF_constraint): Likewise.
12085 (vfml_half, vfml_half_selector): New int attributes.
12086 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12088 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12089 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12091 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12092 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12093 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12094 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12096 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12097 Document new effective target and option set.
12099 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12101 * config/arm/arm-cpus.in (armv8_4): New feature.
12102 (ARMv8_4a): New fgroup.
12103 (armv8.4-a): New arch.
12104 * config/arm/arm-tables.opt: Regenerate.
12105 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12106 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12107 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12108 Add matching rules for -march=armv8.4-a and extensions.
12109 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12111 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12114 * config/rx/rx.md (BW): New mode attribute.
12115 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12117 2018-01-11 Richard Biener <rguenther@suse.de>
12119 PR tree-optimization/83435
12120 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12121 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12122 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12124 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12130 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12131 (aarch64_print_address_internal): Use it to check for a zero offset.
12133 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12134 Alan Hayward <alan.hayward@arm.com>
12135 David Sherwood <david.sherwood@arm.com>
12137 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12138 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12139 Return a poly_int64 rather than a HOST_WIDE_INT.
12140 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12141 rather than a HOST_WIDE_INT.
12142 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12143 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12144 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12145 final_offset from HOST_WIDE_INT to poly_int64.
12146 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12147 to_constant when getting the number of units in an Advanced SIMD
12149 (aarch64_builtin_vectorized_function): Check for a constant number
12151 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12153 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12154 attribute instead of GET_MODE_NUNITS.
12155 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12156 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12157 GET_MODE_SIZE for fixed-size registers.
12158 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12159 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12160 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12161 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12162 (aarch64_print_operand, aarch64_print_address_internal)
12163 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12164 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12165 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12166 Handle polynomial GET_MODE_SIZE.
12167 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12168 wider than SImode without modification.
12169 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12170 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12171 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12172 passing and returning SVE modes.
12173 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12174 rather than GEN_INT.
12175 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12176 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12177 (aarch64_allocate_and_probe_stack_space): Likewise.
12178 (aarch64_layout_frame): Cope with polynomial offsets.
12179 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12180 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12181 polynomial offsets.
12182 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12183 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12184 poly_int64 rather than a HOST_WIDE_INT.
12185 (aarch64_get_separate_components, aarch64_process_components)
12186 (aarch64_expand_prologue, aarch64_expand_epilogue)
12187 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12188 (aarch64_anchor_offset): New function, split out from...
12189 (aarch64_legitimize_address): ...here.
12190 (aarch64_builtin_vectorization_cost): Handle polynomial
12191 TYPE_VECTOR_SUBPARTS.
12192 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12194 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12195 number of elements from the PARALLEL rather than the mode.
12196 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12197 rather than GET_MODE_BITSIZE.
12198 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12199 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12200 (aarch64_expand_vec_perm_const_1): Handle polynomial
12201 d->perm.length () and d->perm elements.
12202 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12203 Apply to_constant to d->perm elements.
12204 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12205 polynomial CONST_VECTOR_NUNITS.
12206 (aarch64_move_pointer): Take amount as a poly_int64 rather
12208 (aarch64_progress_pointer): Avoid temporary variable.
12209 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12210 the mode attribute instead of GET_MODE.
12212 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12213 Alan Hayward <alan.hayward@arm.com>
12214 David Sherwood <david.sherwood@arm.com>
12216 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12217 x exists before using it.
12218 (aarch64_add_constant_internal): Rename to...
12219 (aarch64_add_offset_1): ...this. Replace regnum with separate
12220 src and dest rtxes. Handle the case in which they're different,
12221 including when the offset is zero. Replace scratchreg with an rtx.
12222 Use 2 additions if there is no spare register into which we can
12223 move a 16-bit constant.
12224 (aarch64_add_constant): Delete.
12225 (aarch64_add_offset): Replace reg with separate src and dest
12226 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12227 Use aarch64_add_offset_1.
12228 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12229 an rtx rather than an int. Take the delta as a poly_int64
12230 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12231 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12232 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12233 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12234 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12235 and aarch64_add_sp.
12236 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12237 aarch64_add_constant.
12239 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12241 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12242 Use scalar_float_mode.
12244 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12246 * config/aarch64/aarch64-simd.md
12247 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12248 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12249 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12250 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12251 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12252 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12253 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12254 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12255 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12256 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12258 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12261 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12262 targ_options->x_arm_arch_string is non NULL.
12264 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12266 * config/aarch64/aarch64.h
12267 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12269 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12272 * expmed.c (emit_store_flag_force): Swap if const op0
12273 and change VOIDmode to mode of op0.
12275 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12277 PR rtl-optimization/83761
12278 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12279 than bytes to mode_for_size.
12281 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12283 PR middle-end/83189
12284 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12285 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12288 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12290 PR middle-end/83575
12291 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12292 when in layout mode.
12293 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12294 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12297 2018-01-10 Michael Collison <michael.collison@arm.com>
12299 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12300 * config/aarch64/aarch64-option-extension.def: Add
12301 AARCH64_OPT_EXTENSION of 'fp16fml'.
12302 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12303 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12304 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12305 * config/aarch64/constraints.md (Ui7): New constraint.
12306 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12307 (VFMLA_SEL_W): Ditto.
12310 (VFMLA16_LOW): New int iterator.
12311 (VFMLA16_HIGH): Ditto.
12312 (UNSPEC_FMLAL): New unspec.
12313 (UNSPEC_FMLSL): Ditto.
12314 (UNSPEC_FMLAL2): Ditto.
12315 (UNSPEC_FMLSL2): Ditto.
12316 (f16mac): New code attribute.
12317 * config/aarch64/aarch64-simd-builtins.def
12318 (aarch64_fmlal_lowv2sf): Ditto.
12319 (aarch64_fmlsl_lowv2sf): Ditto.
12320 (aarch64_fmlalq_lowv4sf): Ditto.
12321 (aarch64_fmlslq_lowv4sf): Ditto.
12322 (aarch64_fmlal_highv2sf): Ditto.
12323 (aarch64_fmlsl_highv2sf): Ditto.
12324 (aarch64_fmlalq_highv4sf): Ditto.
12325 (aarch64_fmlslq_highv4sf): Ditto.
12326 (aarch64_fmlal_lane_lowv2sf): Ditto.
12327 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12328 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12329 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12330 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12331 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12332 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12333 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12334 (aarch64_fmlal_lane_highv2sf): Ditto.
12335 (aarch64_fmlsl_lane_highv2sf): Ditto.
12336 (aarch64_fmlal_laneq_highv2sf): Ditto.
12337 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12338 (aarch64_fmlalq_lane_highv4sf): Ditto.
12339 (aarch64_fmlsl_lane_highv4sf): Ditto.
12340 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12341 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12342 * config/aarch64/aarch64-simd.md:
12343 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12344 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12345 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12346 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12347 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12348 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12349 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12350 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12351 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12352 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12353 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12354 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12355 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12356 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12357 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12358 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12359 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12360 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12361 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12362 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12363 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12364 (vfmlsl_low_u32): Ditto.
12365 (vfmlalq_low_u32): Ditto.
12366 (vfmlslq_low_u32): Ditto.
12367 (vfmlal_high_u32): Ditto.
12368 (vfmlsl_high_u32): Ditto.
12369 (vfmlalq_high_u32): Ditto.
12370 (vfmlslq_high_u32): Ditto.
12371 (vfmlal_lane_low_u32): Ditto.
12372 (vfmlsl_lane_low_u32): Ditto.
12373 (vfmlal_laneq_low_u32): Ditto.
12374 (vfmlsl_laneq_low_u32): Ditto.
12375 (vfmlalq_lane_low_u32): Ditto.
12376 (vfmlslq_lane_low_u32): Ditto.
12377 (vfmlalq_laneq_low_u32): Ditto.
12378 (vfmlslq_laneq_low_u32): Ditto.
12379 (vfmlal_lane_high_u32): Ditto.
12380 (vfmlsl_lane_high_u32): Ditto.
12381 (vfmlal_laneq_high_u32): Ditto.
12382 (vfmlsl_laneq_high_u32): Ditto.
12383 (vfmlalq_lane_high_u32): Ditto.
12384 (vfmlslq_lane_high_u32): Ditto.
12385 (vfmlalq_laneq_high_u32): Ditto.
12386 (vfmlslq_laneq_high_u32): Ditto.
12387 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12388 (AARCH64_FL_FOR_ARCH8_4): New.
12389 (AARCH64_ISA_F16FML): New ISA flag.
12390 (TARGET_F16FML): New feature flag for fp16fml.
12391 (doc/invoke.texi): Document new fp16fml option.
12393 2018-01-10 Michael Collison <michael.collison@arm.com>
12395 * config/aarch64/aarch64-builtins.c:
12396 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12397 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12398 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12399 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12400 (AARCH64_ISA_SHA3): New ISA flag.
12401 (TARGET_SHA3): New feature flag for sha3.
12402 * config/aarch64/iterators.md (sha512_op): New int attribute.
12403 (CRYPTO_SHA512): New int iterator.
12404 (UNSPEC_SHA512H): New unspec.
12405 (UNSPEC_SHA512H2): Ditto.
12406 (UNSPEC_SHA512SU0): Ditto.
12407 (UNSPEC_SHA512SU1): Ditto.
12408 * config/aarch64/aarch64-simd-builtins.def
12409 (aarch64_crypto_sha512hqv2di): New builtin.
12410 (aarch64_crypto_sha512h2qv2di): Ditto.
12411 (aarch64_crypto_sha512su0qv2di): Ditto.
12412 (aarch64_crypto_sha512su1qv2di): Ditto.
12413 (aarch64_eor3qv8hi): Ditto.
12414 (aarch64_rax1qv2di): Ditto.
12415 (aarch64_xarqv2di): Ditto.
12416 (aarch64_bcaxqv8hi): Ditto.
12417 * config/aarch64/aarch64-simd.md:
12418 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12419 (aarch64_crypto_sha512su0qv2di): Ditto.
12420 (aarch64_crypto_sha512su1qv2di): Ditto.
12421 (aarch64_eor3qv8hi): Ditto.
12422 (aarch64_rax1qv2di): Ditto.
12423 (aarch64_xarqv2di): Ditto.
12424 (aarch64_bcaxqv8hi): Ditto.
12425 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12426 (vsha512h2q_u64): Ditto.
12427 (vsha512su0q_u64): Ditto.
12428 (vsha512su1q_u64): Ditto.
12429 (veor3q_u16): Ditto.
12430 (vrax1q_u64): Ditto.
12431 (vxarq_u64): Ditto.
12432 (vbcaxq_u16): Ditto.
12433 * config/arm/types.md (crypto_sha512): New type attribute.
12434 (crypto_sha3): Ditto.
12435 (doc/invoke.texi): Document new sha3 option.
12437 2018-01-10 Michael Collison <michael.collison@arm.com>
12439 * config/aarch64/aarch64-builtins.c:
12440 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12441 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12442 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12443 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12444 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12445 (AARCH64_ISA_SM4): New ISA flag.
12446 (TARGET_SM4): New feature flag for sm4.
12447 * config/aarch64/aarch64-simd-builtins.def
12448 (aarch64_sm3ss1qv4si): Ditto.
12449 (aarch64_sm3tt1aq4si): Ditto.
12450 (aarch64_sm3tt1bq4si): Ditto.
12451 (aarch64_sm3tt2aq4si): Ditto.
12452 (aarch64_sm3tt2bq4si): Ditto.
12453 (aarch64_sm3partw1qv4si): Ditto.
12454 (aarch64_sm3partw2qv4si): Ditto.
12455 (aarch64_sm4eqv4si): Ditto.
12456 (aarch64_sm4ekeyqv4si): Ditto.
12457 * config/aarch64/aarch64-simd.md:
12458 (aarch64_sm3ss1qv4si): Ditto.
12459 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12460 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12461 (aarch64_sm4eqv4si): Ditto.
12462 (aarch64_sm4ekeyqv4si): Ditto.
12463 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12464 (sm3part_op): Ditto.
12465 (CRYPTO_SM3TT): Ditto.
12466 (CRYPTO_SM3PART): Ditto.
12467 (UNSPEC_SM3SS1): New unspec.
12468 (UNSPEC_SM3TT1A): Ditto.
12469 (UNSPEC_SM3TT1B): Ditto.
12470 (UNSPEC_SM3TT2A): Ditto.
12471 (UNSPEC_SM3TT2B): Ditto.
12472 (UNSPEC_SM3PARTW1): Ditto.
12473 (UNSPEC_SM3PARTW2): Ditto.
12474 (UNSPEC_SM4E): Ditto.
12475 (UNSPEC_SM4EKEY): Ditto.
12476 * config/aarch64/constraints.md (Ui2): New constraint.
12477 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12478 * config/arm/types.md (crypto_sm3): New type attribute.
12479 (crypto_sm4): Ditto.
12480 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12481 (vsm3tt1aq_u32): Ditto.
12482 (vsm3tt1bq_u32): Ditto.
12483 (vsm3tt2aq_u32): Ditto.
12484 (vsm3tt2bq_u32): Ditto.
12485 (vsm3partw1q_u32): Ditto.
12486 (vsm3partw2q_u32): Ditto.
12487 (vsm4eq_u32): Ditto.
12488 (vsm4ekeyq_u32): Ditto.
12489 (doc/invoke.texi): Document new sm4 option.
12491 2018-01-10 Michael Collison <michael.collison@arm.com>
12493 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12494 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12495 (AARCH64_FL_FOR_ARCH8_4): New.
12496 (AARCH64_FL_V8_4): New flag.
12497 (doc/invoke.texi): Document new armv8.4-a option.
12499 2018-01-10 Michael Collison <michael.collison@arm.com>
12501 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12502 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12503 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12504 * config/aarch64/aarch64-option-extension.def: Add
12505 AARCH64_OPT_EXTENSION of 'sha2'.
12506 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12507 (crypto): Disable sha2 and aes if crypto disabled.
12508 (crypto): Enable aes and sha2 if enabled.
12509 (simd): Disable sha2 and aes if simd disabled.
12510 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12512 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12513 (TARGET_SHA2): New feature flag for sha2.
12514 (TARGET_AES): New feature flag for aes.
12515 * config/aarch64/aarch64-simd.md:
12516 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12517 conditional on TARGET_AES.
12518 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12519 (aarch64_crypto_sha1hsi): Make pattern conditional
12521 (aarch64_crypto_sha1hv4si): Ditto.
12522 (aarch64_be_crypto_sha1hv4si): Ditto.
12523 (aarch64_crypto_sha1su1v4si): Ditto.
12524 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12525 (aarch64_crypto_sha1su0v4si): Ditto.
12526 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12527 (aarch64_crypto_sha256su0v4si): Ditto.
12528 (aarch64_crypto_sha256su1v4si): Ditto.
12529 (doc/invoke.texi): Document new aes and sha2 options.
12531 2018-01-10 Martin Sebor <msebor@redhat.com>
12533 PR tree-optimization/83781
12534 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12537 2018-01-11 Martin Sebor <msebor@gmail.com>
12538 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12540 PR tree-optimization/83501
12541 PR tree-optimization/81703
12543 * tree-ssa-strlen.c (get_string_cst): Rename...
12544 (get_string_len): ...to this. Handle global constants.
12545 (handle_char_store): Adjust.
12547 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12548 Jim Wilson <jimw@sifive.com>
12550 * config/riscv/riscv-protos.h (riscv_output_return): New.
12551 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12552 (riscv_attribute_table, riscv_output_return),
12553 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12554 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12555 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12557 (riscv_expand_prologue): Add early return for naked function.
12558 (riscv_expand_epilogue): Likewise.
12559 (riscv_function_ok_for_sibcall): Return false for naked function.
12560 (riscv_set_current_function): New.
12561 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12562 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12563 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12564 * doc/extend.texi (RISC-V Function Attributes): New.
12566 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12568 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12569 check for 128-bit long double before checking TCmode.
12570 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12571 128-bit long doubles before checking TFmode or TCmode.
12572 (FLOAT128_IBM_P): Likewise.
12574 2018-01-10 Martin Sebor <msebor@redhat.com>
12576 PR tree-optimization/83671
12577 * builtins.c (c_strlen): Unconditionally return zero for the empty
12579 Use -Warray-bounds for warnings.
12580 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12581 for non-constant array indices with COMPONENT_REF, arrays of
12582 arrays, and pointers to arrays.
12583 (gimple_fold_builtin_strlen): Determine and set length range for
12584 non-constant character arrays.
12586 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12588 PR middle-end/81897
12589 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12592 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12594 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12596 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12599 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12600 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12601 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12602 indexed_or_indirect_operand predicate.
12603 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12604 (*vsx_le_perm_load_v8hi): Likewise.
12605 (*vsx_le_perm_load_v16qi): Likewise.
12606 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12607 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12608 (*vsx_le_perm_store_v8hi): Likewise.
12609 (*vsx_le_perm_store_v16qi): Likewise.
12610 (eight unnamed splitters): Likewise.
12612 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12614 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12615 * config/rs6000/emmintrin.h: Likewise.
12616 * config/rs6000/mmintrin.h: Likewise.
12617 * config/rs6000/xmmintrin.h: Likewise.
12619 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12622 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12624 * tree.c (tree_nop_conversion): Return true for location wrapper
12626 (maybe_wrap_with_location): New function.
12627 (selftest::check_strip_nops): New function.
12628 (selftest::test_location_wrappers): New function.
12629 (selftest::tree_c_tests): Call it.
12630 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12631 (maybe_wrap_with_location): New decl.
12632 (EXPR_LOCATION_WRAPPER_P): New macro.
12633 (location_wrapper_p): New inline function.
12634 (tree_strip_any_location_wrapper): New inline function.
12636 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12639 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12640 stack_realign_offset for the largest alignment of stack slot
12642 (ix86_find_max_used_stack_alignment): New function.
12643 (ix86_finalize_stack_frame_flags): Use it. Set
12644 max_used_stack_alignment if we don't realign stack.
12645 * config/i386/i386.h (machine_function): Add
12646 max_used_stack_alignment.
12648 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12650 * config/arm/arm.opt (-mbranch-cost): New option.
12651 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12654 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12657 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12658 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12660 2018-01-10 Richard Biener <rguenther@suse.de>
12663 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12664 early out so it also covers the case where we have a non-NULL
12667 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12669 PR tree-optimization/83753
12670 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12671 for non-strided grouped accesses if the number of elements is 1.
12673 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12676 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12677 * i386.h (TARGET_USE_GATHER): Define.
12678 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12680 2018-01-10 Martin Liska <mliska@suse.cz>
12683 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12684 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12686 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12687 CLEANUP_NO_PARTITIONING is not set.
12689 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12691 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12692 for vectors, as a partial revert of r254296.
12693 * rtl.h (const_vec_p): Delete.
12694 (const_vec_duplicate_p): Don't test for vector CONSTs.
12695 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12696 * expmed.c (make_tree): Likewise.
12699 * common.md (E, F): Use CONSTANT_P instead of checking for
12701 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12702 checking for CONST_VECTOR.
12704 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12706 PR middle-end/83575
12707 * predict.c (force_edge_cold): Handle in more sane way edges
12708 with no prediction.
12710 2018-01-09 Carl Love <cel@us.ibm.com>
12712 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12714 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12715 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12716 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12717 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12718 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12719 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12720 * config/rs6000/rs6000-protos.h: Add extern defition for
12721 rs6000_generate_float2_double_code.
12722 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12724 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12725 (float2_v2df): Add define_expand.
12727 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12730 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12731 op_mode in the force_to_mode call.
12733 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12735 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12736 instead of checking each element individually.
12737 (aarch64_evpc_uzp): Likewise.
12738 (aarch64_evpc_zip): Likewise.
12739 (aarch64_evpc_ext): Likewise.
12740 (aarch64_evpc_rev): Likewise.
12741 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12742 instead of checking each element individually. Return true without
12744 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12745 whether all selected elements come from the same input, instead of
12746 checking each element individually. Remove calls to gen_rtx_REG,
12747 start_sequence and end_sequence and instead assert that no rtl is
12750 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12752 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12753 order of HIGH and CONST checks.
12755 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12757 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12758 if the destination isn't an SSA_NAME.
12760 2018-01-09 Richard Biener <rguenther@suse.de>
12762 PR tree-optimization/83668
12763 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12765 (canonicalize_loop_form): ... here, renamed from ...
12766 (canonicalize_loop_closed_ssa_form): ... this and amended to
12767 swap successor edges for loop exit blocks to make us use
12768 the RPO order we need for initial schedule generation.
12770 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12772 PR tree-optimization/64811
12773 * match.pd: When optimizing comparisons with Inf, avoid
12774 introducing or losing exceptions from comparisons with NaN.
12776 2018-01-09 Martin Liska <mliska@suse.cz>
12779 * asan.c (shadow_mem_size): Add gcc_assert.
12781 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12783 Don't save registers in main().
12786 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12787 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12788 * config/avr/avr.c (avr_set_current_function): Don't error if
12789 naked, OS_task or OS_main are specified at the same time.
12790 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12792 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12794 * common/config/avr/avr-common.c (avr_option_optimization_table):
12795 Switch on -mmain-is-OS_task for optimizing compilations.
12797 2018-01-09 Richard Biener <rguenther@suse.de>
12799 PR tree-optimization/83572
12800 * graphite.c: Include cfganal.h.
12801 (graphite_transform_loops): Connect infinite loops to exit
12802 and remove fake edges at the end.
12804 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12806 * ipa-inline.c (edge_badness): Revert accidental checkin.
12808 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12811 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12812 symbols; not inline clones.
12814 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12817 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12818 hard registers. Formatting fixes.
12820 PR preprocessor/83722
12821 * gcc.c (try_generate_repro): Pass
12822 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12823 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12826 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12827 Kito Cheng <kito.cheng@gmail.com>
12829 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12830 (riscv_leaf_function_p): Delete.
12831 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12833 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12835 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12837 (do_ifelse): New function.
12838 (do_isel): New function.
12839 (do_sub3): New function.
12840 (do_add3): New function.
12841 (do_load_mask_compare): New function.
12842 (do_overlap_load_compare): New function.
12843 (expand_compare_loop): New function.
12844 (expand_block_compare): Call expand_compare_loop() when appropriate.
12845 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12846 option description.
12847 (-mblock-compare-inline-loop-limit): New option.
12849 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12852 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12853 Reverse order of second and third operands in first alternative.
12854 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12855 of first and second elements in UNSPEC_VPERMR vector.
12856 (altivec_expand_vec_perm_le): Likewise.
12858 2018-01-08 Jeff Law <law@redhat.com>
12860 PR rtl-optimizatin/81308
12861 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12862 (process_switch): If group_case_labels makes a change, then set
12864 (pass_convert_switch::execute): If a switch is converted, then
12865 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12867 PR rtl-optimization/81308
12868 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12871 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12873 PR target/83663 - Revert r255946
12874 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12875 generation for cases where splatting a value is not useful.
12876 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12877 across a vec_duplicate and a paradoxical subreg forming a vector
12878 mode to a vec_concat.
12880 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12882 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12883 -march=armv8.3-a variants.
12884 * config/arm/t-multilib: Likewise.
12885 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12887 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12889 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12891 (cceq_ior_compare_complement): Give it a name so I can use it, and
12892 change boolean_or_operator predicate to boolean_operator so it can
12893 be used to generate a crand.
12894 (eqne): New code iterator.
12895 (bd/bd_neg): New code_attrs.
12896 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12897 a single define_insn.
12898 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12899 decrement (bdnzt/bdnzf/bdzt/bdzf).
12900 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12901 with the new names of the branch decrement patterns, and added the
12902 names of the branch decrement conditional patterns.
12904 2018-01-08 Richard Biener <rguenther@suse.de>
12906 PR tree-optimization/83563
12907 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12910 2018-01-08 Richard Biener <rguenther@suse.de>
12912 PR middle-end/83713
12913 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12915 2018-01-08 Richard Biener <rguenther@suse.de>
12917 PR tree-optimization/83685
12918 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12919 references to abnormals.
12921 2018-01-08 Richard Biener <rguenther@suse.de>
12924 * dwarf2out.c (output_indirect_strings): Handle empty
12925 skeleton_debug_str_hash.
12926 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12928 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12930 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12931 (emit_store_direct): Likewise.
12932 (arc_trampoline_adjust_address): Likewise.
12933 (arc_asm_trampoline_template): New function.
12934 (arc_initialize_trampoline): Use asm_trampoline_template.
12935 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12936 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12937 * config/arc/arc.md (flush_icache): Delete pattern.
12939 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12941 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12942 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12945 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12948 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12949 by not USED_FOR_TARGET.
12950 (make_pass_resolve_sw_modes): Likewise.
12952 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12954 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12957 2018-01-08 Richard Biener <rguenther@suse.de>
12959 PR middle-end/83580
12960 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12962 2018-01-08 Richard Biener <rguenther@suse.de>
12964 PR middle-end/83517
12965 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12967 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12969 PR middle-end/81897
12970 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12971 basic blocks with a small number of successors.
12972 (convert_control_dep_chain_into_preds): Improve handling of
12974 (dump_predicates): Split apart into...
12975 (dump_pred_chain): ...here...
12976 (dump_pred_info): ...and here.
12977 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12978 (can_chain_union_be_invalidated_p): Improve check for invalidation
12980 (uninit_uses_cannot_happen): Avoid unnecessary if
12981 convert_control_dep_chain_into_preds yielded nothing.
12983 2018-01-06 Martin Sebor <msebor@redhat.com>
12985 PR tree-optimization/83640
12986 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12987 subtracting negative offset from size.
12988 (builtin_access::overlap): Adjust offset bounds of the access to fall
12989 within the size of the object if possible.
12991 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12993 PR rtl-optimization/83699
12994 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12995 extract_bit_field_as_subreg to cases in which the extracted
12996 value is also a vector.
12998 * lra-constraints.c (process_alt_operands): Test for the equivalence
12999 substitutions when detecting a possible reload cycle.
13001 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13004 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13005 by default if flag_selective_schedling{,2}. Formatting fixes.
13007 PR rtl-optimization/83682
13008 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13009 if it has non-VECTOR_MODE element mode.
13010 (vec_duplicate_p): Likewise.
13012 PR middle-end/83694
13013 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13014 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13016 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13019 * config/i386/i386-builtin.def
13020 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13021 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13022 Require also OPTION_MASK_ISA_AVX512F in addition to
13023 OPTION_MASK_ISA_GFNI.
13024 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13025 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13026 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13027 to OPTION_MASK_ISA_GFNI.
13028 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13029 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13030 OPTION_MASK_ISA_AVX512BW.
13031 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13032 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13033 addition to OPTION_MASK_ISA_GFNI.
13034 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13035 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13036 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13037 to OPTION_MASK_ISA_GFNI.
13038 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13039 a requirement for all ISAs rather than any of them with a few
13041 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13043 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13044 bitmasks to be enabled with 3 exceptions, instead of requiring any
13045 enabled ISA with lots of exceptions.
13046 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13047 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13048 Change avx512bw in isa attribute to avx512f.
13049 * config/i386/sgxintrin.h: Add license boilerplate.
13050 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13051 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13052 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13053 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13055 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13056 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13057 temporarily sse2 rather than sse if not enabled already.
13060 * config/i386/sse.md (VI248_VLBW): Rename to ...
13061 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13062 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13063 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13064 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13065 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13066 mode iterator instead of VI248_VLBW.
13068 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13070 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13071 (record_modified): Skip clobbers; add debug output.
13072 (param_change_prob): Use sreal frequencies.
13074 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13076 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13077 punt for user-aligned variables.
13079 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13081 * tree-chrec.c (chrec_contains_symbols): Return true for
13084 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13087 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13088 of (x|y) == x for BICS pattern.
13090 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13092 PR tree-optimization/83605
13093 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13094 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13097 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13099 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13100 * config/epiphany/rtems.h: New file.
13102 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13103 Uros Bizjak <ubizjak@gmail.com>
13106 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13107 QIreg_operand instead of register_operand predicate.
13108 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13109 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13110 comments instead of -fmitigate[-_]rop.
13112 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13115 * cgraphunit.c (symbol_table::compile): Switch to text_section
13116 before calling assembly_start debug hook.
13117 * run-rtl-passes.c (run_rtl_passes): Likewise.
13120 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13122 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13123 range_int_cst_p rather than !symbolic_range_p before calling
13124 extract_range_from_multiplicative_op_1.
13126 2018-01-04 Jeff Law <law@redhat.com>
13128 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13129 redundant test in assertion.
13131 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13133 * doc/rtl.texi: Document machine_mode wrapper classes.
13135 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13137 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13138 using tree_to_uhwi.
13140 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13142 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13143 the VEC_PERM_EXPR fold to fail.
13145 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13148 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13149 to switched_sections.
13151 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13154 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13155 test for d.testing.
13157 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13160 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13161 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13163 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13166 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13167 is BLKmode and bitpos not zero or mode change is needed.
13169 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13172 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13175 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13178 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13179 instead of MULT rtx. Update all corresponding splitters.
13180 (*saddl_se): Ditto.
13181 (*ssub<modesuffix>): Ditto.
13182 (*ssubl_se): Ditto.
13183 (*cmp_sadd_di): Update split patterns.
13184 (*cmp_sadd_si): Ditto.
13185 (*cmp_sadd_sidi): Ditto.
13186 (*cmp_ssub_di): Ditto.
13187 (*cmp_ssub_si): Ditto.
13188 (*cmp_ssub_sidi): Ditto.
13189 * config/alpha/predicates.md (const23_operand): New predicate.
13190 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13191 Look for ASHIFT, not MULT inner operand.
13192 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13194 2018-01-04 Martin Liska <mliska@suse.cz>
13196 PR gcov-profile/83669
13197 * gcov.c (output_intermediate_file): Add version to intermediate
13199 * doc/gcov.texi: Document new field 'version' in intermediate
13200 file format. Fix location of '-k' option of gcov command.
13202 2018-01-04 Martin Liska <mliska@suse.cz>
13205 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13207 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13209 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13211 2018-01-03 Martin Sebor <msebor@redhat.com>
13213 PR tree-optimization/83655
13214 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13215 checking calls with invalid arguments.
13217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13219 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13220 (vectorizable_mask_load_store): Delete.
13221 (vectorizable_call): Return false for masked loads and stores.
13222 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13223 instead of gimple_assign_rhs1.
13224 (vectorizable_load): Handle IFN_MASK_LOAD.
13225 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13229 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13231 (vectorizable_mask_load_store): ...here.
13232 (vectorizable_load): ...and here.
13234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13236 * tree-vect-stmts.c (vect_build_all_ones_mask)
13237 (vect_build_zero_merge_argument): New functions, split out from...
13238 (vectorizable_load): ...here.
13240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13242 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13244 (vectorizable_mask_load_store): ...here.
13245 (vectorizable_store): ...and here.
13247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13249 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13251 (vectorizable_mask_load_store): ...here.
13253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13255 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13256 (vect_model_store_cost): Take a vec_load_store_type instead of a
13258 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13259 (vect_model_store_cost): Take a vec_load_store_type instead of a
13261 (vectorizable_mask_load_store): Update accordingly.
13262 (vectorizable_store): Likewise.
13263 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13267 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13268 IFN_MASK_LOAD calls here rather than...
13269 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13272 Alan Hayward <alan.hayward@arm.com>
13273 David Sherwood <david.sherwood@arm.com>
13275 * expmed.c (extract_bit_field_1): For vector extracts,
13276 fall back to extract_bit_field_as_subreg if vec_extract
13279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13280 Alan Hayward <alan.hayward@arm.com>
13281 David Sherwood <david.sherwood@arm.com>
13283 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13284 they are variable or constant sized.
13285 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13286 slots for constant-sized data.
13288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13289 Alan Hayward <alan.hayward@arm.com>
13290 David Sherwood <david.sherwood@arm.com>
13292 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13293 handling COND_EXPRs with boolean comparisons, try to find a better
13294 basis for the mask type than the boolean itself.
13296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13298 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13299 is calculated and how it can be overridden.
13300 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13301 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13303 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13307 Alan Hayward <alan.hayward@arm.com>
13308 David Sherwood <david.sherwood@arm.com>
13310 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13311 Remove the mode argument.
13312 (aarch64_simd_valid_immediate): Remove the mode and inverse
13314 * config/aarch64/iterators.md (bitsize): New iterator.
13315 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13316 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13317 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13318 aarch64_simd_valid_immediate.
13319 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13320 (aarch64_reg_or_bic_imm): Likewise.
13321 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13322 with an insn_type enum and msl with a modifier_type enum.
13323 Replace element_width with a scalar_mode. Change the shift
13324 to unsigned int. Add constructors for scalar_float_mode and
13325 scalar_int_mode elements.
13326 (aarch64_vect_float_const_representable_p): Delete.
13327 (aarch64_can_const_movi_rtx_p)
13328 (aarch64_simd_scalar_immediate_valid_for_move)
13329 (aarch64_simd_make_constant): Update call to
13330 aarch64_simd_valid_immediate.
13331 (aarch64_advsimd_valid_immediate_hs): New function.
13332 (aarch64_advsimd_valid_immediate): Likewise.
13333 (aarch64_simd_valid_immediate): Remove mode and inverse
13334 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13335 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13336 and aarch64_float_const_representable_p on the result.
13337 (aarch64_output_simd_mov_immediate): Remove mode argument.
13338 Update call to aarch64_simd_valid_immediate and use of
13339 simd_immediate_info.
13340 (aarch64_output_scalar_simd_mov_immediate): Update call
13343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13344 Alan Hayward <alan.hayward@arm.com>
13345 David Sherwood <david.sherwood@arm.com>
13347 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13348 (mode_nunits): Likewise CONST_MODE_NUNITS.
13349 * machmode.def (ADJUST_NUNITS): Document.
13350 * genmodes.c (mode_data::need_nunits_adj): New field.
13351 (blank_mode): Update accordingly.
13352 (adj_nunits): New variable.
13353 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13355 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13356 listed in adj_nunits.
13357 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13358 listed in adj_nunits. Don't emit case statements for such modes.
13359 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13360 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13361 nothing if adj_nunits is nonnull.
13362 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13363 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13364 (emit_mode_fbit): Update use of print_maybe_const_decl.
13365 (emit_move_size): Likewise. Treat the array as non-const
13367 (emit_mode_adjustments): Handle adj_nunits.
13369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13371 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13372 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13373 (VECTOR_MODES): Use it.
13374 (make_vector_modes): Take the prefix as an argument.
13376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13377 Alan Hayward <alan.hayward@arm.com>
13378 David Sherwood <david.sherwood@arm.com>
13380 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13381 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13382 for MODE_VECTOR_BOOL.
13383 * machmode.def (VECTOR_BOOL_MODE): Document.
13384 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13385 (make_vector_bool_mode): New function.
13386 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13388 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13389 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13391 * stor-layout.c (int_mode_for_mode): Likewise.
13392 * tree.c (build_vector_type_for_mode): Likewise.
13393 * varasm.c (output_constant_pool_2): Likewise.
13394 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13395 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13396 for MODE_VECTOR_BOOL.
13397 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13398 of mode class checks.
13399 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13400 instead of a list of mode class checks.
13401 (expand_vector_scalar_condition): Likewise.
13402 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13405 Alan Hayward <alan.hayward@arm.com>
13406 David Sherwood <david.sherwood@arm.com>
13408 * machmode.h (mode_size): Change from unsigned short to
13410 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13411 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13412 or if measurement_type is not polynomial.
13413 (fixed_size_mode::includes_p): Check for constant-sized modes.
13414 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13415 return a poly_uint16 rather than an unsigned short.
13416 (emit_mode_size): Change the type of mode_size from unsigned short
13417 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13418 (emit_mode_adjustments): Cope with polynomial vector sizes.
13419 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13421 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13423 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13424 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13425 * caller-save.c (setup_save_areas): Likewise.
13426 (replace_reg_with_saved_mem): Likewise.
13427 * calls.c (emit_library_call_value_1): Likewise.
13428 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13429 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13430 (gen_lowpart_for_combine): Likewise.
13431 * convert.c (convert_to_integer_1): Likewise.
13432 * cse.c (equiv_constant, cse_insn): Likewise.
13433 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13434 (cselib_subst_to_values): Likewise.
13435 * dce.c (word_dce_process_block): Likewise.
13436 * df-problems.c (df_word_lr_mark_ref): Likewise.
13437 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13438 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13439 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13440 (rtl_for_decl_location): Likewise.
13441 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13442 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13443 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13444 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13445 (expand_expr_real_1): Likewise.
13446 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13447 (pad_below): Likewise.
13448 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13449 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13450 * ira.c (get_subreg_tracking_sizes): Likewise.
13451 * ira-build.c (ira_create_allocno_objects): Likewise.
13452 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13453 (ira_sort_regnos_for_alter_reg): Likewise.
13454 * ira-costs.c (record_operand_costs): Likewise.
13455 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13456 (resolve_simple_move): Likewise.
13457 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13458 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13459 (lra_constraints): Likewise.
13460 (CONST_POOL_OK_P): Reject variable-sized modes.
13461 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13462 (add_pseudo_to_slot, lra_spill): Likewise.
13463 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13464 * optabs-query.c (get_best_extraction_insn): Likewise.
13465 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13466 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13467 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13468 * recog.c (offsettable_address_addr_space_p): Likewise.
13469 * regcprop.c (maybe_mode_change): Likewise.
13470 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13471 * regrename.c (build_def_use): Likewise.
13472 * regstat.c (dump_reg_info): Likewise.
13473 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13474 (find_reloads, find_reloads_subreg_address): Likewise.
13475 * reload1.c (eliminate_regs_1): Likewise.
13476 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13477 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13478 (simplify_binary_operation_1, simplify_subreg): Likewise.
13479 * targhooks.c (default_function_arg_padding): Likewise.
13480 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13481 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13482 (verify_gimple_assign_ternary): Likewise.
13483 * tree-inline.c (estimate_move_cost): Likewise.
13484 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13485 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13486 (get_address_cost_ainc): Likewise.
13487 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13488 (vect_supportable_dr_alignment): Likewise.
13489 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13490 (vectorizable_reduction): Likewise.
13491 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13492 (vectorizable_operation, vectorizable_load): Likewise.
13493 * tree.c (build_same_sized_truth_vector_type): Likewise.
13494 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13495 * var-tracking.c (emit_note_insn_var_location): Likewise.
13496 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13497 (ADDR_VEC_ALIGN): Likewise.
13499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13500 Alan Hayward <alan.hayward@arm.com>
13501 David Sherwood <david.sherwood@arm.com>
13503 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13505 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13506 or if measurement_type is polynomial.
13507 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13508 * combine.c (make_extraction): Likewise.
13509 * dse.c (find_shift_sequence): Likewise.
13510 * dwarf2out.c (mem_loc_descriptor): Likewise.
13511 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13512 (extract_bit_field, extract_low_bits): Likewise.
13513 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13514 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13515 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13516 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13517 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13518 * reload.c (find_reloads): Likewise.
13519 * reload1.c (alter_reg): Likewise.
13520 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13521 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13522 * tree-if-conv.c (predicate_mem_writes): Likewise.
13523 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13524 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13525 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13526 * valtrack.c (dead_debug_insert_temp): Likewise.
13527 * varasm.c (mergeable_constant_section): Likewise.
13528 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13531 Alan Hayward <alan.hayward@arm.com>
13532 David Sherwood <david.sherwood@arm.com>
13534 * expr.c (expand_assignment): Cope with polynomial mode sizes
13535 when assigning to a CONCAT.
13537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13538 Alan Hayward <alan.hayward@arm.com>
13539 David Sherwood <david.sherwood@arm.com>
13541 * machmode.h (mode_precision): Change from unsigned short to
13543 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13545 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13546 or if measurement_type is not polynomial.
13547 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13548 in which the mode is already known to be a scalar_int_mode.
13549 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13550 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13552 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13553 for GET_MODE_PRECISION.
13554 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13555 for GET_MODE_PRECISION.
13556 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13558 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13559 (expand_field_assignment, make_extraction): Likewise.
13560 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13561 (get_last_value): Likewise.
13562 * convert.c (convert_to_integer_1): Likewise.
13563 * cse.c (cse_insn): Likewise.
13564 * expr.c (expand_expr_real_1): Likewise.
13565 * lra-constraints.c (simplify_operand_subreg): Likewise.
13566 * optabs-query.c (can_atomic_load_p): Likewise.
13567 * optabs.c (expand_atomic_load): Likewise.
13568 (expand_atomic_store): Likewise.
13569 * ree.c (combine_reaching_defs): Likewise.
13570 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13571 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13572 * tree.h (type_has_mode_precision_p): Likewise.
13573 * ubsan.c (instrument_si_overflow): Likewise.
13575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13576 Alan Hayward <alan.hayward@arm.com>
13577 David Sherwood <david.sherwood@arm.com>
13579 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13580 polynomial numbers of units.
13581 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13582 (valid_vector_subparts_p): New function.
13583 (build_vector_type): Remove temporary shim and take the number
13584 of units as a poly_uint64 rather than an int.
13585 (build_opaque_vector_type): Take the number of units as a
13586 poly_uint64 rather than an int.
13587 * tree.c (build_vector_from_ctor): Handle polynomial
13588 TYPE_VECTOR_SUBPARTS.
13589 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13590 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13591 (build_vector_from_val): If the number of units is variable,
13592 use build_vec_duplicate_cst for constant operands and
13593 VEC_DUPLICATE_EXPR otherwise.
13594 (make_vector_type): Remove temporary is_constant ().
13595 (build_vector_type, build_opaque_vector_type): Take the number of
13596 units as a poly_uint64 rather than an int.
13597 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13599 * cfgexpand.c (expand_debug_expr): Likewise.
13600 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13601 (store_constructor, expand_expr_real_1): Likewise.
13602 (const_scalar_mask_from_tree): Likewise.
13603 * fold-const-call.c (fold_const_reduction): Likewise.
13604 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13605 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13606 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13607 (fold_relational_const): Likewise.
13608 (native_interpret_vector): Likewise. Change the size from an
13609 int to an unsigned int.
13610 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13611 TYPE_VECTOR_SUBPARTS.
13612 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13613 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13614 duplicating a non-constant operand into a variable-length vector.
13615 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13616 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13617 * ipa-icf.c (sem_variable::equals): Likewise.
13618 * match.pd: Likewise.
13619 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13620 * print-tree.c (print_node): Likewise.
13621 * stor-layout.c (layout_type): Likewise.
13622 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13623 * tree-cfg.c (verify_gimple_comparison): Likewise.
13624 (verify_gimple_assign_binary): Likewise.
13625 (verify_gimple_assign_ternary): Likewise.
13626 (verify_gimple_assign_single): Likewise.
13627 * tree-pretty-print.c (dump_generic_node): Likewise.
13628 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13629 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13630 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13631 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13632 (vect_shift_permute_load_chain): Likewise.
13633 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13634 (expand_vector_condition, optimize_vector_constructor): Likewise.
13635 (lower_vec_perm, get_compute_type): Likewise.
13636 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13637 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13638 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13639 (vect_recog_mask_conversion_pattern): Likewise.
13640 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13641 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13642 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13643 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13644 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13645 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13646 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13647 (supportable_widening_operation): Likewise.
13648 (supportable_narrowing_operation): Likewise.
13649 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13651 * varasm.c (output_constant): Likewise.
13653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13654 Alan Hayward <alan.hayward@arm.com>
13655 David Sherwood <david.sherwood@arm.com>
13657 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13658 so that both the length == 3 and length != 3 cases set up their
13659 own permute vectors. Add comments explaining why we know the
13660 number of elements is constant.
13661 (vect_permute_load_chain): Likewise.
13663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13664 Alan Hayward <alan.hayward@arm.com>
13665 David Sherwood <david.sherwood@arm.com>
13667 * machmode.h (mode_nunits): Change from unsigned char to
13669 (ONLY_FIXED_SIZE_MODES): New macro.
13670 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13671 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13672 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13674 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13675 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13676 or if measurement_type is not polynomial.
13677 * genmodes.c (ZERO_COEFFS): New macro.
13678 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13680 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13681 Use ZERO_COEFFS when emitting initializers.
13682 * data-streamer.h (bp_pack_poly_value): New function.
13683 (bp_unpack_poly_value): Likewise.
13684 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13685 for GET_MODE_NUNITS.
13686 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13687 for GET_MODE_NUNITS.
13688 * tree.c (make_vector_type): Remove temporary shim and make
13689 the real function take the number of units as a poly_uint64
13690 rather than an int.
13691 (build_vector_type_for_mode): Handle polynomial nunits.
13692 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13693 * emit-rtl.c (const_vec_series_p_1): Likewise.
13694 (gen_rtx_CONST_VECTOR): Likewise.
13695 * fold-const.c (test_vec_duplicate_folding): Likewise.
13696 * genrecog.c (validate_pattern): Likewise.
13697 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13698 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13699 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13700 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13701 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13702 * rtlanal.c (subreg_get_info): Likewise.
13703 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13704 (vect_grouped_load_supported): Likewise.
13705 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13706 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13707 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13708 (simplify_const_unary_operation, simplify_binary_operation_1)
13709 (simplify_const_binary_operation, simplify_ternary_operation)
13710 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13711 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13712 instead of CONST_VECTOR_NUNITS.
13713 * varasm.c (output_constant_pool_2): Likewise.
13714 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13715 explicit-encoded elements in the XVEC for variable-length vectors.
13717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13719 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13722 Alan Hayward <alan.hayward@arm.com>
13723 David Sherwood <david.sherwood@arm.com>
13725 * coretypes.h (fixed_size_mode): Declare.
13726 (fixed_size_mode_pod): New typedef.
13727 * builtins.h (target_builtins::x_apply_args_mode)
13728 (target_builtins::x_apply_result_mode): Change type to
13729 fixed_size_mode_pod.
13730 * builtins.c (apply_args_size, apply_result_size, result_vector)
13731 (expand_builtin_apply_args_1, expand_builtin_apply)
13732 (expand_builtin_return): Update accordingly.
13734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13736 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13737 * cselib.c (cselib_hash_rtx): Likewise.
13738 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13739 CONST_VECTOR encoding.
13741 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13742 Jeff Law <law@redhat.com>
13745 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13746 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13747 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13748 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13751 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13752 explicitly probe *sp in a noreturn function if there were any callee
13753 register saves or frame pointer is needed.
13755 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13758 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13759 BLKmode for ternary, binary or unary expressions.
13762 * var-tracking.c (delete_vta_debug_insn): New inline function.
13763 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13764 insns from get_insns () to NULL instead of each bb separately.
13765 Use delete_vta_debug_insn. No longer static.
13766 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13767 delete_vta_debug_insns callers.
13768 * rtl.h (delete_vta_debug_insns): Declare.
13769 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13770 instead of variable_tracking_main.
13772 2018-01-03 Martin Sebor <msebor@redhat.com>
13774 PR tree-optimization/83603
13775 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13776 arguments past the endof the argument list in functions declared
13777 without a prototype.
13778 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13779 Avoid checking when arguments are null.
13781 2018-01-03 Martin Sebor <msebor@redhat.com>
13784 * doc/extend.texi (attribute const): Fix a typo.
13785 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13786 issuing -Wsuggest-attribute for void functions.
13788 2018-01-03 Martin Sebor <msebor@redhat.com>
13790 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13791 offset_int::from instead of wide_int::to_shwi.
13792 (maybe_diag_overlap): Remove assertion.
13793 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13794 * gimple-ssa-sprintf.c (format_directive): Same.
13795 (parse_directive): Same.
13796 (sprintf_dom_walker::compute_format_length): Same.
13797 (try_substitute_return_value): Same.
13799 2018-01-03 Jeff Law <law@redhat.com>
13801 PR middle-end/83654
13802 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13803 non-constant residual for zero at runtime and avoid probing in
13804 that case. Reorganize code for trailing problem to mirror handling
13807 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13809 PR tree-optimization/83501
13810 * tree-ssa-strlen.c (get_string_cst): New.
13811 (handle_char_store): Call get_string_cst.
13813 2018-01-03 Martin Liska <mliska@suse.cz>
13815 PR tree-optimization/83593
13816 * tree-ssa-strlen.c: Include tree-cfg.h.
13817 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13818 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13819 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13821 (strlen_dom_walker::before_dom_children): Call
13822 gimple_purge_dead_eh_edges. Dump tranformation with details
13824 (strlen_dom_walker::before_dom_children): Update call by adding
13825 new argument cleanup_eh.
13826 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13828 2018-01-03 Martin Liska <mliska@suse.cz>
13831 * cif-code.def (VARIADIC_THUNK): New enum value.
13832 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13835 2018-01-03 Jan Beulich <jbeulich@suse.com>
13837 * sse.md (mov<mode>_internal): Tighten condition for when to use
13838 vmovdqu<ssescalarsize> for TI and OI modes.
13840 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13842 Update copyright years.
13844 2018-01-03 Martin Liska <mliska@suse.cz>
13847 * ipa-visibility.c (function_and_variable_visibility): Skip
13848 functions with noipa attribure.
13850 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13852 * gcc.c (process_command): Update copyright notice dates.
13853 * gcov-dump.c (print_version): Ditto.
13854 * gcov.c (print_version): Ditto.
13855 * gcov-tool.c (print_version): Ditto.
13856 * gengtype.c (create_file): Ditto.
13857 * doc/cpp.texi: Bump @copying's copyright year.
13858 * doc/cppinternals.texi: Ditto.
13859 * doc/gcc.texi: Ditto.
13860 * doc/gccint.texi: Ditto.
13861 * doc/gcov.texi: Ditto.
13862 * doc/install.texi: Ditto.
13863 * doc/invoke.texi: Ditto.
13865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13867 * vector-builder.h (vector_builder::m_full_nelts): Change from
13868 unsigned int to poly_uint64.
13869 (vector_builder::full_nelts): Update prototype accordingly.
13870 (vector_builder::new_vector): Likewise.
13871 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13872 (vector_builder::operator ==): Likewise.
13873 (vector_builder::finalize): Likewise.
13874 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13875 Take the number of elements as a poly_uint64 rather than an
13877 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13878 from unsigned int to poly_uint64.
13879 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13880 (vec_perm_indices::new_vector): Likewise.
13881 (vec_perm_indices::length): Likewise.
13882 (vec_perm_indices::nelts_per_input): Likewise.
13883 (vec_perm_indices::input_nelts): Likewise.
13884 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13885 number of elements per input as a poly_uint64 rather than an
13886 unsigned int. Use the original encoding for variable-length
13887 vectors, rather than clamping each individual element.
13888 For the second and subsequent elements in each pattern,
13889 clamp the step and base before clamping their sum.
13890 (vec_perm_indices::series_p): Handle polynomial element counts.
13891 (vec_perm_indices::all_in_range_p): Likewise.
13892 (vec_perm_indices_to_tree): Likewise.
13893 (vec_perm_indices_to_rtx): Likewise.
13894 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13895 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13896 (tree_vector_builder::new_binary_operation): Handle polynomial
13897 element counts. Return false if we need to know the number
13898 of elements at compile time.
13899 * fold-const.c (fold_vec_perm): Punt if the number of elements
13900 isn't known at compile time.
13902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13904 * vec-perm-indices.h (vec_perm_builder): Change element type
13905 from HOST_WIDE_INT to poly_int64.
13906 (vec_perm_indices::element_type): Update accordingly.
13907 (vec_perm_indices::clamp): Handle polynomial element_types.
13908 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13909 (vec_perm_indices::all_in_range_p): Likewise.
13910 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13912 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13913 polynomial vec_perm_indices element types.
13914 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13915 * fold-const.c (fold_vec_perm): Likewise.
13916 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13917 * tree-vect-generic.c (lower_vec_perm): Likewise.
13918 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13919 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13920 element type to HOST_WIDE_INT.
13922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13923 Alan Hayward <alan.hayward@arm.com>
13924 David Sherwood <david.sherwood@arm.com>
13926 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13927 rather than an int. Use plus_constant.
13928 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13929 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13936 a HOST_WIDE_INT to a poly_int64.
13938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13939 Alan Hayward <alan.hayward@arm.com>
13940 David Sherwood <david.sherwood@arm.com>
13942 * calls.c (load_register_parameters): Cope with polynomial
13943 mode sizes. Require a constant size for BLKmode parameters
13944 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13945 forces a parameter to be padded at the lsb end in order to
13946 fill a complete number of words, require the parameter size
13947 to be ordered wrt UNITS_PER_WORD.
13949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13950 Alan Hayward <alan.hayward@arm.com>
13951 David Sherwood <david.sherwood@arm.com>
13953 * reload1.c (spill_stack_slot_width): Change element type
13954 from unsigned int to poly_uint64_pod.
13955 (alter_reg): Treat mode sizes as polynomial.
13957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13958 Alan Hayward <alan.hayward@arm.com>
13959 David Sherwood <david.sherwood@arm.com>
13961 * reload.c (complex_word_subreg_p): New function.
13962 (reload_inner_reg_of_subreg, push_reload): Use it.
13964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13965 Alan Hayward <alan.hayward@arm.com>
13966 David Sherwood <david.sherwood@arm.com>
13968 * lra-constraints.c (process_alt_operands): Reject matched
13969 operands whose sizes aren't ordered.
13970 (match_reload): Refer to this check here.
13972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13973 Alan Hayward <alan.hayward@arm.com>
13974 David Sherwood <david.sherwood@arm.com>
13976 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13977 that the mode size is in the set {1, 2, 4, 8, 16}.
13979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13984 Use plus_constant instead of gen_rtx_PLUS.
13986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13987 Alan Hayward <alan.hayward@arm.com>
13988 David Sherwood <david.sherwood@arm.com>
13990 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13991 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13992 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13993 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13994 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13995 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13996 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13997 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13998 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13999 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14001 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14002 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14003 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14004 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14005 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14006 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14007 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14008 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14009 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14010 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14012 * expr.c (emit_move_resolve_push): Treat the input and result
14013 of PUSH_ROUNDING as a poly_int64.
14014 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14015 (emit_push_insn): Likewise.
14016 * lra-eliminations.c (mark_not_eliminable): Likewise.
14017 * recog.c (push_operand): Likewise.
14018 * reload1.c (elimination_effects): Likewise.
14019 * rtlanal.c (nonzero_bits1): Likewise.
14020 * calls.c (store_one_arg): Likewise. Require the padding to be
14021 known at compile time.
14023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14024 Alan Hayward <alan.hayward@arm.com>
14025 David Sherwood <david.sherwood@arm.com>
14027 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14028 Use plus_constant instead of gen_rtx_PLUS.
14030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14031 Alan Hayward <alan.hayward@arm.com>
14032 David Sherwood <david.sherwood@arm.com>
14034 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14035 rather than an int.
14037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14038 Alan Hayward <alan.hayward@arm.com>
14039 David Sherwood <david.sherwood@arm.com>
14041 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14042 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14043 via stack temporaries. Treat the mode size as polynomial too.
14045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14046 Alan Hayward <alan.hayward@arm.com>
14047 David Sherwood <david.sherwood@arm.com>
14049 * expr.c (expand_expr_real_2): When handling conversions involving
14050 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14051 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14052 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14061 Alan Hayward <alan.hayward@arm.com>
14062 David Sherwood <david.sherwood@arm.com>
14064 * combine.c (can_change_dest_mode): Handle polynomial
14065 REGMODE_NATURAL_SIZE.
14066 * expmed.c (store_bit_field_1): Likewise.
14067 * expr.c (store_constructor): Likewise.
14068 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14069 and polynomial REGMODE_NATURAL_SIZE.
14070 (gen_lowpart_common): Likewise.
14071 * reginfo.c (record_subregs_of_mode): Likewise.
14072 * rtlanal.c (read_modify_subreg_p): Likewise.
14074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14075 Alan Hayward <alan.hayward@arm.com>
14076 David Sherwood <david.sherwood@arm.com>
14078 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14079 numbers of elements.
14081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14082 Alan Hayward <alan.hayward@arm.com>
14083 David Sherwood <david.sherwood@arm.com>
14085 * match.pd: Cope with polynomial numbers of vector elements.
14087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14088 Alan Hayward <alan.hayward@arm.com>
14089 David Sherwood <david.sherwood@arm.com>
14091 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14092 in a POINTER_PLUS_EXPR.
14094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14095 Alan Hayward <alan.hayward@arm.com>
14096 David Sherwood <david.sherwood@arm.com>
14098 * omp-simd-clone.c (simd_clone_subparts): New function.
14099 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14100 (ipa_simd_modify_function_body): Likewise.
14102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14103 Alan Hayward <alan.hayward@arm.com>
14104 David Sherwood <david.sherwood@arm.com>
14106 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14107 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14108 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14109 (expand_vector_condition, vector_element): Likewise.
14110 (subparts_gt): New function.
14111 (get_compute_type): Use subparts_gt.
14112 (count_type_subparts): Delete.
14113 (expand_vector_operations_1): Use subparts_gt instead of
14114 count_type_subparts.
14116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14117 Alan Hayward <alan.hayward@arm.com>
14118 David Sherwood <david.sherwood@arm.com>
14120 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14121 (vect_compile_time_alias): ...this new function. Do the calculation
14122 on poly_ints rather than trees.
14123 (vect_prune_runtime_alias_test_list): Update call accordingly.
14125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14126 Alan Hayward <alan.hayward@arm.com>
14127 David Sherwood <david.sherwood@arm.com>
14129 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14131 (vect_schedule_slp_instance): Likewise.
14133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14134 Alan Hayward <alan.hayward@arm.com>
14135 David Sherwood <david.sherwood@arm.com>
14137 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14138 constant and extern definitions for variable-length vectors.
14139 (vect_get_constant_vectors): Note that the number of units
14140 is known to be constant.
14142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14143 Alan Hayward <alan.hayward@arm.com>
14144 David Sherwood <david.sherwood@arm.com>
14146 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14147 of units as polynomial. Choose between WIDE and NARROW based
14150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14151 Alan Hayward <alan.hayward@arm.com>
14152 David Sherwood <david.sherwood@arm.com>
14154 * tree-vect-stmts.c (simd_clone_subparts): New function.
14155 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14158 Alan Hayward <alan.hayward@arm.com>
14159 David Sherwood <david.sherwood@arm.com>
14161 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14162 vectors as polynomial. Use build_index_vector for
14163 IFN_GOMP_SIMD_LANE.
14165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14166 Alan Hayward <alan.hayward@arm.com>
14167 David Sherwood <david.sherwood@arm.com>
14169 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14170 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14171 for variable-length vectors.
14172 (vectorizable_mask_load_store): Treat the number of units as
14173 polynomial, asserting that it is constant if the condition has
14174 already been enforced.
14175 (vectorizable_store, vectorizable_load): 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-loop.c (vectorizable_live_operation): Treat the number
14182 of units as polynomial. Punt if we can't tell at compile time
14183 which vector contains the final result.
14185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14186 Alan Hayward <alan.hayward@arm.com>
14187 David Sherwood <david.sherwood@arm.com>
14189 * tree-vect-loop.c (vectorizable_induction): Treat the number
14190 of units as polynomial. Punt on SLP inductions. Use an integer
14191 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14192 cast of such a series for variable-length floating-point
14195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14196 Alan Hayward <alan.hayward@arm.com>
14197 David Sherwood <david.sherwood@arm.com>
14199 * tree.h (build_index_vector): Declare.
14200 * tree.c (build_index_vector): New function.
14201 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14202 of units as polynomial, forcibly converting it to a constant if
14203 vectorizable_reduction has already enforced the condition.
14204 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14205 to create a {1,2,3,...} vector.
14206 (vectorizable_reduction): Treat the number of units as polynomial.
14207 Choose vectype_in based on the largest scalar element size rather
14208 than the smallest number of units. Enforce the restrictions
14211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14212 Alan Hayward <alan.hayward@arm.com>
14213 David Sherwood <david.sherwood@arm.com>
14215 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14216 number of units as polynomial.
14218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14219 Alan Hayward <alan.hayward@arm.com>
14220 David Sherwood <david.sherwood@arm.com>
14222 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14223 * target.def (autovectorize_vector_sizes): Return the vector sizes
14224 by pointer, using vector_sizes rather than a bitmask.
14225 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14226 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14227 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14229 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14230 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14231 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14232 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14233 * omp-general.c (omp_max_vf): Likewise.
14234 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14235 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14236 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14237 * tree-vect-slp.c (vect_slp_bb): Likewise.
14238 * doc/tm.texi: Regenerate.
14239 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14241 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14242 the vector size as a poly_uint64 rather than an unsigned int.
14243 (current_vector_size): Change from an unsigned int to a poly_uint64.
14244 (get_vectype_for_scalar_type): Update accordingly.
14245 * tree.h (build_truth_vector_type): Take the size and number of
14246 units as a poly_uint64 rather than an unsigned int.
14247 (build_vector_type): Add a temporary overload that takes
14248 the number of units as a poly_uint64 rather than an unsigned int.
14249 * tree.c (make_vector_type): Likewise.
14250 (build_truth_vector_type): Take the number of units as a poly_uint64
14251 rather than an unsigned int.
14253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14254 Alan Hayward <alan.hayward@arm.com>
14255 David Sherwood <david.sherwood@arm.com>
14257 * target.def (get_mask_mode): Take the number of units and length
14258 as poly_uint64s rather than unsigned ints.
14259 * targhooks.h (default_get_mask_mode): Update accordingly.
14260 * targhooks.c (default_get_mask_mode): Likewise.
14261 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14262 * doc/tm.texi: Regenerate.
14264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14265 Alan Hayward <alan.hayward@arm.com>
14266 David Sherwood <david.sherwood@arm.com>
14268 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14269 * omp-general.c (omp_max_vf): Likewise.
14270 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14271 (expand_omp_simd): Handle polynomial safelen.
14272 * omp-low.c (omplow_simd_context): Add a default constructor.
14273 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14274 (lower_rec_simd_input_clauses): Update accordingly.
14275 (lower_rec_input_clauses): Likewise.
14277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14278 Alan Hayward <alan.hayward@arm.com>
14279 David Sherwood <david.sherwood@arm.com>
14281 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14282 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14283 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14284 (vect_analyze_slp_cost): Likewise.
14285 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14286 (vect_model_load_cost): Likewise.
14288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14289 Alan Hayward <alan.hayward@arm.com>
14290 David Sherwood <david.sherwood@arm.com>
14292 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14293 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14294 from an unsigned int * to a poly_uint64_pod *.
14295 (calculate_unrolling_factor): New function.
14296 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14299 Alan Hayward <alan.hayward@arm.com>
14300 David Sherwood <david.sherwood@arm.com>
14302 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14303 from an unsigned int to a poly_uint64.
14304 (_loop_vec_info::slp_unrolling_factor): Likewise.
14305 (_loop_vec_info::vectorization_factor): Change from an int
14307 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14308 (vect_get_num_vectors): New function.
14309 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14310 (vect_get_num_copies): Use vect_get_num_vectors.
14311 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14312 to an unsigned int *.
14313 (vect_analyze_data_refs): Change min_vf from an int * to a
14315 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14316 than an unsigned HOST_WIDE_INT.
14317 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14318 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14319 to an unsigned int *.
14320 (vect_analyze_data_ref_dependences): Likewise.
14321 (vect_compute_data_ref_alignment): Handle polynomial vf.
14322 (vect_enhance_data_refs_alignment): Likewise.
14323 (vect_prune_runtime_alias_test_list): Likewise.
14324 (vect_shift_permute_load_chain): Likewise.
14325 (vect_supportable_dr_alignment): Likewise.
14326 (dependence_distance_ge_vf): Take the vectorization factor as a
14327 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14328 (vect_analyze_data_refs): Change min_vf from an int * to a
14330 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14331 vfm1 as a poly_uint64 rather than an int. Make the same change
14332 for the returned bound_scalar.
14333 (vect_gen_vector_loop_niters): Handle polynomial vf.
14334 (vect_do_peeling): Likewise. Update call to
14335 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14336 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14338 * tree-vect-loop.c (vect_determine_vectorization_factor)
14339 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14340 (vect_get_known_peeling_cost): Likewise.
14341 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14342 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14343 (vect_transform_loop): Likewise. Use the lowest possible VF when
14344 updating the upper bounds of the loop.
14345 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14346 rather than an int.
14347 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14348 polynomial unroll factors.
14349 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14350 (vect_make_slp_decision): Likewise.
14351 (vect_supported_load_permutation_p): Likewise, and polynomial
14353 (vect_analyze_slp_cost): Handle polynomial vf.
14354 (vect_slp_analyze_node_operations): Likewise.
14355 (vect_slp_analyze_bb_1): Likewise.
14356 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14357 than an unsigned HOST_WIDE_INT.
14358 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14359 (vectorizable_load): Handle polynomial vf.
14360 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14362 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14365 Alan Hayward <alan.hayward@arm.com>
14366 David Sherwood <david.sherwood@arm.com>
14368 * match.pd: Handle bit operations involving three constants
14369 and try to fold one pair.
14371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14373 * tree-vect-loop-manip.c: Include gimple-fold.h.
14374 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14375 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14376 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14377 Add a path that uses a step of VF instead of 1, but disable it
14379 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14380 and niters_no_overflow parameters. Update calls to
14381 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14382 Create a new SSA name if the latter choses to use a ste other
14383 than zero, and return it via niters_vector_mult_vf_var.
14384 * tree-vect-loop.c (vect_transform_loop): Update calls to
14385 vect_do_peeling, vect_gen_vector_loop_niters and
14386 slpeel_make_loop_iterate_ntimes.
14387 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14388 (vect_gen_vector_loop_niters): Update declarations after above changes.
14390 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14392 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14393 128-bit round to integer instructions.
14394 (ceil<mode>2): Likewise.
14395 (btrunc<mode>2): Likewise.
14396 (round<mode>2): Likewise.
14398 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14400 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14401 unaligned VSX load/store on P8/P9.
14402 (expand_block_clear): Allow the use of unaligned VSX
14403 load/store on P8/P9.
14405 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14407 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14409 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14410 swap associated with both a load and a store.
14412 2018-01-02 Andrew Waterman <andrew@sifive.com>
14414 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14415 * config/riscv/riscv.md (clear_cache): Use it.
14417 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14419 * web.c: Remove out-of-date comment.
14421 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14423 * expr.c (fixup_args_size_notes): Check that any existing
14424 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14425 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14426 (emit_single_push_insn): ...here.
14428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14430 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14431 (const_vector_encoded_nelts): New function.
14432 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14433 (const_vector_int_elt, const_vector_elt): Declare.
14434 * emit-rtl.c (const_vector_int_elt_1): New function.
14435 (const_vector_elt): Likewise.
14436 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14437 of CONST_VECTOR_ELT.
14439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14441 * expr.c: Include rtx-vector-builder.h.
14442 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14443 directly on the tree encoding.
14444 (const_vector_from_tree): Likewise.
14445 * optabs.c: Include rtx-vector-builder.h.
14446 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14447 sequence of "u" values.
14448 * vec-perm-indices.c: Include rtx-vector-builder.h.
14449 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14450 directly on the vec_perm_indices encoding.
14452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14454 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14455 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14456 * rtx-vector-builder.h: New file.
14457 * rtx-vector-builder.c: Likewise.
14458 * rtl.h (rtx_def::u2): Add a const_vector field.
14459 (CONST_VECTOR_NPATTERNS): New macro.
14460 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14461 (CONST_VECTOR_DUPLICATE_P): Likewise.
14462 (CONST_VECTOR_STEPPED_P): Likewise.
14463 (CONST_VECTOR_ENCODED_ELT): Likewise.
14464 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14465 (unwrap_const_vec_duplicate): Likewise.
14466 (const_vec_series_p): Check for a non-duplicated vector encoding.
14467 Say that the function only returns true for integer vectors.
14468 * emit-rtl.c: Include rtx-vector-builder.h.
14469 (gen_const_vec_duplicate_1): Delete.
14470 (gen_const_vector): Call gen_const_vec_duplicate instead of
14471 gen_const_vec_duplicate_1.
14472 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14473 (gen_const_vec_duplicate): Use rtx_vector_builder.
14474 (gen_const_vec_series): Likewise.
14475 (gen_rtx_CONST_VECTOR): Likewise.
14476 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14477 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14478 Build a new vector rather than modifying a CONST_VECTOR in-place.
14479 (handle_special_swappables): Update call accordingly.
14480 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14481 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14482 Build a new vector rather than modifying a CONST_VECTOR in-place.
14483 (handle_special_swappables): Update call accordingly.
14485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14487 * simplify-rtx.c (simplify_const_binary_operation): Use
14488 CONST_VECTOR_ELT instead of XVECEXP.
14490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14492 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14493 the selector elements to be different from the data elements
14494 if the selector is a VECTOR_CST.
14495 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14496 ssizetype for the selector.
14498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14500 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14501 before testing each element individually.
14502 * tree-vect-generic.c (lower_vec_perm): Likewise.
14504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14506 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14507 * selftest-run-tests.c (selftest::run_tests): Call it.
14508 * vector-builder.h (vector_builder::operator ==): New function.
14509 (vector_builder::operator !=): Likewise.
14510 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14511 (vec_perm_indices::all_from_input_p): New function.
14512 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14513 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14514 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14515 instead of reading the VECTOR_CST directly. Detect whether both
14516 vector inputs are the same before constructing the vec_perm_indices,
14517 and update the number of inputs argument accordingly. Use the
14518 utility functions added above. Only construct sel2 if we need to.
14520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14522 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14523 the broadcast of the low byte.
14524 (expand_mult_highpart): Use an explicit encoding for the permutes.
14525 * optabs-query.c (can_mult_highpart_p): Likewise.
14526 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14527 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14528 (vectorizable_bswap): Likewise.
14529 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14530 explicit encoding for the power-of-2 permutes.
14531 (vect_permute_store_chain): Likewise.
14532 (vect_grouped_load_supported): Likewise.
14533 (vect_permute_load_chain): Likewise.
14535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14537 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14538 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14539 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14540 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14541 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14542 (vect_gen_perm_mask_any): Likewise.
14544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14546 * int-vector-builder.h: New file.
14547 * vec-perm-indices.h: Include int-vector-builder.h.
14548 (vec_perm_indices): Redefine as an int_vector_builder.
14549 (auto_vec_perm_indices): Delete.
14550 (vec_perm_builder): Redefine as a stand-alone class.
14551 (vec_perm_indices::vec_perm_indices): New function.
14552 (vec_perm_indices::clamp): Likewise.
14553 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14554 (vec_perm_indices::new_vector): New function.
14555 (vec_perm_indices::new_expanded_vector): Update for new
14556 vec_perm_indices class.
14557 (vec_perm_indices::rotate_inputs): New function.
14558 (vec_perm_indices::all_in_range_p): Operate directly on the
14559 encoded form, without computing elided elements.
14560 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14561 encoding. Update for new vec_perm_indices class.
14562 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14563 the given vec_perm_builder.
14564 (expand_vec_perm_var): Update vec_perm_builder constructor.
14565 (expand_mult_highpart): Use vec_perm_builder instead of
14566 auto_vec_perm_indices.
14567 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14568 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14569 or double series encoding as appropriate.
14570 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14571 vec_perm_indices instead of auto_vec_perm_indices.
14572 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14573 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14574 (vect_permute_store_chain): Likewise.
14575 (vect_grouped_load_supported): Likewise.
14576 (vect_permute_load_chain): Likewise.
14577 (vect_shift_permute_load_chain): Likewise.
14578 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14579 (vect_transform_slp_perm_load): Likewise.
14580 (vect_schedule_slp_instance): Likewise.
14581 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14582 (vectorizable_mask_load_store): Likewise.
14583 (vectorizable_bswap): Likewise.
14584 (vectorizable_store): Likewise.
14585 (vectorizable_load): Likewise.
14586 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14587 vec_perm_indices instead of auto_vec_perm_indices. Use
14588 tree_to_vec_perm_builder to read the vector from a tree.
14589 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14590 vec_perm_builder instead of a vec_perm_indices.
14591 (have_whole_vector_shift): Use vec_perm_builder and
14592 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14593 truncation to calc_vec_perm_mask_for_shift.
14594 (vect_create_epilog_for_reduction): Likewise.
14595 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14596 from auto_vec_perm_indices to vec_perm_indices.
14597 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14598 instead of changing individual elements.
14599 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14600 the vector in d.perm.
14601 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14602 from auto_vec_perm_indices to vec_perm_indices.
14603 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14604 instead of changing individual elements.
14605 (arm_vectorize_vec_perm_const): Use new_vector to install
14606 the vector in d.perm.
14607 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14608 Update vec_perm_builder constructor.
14609 (rs6000_expand_interleave): Likewise.
14610 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14611 (rs6000_expand_interleave): Likewise.
14613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14615 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14616 to qimode could truncate the indices.
14617 * optabs.c (expand_vec_perm_var): Likewise.
14619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14621 * Makefile.in (OBJS): Add vec-perm-indices.o.
14622 * vec-perm-indices.h: New file.
14623 * vec-perm-indices.c: Likewise.
14624 * target.h (vec_perm_indices): Replace with a forward class
14626 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14627 * optabs.h: Include vec-perm-indices.h.
14628 (expand_vec_perm): Delete.
14629 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14630 (expand_vec_perm_const): Declare.
14631 * target.def (vec_perm_const_ok): Replace with...
14632 (vec_perm_const): ...this new hook.
14633 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14634 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14635 * doc/tm.texi: Regenerate.
14636 * optabs.def (vec_perm_const): Delete.
14637 * doc/md.texi (vec_perm_const): Likewise.
14638 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14639 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14640 expand_vec_perm for constant permutation vectors. Assert that
14641 the mode of variable permutation vectors is the integer equivalent
14642 of the mode that is being permuted.
14643 * optabs-query.h (selector_fits_mode_p): Declare.
14644 * optabs-query.c: Include vec-perm-indices.h.
14645 (selector_fits_mode_p): New function.
14646 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14647 is defined, instead of checking whether the vec_perm_const_optab
14648 exists. Use targetm.vectorize.vec_perm_const instead of
14649 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14650 fit in the vector mode before using a variable permute.
14651 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14652 vec_perm_indices instead of an rtx.
14653 (expand_vec_perm): Replace with...
14654 (expand_vec_perm_const): ...this new function. Take the selector
14655 as a vec_perm_indices rather than an rtx. Also take the mode of
14656 the selector. Update call to shift_amt_for_vec_perm_mask.
14657 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14658 Use vec_perm_indices::new_expanded_vector to expand the original
14659 selector into bytes. Check whether the indices fit in the vector
14660 mode before using a variable permute.
14661 (expand_vec_perm_var): Make global.
14662 (expand_mult_highpart): Use expand_vec_perm_const.
14663 * fold-const.c: Includes vec-perm-indices.h.
14664 * tree-ssa-forwprop.c: Likewise.
14665 * tree-vect-data-refs.c: Likewise.
14666 * tree-vect-generic.c: Likewise.
14667 * tree-vect-loop.c: Likewise.
14668 * tree-vect-slp.c: Likewise.
14669 * tree-vect-stmts.c: Likewise.
14670 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14672 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14673 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14674 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14675 (aarch64_vectorize_vec_perm_const): ...this new function.
14676 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14677 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14678 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14679 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14680 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14681 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14682 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14684 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14685 check for NEON modes.
14686 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14687 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14688 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14689 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14691 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14692 the old VEC_PERM_CONST conditions.
14693 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14694 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14695 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14696 (ia64_vectorize_vec_perm_const_ok): Merge into...
14697 (ia64_vectorize_vec_perm_const): ...this new function.
14698 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14699 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14700 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14701 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14702 * config/mips/mips.c (mips_expand_vec_perm_const)
14703 (mips_vectorize_vec_perm_const_ok): Merge into...
14704 (mips_vectorize_vec_perm_const): ...this new function.
14705 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14706 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14707 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14708 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14709 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14710 (rs6000_expand_vec_perm_const): Delete.
14711 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14713 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14714 (altivec_expand_vec_perm_const_le): Take each operand individually.
14715 Operate on constant selectors rather than rtxes.
14716 (altivec_expand_vec_perm_const): Likewise. Update call to
14717 altivec_expand_vec_perm_const_le.
14718 (rs6000_expand_vec_perm_const): Delete.
14719 (rs6000_vectorize_vec_perm_const_ok): Delete.
14720 (rs6000_vectorize_vec_perm_const): New function.
14721 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14722 an element count and rtx array.
14723 (rs6000_expand_extract_even): Update call accordingly.
14724 (rs6000_expand_interleave): Likewise.
14725 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14726 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14727 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14728 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14729 (rs6000_expand_vec_perm_const): Delete.
14730 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14731 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14732 (altivec_expand_vec_perm_const_le): Take each operand individually.
14733 Operate on constant selectors rather than rtxes.
14734 (altivec_expand_vec_perm_const): Likewise. Update call to
14735 altivec_expand_vec_perm_const_le.
14736 (rs6000_expand_vec_perm_const): Delete.
14737 (rs6000_vectorize_vec_perm_const_ok): Delete.
14738 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14739 reference to the SPE evmerge intructions.
14740 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14741 an element count and rtx array.
14742 (rs6000_expand_extract_even): Update call accordingly.
14743 (rs6000_expand_interleave): Likewise.
14744 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14745 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14747 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14751 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14752 vector mode and that that mode matches the mode of the data
14754 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14755 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14756 directly using expand_vec_perm_1 when forcing selectors into
14758 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14762 * optabs-query.h (can_vec_perm_p): Delete.
14763 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14764 * optabs-query.c (can_vec_perm_p): Split into...
14765 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14766 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14767 particular selector is valid.
14768 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14769 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14770 (vect_grouped_load_supported): Likewise.
14771 (vect_shift_permute_load_chain): Likewise.
14772 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14773 (vect_transform_slp_perm_load): Likewise.
14774 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14775 (vectorizable_bswap): Likewise.
14776 (vect_gen_perm_mask_checked): Likewise.
14777 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14778 implementations of variable permutation vectors into account
14779 when deciding which selector to use.
14780 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14781 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14782 with a false third argument.
14783 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14784 to test whether the constant selector is valid and can_vec_perm_var_p
14785 to test whether a variable selector is valid.
14787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14789 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14790 * optabs-query.c (can_vec_perm_p): Likewise.
14791 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14792 instead of vec_perm_indices.
14793 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14794 (vect_gen_perm_mask_checked): Likewise,
14795 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14796 (vect_gen_perm_mask_checked): Likewise,
14798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14800 * optabs-query.h (qimode_for_vec_perm): Declare.
14801 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14802 (qimode_for_vec_perm): ...this new function.
14803 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14805 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14807 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14808 does not have a conditional at the top.
14810 2018-01-02 Richard Biener <rguenther@suse.de>
14812 * ipa-inline.c (big_speedup_p): Fix expression.
14814 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14817 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14820 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14824 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14825 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14826 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14827 cond_taken_branch_cost 3->4.
14829 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14831 PR tree-optimization/83581
14832 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14833 TODO_cleanup_cfg if any changes have been made.
14835 PR middle-end/83608
14836 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14837 convert_modes if target mode has the right side, but different mode
14840 PR middle-end/83609
14841 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14842 last argument when extracting from CONCAT. If either from_real or
14843 from_imag is NULL, use expansion through memory. If result is not
14844 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14845 the parts directly to inner mode, if even that fails, use expansion
14848 PR middle-end/83623
14849 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14850 check for bswap in mode rather than HImode and use that in expand_unop
14853 Copyright (C) 2018 Free Software Foundation, Inc.
14855 Copying and distribution of this file, with or without modification,
14856 are permitted in any medium without royalty provided the copyright
14857 notice and this notice are preserved.