1 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
7 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8 Cesar Philippidis <cesar@codesourcery.com>
11 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12 when emitting error on private/firstprivate reductions.
13 * omp-low.c (lower_omp_target): Avoid reference-type processing
14 on pointers for firstprivate clause.
16 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
18 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
21 * config/aarch64/aarch64-simd.md
22 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
23 (aarch64_ld1_x3_<mode>): Likewise
24 (aarch64_st1x2<VALLDIF:mode>): Likewise
25 (aarch64_st1_x2_<mode>): Likewise
26 (aarch64_st1x3<VALLDIF:mode>): Likewise
27 (aarch64_st1_x3_<mode>): Likewise
28 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
29 (vld1_s8_x3): Likewise.
30 (vld1_u16_x3): Likewise.
31 (vld1_s16_x3): Likewise.
32 (vld1_u32_x3): Likewise.
33 (vld1_s32_x3): Likewise.
34 (vld1_u64_x3): Likewise.
35 (vld1_s64_x3): Likewise.
36 (vld1_f16_x3): Likewise.
37 (vld1_f32_x3): Likewise.
38 (vld1_f64_x3): Likewise.
39 (vld1_p8_x3): Likewise.
40 (vld1_p16_x3): Likewise.
41 (vld1_p64_x3): Likewise.
42 (vld1q_u8_x3): Likewise.
43 (vld1q_s8_x3): Likewise.
44 (vld1q_u16_x3): Likewise.
45 (vld1q_s16_x3): Likewise.
46 (vld1q_u32_x3): Likewise.
47 (vld1q_s32_x3): Likewise.
48 (vld1q_u64_x3): Likewise.
49 (vld1q_s64_x3): Likewise.
50 (vld1q_f16_x3): Likewise.
51 (vld1q_f32_x3): Likewise.
52 (vld1q_f64_x3): Likewise.
53 (vld1q_p8_x3): Likewise.
54 (vld1q_p16_x3): Likewise.
55 (vld1q_p64_x3): Likewise.
56 (vst1_s64_x2): Likewise.
57 (vst1_u64_x2): Likewise.
58 (vst1_f64_x2): Likewise.
59 (vst1_s8_x2): Likewise.
60 (vst1_p8_x2): Likewise.
61 (vst1_s16_x2): Likewise.
62 (vst1_p16_x2): Likewise.
63 (vst1_s32_x2): Likewise.
64 (vst1_u8_x2): Likewise.
65 (vst1_u16_x2): Likewise.
66 (vst1_u32_x2): Likewise.
67 (vst1_f16_x2): Likewise.
68 (vst1_f32_x2): Likewise.
69 (vst1_p64_x2): Likewise.
70 (vst1q_s8_x2): Likewise.
71 (vst1q_p8_x2): Likewise.
72 (vst1q_s16_x2): Likewise.
73 (vst1q_p16_x2): Likewise.
74 (vst1q_s32_x2): Likewise.
75 (vst1q_s64_x2): Likewise.
76 (vst1q_u8_x2): Likewise.
77 (vst1q_u16_x2): Likewise.
78 (vst1q_u32_x2): Likewise.
79 (vst1q_u64_x2): Likewise.
80 (vst1q_f16_x2): Likewise.
81 (vst1q_f32_x2): Likewise.
82 (vst1q_f64_x2): Likewise.
83 (vst1q_p64_x2): Likewise.
84 (vst1_s64_x3): Likewise.
85 (vst1_u64_x3): Likewise.
86 (vst1_f64_x3): Likewise.
87 (vst1_s8_x3): Likewise.
88 (vst1_p8_x3): Likewise.
89 (vst1_s16_x3): Likewise.
90 (vst1_p16_x3): Likewise.
91 (vst1_s32_x3): Likewise.
92 (vst1_u8_x3): Likewise.
93 (vst1_u16_x3): Likewise.
94 (vst1_u32_x3): Likewise.
95 (vst1_f16_x3): Likewise.
96 (vst1_f32_x3): Likewise.
97 (vst1_p64_x3): Likewise.
98 (vst1q_s8_x3): Likewise.
99 (vst1q_p8_x3): Likewise.
100 (vst1q_s16_x3): Likewise.
101 (vst1q_p16_x3): Likewise.
102 (vst1q_s32_x3): Likewise.
103 (vst1q_s64_x3): Likewise.
104 (vst1q_u8_x3): Likewise.
105 (vst1q_u16_x3): Likewise.
106 (vst1q_u32_x3): Likewise.
107 (vst1q_u64_x3): Likewise.
108 (vst1q_f16_x3): Likewise.
109 (vst1q_f32_x3): Likewise.
110 (vst1q_f64_x3): Likewise.
111 (vst1q_p64_x3): Likewise.
113 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
115 * config/msp430/msp430.c (msp430_output_labelref): Prepend
116 user_label_prefix to name.
118 * tree-core.h: Update comment about the format of NAME string
119 passed to handler in attribute_spec.
121 * config/msp430/msp430.md: Remove erroneous subreg expression from
122 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
123 zero_extend{q,h}isi2.
125 2018-05-30 Borislav Petkov <bp@suse.de>
127 * doc/extend.texi: Document some architecture specific
128 constraints and sort entries.
130 2018-05-30 Martin Sebor <msebor@redhat.com>
133 * builtins.c (expand_builtin_stpcpy_1): New function.
134 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
135 only if the former succeeds.
137 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
139 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
142 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
144 * doc/invoke.texi (-flinker-output): Document
146 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
148 * passes.c (ipa_write_summaries): Only modify statements if body
150 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
151 incrementally linking.
152 (ipa_passes): Likewise.
153 * lto-cgraph.c (lto_output_node): When incrementally linking do not
154 pass down resolution info.
155 * common.opt (flag_incremental_link): Update info.
156 * gcc.c (plugin specs): Turn flinker-output=* to
157 -plugin-opt=-linker-output-known
158 * toplev.c (compile_file): Also cut compilation when doing incremental
160 * flag-types. (enum lto_partition_model): Add
161 LTO_LINKER_OUTPUT_NOLTOREL.
162 (invoke.texi): Add -flinker-output docs.
163 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
164 link same way as WPA; do not stream in dead initializers.
166 * dwarf2out.c (dwarf2out_die_ref_for_decl,
167 darf2out_register_external_decl): Support incremental link.
169 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
171 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
173 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
175 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
176 it down to simple_object_copy_lto_debug_sections.
177 (run_gcc): Determine incremental LTO link time and configure
178 lto1 into non-wpa mode, disable renaming of debug sections.
180 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
182 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
183 descriptions of various incorrectly documented functions.
185 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
188 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
189 address check not strict.
191 2018-05-30 Richard Biener <rguenther@suse.de>
193 PR tree-optimization/85964
194 * tracer.c (better_p): Drop initialized count check, we only
195 call the function with initialized counts now.
196 (find_best_successor): Do find a best edge if one
197 has uninitialized count.
198 (find_best_predecessor): Likewise. Do BB frequency check only
199 if count is initialized.
201 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
203 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
204 (aarch64_ldrstr_offset_compare): New.
205 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
206 load/store orderings.
207 (aarch64_gen_adjusted_ldpstp): Likewise.
209 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
211 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
212 Check for subset of GENERAL_REGS and FP_REGS.
213 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
216 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
218 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
219 and wi::to_poly_offset. Add the current offset and then check
220 whether the sum fits, rather than using an unchecked addition of
221 a checked term. Check for a shwi rather than a uhwi.
222 * expr.c (get_bit_range): Use tree_to_poly_uint64.
223 (store_constructor): Use poly_int_tree_p.
224 (expand_expr_real_1): Likewise.
225 * function.c (assign_temp): Likewise.
226 * fold-const.c (const_binop): Use poly_int_tree_p and
228 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
230 * ipa-icf-gimple.c (func_checker::compare_operand): Use
231 to_poly_offset for MEM offsets.
232 * ipa-icf.c (sem_variable::equals): Likewise.
233 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
234 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
235 wi::to_poly_offset for BIT_FIELD_REF offsets.
236 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
238 * var-tracking.c (emit_note_insn_var_location): Use
241 2018-05-29 Jim Wilson <jimw@sifive.com>
243 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
245 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
248 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
249 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
251 (sse4_1_round<mode>2): Use nonimmediate_operand
252 for operand 1 predicate.
254 2018-05-29 Martin Sebor <msebor@redhat.com>
255 Richard Biener <rguenther@suse.de>
258 * calls.c (get_size_range): Call determine_value_range instead
260 * tree-vrp.h (determine_value_range): Declared new function.
261 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
263 2018-05-29 Richard Biener <rguenther@suse.de>
265 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
266 sure to use non-pattern stmts for get_earlier_stmt arguments.
267 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
268 called on pattern stmts.
269 (get_later_stmt): Likewise.
271 2018-05-29 Martin Liska <mliska@suse.cz>
273 PR gcov-profile/85759
274 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
277 2018-05-29 Jakub Jelinek <jakub@redhat.com>
279 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
281 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
285 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
286 VEC_PACK_FLOAT_EXPR): New tree codes.
287 * tree-pretty-print.c (op_code_prio): Handle
288 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
289 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
290 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
291 * tree-inline.c (estimate_operator_cost): Likewise.
292 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
293 * fold-const.c (const_binop): Likewise.
294 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
295 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
296 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
297 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
298 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
299 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
300 * expr.c (expand_expr_real_2): Likewise.
301 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
302 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
303 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
305 * optabs.c (expand_widen_pattern_expr): For
306 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
307 sign from result type rather than operand's type.
308 (expand_binop_directly): For vec_packu_float_optab and
309 vec_packs_float_optab allow result type to be different from operand's
311 * optabs-tree.c (optab_for_tree_code): Handle
312 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
313 VEC_PACK_FLOAT_EXPR. Formatting fixes.
314 * tree-vect-generic.c (expand_vector_operations_1): Handle
315 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
317 * tree-vect-stmts.c (supportable_widening_operation): Handle
319 (supportable_narrowing_operation): Handle FLOAT_EXPR.
320 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
321 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
322 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
323 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
325 (vec_pack<floatprefix>_float_<mode>): New expander.
326 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
328 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
329 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
330 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
331 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
332 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
334 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
335 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
336 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
337 VEC_PACK_FLOAT_EXPR): Document.
339 2018-05-29 Richard Biener <rguenther@suse.de>
341 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
343 (stmt_vec_info_vec): Make pointer.
344 (init_stmt_vec_info_vec): Remove.
345 (free_stmt_vec_info_vec): Likewise.
346 (set_stmt_vec_info_vec): New function.
347 (free_stmt_vec_infos): Likewise.
348 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
349 (set_vinfo_for_stmt): Likewise.
350 (get_earlier_stmt): Likewise.
351 (get_later_stmt): Likewise.
352 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
353 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
354 (vec_info::~vec_info): Free stmt_vec_infos.
355 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
356 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
357 (pass_slp_vectorize::execute): Likewise.
358 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
359 (free_stmt_vec_info_vec): Likewise.
360 (set_stmt_vec_info_vec): New function.
361 (free_stmt_vec_infos): Likewise.
362 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
363 the global stmt_vec_info_vec.
364 * tree-parloops.c (gather_scalar_reductions): Use
365 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
368 2018-05-29 Richard Biener <rguenther@suse.de>
370 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
372 2018-05-29 Martin Liska <mliska@suse.cz>
373 David Malcolm <dmalcolm@redhat.com>
375 * vec.c (test_reverse): New.
376 (vec_c_tests): Add new test.
377 * vec.h (vl_ptr>::reverse): New function.
379 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
381 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
383 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
386 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
388 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
390 2018-05-28 Richard Biener <rguenther@suse.de>
392 PR tree-optimization/85933
393 * tree-vect-data-refs.c (vect_record_base_alignments): Only
394 look at stmts marked as vectorizable.
396 2018-05-28 Richard Biener <rguenther@suse.de>
398 PR tree-optimization/85934
399 * tree-vect-generic.c (expand_vector_operations_1): Hoist
400 vector boolean check before scalar optimization.
402 2018-05-28 Jakub Jelinek <jakub@redhat.com>
404 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
407 2018-05-28 Mark Wielaard <mark@klomp.org>
409 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
410 if it is an expression containing a minus sign.
412 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
414 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
416 2018-05-27 Paul Koning <ni1d@arrl.net>
418 * config/pdp11/pdp11.md (truncsihi2): Remove.
420 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
421 Chung-Ju Wu <jasonwucj@gmail.com>
423 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
425 (unaligned_store_dw): Ditto.
426 * config/nds32/nds32-memory-manipulation.c
427 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
428 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
429 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
430 (emit_setmem_word_loop): Rename to ...
431 (emit_setmem_doubleword_loop): ... this.
432 (nds32_gen_dup_4_byte_to_word_value): New function.
433 (nds32_gen_dup_8_byte_to_double_word_value): New function.
434 (nds32_expand_setmem_loop): Refine implementation.
435 (nds32_expand_setmem_loop_v3m): Ditto.
436 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
439 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
441 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
443 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
445 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
446 (nds32_init_machine_status): Initialize machine->attr_naked_p and
447 machine->attr_no_prologue_p.
448 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
449 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
450 (nds32_expand_epilogue): Consider attr_naked_p.
451 (nds32_expand_epilogue_v3pop): Likewise.
452 (nds32_can_use_return_insn): Likewise.
453 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
454 attr_no_prologue_p fields.
455 * config/nds32/nds32.opt (mret-in-naked-func): New option.
457 2018-05-27 Jakub Jelinek <jakub@redhat.com>
460 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
463 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
465 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
467 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
469 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
471 (*<floatsuffix>floatv2div2sf2): Rename to ...
472 (*float<floatunssuffix>v2div2sf2): ... this.
473 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
474 (float<floatunssuffix>v2div2sf2_mask): ... this.
475 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
476 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
477 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
479 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
481 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
483 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
485 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
487 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
489 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
490 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
491 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
492 gen_ufix_truncv8dfv8si2.
493 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
494 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
495 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
496 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
497 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
498 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
499 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
500 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
502 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
506 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
508 2018-05-25 Jim Wilson <jimw@sifive.com>
510 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
511 * config/riscv/riscv.c (struct machine_function): Add
512 interrupt_handler_p and attribute_checked_p fields.
513 (riscv_attribute_table): Add interrupt.
514 (riscv_interrupt_type_p): New.
515 (riscv_save_reg_p): Save extra regs for interrupt handler.
516 (riscv_use_save_libcall): Return false for interrupt handler.
517 (riscv_first_stack_step): Add forward declaration.
518 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
519 for interrupt handler with large frame. Use it for saved reg list.
520 (riscv_expand_prologue): Move flag_stack_usage_info support to
521 eliminate duplication.
522 (riscv_expand_epilogue): Generate mret for interrupt handler.
523 (riscv_epilogue_uses): New.
524 (riscv_can_use_return_insn): Return false for interrupt handler.
525 (riscv_function_ok_for_sibcall): Likewise.
526 (riscv_set_current_function): Add interrupt handler support.
527 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
528 * config/riscv/riscv.md (UNSPECV_MRET): New.
530 (riscv_frflags, riscv_fsflags): Use tab after opcode.
532 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
534 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
536 PR tree-optimization/85712
537 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
538 this candidate has already been replaced in-situ by a copy.
540 2018-05-25 Jason Merrill <jason@redhat.com>
542 PR c++/80485 - inline function non-zero address.
543 * symtab.c (nonzero_address): Check DECL_COMDAT.
545 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
548 * config/alpha/alpha.md (ashlsi3): New insn pattern.
549 (*ashlsi_se): Rename from *ashldi_se. Define as sign
550 extension of SImode operation. Use const123_operand predicate.
552 (*saddl_se_1): Ditto.
554 (*ssubl_se_1): Ditto.
555 * config/alpha/predicates.md (const123_operand): New predicate.
556 * config/alpha/constraints.md (P): Use IN_RANGE.
558 2018-05-25 Richard Biener <rguenther@suse.de>
560 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
562 (ref_maybe_used_by_stmt_p): Likewise.
563 (stmt_may_clobber_ref_p): Likewise.
564 (stmt_may_clobber_ref_p_1): Likewise.
565 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
567 (ref_maybe_used_by_stmt_p): Likewise.
568 (stmt_may_clobber_ref_p): Likewise.
569 (stmt_may_clobber_ref_p_1): Likewise.
570 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
571 the alias oracle to disambiguate DRs with stmts DR analysis
573 (vect_analyze_data_refs): Do not give up on not analyzable
574 DRs for BB vectorization. Remove code truncating the dataref
577 2018-05-25 Jakub Jelinek <jakub@redhat.com>
580 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
581 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
582 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
584 2018-05-25 Richard Biener <rguenther@suse.de>
586 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
587 function, combining stmt data ref gathering and fatal analysis
589 (vect_analyze_data_refs): Remove now redudnant code and simplify.
590 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
591 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
592 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
593 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
595 2018-05-25 Bin Cheng <bin.cheng@arm.com>
597 PR tree-optimization/85720
598 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
599 SCC if all partitions are builtins.
600 (version_loop_by_alias_check): New parameter. Generate cancelable
601 runtime alias check if all partitions are builtins.
602 (distribute_loop): Update call to above function.
604 2018-05-25 Bin Cheng <bin.cheng@arm.com>
606 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
607 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
608 (parm_default_def_partition_arg): Ditto.
609 (set_parm_default_def_partition): Ditto.
610 (get_parm_default_def_partitions): Ditto and make it static.
611 (get_undefined_value_partitions): Ditto and make it static.
612 (remove_ssa_form): Refactor call to init_var_map here.
613 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
614 computation for loop region.
615 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
616 (register_default_def): Delete.
617 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
618 (parm_default_def_partition_arg): Ditto.
619 (set_parm_default_def_partition): Ditto.
620 (get_parm_default_def_partitions): Ditto and make it static.
621 (get_undefined_value_partitions): Ditto and make it static.
622 (coalesce_with_default, coalesce_with_default): Update comment.
623 (create_coalesce_list_for_region): New func factored out from
624 create_outofssa_var_map.
625 (populate_coalesce_list_for_outofssa): New func factored out from
626 create_outofssa_var_map and coalesce_ssa_name.
627 (create_outofssa_var_map): Delete.
628 (coalesce_ssa_name): Refactor to support live range computation.
629 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
630 (get_parm_default_def_partitions): Delete.
631 (get_undefined_value_partitions): Ditto.
632 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
633 computation for loop region.
634 (new_tree_live_info, loe_visit_block): Ditto.
635 (live_worklist, set_var_live_on_entry): Ditto.
636 (calculate_live_on_exit, verify_live_on_entry): Ditto.
637 * tree-ssa-live.h (struct _var_map): New fields.
638 (init_var_map): Change decl.
639 (region_contains_p): New.
641 2018-05-25 Bin Cheng <bin.cheng@arm.com>
643 * tree-ssa-live.h (live_merge_and_clear): Delete.
645 2018-05-25 Richard Biener <rguenther@suse.de>
648 * tree-dump.c (dequeue_and_dump): Remove access to removed
649 operand 2 of a SWITCH_EXPR.
651 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
653 * doc/sourcebuild.texi (vect_double_cond_arith): Include
654 multiplication and division.
655 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
656 (cond_udiv@var{m}, cond_umod@var{m}): Document.
657 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
658 (cond_udiv_optab, cond_umod_optab): New optabs.
659 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
660 (IFN_COND_RDIV): New internal functions.
661 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
662 TRUNC_MOD_EXPR and RDIV_EXPR.
663 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
664 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
666 (SVE_INT_BINARY): Include mult.
667 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
668 (optab, sve_int_op): Handle mult.
669 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
671 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
672 for SVE_INT_BINARY_SD.
674 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
676 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
677 (optab, sve_int_op): Handle div and udiv.
678 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
679 for SVE_INT_BINARY_SD.
680 (*<optab><mode>3): New insn for the same.
682 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
684 * tree-vect-patterns.c: Include predict.h.
685 (vect_recog_divmod_pattern): Restrict check for division support
686 to when optimizing for size.
688 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
690 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
691 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
692 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
693 (gimple_match_op::set_op): Likewise.
694 (gimple_resimplify4): Declare.
695 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
696 (expr::gen_transform): Likewise.
697 (decision_tree::gen): Generate a simplification routine for 4 operands.
698 * gimple-match-head.c (gimple_simplify): Add an overload for
699 4 operands. In the top-level function, handle up to 4 call
700 arguments and call gimple_resimplify4.
701 (gimple_resimplify4): New function.
702 (build_call_internal): Pass a fourth operand.
703 (maybe_push_to_seq): Likewise.
704 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
705 Fold VEC_COND_EXPRs of an operation and a default value into
706 an IFN_COND_* function if possible.
707 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
709 (SVE_COND_FP_BINARY): Include them.
710 (optab, sve_fp_op): Handle them.
711 (SVE_INT_BINARY_REV): New code iterator.
712 (SVE_COND_FP_BINARY_REV): New int iterator.
713 (commutative): New int attribute.
714 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
716 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
718 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
719 (*cond_<optab><mode>): New patterns for reversed operands.
721 2018-05-25 Richard Biener <rguenther@suse.de>
723 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
724 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
725 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
726 (STMT_VINFO_GROUPED_ACCESS): Adjust.
727 * tree-vect-data-refs.c (everywhere): Adjust users.
728 * tree-vect-loop.c (everywhere): Likewise.
729 * tree-vect-slp.c (everywhere): Likewise.
730 * tree-vect-stmts.c (everywhere): Likewise.
731 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
733 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
735 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
737 (gcc_cv_as_section_exclude): ... this.
738 Try Solaris as #exclude syntax.
739 * configure: Regenerate.
740 * config.in: Regenerate.
741 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
743 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
744 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
746 * varasm.c (default_elf_asm_named_section): Don't check if
747 HAVE_GAS_SECTION_EXCLUDE is defined.
749 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
751 * doc/md.texi: Update the documentation of the cond_* optabs
752 to mention the new final operand. Fix GET_MODE_NUNITS call.
753 Describe the scalar case too.
754 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
755 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
757 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
758 (get_conditional_internal_fn): Update comment.
759 * tree-vect-loop.c (vectorizable_reduction): Pass the original
760 accumulator value as a final argument to conditional functions.
761 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
762 a define_expand and add an "else" operand. Assert for now that
763 the else operand is equal to operand 2. Use SVE_INT_BINARY and
764 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
765 (*cond_<optab><mode>): New patterns.
766 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
767 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
768 (UNSPEC_COND_EOR): Delete.
769 (optab): Remove associated mappings.
770 (SVE_INT_BINARY): New code iterator.
771 (sve_int_op): Remove int attribute and add "minus" to the code
773 (SVE_COND_INT_OP): Delete.
774 (SVE_COND_FP_OP): Rename to...
775 (SVE_COND_FP_BINARY): ...this.
777 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
779 * optabs.c (can_reuse_operands_p): New function.
780 (maybe_legitimize_operands): Try to reuse the results for
783 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
785 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
786 Add {q} suffix to insn mnemonic.
788 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
790 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
791 (msp430_warn_func_return): New.
793 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
795 * fold-const.c (tree_nonzero_bits): New function.
796 * fold-const.h (tree_nonzero_bits): Likewise.
797 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
798 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
800 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
804 * varasm.c (assemble_alias): Check ifunc_resolver only on
807 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
810 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
811 when memory input operand is handled.
813 2018-05-24 Luis Machado <luis.machado@linaro.org>
815 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
817 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
819 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
821 * match.pd: Delay FMA folds until after vectorization.
823 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
826 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
827 address check not strict.
829 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
831 * gimple-match.h (gimple_match_op): New class.
832 (mprts_hook): Replace parameters with a gimple_match_op *.
833 (maybe_build_generic_op): Likewise.
834 (gimple_simplified_result_is_gimple_val): Replace parameters with
835 a const gimple_match_op *.
836 (gimple_simplify): Replace code_helper * and tree * parameters with
837 a gimple_match_op * parameter.
838 (gimple_resimplify1): Replace code_helper *, tree and tree *
839 parameters with a gimple_match_op * parameter.
840 (gimple_resimplify2): Likewise.
841 (gimple_resimplify3): Likewise.
842 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
843 parameters with a gimple_match_op * parameter.
844 * gimple-match-head.c (gimple_simplify): Change prototypes of
845 auto-generated functions to take a gimple_match_op * instead of
846 separate code_helper * and tree * parameters. Make the same
847 change in the top-level overload and update calls to the
848 gimple_resimplify routines. Update calls to the auto-generated
849 functions and to maybe_push_res_to_seq in the publicly-facing
850 operation-specific gimple_simplify overloads.
851 (gimple_match_op::MAX_NUM_OPS): Define.
852 (gimple_resimplify1): Replace rcode and ops with a single res_op
853 parameter. Update call to gimple_simplify.
854 (gimple_resimplify2): Likewise.
855 (gimple_resimplify3): Likewise.
856 (mprts_hook): Replace parameters with a gimple_match_op *.
857 (maybe_build_generic_op): Likewise.
858 (build_call_internal): Replace type, nargs and ops with
860 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
861 with a single gimple_match_op *. Update calls to mprts_hook,
862 build_call_internal and gimple_simplified_result_is_gimple_val.
863 Factor out code that is common to the tree_code and combined_fn cases.
864 * genmatch.c (expr::gen_transform): Replace tem_code and
865 tem_ops with a gimple_match_op called tem_op. Update calls
866 to the gimple_resimplify functions and maybe_push_res_to_seq.
867 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
868 res_ops. Update call to the gimple_resimplify functions.
869 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
870 (decision_tree::gen): Make the functions take a gimple_match_op *
871 called res_op instead of separate res_code and res_ops parameters.
872 Update call accordingly.
873 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
874 and ops with a single res_op parameter. Update calls to
875 maybe_build_generic_op and maybe_push_res_to_seq.
876 (fold_stmt_1): Update calls to gimple_simplify and
877 replace_stmt_with_simplification.
878 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
879 and gimple_simplified_result_is_gimple_val.
880 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
882 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
883 with a gimple_match_op *.
884 (vn_nary_build_or_lookup): Likewise. Update call to
885 vn_nary_build_or_lookup_1.
886 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
887 gimple_match_op *. Update calls to the gimple_resimplify routines
888 and to gimple_simplified_result_is_gimple_val.
889 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
890 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
891 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
892 (visit_nary_op): Likewise.
893 (visit_reference_op_load): Likewise.
895 2018-05-23 Luis Machado <luis.machado@linaro.org>
897 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
898 modifier for printing the step amount.
900 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
903 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
906 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
908 * doc/sourcebuild.texi (Endianness): New subsubsection.
910 2018-05-23 Luis Machado <luis.machado@linaro.org>
912 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
913 <prefetch_dynamic_strides>: New const bool field.
914 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
915 prefetch_dynamic_strides.
916 (exynosm1_prefetch_tune): Likewise.
917 (thunderxt88_prefetch_tune): Likewise.
918 (thunderx_prefetch_tune): Likewise.
919 (thunderx2t99_prefetch_tune): Likewise.
920 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
922 (aarch64_override_options_internal): Update to set
923 PARAM_PREFETCH_DYNAMIC_STRIDES.
924 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
925 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
926 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
927 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
928 prefetch-dynamic-strides setting.
930 2018-05-23 Luis Machado <luis.machado@linaro.org>
932 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
933 <minimum_stride>: New const int field.
934 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
935 minimum_stride field defaulting to -1.
936 (exynosm1_prefetch_tune): Likewise.
937 (thunderxt88_prefetch_tune): Likewise.
938 (thunderx_prefetch_tune): Likewise.
939 (thunderx2t99_prefetch_tune): Likewise.
940 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
941 <default_opt_level>: Set to 3.
942 (aarch64_override_options_internal): Update to set
943 PARAM_PREFETCH_MINIMUM_STRIDE.
944 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
945 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
946 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
947 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
948 stride is constant and is below the minimum stride threshold.
950 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
952 * config/arm/arm-cpus.in (mode26): Delete.
953 (armv4): Delete mode26 reference.
954 * config/arm/arm.c (arm_configure_build_target): Delete use of
957 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
959 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
961 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
962 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
964 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
965 (floatunsdidf2): Ditto.
967 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
969 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
970 (fixuns_trunc<mode>si2_avx512f): Ditto.
971 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
972 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
973 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
975 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
977 PR rtl-optimization/79985
978 * df-scan.c (df_insn_refs_collect): Remove special case for
979 global registers and asm statements.
981 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
983 * extend.texi (Global Register Variables): Rewrite the bullet list.
984 Note that the register is available for allocation. Note that access
985 via inline asm must use constraints. Add note about async-signal
986 handlers. Remove paragraph about automagic register selection.
988 2018-05-23 Richard Biener <rguenther@suse.de>
990 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
991 of fixed offset from memset VN.
993 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
995 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
997 (alloc_cand_and_find_basis): Initialize first_interp field.
998 (slsr_process_mul): Modify first_interp field.
999 (slsr_process_add): Likewise.
1000 (slsr_process_cast): Modify first_interp field for each new
1002 (slsr_process_copy): Likewise.
1003 (dump_candidate): Dump first_interp field.
1004 (replace_mult_candidate): Process all interpretations, not just
1006 (replace_rhs_if_not_dup): Likewise.
1007 (replace_one_candidate): Likewise.
1009 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1011 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1013 (aarch64_needs_frame_chain): New function.
1014 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1016 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1019 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1020 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1021 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1022 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1023 as true for strict-align.
1024 (aarch64_can_inline_p): Perform checks even when callee has no
1025 attributes to check for strict alignment.
1026 * doc/extend.texi (AArch64 Function Attributes): Document
1028 * doc/invoke.texi: (AArch64 Options): Likewise.
1030 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1032 PR tree-optimization/85853
1033 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1034 the handling of the root of the node to...
1035 (vect_slp_analyze_node_operations_1): ...this new function,
1036 and run the whole thing with the child nodes' def types
1037 set according to their SLP node's def type.
1039 2018-05-23 Richard Biener <rguenther@suse.de>
1042 * tree-data-ref.c (create_runtime_alias_checks): Defer
1043 and ignore overflow warnings.
1045 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1047 PR tree-optimization/85822
1048 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1051 2018-05-23 Richard Biener <rguenther@suse.de>
1053 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1054 memset constants via native_interpret_expr.
1056 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1059 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1061 (cgraph_node::create_alias): Likewise.
1062 (cgraph_node::get_availability): Check ifunc_resolver instead
1063 of looking up ifunc attribute.
1064 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1065 * varasm.c (do_assemble_alias): Likewise.
1066 (assemble_alias): Likewise.
1067 (default_binds_local_p_3): Likewise.
1068 * cgraph.h (cgraph_node): Add ifunc_resolver.
1069 (cgraph_node::only_called_directly_or_aliased_p): Return false
1071 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1073 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1074 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1075 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1076 instead of looking up ifunc attribute.
1078 2018-05-22 Luis Machado <luis.machado@linaro.org>
1080 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1082 2018-05-22 Martin Sebor <msebor@redhat.com>
1085 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1086 only when expasion succeeds.
1087 (expand_builtin_strcmp): Same.
1088 (expand_builtin_strncmp): Same.
1090 2018-05-22 Martin Sebor <msebor@redhat.com>
1092 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1094 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1095 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1097 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1098 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1100 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1101 Add check for legitimate_address.
1102 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1103 (aarch64_swap_ldrstr_operands): New.
1104 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1107 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1108 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1110 * config/aarch64/aarch64.md: New patterns to generate stp
1112 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1113 single words and double words.
1114 (load_pair_sw, load_pair_dw): Likewise.
1115 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1117 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1119 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1120 for different mode ldpstp and add peephole for merged zero stores.
1122 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1124 (aarch64_gen_store_pair): Rename calls to match new patterns.
1125 (aarch64_gen_load_pair): Rename calls to match new patterns.
1126 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1127 (load_pair<DREG:mode><DREG2:mode>): ... This.
1128 (store_pair<mode>): Rename to...
1129 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1130 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1132 (V_INT_EQUIV): Handle SImode.
1133 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1136 2018-05-22 Martin Sebor <msebor@redhat.com>
1139 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1140 or ajust the presumed bound on an operation to avoid unnecessary
1143 2018-05-22 Martin Sebor <msebor@redhat.com>
1145 PR tree-optimization/85826
1146 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1147 assuming that a DECL necesarily has a constant size.
1149 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1152 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1154 2018-05-22 Richard Biener <rguenther@suse.de>
1156 PR tree-optimization/85834
1157 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1158 non-constant and non-zero memset arguments.
1160 2018-05-22 Martin Liska <mliska@suse.cz>
1163 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1165 2018-05-22 Richard Biener <rguenther@suse.de>
1167 PR tree-optimization/85863
1168 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1169 comparisons when vectype is specified.
1170 (vectorizable_condition): Do not specify vectype for
1171 vect_is_simple_cond when SLP vectorizing.
1173 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1176 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1177 define __ibm128 as long double.
1178 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1179 as a distinct type when IEEE 128-bit support is enabled.
1180 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1181 128-bit types to use the correct functions.
1182 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1183 convert between 128-bit floating point types that have different
1184 modes but the same representation, instead of using gen_lowpart to
1186 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1188 (IFKF_reg): New attributes to give the register constraints for
1190 (extend<mode>tf2_internal): New insns to mark an explicit
1191 conversion between 128-bit floating point types that have a
1192 different mode but share the same representation.
1194 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1196 PR tree-optimization/85814
1197 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1198 a null return from get_strinfo when unsharing the next
1199 strinfo in the chain.
1201 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1204 * varasm.c (weak_finish): Clean up weak_decls.
1206 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1208 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1209 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1210 UNSPEC_UADALP values.
1211 * config/aarch64/iterators.md (ABAL): New int iterator.
1214 (sur): Add mappings for the above.
1215 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1217 (aarch64_<sur>abal<mode>_4): Likewise.
1218 (aarch64_<sur>adalp<mode>_3): Likewise.
1219 (<sur>sadv16qi): New define_expand.
1221 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1223 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1224 (*movdf_internal): Ditto.
1225 (*rcpsf2_sse): Ditto.
1226 (*rsqrtsf2_sse): Ditto.
1227 (*sqrt<mode>2_sse): Ditto.
1229 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1231 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1233 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1234 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1235 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1236 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1238 * config/aarch64/arm_neon.h: Likewise.
1239 * config/aarch64/iterators.md (VQ_I): New.
1241 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1243 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1245 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1248 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1250 * config/nds32/constraints.md (S): New constraint.
1251 * config/nds32/nds32.md (call_internal): Use constraint S.
1252 (call_value_internal): Likewise.
1253 (sibcall_internal): Likewise.
1254 (sibcall_value_internal): Likewise.
1256 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1257 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1262 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1263 Chung-Ju Wu <jasonwucj@gmail.com>
1265 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1266 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1267 (nds32_rtx_costs_impl): Simplify.
1268 (nds32_address_cost_impl): Simplify.
1269 (nds32_init_rtx_costs): New function.
1270 (nds32_rtx_costs_speed_prefer): Likewise.
1271 (nds32_rtx_costs_size_prefer): Likewise.
1272 (nds32_address_cost_speed_prefer): Likewise.
1273 (nds32_address_cost_speed_fwprop): Likewise.
1274 (nds32_address_cost_size_prefer): Likewise.
1275 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1276 * config/nds32/nds32.c (nds32_option_override): Use
1277 nds32_init_rtx_costs function.
1279 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1281 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1282 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1283 (TARGET_PIPELINE_N8): Likewise.
1284 (TARGET_PIPELINE_N10): Likewise.
1285 (TARGET_PIPELINE_N13): Likewise.
1286 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1288 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1290 * config/nds32/nds32-fpu.md: Update copyright year.
1292 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1294 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1296 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1298 * config/nds32/nds32.c
1299 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1300 * config/nds32/nds32.opt (minline-asm-r15): New option.
1302 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1304 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1306 * config/nds32/nds32.md (abssi2): New pattern.
1308 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1310 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1311 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1312 Merge insn pattern from sse_cvtsi2ss<round_name> and
1313 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1314 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1315 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1316 using SWI48 mode iterator.
1317 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1318 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1319 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1320 pattern from sse_cvttss2si<round_saeonly_name>
1321 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1322 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1323 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1324 using SWI48 mode iterator.
1325 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1326 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1327 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1328 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1329 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1330 using SWI48 mode iterator.
1331 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1332 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1333 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1334 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1335 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1336 SWI48 mode iterator.
1337 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1338 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1339 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1340 pattern from sse_cvttsd2si<round_saeonly_name>
1341 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1343 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1345 * config/nds32/nds32-md-auxiliary.c
1346 (nds32_valid_smw_lwm_base_p): Refine.
1347 (nds32_output_smw_single_word): Refine.
1348 (nds32_output_smw_double_word): New.
1349 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1351 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1354 (nds32_output_stack_pop): Refine.
1355 (nds32_expand_unaligned_load): Refine.
1356 (nds32_expand_unaligned_store): Refine.
1358 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1359 Chung-Ju Wu <jasonwucj@gmail.com>
1361 * config/nds32/constants.md: Add TP_REGNUM constant.
1362 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1363 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1365 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1366 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1367 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1368 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1369 and PIC code generation.
1370 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1372 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1374 * config/nds32/nds32.md: Support TLS and PIC.
1375 * config/nds32/nds32.c: Support TLS and PIC.
1376 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1377 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1380 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1382 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1383 mode with E_ prefix.
1385 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1386 Chung-Ju Wu <jasonwucj@gmail.com>
1388 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1389 * config/nds32/nds32-md-auxiliary.c
1390 (symbolic_reference_mentioned_p): New.
1391 (nds32_legitimize_ict_address): New.
1392 (nds32_expand_ict_move): New.
1393 (nds32_indirect_call_referenced_p): New.
1394 (nds32_symbol_binds_local_p): Delete.
1395 (nds32_long_call_p): Modify.
1396 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1397 * config/nds32/nds32-protos.h
1398 (symbolic_reference_mentioned_p): Declare.
1399 (nds32_legitimize_ict_address): Declare.
1400 (nds32_expand_ict_move): Declare.
1401 (nds32_indirect_call_referenced_p): Declare.
1402 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1403 (nds32_relax_group): Use nds32_ict_const_p as condition.
1404 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1405 (nds32_asm_file_start): Output ict_model directive in asm code.
1406 (nds32_legitimate_address_p): Consider indirect call.
1407 (nds32_print_operand): Consider indirect call.
1408 (nds32_print_operand_address): Consider indirect call.
1409 (nds32_insert_attributes): Handle "indirect_call" attribute.
1410 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1411 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1412 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1413 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1414 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1415 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1416 (TARGET_ICT_MODEL_SMALL): Define.
1417 (TARGET_ICT_MODEL_LARGE): Define.
1418 * config/nds32/nds32.md (movsi): Consider ict model.
1419 (call, call_value): Consider ict model.
1420 (sibcall, sibcall_value): Consider ict model.
1421 * config/nds32/nds32.opt (mict-model): New option.
1422 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1425 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1426 Monk Chiang <sh.chiang04@gmail.com>
1427 Jim Wilson <jimw@sifive.com>
1429 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1430 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1431 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1432 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1433 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1434 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1435 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1436 compute save_libcall_adjustment properly.
1437 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1438 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1439 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1440 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1441 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1442 (ABI_SPEC): Handle mabi=ilp32e.
1443 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1444 (RVE): Add RVE mask.
1445 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1446 <-march>: Add rv32e as an example.
1448 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1451 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1452 (intra_create_variable_infos): Handle C++ constructors.
1454 2018-05-18 Martin Liska <mliska@suse.cz>
1456 * passes.def: Remove a redundant pass.
1458 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1461 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1463 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1465 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1467 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1469 (armv2, armv2a, armv3, armv3m): Delete architectures.
1470 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1471 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1472 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1474 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1475 (*mulsidi3adddi): Likewise.
1476 (mulsidi3): Likewise.
1477 (*mulsidi3_nov6): Likewise.
1478 (umulsidi3): Likewise.
1479 (umulsidi3_nov6): Likewise.
1480 (umaddsidi4): Likewise.
1481 (*umulsidi3adddi): Likewise.
1482 (smulsi3_highpart): Likewise.
1483 (*smulsi3_highpart_nov6): Likewise.
1484 (umulsi3_highpart): Likewise.
1485 (*umulsi3_highpart_nov6): Likewise.
1486 * config/arm/arm.h (arm_arch3m): Delete.
1487 * config/arm/arm.c (arm_arch3m): Delete.
1488 (arm_option_override_internal): Update armv3-related comment.
1489 (arm_configure_build_target): Delete use of isa_bit_mode32.
1490 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1491 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1492 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1494 (mulusa3): Likewise.
1495 * config/arm/arm-protos.h (arm_arch3m): Delete.
1496 * config/arm/arm-tables.opt: Regenerate.
1497 * config/arm/arm-tune.md: Likewise.
1498 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1499 deleted architectures.
1501 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1503 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1504 (armv5t, armv5te): New features.
1505 (ARMv5, ARMv5e): Delete fgroups.
1506 (ARMv5t, ARMv5te): Adjust for above changes.
1508 (armv5, armv5e): Delete arches.
1509 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1511 (*call_reg_arm): Likewise.
1512 (*call_value_reg_armv5): Likewise.
1513 (*call_value_reg_arm): Likewise.
1514 (*call_symbol): Likewise.
1515 (*call_value_symbol): Likewise.
1516 (*sibcall_insn): Likewise.
1517 (*sibcall_value_insn): Likewise.
1519 (prefetch): Likewise.
1520 (define_split and define_peephole2 dependent on arm_arch5):
1522 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1524 (TARGET_ARM_QBIT): Likewise.
1525 (TARGET_DSP_MULTIPLY): Likewise.
1526 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1527 (arm_arch5, arm_arch5e): Delete.
1528 (arm_arch5t, arm_arch5te): Declare.
1529 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1530 (arm_arch5t): Declare.
1531 (arm_option_reconfigure_globals): Update for the above.
1532 (arm_options_perform_arch_sanity_checks): Update comment, replace
1533 use of arm_arch5 with arm_arch5t.
1534 (use_return_insn): Likewise.
1535 (arm_emit_call_insn): Likewise.
1536 (output_return_instruction): Likewise.
1537 (arm_final_prescan_insn): Likewise.
1538 (arm_coproc_builtin_available): Likewise.
1539 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1540 arm_arch5e with arm_arch5t and arm_arch5te.
1541 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1542 (arm_arch5t, arm_arch5te): Declare.
1543 * config/arm/arm-tables.opt: Regenerate.
1544 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1545 * config/arm/t-multilib: Likewise.
1546 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1547 instead of arm_arch5.
1548 (*call_reg_thumb1): Likewise.
1549 (*call_value_reg_thumb1_v5): Likewise.
1550 (*call_value_reg_thumb1): Likewise.
1551 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1553 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1555 2018-05-18 Martin Liska <mliska@suse.cz>
1557 PR gcov-profile/84846
1558 * doc/gcov.texi: Document -t option of gcov tool.
1560 2018-05-18 Martin Liska <mliska@suse.cz>
1562 PR gcov-profile/84846
1563 * gcov.c (print_usage): Add new -t option.
1564 (process_args): Handle the option.
1565 (generate_results): Use stdout as output when requested by
1568 2018-05-18 Martin Liska <mliska@suse.cz>
1570 PR gcov-profile/84846
1571 * coverage.c (coverage_init): Write PWD to .gcno file.
1572 * doc/gcov.texi: Document how working directory is printed.
1573 * gcov-dump.c (dump_gcov_file): Print PWD.
1574 * gcov.c (output_intermediate_file): Likewise.
1575 (read_graph_file): Read PWD string.
1576 (output_lines): Print PWD.
1578 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1581 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1582 for retval and return false if all args to phi are zero.
1584 2018-05-18 Richard Biener <rguenther@suse.de>
1586 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1588 (evrp_dom_walker::before_dom_children): Call it.
1590 2018-05-18 Richard Biener <rguenther@suse.de>
1592 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1593 results when processing array refs with variable index.
1595 2018-05-18 Toon Moene <toon@moene.org>
1597 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1598 directly after that of -floop-interchange. Indicate that both
1599 options are enabled by default when specifying -O3.
1601 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1603 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1604 iterator. Delete separate integer-mode vec_set<mode> expander.
1605 (aarch64_simd_vec_setv2di): Delete.
1606 (vec_setv2di): Delete.
1607 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1608 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1609 the "w, r" alternative.
1611 2018-05-18 Martin Liska <mliska@suse.cz>
1613 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1614 * tree-pass.h (make_pass_lower_switch_O0): New function.
1615 * tree-switch-conversion.c (node_has_low_bound): Remove.
1616 (node_has_high_bound): Likewise.
1617 (node_is_bounded): Likewise.
1618 (class pass_lower_switch): Make it a template type and create
1620 (pass_lower_switch::execute): Add template argument.
1621 (make_pass_lower_switch): New function.
1622 (make_pass_lower_switch_O0): New function.
1623 (do_jump_if_equal): Remove.
1624 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1625 all the hard work to tree optimization passes.
1627 2018-05-18 Martin Liska <mliska@suse.cz>
1629 * dbgcnt.c (limit_low): Renamed from limit.
1630 (limit_high): New variable.
1631 (dbg_cnt_is_enabled): Check for upper limit.
1632 (dbg_cnt): Adjust dumping.
1633 (dbg_cnt_set_limit_by_index): Add new argument for high
1635 (dbg_cnt_set_limit_by_name): Likewise.
1636 (dbg_cnt_process_single_pair): Parse new format.
1637 (dbg_cnt_process_opt): Use strtok.
1638 (dbg_cnt_list_all_counters): Remove 'value' and add
1640 * doc/invoke.texi: Document changes.
1642 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1644 * doc/sourcebuild.texi (scalar_all_fma): Document.
1645 * tree.def (FMA_EXPR): Delete.
1646 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1647 * internal-fn.c (ternary_direct): New macro.
1648 (expand_ternary_optab_fn): Likewise.
1649 (direct_ternary_optab_supported_p): Likewise.
1650 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1651 * builtins.c (fold_builtin_fma): Delete.
1652 (fold_builtin_3): Don't call it.
1653 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1654 * expr.c (expand_expr_real_2): Likewise.
1655 * fold-const.c (operand_equal_p): Likewise.
1656 (fold_ternary_loc): Likewise.
1657 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1658 * gimple.c (DEFTREECODE): Likewise.
1659 * gimplify.c (gimplify_expr): Likewise.
1660 * optabs-tree.c (optab_for_tree_code): Likewise.
1661 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1662 * tree-eh.c (operation_could_trap_p): Likewise.
1663 (stmt_could_throw_1_p): Likewise.
1664 * tree-inline.c (estimate_operator_cost): Likewise.
1665 * tree-pretty-print.c (dump_generic_node): Likewise.
1666 (op_code_prio): Likewise.
1667 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1668 * tree-ssa-operands.c (get_expr_operands): Likewise.
1669 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1670 * fold-const-call.h (fold_fma): Delete.
1671 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1672 CFN_FNMA and CFN_FNMS.
1674 * genmatch.c (combined_fn): New enum.
1675 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1676 (commutative_op): New function.
1677 (commutate): Use it. Handle more than 2 operands.
1678 (dt_operand::gen_gimple_expr): Use commutative_op.
1679 (parser::parse_expr): Allow :c to be used with non-binary
1680 operators if the commutative operand is known.
1681 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1682 CFN_FMS, CFN_FNMA and CFN_FNMS.
1683 (backprop::process_assign_use): Remove FMA_EXPR handling.
1684 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1685 (gen_hsa_fma): New function.
1686 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1687 IFN_FNMA and IFN_FNMS.
1688 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1689 * gimple-fold.h (follow_all_ssa_edges): Declare.
1690 * gimple-fold.c (follow_all_ssa_edges): New function.
1691 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1692 gimple_build interface and use follow_all_ssa_edges to fold the result.
1693 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1694 instead of checking for optabs directly.
1695 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1696 rather than FMA_EXPRs.
1697 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1698 call to IFN_FMA instead of an FMA_EXPR.
1700 2018-05-17 Jim Wilson <jimw@sifive.com>
1702 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1703 sign extended promoted subreg, and the range does not have the sign bit
1704 set, then do a sign extend.
1706 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1707 test, check for sign extended subreg and/or constant operands, and
1708 do a sign extend in that case.
1710 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1712 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1713 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1715 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1716 Change logics_shift_reg to logics_shift_imm.
1717 (thunderx2t99_fp_loadpair_basic): Delete.
1718 (thunderx2t99_fp_storepair_basic): Delete.
1719 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1720 (thunderx2t99_asimd_polynomial): Delete.
1721 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1722 and neon_fp_mul_d_scalar_q.
1723 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1724 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1725 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1726 (thunderx2t99_asimd_lut): Add missing tbl types.
1727 (thunderx2t99_asimd_ext): Delete.
1728 (thunderx2t99_asimd_load1_1_mult): Delete.
1729 (thunderx2t99_asimd_load1_2_mult): Delete.
1730 (thunderx2t99_asimd_load1_ldp): New.
1731 (thunderx2t99_asimd_load1): New.
1732 (thunderx2t99_asimd_load2): Add missing *load2* types.
1733 (thunderx2t99_asimd_load3): New.
1734 (thunderx2t99_asimd_load4): New.
1735 (thunderx2t99_asimd_store1_1_mult): Delete.
1736 (thunderx2t99_asimd_store1_2_mult): Delete.
1737 (thunderx2t99_asimd_store2_mult): Delete.
1738 (thunderx2t99_asimd_store2_onelane): Delete.
1739 (thunderx2t99_asimd_store_stp): New.
1740 (thunderx2t99_asimd_store1): New.
1741 (thunderx2t99_asimd_store2): New.
1742 (thunderx2t99_asimd_store3): New.
1743 (thunderx2t99_asimd_store4): New.
1745 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1747 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1748 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1750 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1751 Segher Boessenkool <segher@kernel.crashing.org>
1754 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1757 2018-05-17 Richard Biener <rguenther@suse.de>
1759 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1760 for pruning loop and prune defs feeding only already visited PHIs.
1762 2018-05-17 Richard Biener <rguenther@suse.de>
1764 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1766 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1767 Richard Biener <rguenther@suse.de>
1769 PR tree-optimization/85793
1770 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1771 for VMAT_ELEMENTWISE.
1773 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1775 * internal-fn.h (lookup_internal_fn): Declare
1776 * internal-fn.c (lookup_internal_fn): New function.
1777 * gimple.c (gimple_build_call_from_tree): Handle calls to
1779 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1780 internal function names.
1781 * tree-pretty-print.c (dump_generic_node): Likewise.
1782 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1784 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1786 * gimple-fold.h (gimple_build): Make the function forms take
1787 combined_fn rather than built_in_function.
1788 (gimple_simplify): Likewise.
1789 * gimple-match-head.c (gimple_simplify): Likewise.
1790 * gimple-fold.c (gimple_build): Likewise.
1791 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1792 rather than gimple_build_call_internal.
1793 (get_initial_defs_for_reduction): Likewise.
1794 (vect_create_epilog_for_reduction): Likewise.
1795 (vectorizable_live_operation): Likewise.
1797 2018-05-17 Martin Liska <mliska@suse.cz>
1799 * gimple-ssa-sprintf.c (format_directive): Do not use
1800 space in between 'G_' and '('.
1802 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1805 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1806 even if the mask is not all ones.
1809 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1811 (ix86_gimple_fold_builtin): Likewise.
1814 * config/i386/i386.c: Include tree-vector-builder.h.
1815 (ix86_vector_shift_count): New function.
1816 (ix86_fold_builtin): Fold shift builtins by scalar count.
1817 (ix86_gimple_fold_builtin): Likewise.
1819 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1820 _mm512_setzero): New intrinsics.
1822 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1823 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1825 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1826 code generation for cases where splatting a value is not useful.
1827 * simplify-rtx.c (simplify_ternary_operation): Simplify
1828 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1829 mode to a vec_concat.
1831 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1833 * config.gcc: Support "goldmont-plus".
1834 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1836 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1837 PROCESSOR_GOLDMONT_PLUS.
1838 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1839 (processor_target_table): Add "goldmont-plus".
1840 (PTA_GOLDMONT_PLUS): Define.
1841 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1842 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1843 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1844 (fold_builtin_cpu): Add "goldmont-plus".
1845 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1846 (ix86_option_override_internal): Add "goldmont-plus".
1847 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1848 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1849 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1850 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1852 2018-05-17 Richard Biener <rguenther@suse.de>
1854 PR tree-optimization/85757
1855 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1856 remove defs that only feed that PHI from further processing.
1858 2018-05-16 Jim Wilson <jimw@sifive.com>
1860 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1862 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1864 2018-05-16 Mark Wielaard <mark@klomp.org>
1866 * dwarf2out.c (count_index_strings): New function.
1867 (output_indirect_strings): Call count_index_strings and generate
1868 header for dwarf_version >= 5.
1870 2018-05-16 Mark Wielaard <mark@klomp.org>
1872 * dwarf2out.c (dwarf_FORM): New function.
1873 (set_indirect_string): Use dwarf_FORM.
1874 (reset_indirect_string): Likewise.
1875 (size_of_die): Likewise.
1876 (value_format): Likewise.
1877 (output_die): Likewise.
1878 (add_skeleton_AT_string): Likewise.
1879 (output_macinfo_op): Likewise.
1880 (index_string): Likewise.
1881 (output_index_string_offset): Likewise.
1882 (output_index_string): Likewise.
1883 (count_index_strings): Likewise.
1885 2018-05-16 Carl Love <cel@us.ibm.com>
1887 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1888 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1890 2018-05-16 Martin Jambor <mjambor@suse.cz>
1892 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1893 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1895 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1897 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1898 (fnma<mode>4): Likewise.
1899 (fms<mode>4): Likewise.
1900 (fnms<mode>4): Likewise.
1901 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1902 (aarch64_fnma<mode>4): Likewise.
1903 (aarch64_fms<mode>4): Likewise.
1904 (aarch64_fnms<mode>4): Likewise.
1905 (aarch64_fnmadd<mode>4): Likewise.
1907 2018-05-16 Jason Merrill <jason@redhat.com>
1909 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1911 2018-05-16 Richard Biener <rguenther@suse.de>
1913 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1914 (dump_stmt_cost): Declare.
1915 (add_stmt_cost): Dump cost we add.
1916 (add_stmt_costs): New function.
1917 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1919 (vect_analyze_stmt): Adjust prototype.
1920 (vectorizable_condition): Likewise.
1921 (vectorizable_live_operation): Likewise.
1922 (vectorizable_reduction): Likewise.
1923 (vectorizable_induction): Likewise.
1924 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1925 cost vector to pass to vectorizable_ and record afterwards.
1926 (vect_model_reduction_cost): Take cost vector argument and adjust.
1927 (vect_model_induction_cost): Likewise.
1928 (vectorizable_reduction): Likewise.
1929 (vectorizable_induction): Likewise.
1930 (vectorizable_live_operation): Likewise.
1931 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1932 SLP_TREE_NUMBER_OF_VEC_STMTS.
1933 (vect_analyze_slp_cost_1): Remove.
1934 (vect_analyze_slp_cost): Likewise.
1935 (vect_slp_analyze_node_operations): Take visited args and
1936 a target cost vector. Avoid processing already visited stmt sets.
1937 (vect_slp_analyze_operations): Use a local cost vector to gather
1938 costs and register those of non-discarded instances.
1939 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1940 (vect_schedule_slp_instance): Remove copying of
1941 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1943 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1944 adding cost. Record cost entry location.
1945 (vect_prologue_cost_for_slp_op): Function to compute cost of
1946 a constant or invariant generated for SLP vect in the prologue,
1947 split out from vect_analyze_slp_cost_1.
1948 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1949 (vect_model_promotion_demotion_cost): Likewise.
1950 (vect_model_store_cost): Likewise, make static.
1951 (vect_model_load_cost): Likewise.
1952 (vectorizable_bswap): Add cost vector arg and adjust.
1953 (vectorizable_call): Likewise.
1954 (vectorizable_simd_clone_call): Likewise.
1955 (vectorizable_conversion): Likewise.
1956 (vectorizable_assignment): Likewise.
1957 (vectorizable_shift): Likewise.
1958 (vectorizable_operation): Likewise.
1959 (vectorizable_store): Likewise.
1960 (vectorizable_load): Likewise.
1961 (vectorizable_condition): Likewise.
1962 (vectorizable_comparison): Likewise.
1963 (can_vectorize_live_stmts): Likewise.
1964 (vect_analyze_stmt): Likewise.
1965 (vect_transform_stmt): Adjust calls to vectorizable_*.
1966 * tree-vectorizer.c: Include gimple-pretty-print.h.
1967 (dump_stmt_cost): New function.
1969 2018-05-16 Richard Biener <rguenther@suse.de>
1971 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1972 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1973 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1974 (check_name): New callback.
1975 (dse_classify_store): Track cycles via a visited bitmap of PHI
1976 defs and simplify handling of in-loop and across loop dead stores
1977 and properly fail for loop-variant refs. Handle byte-tracking with
1978 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1981 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1983 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1984 (vect_get_mask_type_for_stmt): Likewise.
1985 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1987 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1988 to determine the statement's vector type and the vector type that
1989 should be used for calculating nunits. Deal with cases in which
1990 the type has to be deferred.
1991 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1992 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1993 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1994 (vect_determine_vf_for_stmt): New functions, split out from...
1995 (vect_determine_vectorization_factor): ...here.
1996 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1997 (vect_get_mask_type_for_stmt): New functions, split out from
1998 vect_determine_vectorization_factor.
2000 2018-05-16 Richard Biener <rguenther@suse.de>
2002 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2003 verify the [VEC_]COND_EXPR embedded comparison.
2005 2018-05-15 Martin Sebor <msebor@redhat.com>
2007 PR tree-optimization/85753
2008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2009 RECORD_TYPE in addition to ARRAY_TYPE.
2011 2018-05-15 Martin Sebor <msebor@redhat.com>
2014 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2016 2018-05-15 Richard Biener <rguenther@suse.de>
2018 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2019 add by_clobber_p one. Change algorithm to collect all defs
2020 representing uses we need to walk and try reducing them to
2021 a single one before failing.
2022 (dse_dom_walker::dse_optimize_stmt): Adjust.
2024 2018-05-13 Mark Wielaard <mark@klomp.org>
2026 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2027 (size_of_loc_descr): Likewise.
2028 (output_loc_operands): Likewise.
2029 (output_loc_operands_raw): Likewise.
2030 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2031 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2032 (hash_loc_operands): Likewise.
2033 (compare_loc_operands): Likewise.
2035 2018-05-14 Mark Wielaard <mark@klomp.org>
2037 * dwarf2out.c (count_index_addrs): New function.
2038 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2040 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2042 PR tree-optimization/83648
2043 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2044 return value as malloc candidate.
2046 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2049 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2050 as true in call to suggest_attribute.
2052 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2054 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2055 -mreadonly-in-sdata.
2057 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2059 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2061 (aarch64_crypto_aesd_fused): Likewise.
2063 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2065 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2066 (movsi_aarch64): Likewise.
2067 (load_pairsi): Likewise.
2068 (load_pairdi): Likewise.
2069 (store_pairsi): Likewise.
2070 (store_pairdi): Likewise.
2071 (load_pairsf): Likewise.
2072 (load_pairdf): Likewise.
2073 (store_pairsf): Likewise.
2074 (store_pairdf): Likewise.
2075 (zero_extend): Likewise.
2076 (trunc): Swap alternatives.
2077 (fcvt_target): Add '?' to prefer w over r.
2079 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2082 * config/i386/i386.md: Disallow non-commutative arithmetics in
2083 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2084 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2085 in the peephole2 before it.
2087 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2089 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2090 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2091 (ix86_handle_option): Handle -mcldemote.
2092 * config.gcc: New header.
2093 * config/i386/cldemoteintrin.h: New file.
2094 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2095 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2097 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2098 OPTION_MASK_ISA_CLDEMOTE.
2099 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2100 (ix86_valid_target_attribute_inner_p): Ditto.
2101 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2102 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2103 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2104 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2105 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2107 * config/i386/i386.opt: Add -mcldemote.
2108 * config/i386/x86intrin.h: New header.
2109 * doc/invoke.texi: Add -mcldemote.
2111 2018-05-14 Richard Biener <rguenther@suse.de>
2113 * doc/match-and-simplify.texi: Adjust :s documentation.
2115 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2117 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2118 intended memcpy size.
2119 (REORDER_45): Likewise.
2121 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2123 * sort.cc: New file.
2124 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2125 * vec.c (qsort_chk): Use gcc_qsort.
2126 * Makefile.in (OBJS-libcommon): Add sort.o.
2127 (build/sort.o): New target. Use it...
2128 (BUILD_RTL): ... here, and...
2129 (build/gencfn-macros): ... here, and...
2130 (build/genmatch): ... here.
2132 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2133 Chung-Ju Wu <jasonwucj@gmail.com>
2135 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2136 * config/nds32/nds32-graywolf.md: New file.
2137 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2138 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2140 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2141 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2142 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2143 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2144 * config/nds32/pipelines.md: Include n15 settings.
2146 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2147 Chung-Ju Wu <jasonwucj@gmail.com>
2149 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2150 * config/nds32/nds32-n13.md: New file.
2151 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2152 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2154 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2155 * config/nds32/nds32.md (pipeline_model): Add n13.
2156 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2157 * config/nds32/pipelines.md: Include n13 settings.
2159 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2160 Chung-Ju Wu <jasonwucj@gmail.com>
2162 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2163 * config/nds32/nds32-n10.md: New file.
2164 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2165 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2167 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2168 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2169 * config/nds32/nds32.md (pipeline_model): Add n10.
2170 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2171 * config/nds32/pipelines.md: Include n10 settings.
2173 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2174 Kito Cheng <kito.cheng@gmail.com>
2175 Chung-Ju Wu <jasonwucj@gmail.com>
2177 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2178 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2179 Add enum values for DSP extension instructions.
2180 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2182 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2183 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2185 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2186 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2187 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2188 * config/nds32/nds32-intrinsic.md: Likewise.
2189 * config/nds32/nds32_intrinsic.h: Likewise.
2190 * config/nds32/nds32-md-auxiliary.c: Likewise.
2191 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2192 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2193 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2194 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2195 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2196 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2197 TYPE_DMAC in switch statement.
2198 * config/nds32/nds32.c: New checking and implementation for DSP
2199 extension instructions.
2200 * config/nds32/nds32.h: Likewise.
2201 * config/nds32/nds32.md: Likewise.
2202 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2203 * config/nds32/predicates.md: Implement new predicates for DSP
2206 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2208 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2209 Reformat alternatives and attributes so it is easier to identify
2210 which constraints/attributes go with which instruction.
2211 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2212 (mov<mode>_softfloat32, FMOVE64): Likewise.
2213 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2214 (mov<mode>_softfloat64, FMOVE64): Likewise.
2216 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2218 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2220 (Basic PowerPC Built-in Functions): The new name of the
2221 subsection previously known as "PowerPC Built-in Functions".
2222 (Basic PowerPC Built-in Functions Available on all Configurations):
2224 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2225 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2226 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2227 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2229 2018-05-11 Martin Jambor <mjambor@suse.cz>
2232 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2235 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2238 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2240 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2242 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2243 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2244 (ix86_handle_option): Handle -mwaitpkg.
2245 * config.gcc: New header.
2246 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2247 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2248 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2250 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2251 OPTION_MASK_ISA_WAITPKG.
2252 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2253 (ix86_option_override_internal): Add PTA_WAITPKG.
2254 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2255 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2256 IX86_BUILTIN_TPAUSE.
2257 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2258 __builtin_ia32_umwait and __builtin_ia32_tpause.
2259 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2260 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2261 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2262 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2263 UNSPECV_TPAUSE): New.
2264 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2265 * config/i386/i386.opt: Add -mwaitpkg.
2266 * config/i386/waitpkgintrin.h: New file.
2267 * config/i386/x86intrin.h: New header.
2268 * doc/invoke.texi: Add -mwaitpkg.
2270 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2273 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2275 (cortex-m0): Use armv6s-m isa.
2276 (cortex-m0plus): Likewise.
2277 (cortex-m1): Likewise.
2278 (cortex-m0.small-multiply): Likewise.
2279 (cortex-m0plus.small-multiply): Likewise.
2280 (cortex-m1.small-multiply): Likewise.
2282 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2283 Jakub Jelinek <jakub@redhat.com>
2285 PR tree-optimization/85692
2286 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2287 source permute as well.
2289 2018-05-11 Martin Liska <mliska@suse.cz>
2292 * doc/extend.texi: Document LLVM style format for no_sanitize
2295 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2297 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2298 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2299 (mode_supports_vsx_dform_quad): Likewise.
2300 (mode_supports_vmx_dform): Move these functions to be next to the
2301 other mode_supports functions.
2302 (mode_supports_dq_form): Likewise.
2303 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2304 mode_supports_dq_form.
2305 (reg_offset_addressing_ok_p): Likewise.
2306 (offsettable_ok_by_alignment): Likewise.
2307 (rs6000_legitimate_offset_address_p): Likewise.
2308 (legitimate_lo_sum_address_p): Likewise.
2309 (rs6000_legitimize_address): Likewise.
2310 (rs6000_legitimize_reload_address): Likewise.
2311 (rs6000_secondary_reload_inner): Likewise.
2312 (rs6000_preferred_reload_class): Likewise.
2313 (rs6000_output_move_128bit): Likewise.
2315 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2317 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2318 Generate SImode target register for null target.
2319 <case IX86_BUILTIN_XGETBV>: Ditto.
2320 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2321 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2323 2018-05-10 Carl Love <cel@us.ibm.com>
2325 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2326 dcbtt and dcbtstt if operands[2] is 0.
2328 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2331 * config/i386/sse.md (usadv64qi): New expander.
2333 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2335 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2336 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2337 -maltivec=be support.
2338 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2339 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2340 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2341 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2342 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2343 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2344 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2345 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2346 altivec_vsumsws): Adjust.
2347 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2348 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2349 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2351 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2352 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2353 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2354 (altivec_lve<VI_char>x): Delete expand.
2355 (*altivec_lve<VI_char>x_internal): Rename to...
2356 (altivec_lve<VI_char>x): ... this.
2357 (altivec_lvxl_<mode>): Delete expand.
2358 (*altivec_lvxl_<mode>_internal): Rename to ...
2359 (altivec_lvxl_<mode>): ... this.
2360 (altivec_stvxl_<mode>): Delete expand.
2361 (*altivec_stvxl_<mode>_internal): Rename to ...
2362 (altivec_stvxl_<mode>): ... this.
2363 (altivec_stve<VI_char>x): Delete expand.
2364 (*altivec_stve<VI_char>x_internal): Rename to ...
2365 (altivec_stve<VI_char>x): ... this.
2366 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2367 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2368 reduc_plus_scal_<mode>): Adjust.
2369 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2371 (rs6000_cpu_cpp_builtins): Adjust.
2372 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2373 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2374 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2375 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2376 -maltivec=be support.
2377 (rs6000_split_vec_extract_var): Adjust.
2378 (rs6000_split_v4si_init): Adjust.
2379 (swap_selector_for_mode): Delete.
2380 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2381 altivec_expand_stvex_be): Delete.
2382 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2383 -maltivec=be support.
2384 (rs6000_gimple_fold_builtin): Ditto.
2385 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2387 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2388 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2389 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2390 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2391 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2392 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2393 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2394 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2395 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2396 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2397 anonymous split): Adjust.
2398 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2399 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2401 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2403 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2404 when --with-gxx-include-dir is also specified.
2405 * configure: Regenerate.
2407 2018-05-09 Jim Wilson <jimw@sifive.com>
2410 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2411 * config/riscv/t-withmultilib: New.
2412 * config/riscv/withmultilib.h: New.
2413 * doc/install.texi: Document RISC-V --with-multilib-list support.
2415 2018-05-09 Richard Biener <rguenther@suse.de>
2417 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2419 (vect_bb_vectorization_profitable_p): Adjust. Compute
2420 actual scalar cost using the cost vector and the add_stmt_cost
2423 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2425 PR rtl-optimization/85645
2426 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2427 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2429 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2431 PR rtl-optimization/85645
2432 * shrink-wrap.c (spread_components): Return a boolean saying if
2433 anything was changed.
2434 (try_shrink_wrapping_separate): Iterate spread_components until
2435 nothing changes anymore.
2437 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2439 PR rtl-optimization/85645
2440 * regrename.c (build_def_use): Also kill the chains that include the
2441 destination of a REG_CFA_REGISTER note.
2443 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2445 PR rtl-optimization/85645
2446 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2447 insn that has a REG_CFA_REGISTER note.
2449 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2451 * cfgexpand.c (expand_clobber): New function.
2452 (expand_gimple_stmt_1): Use it.
2453 * tree-vect-stmts.c (vect_clobber_variable): New function,
2455 (vectorizable_simd_clone_call): ...here.
2456 (vectorizable_store): Emit a clobber either side of an
2457 IFN_STORE_LANES sequence.
2458 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2460 2018-05-09 Tom de Vries <tom@codesourcery.com>
2463 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2464 (define_insn "trap_if_false"): Add exit after trap.
2466 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2468 PR rtl-optimization/85638
2469 * bb-reorder.c: Include common/common-target.h.
2470 (create_forwarder_block): New function extracted from...
2471 (fix_up_crossing_landing_pad): ...here. Rename into...
2472 (dw2_fix_up_crossing_landing_pad): ...this.
2473 (sjlj_fix_up_crossing_landing_pad): New function.
2474 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2475 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2476 from both partitions and exit the loop after one iteration.
2478 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2481 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2483 (Basic PowerPC Built-in Functions): The new name of the
2484 subsection previously known as "PowerPC Built-in Functions".
2485 (Basic PowerPC Built-in Functions Available on all Configurations):
2487 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2489 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2491 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2493 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2496 2018-05-08 Jim Wilson <jimw@sifive.com>
2498 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2499 (LD_EMUL_SUFFIX): New.
2500 (LINK_SPEC): Use it.
2502 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2504 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2506 (Basic PowerPC Built-in Functions): The new name of the
2507 subsection previously known as "PowerPC Built-in Functions".
2508 (Basic PowerPC Built-in Functions Available on all Configurations):
2510 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2512 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2514 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2516 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2519 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2522 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2523 after cmpelim optimization.
2525 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2527 * config.gcc: Support "goldmont".
2528 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2529 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2531 * config/i386/i386.c (m_GOLDMONT): Define.
2532 (processor_target_table): Add "goldmont".
2533 (PTA_GOLDMONT): Define.
2534 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2535 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2536 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2537 (fold_builtin_cpu): Add "goldmont".
2538 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2539 (ix86_option_override_internal): Add "goldmont".
2540 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2541 (processor_type): Add PROCESSOR_GOLDMONT.
2542 * config/i386/i386.md: Add CPU "glm".
2543 * config/i386/glm.md: New file.
2544 * config/i386/x86-tune.def: Add m_GOLDMONT.
2545 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2547 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2550 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2552 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2553 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2554 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2557 * config/i386/i386.c (ix86_fold_builtin): Handle
2558 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2561 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2562 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2564 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2567 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2568 (check_arch): Likewise.
2569 (check_fpu): Return the result rather than printing it.
2570 (end arch): Fix operator precedence.
2571 (end cpu): Likewise.
2572 (END): Print the result from check_fpu.
2574 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2575 Alan Hayward <alan.hayward@arm.com>
2576 David Sherwood <david.sherwood@arm.com>
2578 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2579 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2580 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2581 (*fcmuo<mode>_and): New patterns.
2583 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2585 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2586 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2587 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2588 (cmp_op, sve_imm_con): New code attributes.
2589 (SVE_COND_INT_CMP, imm_con): Delete.
2590 (cmp_op): Remove above unspecs from int attribute.
2591 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2593 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2594 comparison-specific unspecs.
2595 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2596 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2597 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2598 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2599 (*vec_fcm<cmp_op><mode>): Rename to...
2600 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2601 (*vec_fcmuo<mode>): Rename to...
2602 (*fcmuo<mode>): ...this and adjust likewise.
2603 (*pred_fcm<cmp_op><mode>): New pattern.
2604 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2605 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2607 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2609 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2610 (aarch64_emit_sve_predicated_cond): New function.
2611 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2612 (aarch64_emit_unspec_cond_or): Replace with...
2613 (aarch64_emit_sve_or_conds): ...this new function. Use
2614 aarch64_emit_sve_ptrue_op for the individual comparisons and
2615 aarch64_emit_binop to OR them together.
2616 (aarch64_emit_inverted_unspec_cond): Replace with...
2617 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2618 aarch64_emit_sve_ptrue_op for the comparison and
2619 aarch64_emit_unop to invert the result.
2620 (aarch64_expand_sve_vec_cmp_float): Update after the above
2621 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2623 2018-05-07 Nathan Sidwell <nathan@acm.org>
2625 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2626 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2627 (Backwards Compatibility): Likewise.
2629 2018-05-07 Luis Machado <luis.machado@linaro.org>
2633 2018-05-07 Luis Machado <luis.machado@linaro.org>
2635 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2636 <prefetch_dynamic_strides>: New const bool field.
2637 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2638 prefetch_dynamic_strides.
2639 (exynosm1_prefetch_tune): Likewise.
2640 (thunderxt88_prefetch_tune): Likewise.
2641 (thunderx_prefetch_tune): Likewise.
2642 (thunderx2t99_prefetch_tune): Likewise.
2643 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2644 (aarch64_override_options_internal): Update to set
2645 PARAM_PREFETCH_DYNAMIC_STRIDES.
2646 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2647 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2648 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2649 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2650 prefetch-dynamic-strides setting.
2652 2018-05-07 Luis Machado <luis.machado@linaro.org>
2654 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2655 <minimum_stride>: New const int field.
2656 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2657 minimum_stride field.
2658 (exynosm1_prefetch_tune): Likewise.
2659 (thunderxt88_prefetch_tune): Likewise.
2660 (thunderx_prefetch_tune): Likewise.
2661 (thunderx2t99_prefetch_tune): Likewise.
2662 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2663 (aarch64_override_options_internal): Update to set
2664 PARAM_PREFETCH_MINIMUM_STRIDE.
2665 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2666 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2667 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2668 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2669 stride is constant and is below the minimum stride threshold.
2671 2018-05-07 Luis Machado <luis.machado@linaro.org>
2673 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2676 2018-05-07 Luis Machado <luis.machado@linaro.org>
2678 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2679 <prefetch_dynamic_strides>: New const bool field.
2680 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2681 prefetch_dynamic_strides.
2682 (exynosm1_prefetch_tune): Likewise.
2683 (thunderxt88_prefetch_tune): Likewise.
2684 (thunderx_prefetch_tune): Likewise.
2685 (thunderx2t99_prefetch_tune): Likewise.
2686 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2687 (aarch64_override_options_internal): Update to set
2688 PARAM_PREFETCH_DYNAMIC_STRIDES.
2689 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2690 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2691 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2692 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2693 prefetch-dynamic-strides setting.
2695 2018-05-07 Luis Machado <luis.machado@linaro.org>
2697 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2698 <minimum_stride>: New const int field.
2699 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2700 minimum_stride field.
2701 (exynosm1_prefetch_tune): Likewise.
2702 (thunderxt88_prefetch_tune): Likewise.
2703 (thunderx_prefetch_tune): Likewise.
2704 (thunderx2t99_prefetch_tune): Likewise.
2705 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2706 (aarch64_override_options_internal): Update to set
2707 PARAM_PREFETCH_MINIMUM_STRIDE.
2708 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2709 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2710 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2711 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2712 stride is constant and is below the minimum stride threshold.
2714 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2717 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2718 the type is addressable. Don't force op into register if it has
2721 2018-05-05 Roland McGrath <mcgrathr@google.com>
2724 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2725 any section for which we don't know a specific type it should have,
2726 regardless of name. Previously this was done only for the exact
2727 names ".init_array", ".fini_array", and ".preinit_array".
2728 (default_elf_asm_named_section): Add comment about
2729 relationship with default_section_type_flags and SECTION_NOTYPE.
2730 (get_section): Don't consider it a type conflict if one side has
2731 SECTION_NOTYPE and the other doesn't, as long as neither has the
2732 SECTION_BSS et al used in the default_section_type_flags logic.
2734 2018-05-05 Tom de Vries <tom@codesourcery.com>
2737 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2738 (workaround_barsyncs): New function.
2739 (nvptx_reorg): Use workaround_barsyncs.
2740 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2741 (define_expand "nvptx_membar_cta"): New define_expand.
2742 (define_insn "*nvptx_membar_cta"): New insn.
2744 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2746 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2747 To improve optimization opportunities.
2748 * builtin-types.def: The new needed builtin types for the above.
2750 2018-05-04 Richard Biener <rguenther@suse.de>
2752 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2753 * gimple-ssa-store-merging.c
2754 (imm_store_chain_info::output_merged_store): Remove redundant create,
2755 release split_store vector contents on failure.
2756 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2757 scalar stmt vector on cache hit.
2759 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2761 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2763 * config.gcc (powerpc-xilinx-eabi*): Remove.
2764 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2766 (fusion_addis_mem_combo_load): Ditto.
2767 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2769 (rs6000_cpu_cpp_builtins): Ditto.
2770 * config/rs6000/rs6000-linux.c
2771 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2772 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2773 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2775 (rs6000_setup_reg_addr_masks): Ditto.
2776 (rs6000_init_hard_regno_mode_ok): Ditto.
2777 (rs6000_option_override_internal): Ditto.
2778 (legitimate_lo_sum_address_p): Ditto.
2779 (rs6000_legitimize_address): Ditto.
2780 (rs6000_legitimize_reload_address): Ditto.
2781 (rs6000_legitimate_address_p): Ditto.
2782 (abi_v4_pass_in_fpr): Ditto.
2783 (setup_incoming_varargs): Ditto.
2784 (rs6000_gimplify_va_arg): Ditto.
2785 (rs6000_split_multireg_move): Ditto.
2786 (rs6000_savres_strategy): Ditto.
2787 (rs6000_emit_prologue_components): Ditto.
2788 (rs6000_emit_epilogue_components): Ditto.
2789 (rs6000_emit_prologue): Ditto.
2790 (rs6000_emit_epilogue): Ditto.
2791 (rs6000_elf_file_end): Ditto.
2792 (rs6000_function_value): Ditto.
2793 (rs6000_libcall_value): Ditto.
2794 * config/rs6000/rs6000.h: Ditto.
2795 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2796 (TARGET_MINMAX): ... this. New.
2797 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2798 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2799 (*movsi_internal1_single): Delete.
2800 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2801 mfpu=, mxilinx-fpu): Delete.
2802 * config/rs6000/singlefp.h: Delete.
2803 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2804 * config/rs6000/t-rs6000: Ditto.
2805 * config/rs6000/t-xilinx: Delete.
2806 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2807 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2808 (VStype_simple): Delete.
2809 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2810 * config/rs6000/xfpu.h: Delete.
2811 * config/rs6000/xfpu.md: Delete.
2812 * config/rs6000/xilinx.h: Delete.
2813 * config/rs6000/xilinx.opt: Delete.
2814 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2815 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2817 2018-05-04 Tom de Vries <tom@codesourcery.com>
2820 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2823 2018-05-04 Richard Biener <rguenther@suse.de>
2826 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2827 (expand_complex_div_wide): Likewise.
2828 (expand_complex_operations_1): Likewise.
2829 (expand_complex_libcall): Preserve EH info of the original stmt.
2830 (tree_lower_complex): Handle removed blocks.
2831 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2832 on complex multiplication and division libcall builtins.
2834 2018-05-04 Richard Biener <rguenther@suse.de>
2837 * fold-const.c (negate_expr_p): Restrict negation of operand
2838 zero of a division to when we know that can happen without
2840 (fold_negate_expr_1): Likewise.
2842 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2845 * real.h (real_nextafter): Declare.
2846 * real.c (real_nextafter): New function.
2847 * fold-const-call.c (fold_const_nextafter): New function.
2848 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2849 CASE_CFN_NEXTTOWARD.
2850 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2851 even when arg1_mode is different from arg0_mode.
2853 2018-05-03 Nathan Sidwell <nathan@acm.org>
2855 * doc/extend.texi (Deprecated Features): Remove
2857 (Backwards Compatibility): Likewise.
2858 * doc/invoke.texi (C++ Language Options): Likewise.
2859 (C++ Dialect Options): Likewise.
2861 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2864 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2865 _mm512_mask_mullox_epi64): New intrinsics.
2867 2018-05-03 Tom de Vries <tom@codesourcery.com>
2870 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2871 dump files): Add offload-tree.
2873 2018-05-03 Richard Biener <rguenther@suse.de>
2875 PR tree-optimization/85615
2876 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2877 to loops not nested in BBs loop father to avoid creating multi-entry
2880 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2882 PR tree-optimization/70291
2883 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2884 arguments. Change return type to tree. Emit libcall as a new
2885 statement rather than replacing existing one when inplace_p is true.
2886 (expand_complex_multiplication_components): New function.
2887 (expand_complex_multiplication): Expand floating-point complex
2888 multiplication using the above.
2889 (expand_complex_division): Rename inner_type parameter to type.
2890 Update expand_complex_libcall call-site.
2891 (expand_complex_operations_1): Update expand_complex_multiplication
2892 and expand_complex_division call-sites.
2894 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2897 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2898 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2899 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2900 the highest significant bit of the shift count mask is clear. In
2901 check whether and[sq]i3 is needed verify that all significant bits
2902 of the shift count other than the highest are set.
2904 2018-05-02 Tom de Vries <tom@codesourcery.com>
2907 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2908 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2909 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2910 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2911 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2912 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2913 __builtin_goacc_parlevel_size.
2915 2018-05-02 Richard Biener <rguenther@suse.de>
2917 PR tree-optimization/85597
2918 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2919 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
2928 2018-05-02 Tom de Vries <tom@codesourcery.com>
2931 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2932 dump files): Add wpa-ipa.
2934 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2936 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2937 powerpc*-*-linux*paired* target.
2938 * config/rs6000/750cl.h: Delete.
2939 * config/rs6000/paired.h: Delete.
2940 * config/rs6000/paired.md: Delete.
2941 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2943 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2944 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2945 comment. Remove paired float support.
2946 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2947 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2949 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2950 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2952 * config/rs6000/rs6000.c: Remove paired float support.
2953 (paired_expand_vector_init, paired_expand_vector_move,
2954 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2955 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2956 paired_expand_builtin, paired_expand_predicate_builtin,
2957 paired_init_builtins): Delete.
2958 * config/rs6000/rs6000.h: Remove paired float support.
2959 * config/rs6000/rs6000.md: Remove paired float support.
2960 (move_from_CR_ov_bit): Delete.
2961 * config/rs6000/rs6000.opt (mpaired): Delete.
2962 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2963 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2965 2018-05-02 Richard Biener <rguenther@suse.de>
2968 * gimplify.c (gimplify_save_expr): When in SSA form allow
2969 SAVE_EXPRs to compute to SSA vars.
2971 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2974 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2975 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2976 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2977 clobber operands[2], instead use a new pseudo. Formatting fixes.
2979 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2981 PR tree-optimization/85586
2982 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2983 exit early for statements in the same group if the accesses are
2986 2018-05-02 Tom de Vries <tom@codesourcery.com>
2989 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2992 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2994 PR tree-optimization/85143
2995 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2997 2018-05-01 Tom de Vries <tom@codesourcery.com>
3000 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3001 not found" error message.
3003 2018-05-01 Tom de Vries <tom@codesourcery.com>
3006 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3007 * vec.c (test_ordered_remove_if): New function.
3008 (vec_c_tests): Call test_ordered_remove_if.
3009 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3010 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3011 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3012 VEC_ORDERED_REMOVE_IF.
3014 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3016 PR tree-optimization/82665
3017 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3018 pointer subtraction where arguments come from a memchr call.
3020 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3022 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3023 --push-state --as-needed and --pop-state instead of --as-needed and
3024 --no-as-needed if ld supports it.
3025 * configure: Regenerated.
3028 * doc/install.texi2html: Replace _002d with - and _002a with * in
3029 generated html files using sed.
3031 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3034 * gcc-rich-location.c (blank_line_before_p): New function.
3035 (use_new_line): New function.
3036 (gcc_rich_location::add_fixit_insert_formatted): New function.
3037 * gcc-rich-location.h
3038 (gcc_rich_location::add_fixit_insert_formatted): New function.
3040 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3042 * selftest.c (assert_streq): Rename "expected" and "actual" to
3043 "val1" and "val2". Extend NULL-handling to cover both inputs
3044 symmetrically, while still requiring both to be non-NULL for a pass.
3045 * selftest.h (assert_streq): Rename "expected" and "actual" to
3047 (ASSERT_EQ): Likewise.
3048 (ASSERT_EQ_AT): Likewise.
3049 (ASSERT_KNOWN_EQ): Likewise.
3050 (ASSERT_KNOWN_EQ_AT): Likewise.
3051 (ASSERT_NE): Likewise.
3052 (ASSERT_MAYBE_NE): Likewise.
3053 (ASSERT_MAYBE_NE_AT): Likewise.
3054 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3055 the assertion to pass.
3056 (ASSERT_STREQ_AT): Likewise.
3058 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3060 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3061 interaction with -pie.
3063 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3065 * selftest.h: Fix alphabetization of per-source-file selftest
3068 2018-04-30 Jason Merrill <jason@redhat.com>
3070 PR c++/61982 - dead stores to destroyed objects.
3071 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3074 2018-04-30 Jason Merrill <jason@redhat.com>
3076 * tree.c (build_clobber): New.
3077 * tree.h: Declare it.
3078 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3080 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3082 * diagnostic-show-locus.c (layout::layout): Update for
3083 location_get_source_line returning a char_span.
3084 (struct char_span): Move to input.h.
3085 (struct correction): Update for fields in char_span becoming
3087 (struct source_line): Update for location_get_source_line
3088 returning a char_span.
3089 (layout::print_line): Likewise.
3090 * edit-context.c (edited_file::print_content): Likewise.
3091 (edited_file::print_diff_hunk): Likewise.
3092 (edited_file::print_run_of_changed_lines): Likewise.
3093 (edited_file::get_num_lines): Likewise.
3094 (edited_line::edited_line): Likewise.
3095 * final.c (asm_show_source): Likewise.
3096 * input.c (location_get_source_line): Convert return type
3097 from const char * to char_span, losing the final "line_len"
3099 (dump_location_info): Update for the above.
3100 (get_substring_ranges_for_loc): Likewise. Use a char_span
3101 when handling the literal within the line.
3102 (test_reading_source_line): Update for location_get_source_line
3103 returning a char_span.
3104 * input.h (class char_span): Move here from
3105 diagnostic-show-locus.c, converting from a struct to a class.
3106 Make data members private.
3107 (char_span::operator bool): New.
3108 (char_span::length): New.
3109 (char_span::get_buffer): New.
3110 (char_span::operator[]): New.
3111 (char_span::subspan): Make const.
3112 (char_span::xstrdup): New.
3113 (location_get_source_line): Convert return type from const char *
3114 to char_span, losing the final "line_size" param.
3116 2018-04-30 Jan Hubicka <jh@suse.cz>
3118 * lto-wrapper.c (ltrans_priorities): New static var.
3119 (cmp_priority): New.
3120 (run_gcc): Read priorities and if doing parallel build order
3121 the Makefile by them.
3123 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3125 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3127 2018-04-30 Richard Biener <rguenther@suse.de>
3129 * tree-cfg.c (verify_address): Remove base argument, add
3130 flag whether to check TREE_ADDRESSABLE and do that.
3131 (verify_expr): Remove.
3132 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3133 (verify_gimple_assign_single): Likewise.
3134 (verify_gimple_switch): Likewise.
3135 (verify_expr_location_1): Dereference tp once. Add (disabled)
3136 piece from verify_expr.
3137 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3139 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3141 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3143 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3145 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3146 (small_data_pattern): Likewise.
3147 (arc_rewrite_small_data): Likewise.
3148 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3149 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3150 (get_symbol_alignment): New function.
3151 (legitimate_small_data_address_p): Likewise.
3152 (legitimate_scaled_address): Update, call
3153 legitimate_small_data_address_p.
3154 (output_sdata): New static variable.
3155 (arc_print_operand): Update how we handle small data operands.
3156 (arc_print_operand_address): Likewise.
3157 (arc_legitimate_address_p): Update, use
3158 legitimate_small_data_address_p.
3159 (arc_rewrite_small_data_p): Remove.
3160 (arc_rewrite_small_data_1): Likewise.
3161 (arc_rewrite_small_data): Likewise.
3162 (small_data_pattern): Likewise.
3163 (compact_sda_memory_operand): Update to use
3164 legitimate_small_data_address_p and get_symbol_alignment.
3165 (prepare_move_operands): Don't rewite sdata pattern.
3166 (prepare_extend_operands): Remove.
3167 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3169 (zero_extendqisi2): Likewise.
3170 (zero_extendhisi2): Likewise.
3171 (extendqihi2): Likewise.
3172 (extendqisi2): Likewise.
3173 (extendhisi2): Likewise.
3177 * config/arc/constraints.md (Usd): Change it to memory constraint.
3179 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3181 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3182 as source of std instructions.
3183 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3184 allow 6-bit constants as source for store instructions.
3185 (movdi_insn): Update instruction pattern to allow 6-bit constants
3186 as source for store instructions.
3188 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3190 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3192 2018-04-30 Nathan Sidwell <nathan@acm.org>
3193 Sandra Loosemore <sandra@codesourcery.com>
3195 * dumpfile.c (dump_open): Allow '-' for stdout.
3196 * doc/invoke.texi (Developer Options): Document dump filename
3197 determination early. Document stdin/stdout selection.
3199 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3201 Microblaze Target: PIC data text relative
3203 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3204 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3206 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3207 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3208 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3209 New addressing mode for data-text relative position indepenedent code.
3210 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3211 'ADDRESS_SYMBOLIC_TXT_REL'.
3212 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3213 (microblaze_legitimate_pic_operand): Exclude function calls from
3214 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3215 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3217 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3218 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3219 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3220 for 'address + offset'.
3221 (microblaze_expand_prologue): Add new function prologue call for
3223 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3224 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3225 table in case of TARGET_PIC_DATA_TEXT_REL.
3226 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3227 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3228 Add new macros 'UNSPEC_TEXT',
3229 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3230 + exclude function calls from 'UNSPEC_PLT' in case of data text
3232 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3233 new target hook for generating address diff vector tables in case of
3235 * doc/tm.texi : Regenerate.
3236 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3237 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3238 of addr diff vector generation.
3239 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3240 target hook definition.
3241 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3242 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3243 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3244 option for data text relative.
3246 2018-04-30 Richard Biener <rguenther@suse.de>
3248 * tree-chrec.h (evolution_function_is_constant_p): Remove
3250 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3252 2018-04-30 Richard Biener <rguenther@suse.de>
3255 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3257 2018-04-30 Richard Biener <rguenther@suse.de>
3259 PR tree-optimization/28364
3260 PR tree-optimization/85275
3261 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3262 copying first exit test.
3264 2018-04-28 Mark Wielaard <mark@klomp.org>
3266 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3268 (dwarf_AT): Handle DW_AT_addr_base.
3269 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3271 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3274 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3275 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3276 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3277 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3279 2018-04-28 Richard Biener <rguenther@suse.de>
3281 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3282 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3283 to reflect use. Only add interesting stmts.
3285 2018-04-27 Martin Jambor <mjambor@suse.cz>
3288 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3289 the jump function allows for passing through aggregate values.
3291 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3293 * input.h (in_system_header_at): Convert from macro to inline
3295 (from_macro_expansion_at): Likewise.
3296 (from_macro_definition_at): Likewise.
3298 2018-04-27 Jeff Law <law@redhat.com>
3300 * config.gcc: Mark tile* targets as deprecated/obsolete.
3302 2018-04-27 Richard Biener <rguenther@suse.de>
3304 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3307 2018-04-27 Richard Biener <rguenther@suse.de>
3309 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3311 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3313 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3314 with Yd constraint. Set "preferred_for_speed" attribute from
3315 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3317 (*movdi_internal): Ditto.
3318 (movti_interunit splitters): Remove
3319 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3320 (movdi_interunit splitters): Ditto.
3321 * config/i386/constraints.md (Ye): Remove.
3322 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3324 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3327 * config/aarch64/constraints.md (Usg): Limit to 31.
3330 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3332 PR tree-optimization/85529
3333 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3334 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3335 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3336 zero extension or masking of the MSB bit.
3337 (optimize_range_tests): Add FIRST_BB argument, pass it through
3338 to optimize_range_tests_var_bound.
3339 (maybe_optimize_range_tests, reassociate_bb): Adjust
3340 optimize_range_tests callers.
3342 2018-04-26 Richard Biener <rguenther@suse.de>
3343 Jakub Jelinek <jakub@redhat.com>
3345 * cgraph.h (symbol_table): Just declare debug method here.
3346 * symtab.c (symbol_table::debug): Define.
3348 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3350 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3352 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3354 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3355 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3356 (*movdi_internal): Substitute Yi and Yj constraint with x
3357 and Ym and Yn constraint with y constraint. Update "isa"
3358 attribute and set "preferred_for_speed" attribute from
3359 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3360 (*movsi_internal): Ditto.
3361 (*movdf_internal): Ditto.
3362 (*movsf_internal): Ditto.
3363 (*zero_extendsidi2): Ditto.
3364 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3365 (sse2_loadld): Ditto.
3366 (*vec_extract<ssevecmodelower>_0): Ditto.
3367 (*vec_extractv4si_0_zext_sse4): Ditto.
3368 (vec_concatv2di): Ditto.
3369 (*vec_dup<mode>): Ditto.
3370 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3371 * config/i386/constraints.md (Yi): Remove.
3376 2018-04-26 Nathan Sidwell <nathan@acm.org>
3378 * dumpfile.c (dump_open): New.
3379 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3380 (dump_finish): Detect stdio/stderr by value not name.
3382 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3384 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3386 2018-04-26 Tom de Vries <tom@codesourcery.com>
3389 * config/nvptx/nvptx.c (verify_neutering_jumps)
3390 (verify_neutering_labels): New function
3391 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3393 2018-04-26 Tom de Vries <tom@codesourcery.com>
3396 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3397 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3400 2018-04-26 Richard Biener <rguenther@suse.de>
3401 Tom de Vries <tom@codesourcery.com>
3404 * lto-streamer-out.c (output_function): Fixup loops if required to match
3405 discovery done in the reader.
3407 2018-04-26 Richard Biener <rguenther@suse.de>
3409 PR tree-optimization/85116
3410 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3411 have a loop exit from the single latch predecessor. Remove
3412 case of header with just condition.
3413 (ch_base::copy_headers): Exclude infinite loops from any
3415 (pass_ch::execute): Record exits.
3417 2018-04-26 Richard Biener <rguenther@suse.de>
3419 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3420 prologue cost vector and pass it to vect_get_load_cost.
3421 (vect_get_peeling_costs_all_drs): Likewise.
3422 (vect_peeling_hash_get_lowest_cost): Likewise.
3423 (vect_enhance_data_refs_alignment): Likewise.
3425 2018-04-26 Richard Biener <rguenther@suse.de>
3428 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3429 checking of integer<->pointer conversions.
3430 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3431 sign-/zero-extending pointer types.
3432 (expand_omp_for_static_chunk): Likewise.
3434 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3435 Jean Lee <xiaoyur347@gmail.com>
3437 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3438 (TARGET_ASAN_SHADOW_OFFSET): Define.
3439 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3440 true for -fsanitize=address.
3442 2018-04-25 Mark Wielaard <mark@klomp.org>
3444 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3447 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3449 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3450 than "alu", remove explicit "memory" and "imm_disp" attributes.
3451 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3454 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3455 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3456 gen_lowpart_no_emit.
3458 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3461 * config/i386/i386.c (ix86_expand_builtin): Change memory
3462 operand to XI, extend p0 to Pmode.
3463 * config/i386/i386.md: Change unspec volatile and operand
3464 1 mode to XI, change operand 0 mode to P.
3466 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3468 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3469 GET_MODE_MASK before any checking.
3470 (nds32_can_use_bset_p): Likewise.
3471 (nds32_can_use_btgl_p): Likewise.
3473 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3475 * config/nds32/nds32-doubleword.md: New define_split pattern for
3476 illegal register number.
3478 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3480 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3482 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3484 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3486 2018-04-25 Richard Biener <rguenther@suse.de>
3488 * lto-streamer.h (LTO_major_version): Bump to 8.
3490 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3492 * BASE-VER: Set to 9.0.0.
3494 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3496 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3497 in __abskf2 and __powikf2.
3499 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3502 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3503 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3504 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3505 Use the above on operand 2. Reindent.
3506 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3508 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3511 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3513 * config/i386/i386.opt (mcet): Removed.
3514 * doc/install.texi: Remove -mcet documentation.
3515 * doc/invoke.texi: Likewise.
3517 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3520 * doc/install.texi: Remove -mcet from bootstrap-cet.
3522 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3525 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3526 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3530 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3531 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3532 containing a CONST_VECTOR.
3534 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3536 * doc/install.texi: Update newlib dependency for nvptx.
3538 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3541 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3542 instead of INTVAL when shifting x left.
3544 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3546 PR tree-optimization/85478
3547 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3548 vect_grouped_store_supported for single element vectors.
3550 2018-04-24 Richard Biener <rguenther@suse.de>
3553 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3554 load cost increase to the case of non-constant step.
3556 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3559 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3560 destination if any_malformed_asm.
3562 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3565 * expr.c (store_field): In the bitfield case, if the value comes from
3566 a function call and is returned in registers by means of a PARALLEL,
3567 do not change the mode of the temporary unless BLKmode and VOIDmode.
3569 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3571 PR rtl-optimization/85423
3572 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3573 dependencies to debug insns when the previous insn is non-debug.
3575 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3577 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3578 enums into a single definition.
3579 (fls): Fix predicates and printing.
3582 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3584 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3585 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3586 and short u6 immediate.
3587 (check_if_valid_sleep_operand): Remove.
3588 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3590 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3592 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3593 flag_always_save_lp condition.
3594 * config/nds32/nds32.opt (malways-save-lp): New option.
3596 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3598 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3599 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3600 * config/nds32/nds32.h
3601 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3602 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3604 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3606 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3607 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3609 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3610 Chung-Ju Wu <jasonwucj@gmail.com>
3612 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3614 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3615 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3617 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3619 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3621 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3623 * config/nds32/nds32-protos.h (nds32_data_alignment,
3624 nds32_local_alignment): Declare.
3625 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3626 nds32_local_alignment): New functions.
3627 (TARGET_CONSTANT_ALIGNMENT): Define.
3628 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3630 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3632 * config/nds32/nds32.c
3633 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3634 (TARGET_MODES_TIEABLE_P): Likewise.
3636 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3638 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3641 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3642 Chung-Ju Wu <jasonwucj@gmail.com>
3644 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3645 for unaligned access.
3646 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3648 * config/nds32/nds32-intrinsic.md: Likewise.
3649 * config/nds32/nds32_intrinsic.h: Likewise.
3650 * config/nds32/nds32.h (nds32_builtins): Likewise.
3651 * config/nds32/nds32.opt (munaligned-access): New option.
3652 * config/nds32/nds32.c (nds32_asm_file_start): Display
3653 flag_unaligned_access status.
3655 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3657 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3658 -mno-relax is present.
3659 * config/riscv/linux.h (LINK_SPEC): Ditto.
3661 2018-04-20 Martin Sebor <msebor@redhat.com>
3664 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3666 (gimple_fold_builtin_stxcpy_chk): Same.
3667 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3669 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3672 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3673 __powikf2 when long double is IEEE 128-bit.
3675 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3677 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3678 step to make sure stack always aligned.
3680 2018-04-20 Carl Love <cel@us.ibm.com>
3683 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3684 size check for arg0.
3686 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3687 Tom de Vries <tom@codesourcery.com>
3690 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3691 Emit insns for calls too.
3692 (nvptx_find_par): Always look for worker-level predecessor insn.
3693 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3695 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3696 (nvptx_process_pars): Propagate frames for calls.
3698 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3701 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3703 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3704 (ix86_handle_option): Don't handle OPT_mibt.
3705 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3707 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3709 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3710 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3711 (ix86_target_macros): Define __CET__ with flag_cf_protection
3712 for -fcf-protection.
3713 * config/i386/i386.c (isa2_opts): Remove -mibt.
3714 * config/i386/i386.h (TARGET_IBT): Removed.
3715 (TARGET_IBT_P): Likewise.
3716 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3717 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3718 * config/i386/i386.opt (mcet): Update help message.
3721 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3722 -mcet as an alias for -mshstk.
3724 2018-04-20 Richard Biener <rguenther@suse.de>
3727 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3728 complexity by forcing a single use of the multiply operand.
3730 2018-04-20 Martin Jambor <mjambor@suse.cz>
3733 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3734 recursion dependency to only apply to non-clones.
3736 2018-04-20 Martin Jambor <mjambor@suse.cz>
3739 * ipa-cp.c (create_specialized_node): Check that clones of
3740 self-recursive edges exist during IPA-CP.
3742 2018-04-19 Toon Moene <toon@moene.org>
3744 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3747 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3749 PR tree-optimization/85467
3750 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3751 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3752 VECTOR_CST element to type.
3754 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3757 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3758 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3759 (builtin_longjmp): Likewise.
3760 (save_stack_nonlocal): New pattern.
3761 (restore_stack_nonlocal): Likewise.
3763 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3766 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3767 Replace ASM_OUTPUT_LABEL with fprintf.
3769 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3772 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3773 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3774 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3775 define __IBT__ and __SHSTK__ for -fcf-protection.
3776 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3778 (ix86_trampoline_init): Likewise.
3779 (x86_output_mi_thunk): Likewise.
3780 (ix86_notrack_prefixed_insn_p): Likewise.
3781 (ix86_option_override_internal): Don't disallow -fcf-protection.
3782 * config/i386/i386.md (rdssp<mode>): Also enable for
3784 (incssp<mode>): Likewise.
3785 (nop_endbr): Likewise.
3786 * config/i386/i386.opt (mcet): Change help message to built-in
3790 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3791 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3792 enable CET built-in functions.
3794 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3796 * common/config/i386/i386-common.c
3797 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3798 OPTION_MASK_ISA_MOVDIRI_UNSET,
3799 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3800 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3801 * config.gcc (movdirintrin.h): New header.
3802 * config/i386/cpuid.h (bit_MOVDIRI,
3803 bit_MOVDIR64B): New bits.
3804 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3806 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3807 (VOID, PVOID, PCVOID)): New function types.
3808 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3809 __builtin_ia32_directstoreu_u64,
3810 __builtin_ia32_movdir64b): New builtins.
3811 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3812 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3814 (ix86_valid_target_attribute_inner_p): Ditto.
3815 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3816 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3817 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3818 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3819 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3820 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3821 (movdiri<mode>, movdir64b_<mode>): New.
3822 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3823 * config/i386/immintrin.h: Include movdirintrin.h.
3824 * config/i386/movdirintrin.h: New file.
3825 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3827 2018-04-19 Richard Biener <rguenther@suse.de>
3830 * cfg.c (clear_bb_flags): When loop state says we have
3831 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3833 2018-04-19 Richard Biener <rguenther@suse.de>
3835 PR tree-optimization/84737
3836 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3837 copying restrict info.
3838 (vect_setup_realignment): Use it.
3839 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3840 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3841 the first DR to all generated stores.
3842 (vectorizable_load): Likewise for loads.
3844 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3846 PR tree-optimization/85446
3847 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3848 the integral and pointer types to have the same precision.
3850 * doc/install.texi: Document --disable-cet being the default and
3853 2018-04-18 Martin Liska <mliska@suse.cz>
3855 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3858 2018-04-18 Martin Liska <mliska@suse.cz>
3861 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3864 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3865 arguments if they are comparable.
3867 2018-04-18 Martin Liska <mliska@suse.cz>
3870 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3873 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3876 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3879 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3880 ENDBR after calling __morestack.
3882 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3885 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3886 by using gcc_base_ver to generate a gcc_driver_version, and use
3887 it when generating GCC_DRIVER_NAME.
3888 * configure: Regenerate.
3890 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3893 * config.gcc: Obsolete powerpc*-*-*spe*.
3895 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3898 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3899 (stabstr_D): Change type of unum from unsigned int to
3900 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3903 2018-04-17 Jim Wilson <jimw@sifive.com>
3906 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3907 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3908 Set arg_pointer_offset after using pretend_args_size.
3910 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3912 PR rtl-optimization/85431
3913 * dse.c (record_store): Ignore zero width stores.
3916 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3917 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3918 __builtin_stack_restore rather than after it.
3919 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3920 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3921 argument instead of virtual_dynamic_stack_rtx.
3923 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3925 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3927 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3928 Add note to error message to explain internal mapping of overloaded
3929 built-in function name to non-overloaded built-in function name.
3930 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3933 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3936 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3937 where the inputs overlap with the output.
3939 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3941 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3942 (=v, v) alternative and explicit "memory" attribute.
3943 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3944 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3946 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3947 "sselog1" type instead of "sselog".
3948 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3949 "sselog". Remove explicit "memory" attribute.
3950 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3951 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3953 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3954 "sselog1" type instead of "sselog". Remove explicit "memory"
3956 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3957 use "sselog1" type instead of "sselog". Remove explicit "memory"
3959 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3960 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3962 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3963 "sselog1" type instead of "sselog". Remove explicit "memory"
3965 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3966 use "sselog1" type instead of "sselog". Remove explicit "memory"
3970 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3973 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3976 2018-04-17 Martin Jambor <mjambor@suse.cz>
3979 * ipa-cp.c (create_specialized_node): Call
3980 expand_all_artificial_thunks if necessary.
3982 2018-04-17 Martin Liska <mliska@suse.cz>
3985 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3986 in message, remote space in between '_G' and '('.
3988 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3991 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3992 avx512f_vmcmp<mode>3<round_saeonly_name>,
3993 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3994 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3995 avx512f_rndscale<mode><round_saeonly_name>,
3996 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3997 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3998 Use %<iptr>2 instead of %2 for -masm=intel.
3999 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4000 avx512f_vcvttss2usi<round_saeonly_name>,
4001 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4003 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4004 avx512f_vcvttsd2usi<round_saeonly_name>,
4005 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4006 Use %q1 instead of %1 for -masm=intel.
4007 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4008 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4009 of %3 for -masm=intel.
4010 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4012 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4014 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4016 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4017 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4019 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4021 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4022 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4023 %g1 and one with %0 and %1.
4024 (avx512er_vmrcp28<mode><round_saeonly_name>,
4025 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4027 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4028 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4029 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4030 of %0 and %{%4%} for -masm=intel.
4031 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4032 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4033 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4034 order of %0 and %{%5%}%{z%} for -masm=intel.
4036 2018-04-17 Jan Hubicka <jh@suse.cz>
4039 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4041 2018-04-17 Martin Liska <mliska@suse.cz>
4044 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4045 for target_clone error message. Make default implementation
4046 clone to be a local declaration.
4047 (separate_attrs): Add new argument and check for an empty
4049 (expand_target_clones): Handle it.
4050 (ipa_target_clone): Make redirection just for target_clones
4053 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4054 Tom de Vries <tom@codesourcery.com>
4057 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4058 tiled basic blocks without omp continue statements.
4060 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4063 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4064 vec_extract expression as having side effects to make sure it gets
4067 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4070 * config/i386/i386.c (get_builtin_code_for_version): Check
4073 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4076 * gcc/config.gcc: Support "skylake".
4077 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4079 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4080 (processor_target_table): Add "skylake".
4081 (ix86_option_override_internal): Add "skylake".
4082 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4083 PROCESSOR_CANNONLAKE.
4084 (get_builtin_code_for_version): Fix priority for
4085 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4086 PROCESSOR_SKYLAKE-AVX512.
4087 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4088 (processor_type): Add PROCESSOR_SKYLAKE.
4090 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4091 Jason Merrill <jason@redhat.com>
4094 * convert.c (convert_to_integer_1): Use direct recursion for
4095 enumeral types and types with a precision less than the number
4096 of bits in their mode.
4098 2018-04-16 Julia Koval <julia.koval@intel.com>
4101 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4102 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4104 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4107 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4108 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4109 and -mno-direct-move.
4111 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4114 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4115 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4116 Ensure negative shifts result in {0}.
4118 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4120 PR rtl-optimization/79916
4121 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4122 regs (if any) to define how to gnerate SD moves when LRA is in
4125 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4127 PR rtl-optimization/85393
4128 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4129 * except.c (expand_dw2_landing_pad_for_region): Make static.
4130 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4131 a label and unconditional jump to old_bb, rather than
4132 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4135 PR rtl-optimization/85376
4136 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4137 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4138 instead of a specific value.
4140 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4141 Bin Cheng <bin.cheng@arm.com>
4143 PR tree-optimization/82965
4144 PR tree-optimization/83991
4145 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4146 by_profile_only parameter.
4147 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4148 information if the loop was predicted to iterate too many times.
4149 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4151 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4154 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4157 2018-04-13 Martin Liska <mliska@suse.cz>
4158 Jakub Jelinek <jakub@redhat.com>
4161 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4162 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4163 * builtins.c (expand_builtin_memory_copy_args): Use
4164 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4165 handle dest_addr == pc_rtx.
4167 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4170 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4171 asked to not generate direct moves.
4172 (fix_trunc<mode>si2_stfiwx): Similar.
4173 (fix_trunc<mode>si2_internal): Similar.
4175 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4178 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4179 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4180 lookup if dest in some wider mode is known to be const0_rtx and
4181 if so, record permanent equivalence for it to be ZERO_EXTEND of
4182 the narrower mode destination.
4184 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4186 * lto-streamer-out.c (output_function): Revert 259346.
4187 * omp-expand.c (expand_oacc_for): Likewise.
4189 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4191 PR rtl-optimization/85354
4192 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4193 * sel-sched.c (sel_global_init): ... here.
4195 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4198 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4199 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4200 mode for PE-COFF targets.
4201 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4202 (i386_pe_asm_lto_end): Likewise.
4203 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4204 (TARGET_ASM_LTO_END): Likewise.
4205 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4206 (i386_pe_asm_lto_start): New function.
4207 (i386_pe_asm_lto_end): Likewise.
4209 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4210 Richard Biener <rguenther@suse.de>
4213 * lto-streamer-out.c (output_function): Fix CFG loop state before
4215 * omp-expand.c (expand_oacc_for): Handle calls to internal
4216 functions like regular functions.
4218 2018-04-12 Richard Biener <rguenther@suse.de>
4221 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4222 for the early LTO debug to properly generate references to it
4223 during DIE emission. Do not re-use that for the skeleton for
4225 (dwarf2out_early_finish): Likewise.
4227 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4230 * config/i386/sse.md
4231 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4232 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4233 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4234 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4235 and output is a reg, avoid creating invalid lowpart subreg, but
4236 instead split into a 512-bit move. Don't split if not AVX512VL,
4237 input is xmm16+ reg and output is a mem.
4238 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4239 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4240 xmm16+ reg and output is a mem.
4242 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4244 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4245 also for flag_dwarf2_cfi_asm.
4247 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4249 PR rtl-optimization/85342
4250 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4251 a bool scalar var inside of the loop instead. Don't try to update
4252 recog_data.operand after failed apply_change_group.
4254 2018-04-12 Tom de Vries <tom@codesourcery.com>
4257 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4258 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4259 array with flexible array member as array without given dimension.
4260 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4261 argument for undefined param to true.
4263 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4266 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4267 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4268 from PowerPC section.
4269 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4270 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4271 help text that is too long.
4272 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4273 help text that is too long.
4274 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4275 help text that is too long.
4277 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4279 * config/alpha/alpha.md (stack_probe_internal): Rename
4280 from "probe_stack". Update all callers.
4282 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4284 PR rtl-optimization/84566
4285 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4286 sched_macro_fuse_insns.
4288 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4291 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4292 (compute_block_dependences): ... from here.
4294 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4296 PR tree-optimization/85331
4297 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4298 from int to HOST_WIDE_INT.
4300 2018-04-11 Martin Jambor <mjambor@suse.cz>
4303 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4304 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4305 not the same as the source val.
4306 (cgraph_edge_brings_value_p): New parameter.
4307 (gather_edges_for_value): Pass destination value to
4308 cgraph_edge_brings_value_p.
4309 (perhaps_add_new_callers): Likewise.
4310 (get_info_about_necessary_edges): Likewise and exclude values brought
4311 only by self-recursive edges.
4312 (create_specialized_node): Redirect only clones of self-calling edges.
4313 (+self_recursive_pass_through_p): New function.
4314 (find_more_scalar_values_for_callers_subset): Use it.
4315 (find_aggregate_values_for_callers_subset): Likewise.
4316 (known_aggs_to_agg_replacement_list): Removed.
4317 (decide_whether_version_node): Re-calculate known constants for all
4318 remaining context clones.
4320 2018-04-11 Richard Biener <rguenther@suse.de>
4323 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4324 from early DWARF output.
4325 (dwarf2out_early_finish): Output line info unconditionally into
4326 early DWARF and add reference to it.
4328 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4331 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4332 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4333 other than V2DFmode using iptr mode attribute.
4334 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4336 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4338 PR rtl-optimization/84659
4339 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4341 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4344 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4346 (output_loc_list): Pass address of a dummy size variable even in the
4347 locview handling loop.
4348 (index_location_lists): Add comment on why skip_loc_list_entry can't
4351 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4354 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4357 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4360 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4361 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4362 and -mstring-compare-inline-limit.
4364 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4367 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4368 for stack clash protection in a register whenever we need it to be in
4371 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4373 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4374 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4376 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4379 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4381 (mlong-double-): Ditto.
4382 * config/rs6000/sysv4.opt (msdata=): Ditto.
4383 (mtls-size=): Ditto.
4385 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4387 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4388 erroneous entries for
4389 "vector int vec_ldl (int, long int *)", and
4390 "vector unsigned int vec_ldl (int, unsigned long int *)".
4391 Add comments and entries for
4392 "vector bool char vec_ldl (int, bool char *)",
4393 "vector bool short vec_ldl (int, bool short *)",
4394 "vector bool int vec_ldl (int, bool int *)",
4395 "vector bool long long vec_ldl (int, bool long long *)",
4396 "vector pixel vec_ldl (int, pixel *)",
4397 "vector long long vec_ldl (int, long long *)",
4398 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4399 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4400 type tree bool_long_long_type_node and correct definition of
4401 bool_V2DI_type_node to make reference to this new type tree.
4402 (rs6000_mangle_type): Replace erroneous reference to
4403 bool_long_type_node with bool_long_long_type_node.
4404 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4405 comments to emphasize sign distinctions for char and int types and
4406 replace RS6000_BTI_bool_long constant with
4407 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4408 use of RS6000_BTI_pixel.
4409 (bool_long_type_node): Remove this macro definition.
4410 (bool_long_long_type_node): New macro definition
4412 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4414 PR rtl-optimization/85300
4415 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4416 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4417 simplify_unary_operation fails.
4419 2018-04-10 Martin Liska <mliska@suse.cz>
4421 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4422 cgraph_edge and ipa_ref.
4424 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4428 * config/i386/sse.md
4429 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4430 computation of the VEC_MERGE selector from mask.
4431 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4432 Fix decoding of the VEC_MERGE selector into mask.
4434 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4436 PR tree-optimization/85286
4437 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4439 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4441 * final.c (final_1): Set insn_last_address as well as
4442 insn_current_address.
4444 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4447 * explow.c (emit_stack_probe): Call validize_mem on memory location
4448 before passing it to gen_probe_stack. Create address operand and
4449 legitimize it for the probe_stack_address case.
4451 2018-04-09 Jan Hubicka <jh@suse.cz>
4454 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4455 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4456 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4458 (free_lang_data): Rebuild type inheritance graph.
4460 2018-04-09 Martin Sebor <msebor@redhat.com>
4462 * invoke.texi (-finline-small-functions): Mention other optimization
4464 (-findirect-inlining, -fpartial-inlining): Same.
4465 (-finline-functions-called-once): Same.
4466 (-freorder-blocks-and-partition): Same.
4468 2018-04-09 Jan Hubicka <jh@suse.cz>
4471 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4472 jumps; choose last target that matches the criteria (i.e.
4473 no partition changes for non-crossing jumps).
4474 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4475 support for redirecting crossing jumps to non-crossing.
4477 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4479 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4480 also for naked functions.
4482 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4484 * config/arc/arc.md (add_shift): New pattern.
4485 (add_shift2): Likewise.
4486 (sub_shift): Likewise.
4487 (sub_shift_cmp0_noout): Likewise.
4488 (compare_si_ashiftsi): Likewise.
4489 (xbfu_cmp0_noout): New combine pattern.
4490 (xbfu_cmp0"): Likewise.
4491 (movsi_set_cc_insn): Place the predicable variant first.
4492 (commutative_binary_cmp0_noout): Remove clobber.
4493 (commutative_binary_cmp0): New pattern.
4494 (noncommutative_binary_cmp0): Likewise.
4495 (noncommutative_binary_cmp0_noout): Likewise.
4496 (noncommutative_binary_comparison_result_used): Removed.
4497 (rsub_cmp0): New pattern.
4498 (rsub_cmp0_noout): Likewise.
4499 (extzvsi): Changed, keep only meaningful variants.
4500 (SQH, SEZ): New iterators.
4501 (SQH_postfix): New mode attribute.
4502 (SEZ_prefix): New code attribute.
4503 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4504 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4505 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4507 (noncommutative_operator): Check the availability of barrel
4510 2018-04-09 Richard Biener <rguenther@suse.de>
4512 PR tree-optimization/85284
4513 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4514 Only use the niter constraining form of simple_iv when the exit
4517 2018-04-09 Tom de Vries <tom@codesourcery.com>
4520 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4521 (define_expand "*memory_barrier"): New define_expand.
4522 (define_insn "memory_barrier"): New insn.
4524 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4526 PR rtl-optimization/80463
4527 PR rtl-optimization/83972
4528 PR rtl-optimization/83480
4530 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4531 correct producer for the insn.
4532 (tidy_control_flow): Fixup seqnos in case of debug insns.
4534 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4536 PR rtl-optimization/83913
4538 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4539 different sched-times when merging exprs.
4541 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4543 PR rtl-optimization/83962
4545 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4546 tidy_fallthru_edge and tidy_control_flow.
4548 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4550 PR rtl-optimization/83530
4552 * sel-sched.c (force_next_insn): New global variable.
4553 (remove_insn_for_debug): When force_next_insn is true, also leave only
4554 next insn in the ready list.
4555 (sel_sched_region): When the region wasn't scheduled, make another pass
4556 over it with force_next_insn set to 1.
4558 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4560 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4562 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4563 for interrupt control.
4564 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4565 functions for interrupt control.
4566 * config/nds32/nds32-intrinsic.md: Likewise.
4567 * config/nds32/nds32_intrinsic.h: Likewise.
4568 * config/nds32/nds32.h (nds32_builtins): Likewise.
4570 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4572 * config/nds32/nds32.c (nds32_init_machine_status,
4573 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4574 strict_aligned_p field.
4575 (nds32_expand_to_rtl_hook): New function.
4576 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4577 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4579 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4580 Chung-Ju Wu <jasonwucj@gmail.com>
4582 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4583 * config/nds32/nds32-n7.md: New file.
4584 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4585 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4587 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4588 * config/nds32/nds32.md (pipeline_model): Add n7.
4589 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4590 * config/nds32/pipelines.md: Include n7 settings.
4592 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4593 Chung-Ju Wu <jasonwucj@gmail.com>
4595 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4596 * config/nds32/nds32-e8.md: New file.
4597 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4600 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4601 * config/nds32/nds32.md (pipeline_model): Add e8.
4602 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4603 * config/nds32/pipelines.md: Include e8 settings.
4605 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4606 Chung-Ju Wu <jasonwucj@gmail.com>
4608 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4609 * config/nds32/nds32-n8.md: New file.
4610 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4611 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4613 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4614 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4615 * config/nds32/nds32.md (pipeline_model): Add n8.
4616 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4617 * config/nds32/pipelines.md: Include n8 settings.
4619 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4620 Chung-Ju Wu <jasonwucj@gmail.com>
4622 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4623 * config/nds32/nds32-n9-2r1w.md: New file.
4624 * config/nds32/nds32-n9-3r2w.md: New file.
4625 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4626 nds32_register_ports): New or modify for cpu n9.
4627 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4629 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4630 * config/nds32/nds32-utils.c: New file.
4631 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4632 TARGET_MUL_SLOW): Define.
4633 * config/nds32/nds32.md (pipeline_model): New attribute.
4634 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4635 New options that support cpu n9.
4636 * config/nds32/pipelines.md: Include n9 settings.
4637 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4639 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4641 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4642 information if necessary.
4643 (output_cond_branch_compare_zero): Likewise.
4644 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4645 (nds32_target_alignment): Refine for alignment.
4646 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4647 (FUNCTION_BOUNDARY): Modify.
4648 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4650 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4652 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4654 * config/nds32/constants.md (unspec_volatile_element): Add values for
4655 TLB operation and data prefetch.
4656 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4657 functions for TLB operation and data prefetch.
4658 * config/nds32/nds32-intrinsic.md: Likewise.
4659 * config/nds32/nds32_intrinsic.h: Likewise.
4660 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4661 (nds32_print_operand): Likewise.
4662 * config/nds32/nds32.h (nds32_builtins): Likewise.
4664 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4665 Andrew Pinski <pinsika@gcc.gnu.org>
4668 * match.pd: Use constant_boolean_node of correct type instead of
4669 boolean_true_node or boolean_false_node for simplifying
4670 pointer comparisons to zero.
4672 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4674 PR tree-optimization/80021
4675 * tree.c (verify_type_variant): Make error call in verify_variant_match
4676 translatable and remove final full stop.
4678 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4680 * config/nds32/constants.md (unspec_volatile_element): Add
4681 UNSPEC_VOLATILE_EH_RETURN.
4682 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4683 nds32_output_stack_pop): Support dwarf exception handling process.
4684 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4685 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4686 exception handling process.
4687 (nds32_compute_stack_frame): Likewise.
4688 (nds32_return_addr_rtx): Likewise.
4689 (nds32_initial_elimination_offset): Likewise.
4690 (nds32_expand_prologue): Likewise.
4691 (nds32_expand_epilogue): Likewise.
4692 (nds32_dynamic_chain_address): New function.
4693 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4695 (DYNAMIC_CHAIN_ADDRESS): Define.
4696 (EH_RETURN_DATA_REGNO): Define.
4697 (EH_RETURN_STACKADJ_RTX): Define.
4698 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4699 patterns for dwarf exception handling.
4701 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4703 * config/nds32/nds32.h: Clean up obsolete macros.
4705 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4707 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4708 Add enum values for particular instructions.
4709 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4710 particular intrinsic functions.
4711 * config/nds32/nds32-intrinsic.md: Likewise.
4712 * config/nds32/nds32_intrinsic.h: Likewise.
4713 * config/nds32/nds32.h (nds32_builtins): Likewise.
4714 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4715 (btst, ave): New patterns for particular instructions.
4717 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4719 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4720 Add enum values for atomic load/store and memory sync.
4721 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4723 * config/nds32/nds32-intrinsic.md: Likewise.
4724 * config/nds32/nds32_intrinsic.h: Likewise.
4725 * config/nds32/nds32.h (nds32_builtins): Likewise.
4727 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4729 PR tree-optimization/85257
4730 * fold-const.c (native_encode_vector): If not all elts could fit
4731 and off is -1, return 0 rather than offset.
4732 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4733 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4734 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4735 adjust buffer in native_interpret_expr call.
4737 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4739 * config/nds32/constants.md (unspec_volatile_element): Add cache
4740 control enum values.
4741 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4742 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4743 * config/nds32/nds32.c (nds32_cctl_names): New.
4744 (nds32_print_operand): Handle cache control register names.
4745 * config/nds32/nds32.h (nds32_builtins): New enum values.
4746 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4748 * config/nds32/nds32.md (type): Add mmu.
4749 * config/nds32/pipelines.md (simple_insn): Add mmu.
4751 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4753 * config/nds32/nds32.md (type): Remove call.
4754 * config/nds32/pipelines.md (simple_insn): Likewise.
4756 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4758 * config/nds32/constants.md (unspec_volatile_element): Add
4759 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4760 UNSPEC_VOLATILE_FMFCFG.
4761 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4762 description for fmfcfg and fmfcsr.
4763 (bdesc_1arg): Add fmtcsr.
4764 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4765 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4766 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4767 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4768 unspec_fmfcfg): New patterns.
4769 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4770 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4771 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4772 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4773 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4774 __nds32__fmfcfg): Define.
4776 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4778 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4779 intrinsic register names.
4780 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4781 intrinsic register enum values and macros.
4783 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4785 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4786 for load/store addressing form.
4787 (nds32_print_operand_address): Likewise.
4789 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4792 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4793 based on LABEL_REF. Remove useless assertion.
4794 (pic_address_needs_scratch): Fix formatting.
4795 (sparc_legitimize_pic_address): Minor tweaks.
4796 (sparc_delegitimize_address): Adjust assertion accordingly.
4797 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4798 into symbolic_operand.
4799 (movsi_high_pic_label_ref): Likewise.
4800 (movsi_lo_sum_pic_label_ref): Likewise.
4801 (movdi_pic_label_ref): Likewise.
4802 (movdi_high_pic_label_ref): Likewise.
4803 (movdi_lo_sum_pic_label_ref): Likewise.
4805 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4807 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4808 custom LIB_SPEC setup.
4810 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4811 Kito Cheng <kito.cheng@gmail.com>
4813 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4814 * config/riscv/freebsd.h: New.
4816 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4818 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4819 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4822 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4823 Kito Cheng <kito.cheng@gmail.com>
4825 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4826 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4827 * config/nds32/nds32-protos.h (nds32_output_call,
4828 nds32_output_return): Declare.
4829 * config/nds32/nds32.md: Refine all the call and return patterns.
4831 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4834 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4835 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4837 PR rtl-optimization/84872
4838 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4839 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4842 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4844 * expr.c (copy_blkmode_to_reg): Revert 254862.
4845 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4847 2018-04-06 Richard Biener <rguenther@suse.de>
4850 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4851 after seeing a component reference with an adjacent field. Treat
4852 refs to arrays at struct end of external decls similar to
4853 refs to unconstrained commons.
4855 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4858 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4859 look through SAVE_EXPRs with non-side-effects argument. Adjust
4861 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4864 2018-04-06 Richard Biener <rguenther@suse.de>
4867 * alias.c (find_base_term): New wrapper around find_base_term
4868 unwinding CSELIB_VAL_PTR changes.
4869 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4872 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4874 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4876 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4877 constant definitions.
4878 ("nop"): lr 0,0 -> nopr r0
4879 ("nop_lr0", "nop_lr1"): New insn definitions.
4881 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4883 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4884 NDS32_V3PUSH_AVAILABLE_P macro.
4886 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4887 Chung-Ju Wu <jasonwucj@gmail.com>
4889 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4890 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4891 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4892 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4893 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4894 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4895 * config/nds32/constraints.md: New constraints and checking for hard
4896 float configuration.
4897 * config/nds32/iterators.md: New mode iterator and attribute for hard
4898 float configuration.
4899 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4901 * config/nds32/nds32-fpu.md: New file.
4902 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4903 deal with hard float code generation.
4904 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4906 (abi_type, float_reg_number): New enum type.
4907 * config/nds32/nds32-predicates.c: New predicates for hard float.
4908 * config/nds32/nds32-protos.h: Declare functions for hard float.
4909 * config/nds32/nds32.c: Implementation for hard float configuration.
4910 * config/nds32/nds32.h: Definitions for hard float configuration.
4911 * config/nds32/nds32.md: Include hard float machine description and
4912 modify patterns for hard float configuration.
4913 * config/nds32/nds32.opt: New options for hard float configuration.
4914 * config/nds32/predicates.md: New predicates for hard float
4917 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4919 * common/config/nds32/nds32-common.c
4920 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4922 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4925 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4926 CONSTRUCTOR_ELT (ctor, ...)->value.
4928 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4931 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4933 2018-04-05 Tom de Vries <tom@codesourcery.com>
4936 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4939 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4940 Kito Cheng <kito.cheng@gmail.com>
4942 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4943 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4944 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4945 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4947 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4948 Kito Cheng <kito.cheng@gmail.com>
4950 * config/nds32/constraints.md (Ufe): New memory constraint.
4951 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4952 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4953 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4955 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4956 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4958 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4960 * config/nds32/nds32.md: Use optimize_size in the condition for
4961 alu-shift instructions.
4963 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4965 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4967 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config/nds32/nds32.md (negsi2): Refine pattern.
4971 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4972 Chung-Ju Wu <jasonwucj@gmail.com>
4974 * config/nds32/iterators.md (shift_rotate): New code iterator.
4975 (shift): New code attribute.
4976 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4977 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4978 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4979 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4980 bit-wise operations.
4981 (andsi3, *andsi3): Ditto.
4982 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4983 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4984 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4985 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4986 nds32_ior_operand, nds32_xor_operand): New predicates.
4988 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4990 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4991 (addsi3, subsi3): ... this.
4993 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4995 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4997 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4999 * config/nds32/nds32.md: Adjust indention.
5001 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5003 * config/nds32/nds32.md (feature): New attribute.
5005 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5007 * config/nds32/nds32.md (subtype): New attribute.
5009 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5012 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5013 expansion to perform a bitwise AND of the argument followed by a
5014 boolean negation of the result.
5016 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5018 PR rtl-optimization/84878
5019 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5020 the basic block. Assert the use reference is not artificial and that
5021 it has an associated insn.
5023 2018-04-04 Michael Matz <matz@suse.de>
5025 * builtins.c (compute_objsize): Pass correct operand
5026 to array_at_struct_end_p.
5028 2018-04-04 Richard Biener <rguenther@suse.de>
5031 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5032 from contexts for DINFO_LEVEL_TERSE and below.
5034 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5036 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5037 resiter_operand condition.
5038 * config/nds32/nds32.md (*move<mode>): Ditto.
5040 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5041 Monk Chiang <sh.chiang04@gmail.com>
5043 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5045 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5047 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5049 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5050 Kito Cheng <kito.cheng@gmail.com>
5052 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5053 nds32_cond_code_str, output_cond_branch,
5054 output_cond_branch_compare_zero, 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: New functions.
5060 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5061 nds32_expand_cstore, nds32_expand_movcc,
5062 nds32_output_cbranchsi4_equality_zero,
5063 nds32_output_cbranchsi4_equality_reg,
5064 nds32_output_cbranchsi4_equality_reg_or_const_int,
5065 nds32_output_cbranchsi4_greater_less_zero): Declare.
5066 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5067 nds32_rimm11s_operand): New predicates.
5068 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5069 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5072 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5074 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5075 * config/nds32/nds32.md: Ditto.
5076 * config/nds32/pipelines.md: Ditto.
5078 2018-04-04 Richard Biener <rguenther@suse.de>
5080 PR tree-optimization/85168
5081 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5082 propagating abnormals.
5084 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5086 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5088 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5089 Kito Cheng <kito.cheng@gmail.com>
5091 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5092 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5093 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5094 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5095 * config/nds32/nds32.md (sibcall_internal): New.
5096 (sibcall_register): Remove.
5097 (sibcall_immediate): Remove.
5098 (sibcall_value_internal): New.
5099 (sibcall_value_register): Remove.
5100 (sibcall_value_immediate): Remove.
5101 * config/nds32/predicates.md (nds32_general_register_operand): New.
5102 (nds32_call_address_operand): New.
5104 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5106 PR rtl-optimization/85167
5107 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5108 bb_defs if *split_p, instead preinitialize it to NULL.
5110 PR tree-optimization/85156
5111 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5112 evaluating the argument multiple times.
5114 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5116 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5118 (_mm_cvtpd_ps): Likewise.
5119 (_mm_cvttpd_epi32): Likewise.
5120 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5121 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5122 vector, pixel, and bool following altivec.h include.
5124 2018-04-03 Martin Sebor <msebor@redhat.com>
5126 * doc/extend.texi (Common Function Attributes): Clarify.
5127 (const attribute): Likewise.
5128 (pure attribute): Likewise.
5130 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5133 * config/i386/i386.c (ix86_expand_vector_set): Use
5134 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5136 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5138 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5139 instructions when changing rounding bits to preserve precision bits
5140 in the x87 control word.
5142 2018-04-03 Martin Liska <mliska@suse.cz>
5144 PR tree-optimization/82491
5145 * rtl.h (strip_offset_and_add): Replace += suboffset with
5146 poly_uint64 () + suboffset.
5148 2018-03-29 Martin Liska <mliska@suse.cz>
5149 Martin Jambor <mjambor@suse.cz>
5152 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5153 param_type is not an integral or pointer type.
5155 2018-04-03 Richard Biener <rguenther@suse.de>
5157 * sese.h (recompute_all_dominators): Remove.
5159 2018-04-02 Martin Sebor <msebor@redhat.com>
5161 * doc/invoke.texi (-Wrestrict): Fix typos.
5163 2018-04-02 Jim Wilson <jimw@sifive.com>
5165 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5166 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5167 (<optab>di3, <optab>si3_extend): Likewise.
5168 (<optab>si3_mask, <optab>si3_mask_1): New.
5169 (<optab>di3_mask, <optab>di3_mask_1): New.
5170 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5171 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5172 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5174 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5176 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5179 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5181 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5182 (nds32_canonicalize_comparison): New function.
5184 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5185 Kito Cheng <kito.cheng@gmail.com>
5186 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5188 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5189 * config/nds32/constants.md (unspec_volatile_element): Add
5190 UNSPEC_VOLATILE_RELAX_GROUP.
5191 * config/nds32/nds32-relax-opt.c: New file.
5192 * config/nds32/nds32-predicates.c
5193 (nds32_symbol_load_store_p): New function.
5194 * config/nds32/nds32-protos.h
5195 (nds32_symbol_load_store_p): Declare function.
5196 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5197 * config/nds32/nds32.c
5198 (nds32_register_pass): New function to register pass.
5199 (nds32_register_passes): New function to register passes.
5200 * config/nds32/nds32.md (relax_group): New pattern.
5201 * config/nds32/nds32.opt (mrelax-hint): New option.
5202 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5204 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5206 * config/nds32/t-nds32: Modify files dependency.
5208 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5210 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5211 (PROFILE_HOOK): Define its implementation.
5213 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5215 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5216 type and 32-bit size.
5218 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5221 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5222 (V_128_256): New mode iterator.
5223 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5224 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5225 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5227 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5228 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5230 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5233 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5234 NaN inputs correctly.
5236 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5239 * config/rs6000/vsx.md (??r): New mode attribute.
5240 (*vsx_mov<mode>_64bit): Use it.
5241 (*vsx_mov<mode>_32bit): Likewise.
5243 2018-03-30 Martin Sebor <msebor@redhat.com>
5245 PR tree-optimization/84818
5246 * builtins.c (check_access): Use warning_n.
5248 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5251 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5253 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5256 2018-03-30 Julia Koval <julia.koval@intel.com>
5259 * x86-tune.def (movx, partial_reg_dependency): Enable for
5262 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5265 * lra-constraints.c (process_alt_operands): Move setting
5266 this_alternative_matches below.
5268 2018-03-29 Martin Liska <mliska@suse.cz>
5271 * doc/invoke.texi: Document how LTO works with debug info.
5272 Describe auto-load support of binutils. Mention 'x86-64'
5273 as valid option value of -march option.
5275 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5277 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5280 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5281 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5282 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5285 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5288 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5289 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5290 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5291 for RS6000_BTM_POWERPC64.
5292 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5293 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5294 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5299 2018-03-28 Carl Love <cel@us.ibm.com>
5302 2017-09-27 Carl Love <cel@us.ibm.com>
5304 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5305 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5306 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5309 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5311 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5312 instead of __vector bool.
5313 (_mm_max_pu8): Likewise.
5314 (_mm_min_pi16): Likewise.
5316 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5319 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5320 (DIVWEUO): Likewise.
5322 (DIVDEUO): Likewise.
5323 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5324 DIVWEUO and DIVDEUO.
5325 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5326 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5327 (div_extend): Likewise.
5328 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5330 (__builtin_divweuo): Likewise.
5331 (__builtin_divdeo): Likewise.
5332 (__builtin_divdeuo): Likewise.
5334 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5337 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5338 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5340 PR tree-optimization/82004
5341 * gimple-match-head.c (optimize_pow_to_exp): New function.
5342 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5343 Don't fold to exp if optimize_pow_to_exp is false.
5345 2018-03-28 Martin Liska <mliska@suse.cz>
5348 * calls.c (initialize_argument_information): Fix trailing space.
5349 * common.opt: Fix typo and provide better explanation for
5350 -fsanitize-coverage option.
5351 * config/i386/i386.opt: Fix typo.
5353 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5354 Martin Liska <mliska@suse.cz>
5357 * gimplify.c (asan_poison_variable): Don't do the check for
5358 gimplify_omp_ctxp here.
5359 (gimplify_decl_expr): Do it here.
5360 (gimplify_target_expr): Likewise.
5362 2018-03-28 Martin Liska <mliska@suse.cz>
5365 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5366 chkp_type_bounds_count if MPX is not enabled.
5368 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5370 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5372 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5375 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5376 function to create the function decl for complex long double
5377 multiply and divide for -mabi=ieeelongdouble.
5378 (init_float128_ieee): Call it.
5380 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5383 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5384 -fcf-protection=branch -mibt.
5385 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5387 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5390 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5392 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5395 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5396 extern array declarations.
5398 2018-03-27 Richard Biener <rguenther@suse.de>
5401 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5402 explicit single_use checks.
5404 2018-03-27 Richard Biener <rguenther@suse.de>
5406 PR tree-optimization/85082
5407 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5410 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5412 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5413 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5414 Turn on fasynchronous-unwind-tables and funwind-tables.
5416 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5419 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5420 (*bmi_blsr_<mode>_ccz): Ditto.
5422 2018-03-26 Tom de Vries <tom@codesourcery.com>
5424 PR tree-optimization/85063
5425 * omp-general.c (offloading_function_p): New function. Factor out
5427 * omp-offload.c (pass_omp_target_link::gate): ... here.
5428 * omp-general.h (offloading_function_p): Declare.
5429 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5430 with attribute omp declare target for offloading functions.
5432 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5434 PR tree-optimization/84005
5435 * tree-data-ref.h (get_base_for_alignment): Declare.
5436 * tree-data-ref.c (get_base_for_alignment_1): New function.
5437 (get_base_for_alignment): Likewise.
5438 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5439 get_base_for_alignment to find a suitable base object, instead
5440 of always using drb->base_address.
5442 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5445 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5446 known size by default.
5448 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5451 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5452 and non BLKmode operands.
5454 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5457 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5458 Clean up attributes.
5460 2018-03-23 Richard Biener <rguenther@suse.de>
5463 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5464 we are going to emit early debug for LTO.
5466 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5469 * function.c (match_asm_constraints_1): Don't optimize if input
5470 doesn't satisfy general_operand predicate for output's mode.
5473 * alias.c (write_dependence_p): Don't require for x_canonicalized
5474 non-VOIDmode if x has VOIDmode.
5477 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5478 just don't try to optimize it rather than assert it never happens.
5480 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5482 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5483 macro expansions for definition of ST_INTERNAL_<mode> and
5484 LD_INTERNAL_<mode> builtins.
5485 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5487 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5489 (altivec_expand_st_builtin): Likewise.
5490 (altivec_expand_builtin): Remove calls to deleted functions.
5491 (rs6000_address_for_altivec): Delete this function.
5492 * config/rs6000/vector.md: Remove expands for
5493 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5495 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5498 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5499 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5500 re-computing once computed.
5501 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5502 (arm_init_machine_status): Initialize
5503 machine->static_chain_stack_bytes.
5505 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5508 * doc/extend.texi: Add four new prototypes for vec_ld.
5509 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5510 definitions for more logical presentation.
5511 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5512 entries for V1TI variants of __builtin_altivec_ld builtin.
5513 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5514 handling of V1TI variant of LVX icode pattern.
5515 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5516 (rs6000_gimple_fold_builtin): Likewise.
5517 (altivec_init_builtins): Add code to define
5518 __builtin_altivec_lvx_v1ti function.
5520 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5523 * function.c (match_asm_constraints_1): Don't do the optimization
5524 if input isn't a REG, SUBREG, MEM or constant.
5526 2018-03-22 Tom de Vries <tom@codesourcery.com>
5528 PR tree-optimization/84956
5529 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5530 bb_has_abnormal_pred.
5532 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5535 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5536 DECL_INITIAL (decl) to decl at the end.
5537 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5540 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5542 * doc/extend.texi (__builtin_tgmath): Document when complex
5543 integer types are treated as _Complex _Float64.
5545 2018-03-21 Tom de Vries <tom@codesourcery.com>
5547 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5549 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5551 PR tree-optimization/84960
5552 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5553 if it is ENTRY block, move them into single succ of ENTRY in that case.
5555 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5557 PR tree-optimization/84811
5558 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5559 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5560 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5561 (known_subrange_p): Update accordingly. Cast each value involved
5562 in the size comparison, rather than casting the result of the
5565 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5567 PR tree-optimization/84982
5568 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5569 by flipping the least significant bit rather than all bits from
5570 bitpos to bitpos + bitsize - 1.
5572 2018-03-21 Nathan Sidwell <nathan@acm.org>
5574 * doc/extend.texi (Deprecated Features): Remove mention of
5575 long-deleted deprecations.
5577 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5580 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5581 * configure: Regenerate.
5583 2018-03-21 Tom de Vries <tom@codesourcery.com>
5585 PR tree-optimization/83126
5586 * tree-parloops.c (num_phis): New function.
5587 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5589 2018-03-21 Nathan Sidwell <nathan@acm.org>
5591 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5592 mention anon-struct/union members and trailing attributes.
5594 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5596 PR tree-optimization/84969
5597 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5598 builtin memset partitions if they set different rhs values.
5600 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5602 PR rtl-optimization/84989
5603 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5604 VEC_DUPLICATE with scalar result mode.
5606 2018-03-21 Martin Liska <mliska@suse.cz>
5609 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5610 not intended return statement.
5612 2018-03-21 Martin Liska <mliska@suse.cz>
5615 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5616 (chkp_find_bound_slots_1): Limit number of iterations.
5618 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5621 * Minor grammar fixes for x86 options.
5623 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5626 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5627 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5629 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5632 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5633 (altivec_lvx_<mode>_1op): Likewise.
5634 (altivec_stvx_<mode>_2op): Likewise.
5635 (altivec_stvx_<mode>_1op): Likewise.
5636 (altivec_lvx_<VM2:mode>): New define_expand.
5637 (altivec_stvx_<VM2:mode>): Likewise.
5638 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5639 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5640 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5641 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5642 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5643 (rs6000_gen_lvx): Likewise.
5644 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5645 (altivec_expand_stv_builtin): Likewise.
5646 (altivec_expand_builtin): Likewise.
5647 * config/rs6000/vector.md: Likewise.
5649 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5652 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5655 2018-03-20 Richard Biener <rguenther@suse.de>
5658 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5659 sign-conversions as zero, fall back to standard scalar_stmt
5662 2018-03-20 Martin Liska <mliska@suse.cz>
5665 * predict.c (rebuild_frequencies): Handle case when we have
5666 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5668 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5671 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5672 flag_section_anchors.
5673 * varasm.c (use_blocks_for_decl_p): Remove hack for
5674 dw2_force_const_mem.
5677 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5679 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5680 be created, use lowpart_subreg of operands[0] rather than operands[0]
5682 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5683 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5684 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5685 and n constraint instead of aarch64_shift_imm_di and Usd.
5686 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5687 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5689 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5692 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5693 to favor GPR over NEON registers.
5694 (<shift>di3_neon): Likewise.
5696 2018-03-20 Tom de Vries <tom@codesourcery.com>
5699 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5700 (nvptx_process_pars): Emit bar.sync asap and alap.
5702 2018-03-20 Tom de Vries <tom@codesourcery.com>
5705 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5706 seen_label if seen_label is already set.
5708 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5711 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5712 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5713 Use 1U instead of 1. Formatting fixes.
5716 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5717 instead of TREE_TYPE (s1) for the return value.
5719 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5721 PR tree-optimization/84946
5722 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5723 bitsize + bitsize in poly_uint64 rather than poly_int64.
5726 * dwarf2asm.c: Include fold-const.c.
5727 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5728 of decl rather than decl itself.
5730 PR rtl-optimization/84643
5731 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5733 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5736 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5737 calling assemble_variable.
5739 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5742 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5743 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5745 2018-03-19 Jim Wilson <jimw@sifive.com>
5748 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5749 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5750 (riscv_first_stack_step): Likewise.
5751 (riscv_option_override): Use STACK_BOUNDARY instead of
5753 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5755 (BIGGEST_ALIGNMENT): Set to 128.
5756 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5757 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5760 2018-03-19 Richard Biener <rguenther@suse.de>
5762 PR tree-optimization/84933
5763 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5764 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5766 2018-03-19 Richard Biener <rguenther@suse.de>
5768 PR tree-optimization/84859
5769 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5770 (cond_if_else_store_replacement): Perform sinking operation on
5771 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5772 Generalize what a BB with a single eligible store is.
5774 2018-03-19 Richard Biener <rguenther@suse.de>
5776 PR tree-optimization/84929
5777 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5778 chrec_is_positive against non-chrec arg.
5780 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5783 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5785 2018-03-18 Martin Liska <mliska@suse.cz>
5787 PR rtl-optimization/84635
5788 * regrename.c (build_def_use): Use matches_mode only when
5791 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5793 PR tree-optimization/84913
5794 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5795 vectorize chains of COND_EXPRs.
5797 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5799 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5801 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5803 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5805 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5807 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5809 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5810 Kito Cheng <kito.cheng@gmail.com>
5812 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5813 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5814 (nds32_adjust_reg_alloc_order): New function.
5815 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5817 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5819 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5820 nds32_print_operand, nds32_print_operand_address): Use
5821 HOST_WIDE_INT_PRINT_DEC instead.
5823 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5825 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5827 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5830 * config/i386/i386.c (initial_ix86_tune_features,
5831 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5833 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5834 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5835 rather than 1u << ix86_tune. Formatting fix.
5836 (ix86_option_override_internal): Change ix86_arch_mask from
5837 unsigned int to unsigned HOST_WIDE_INT, initialize to
5838 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5839 (ix86_function_specific_restore): Likewise.
5841 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5844 * postreload.c (reload_combine_recognize_pattern): Perform
5845 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5846 truncate_int_for_mode the result for the destination's mode.
5849 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5850 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5852 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5855 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5858 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5859 regno_allocno_class_array and sorted_pseudos.
5860 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5861 insns where regno is used.
5863 2018-03-16 Martin Liska <mliska@suse.cz>
5866 * multiple_target.c (create_dispatcher_calls): Redirect
5867 reference in the symbol table.
5869 2018-03-16 Martin Liska <mliska@suse.cz>
5872 * multiple_target.c (create_dispatcher_calls): Redirect also
5875 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5879 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5880 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5881 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5883 2018-03-16 Julia Koval <julia.koval@intel.com>
5885 * doc/invoke.texi (Skylake Server): Add CLWB.
5886 Cannonlake): Remove CLWB.
5888 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5890 PR tree-optimization/84841
5891 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5893 (FLOAT_ONE_CONST_TYPE): Define.
5894 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5895 (sort_by_operand_rank): Put entries with higher constant_type last
5896 rather than first to match comments.
5898 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5900 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5903 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5906 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5907 check and use address of target always.
5909 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5912 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5913 (indirect_thunk_bnd_needed): Likewise.
5914 (indirect_thunks_used): Likewise.
5915 (indirect_thunks_bnd_used): Likewise.
5916 (indirect_return_needed): New.
5917 (indirect_return_bnd_needed): Likewise.
5918 (output_indirect_thunk_function): Add a bool argument for
5920 (output_indirect_thunk_function): Don't generate alias for
5921 function return thunk.
5922 (ix86_code_end): Call output_indirect_thunk_function to generate
5923 function return thunks.
5924 (ix86_output_function_return): Set indirect_return_bnd_needed
5925 and indirect_return_needed instead of indirect_thunk_bnd_needed
5926 and indirect_thunk_needed.
5928 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5930 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5931 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5932 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5934 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5935 Paul Hua <paul.hua.gm@gmail.com>
5938 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5940 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5942 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5943 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5946 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5949 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5950 instead of GET_MODE_SIZE when comparing Units.
5952 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5955 * varasm.c (hash_section): Return an unchangeble hash value
5956 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5957 Return !aarch64_can_use_per_function_literal_pools_p ().
5959 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5962 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5963 rather than address of cmode as last argument to prepare_cmp_insn.
5965 2018-03-15 Julia Koval <julia.koval@intel.com>
5967 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5968 F_AVX512VNNI, F_AVX512BITALG): New.
5970 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5973 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5974 insn for floating-point loads and stores.
5976 2018-03-14 Carl Love <cel@us.ibm.com>
5978 * config/rs6000/rs6000-c.c: Add macro definitions for
5979 ALTIVEC_BUILTIN_VEC_PERMXOR.
5980 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5981 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5982 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5983 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5985 * config/doc/extend.texi: Add prototypes for vec_permxor.
5987 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5990 * diagnostic-show-locus.c (class layout_point): Convert m_line
5991 from int to linenum_type.
5992 (line_span::comparator): Use linenum "compare" function when
5993 comparing line numbers.
5994 (test_line_span): New function.
5995 (layout_range::contains_point): Convert param "row" from int to
5997 (layout_range::intersects_line_p): Likewise.
5998 (layout::will_show_line_p): Likewise.
5999 (layout::print_source_line): Likewise.
6000 (layout::should_print_annotation_line_p): Likewise.
6001 (layout::print_annotation_line): Likewise.
6002 (layout::print_leading_fixits): Likewise.
6003 (layout::annotation_line_showed_range_p): Likewise.
6004 (struct line_corrections): Likewise for field m_row.
6005 (line_corrections::line_corrections): Likewise for param "row".
6006 (layout::print_trailing_fixits): Likewise.
6007 (layout::get_state_at_point): Likewise.
6008 (layout::get_x_bound_for_row): Likewise.
6009 (layout::print_line): Likewise.
6010 (diagnostic_show_locus): Likewise for locals "last_line" and
6012 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6013 * input.c (selftest::test_linenum_comparisons): New function.
6014 (selftest::input_c_tests): Call it.
6015 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6016 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6017 * selftest.h (ASSERT_GT): New macro.
6018 (ASSERT_GT_AT): New macro.
6019 (ASSERT_LT): New macro.
6020 (ASSERT_LT_AT): New macro.
6022 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6024 PR rtl-optimization/84780
6025 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6027 2018-03-14 Martin Liska <mliska@suse.cz>
6029 * tree.c (record_node_allocation_statistics): Use
6030 get_stats_node_kind.
6031 (get_stats_node_kind): New function extracted from
6032 record_node_allocation_statistics.
6033 (free_node): Use get_stats_node_kind.
6035 2018-03-14 Richard Biener <rguenther@suse.de>
6037 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6038 that the value-set of ANTIC_IN doesn't grow.
6041 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6043 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6044 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6046 2018-03-14 Julia Koval <julia.koval@intel.com>
6048 * config.gcc (icelake-client, icelake-server): New.
6050 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6051 (initial_ix86_arch_features): Ditto.
6052 (PTA_SKYLAKE): Add SGX.
6053 (PTA_ICELAKE): Remove.
6054 (PTA_ICELAKE_CLIENT): New.
6055 (PTA_ICELAKE_SERVER): New.
6056 (ix86_option_override_internal): Split up icelake on icelake client and
6058 (get_builtin_code_for_version): Ditto.
6059 (fold_builtin_cpu): Ditto.
6060 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6061 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6062 * config/i386/i386.h (processor_type): Ditto.
6063 * doc/invoke.texi: Ditto.
6065 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6068 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6069 INTEGER_CST offset, add it together with bitpos / 8 and
6070 sign extend based on POINTER_SIZE.
6074 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6077 * config/i386/constraints.md (Yc): New register constraint.
6078 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6079 Use Yc constraint for alternative 2 of operand 0. Remove
6080 preferred_for_speed attribute.
6082 2018-03-14 Richard Biener <rguenther@suse.de>
6084 PR tree-optimization/84830
6085 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6086 with the old one to avoid oscillations.
6088 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6091 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6093 (assign_by_spills): Return a flag of reload assignment failure.
6094 Do not process the reload assignment failures. Do not spill other
6095 reload pseudos if they has the same reg class. Update n if
6097 (lra_assign): Add a return arg. Set up from the result of
6098 assign_by_spills call.
6099 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6100 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6101 usage_insns if it is not NULL.
6102 (spill_hard_reg_in_range): New function.
6103 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6104 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6105 function prototypes.
6106 (lra_assign): Change prototype.
6107 * lra.c (lra): Add code to deal with fails by splitting hard reg
6110 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6112 * config/riscv/riscv.opt (mrelax): New option.
6113 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6114 "norelax" when riscv_mrelax is disabled.
6115 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6117 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6120 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6121 reassociation for int modes.
6123 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6125 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6126 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6128 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6129 * config/aarch64/aarch64-sve.md
6130 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6131 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6132 (*extend<mode><Vwide>2): Rename to...
6133 (aarch64_sve_extend<mode><Vwide>2): ...this.
6134 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6135 renaming the old pattern to...
6136 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6138 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6139 define_expand, renaming the old pattern to...
6140 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6141 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6142 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6143 account when deciding which SVE instruction the optab should use.
6144 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6146 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6148 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6149 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6150 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6151 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6152 (tlsdesc_small_<mode>): Turn a define_expand and use
6153 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6154 (tlsdesc_small_advsimd_<mode>): ...this.
6155 (tlsdesc_small_sve_<mode>): New pattern.
6157 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6159 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6160 (UNSPEC_UMUL_HIGHPART): New constants.
6161 (MUL_HIGHPART): New int iteraor.
6162 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6163 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6165 (*<su>mul<mode>3_highpart): New define_insn.
6167 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6170 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6173 2018-03-13 Martin Liska <mliska@suse.cz>
6176 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6178 (sem_item_optimizer::~sem_item_optimizer): Release it.
6179 (sem_item_optimizer::merge_classes): Register variable aliases.
6180 (sem_item_optimizer::fixup_pt_set): New function.
6181 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6182 * ipa-icf.h: Declare new variables and functions.
6184 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6187 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6188 integer_pow2p@2 and test integer_pow2p in condition.
6189 (A < 0 ? C : 0): Similarly for @1.
6192 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6193 characters starting at p contain '\0' character, don't look beyond
6197 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6198 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6201 * reg-stack.c (change_stack): Change update_end var from int to
6202 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6203 also call set_block_for_insn on the newly added insns and rescan.
6206 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6207 on the last operand.
6210 * tree.c (stabilize_reference_1): Return save_expr (e) for
6211 STATEMENT_LIST even if it doesn't have side-effects.
6213 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6215 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6217 2018-03-12 Renlin Li <renlin.li@arm.com>
6219 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6220 aarch64_output_scalar_simd_mov_immediate.
6222 2018-03-12 Martin Sebor <msebor@redhat.com>
6224 PR tree-optimization/83456
6225 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6226 for perfectly overlapping calls to memcpy.
6227 (gimple_fold_builtin_memory_chk): Same.
6228 (gimple_fold_builtin_strcpy): Handle no-warning.
6229 (gimple_fold_builtin_stxcpy_chk): Same.
6230 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6232 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6234 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6235 parameter. Use it for SFmode.
6236 (rs6000_function_arg_advance_1): Adjust.
6237 (rs6000_function_arg): Adjust.
6238 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6240 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6242 PR rtl-optimization/84169
6243 PR rtl-optimization/84780
6244 * combine.c (can_combine_p): Check for a 2-insn combination whether
6245 the destination register is used between the two insns, too.
6247 2018-03-12 Richard Biener <rguenther@suse.de>
6249 PR tree-optimization/84803
6250 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6251 for refs DR analysis didn't process.
6253 2018-03-12 Richard Biener <rguenther@suse.de>
6255 PR tree-optimization/84777
6256 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6257 force-vectorize loops ignore whether we are optimizing for size.
6259 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6261 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6262 (TARGET_MD_ASM_ADJUST): Define.
6264 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6265 Kito Cheng <kito.cheng@gmail.com>
6266 Chung-Ju Wu <jasonwucj@gmail.com>
6268 * config/nds32/nds32.c (nds32_compute_stack_frame,
6269 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6270 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6271 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6272 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6273 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6274 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6275 * config/nds32/nds32.md (prologue, epilogue): Use macro
6276 NDS32_V3PUSH_AVAILABLE_P to do checking.
6278 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6281 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6282 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6283 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6284 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6285 addition of most attributes on !orig_type_die or the attribute not
6286 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6288 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6289 Chung-Ju Wu <jasonwucj@gmail.com>
6291 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6293 * config/nds32/nds32.opt (mvh): New option.
6295 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6296 Chung-Ju Wu <jasonwucj@gmail.com>
6298 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6300 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6301 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6304 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6305 Chung-Ju Wu <jasonwucj@gmail.com>
6307 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6309 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6310 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6312 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6313 Kito Cheng <kito.cheng@gmail.com>
6314 Chung-Ju Wu <jasonwucj@gmail.com>
6316 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6317 UNSPEC_FFMISM and UNSPEC_FLMISM.
6318 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6319 for ffb, ffmism and flmism.
6320 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6321 (unspec_ffmism): Ditto.
6322 (unspec_flmism): Ditto.
6323 (nds32_expand_builtin_impl): Check if string extension is available.
6324 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6325 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6327 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6330 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6333 * lra-assigns.c (assign_by_spills): Return a flag of reload
6334 assignment failure. Do not process the reload assignment
6335 failures. Do not spill other reload pseudos if they has the same
6337 (lra_assign): Add a return arg. Set up from the result of
6338 assign_by_spills call.
6339 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6340 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6341 usage_insns if it is not NULL.
6342 (spill_hard_reg_in_range): New function.
6343 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6344 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6345 function prototypes.
6346 (lra_assign): Change prototype.
6347 * lra.c (lra): Add code to deal with fails by splitting hard reg
6350 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6353 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6355 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6358 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6359 (dw_val_node): Add val_symbolic_view.
6360 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6361 (symview_upper_bound): New.
6362 (new_line_info_table): Initialize symviews_since_reset.
6363 (dwarf2out_source_line): Count symviews_since_reset and set
6364 symview_upper_bound.
6365 (dw_val_equal_p): Handle symview.
6366 (add_AT_symview): New.
6367 (print_dw_val): Handle symview.
6368 (attr_checksum, attr_checksum_ordered): Likewise.
6369 (same_dw_val_p, size_of_die): Likewise.
6370 (value_format, output_die): Likewise.
6371 (add_high_low_attributes): Use add_AT_symview for entry_view.
6372 (dwarf2out_finish): Reset symview_upper_bound, clear
6375 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6378 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6379 Add strict argument and use it.
6380 (rs6000_split_multireg_move): Update for new strict argument.
6381 (mem_operand_gpr): Disallow all non-offsettable addresses.
6382 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6384 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6387 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6388 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6389 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6392 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6393 decl, use remap_type if we want to use the type.
6395 2018-03-09 Martin Sebor <msebor@redhat.com>
6397 PR tree-optimization/84526
6398 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6400 (builtin_access::generic_overlap): Be prepared to handle non-array
6403 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6405 PR rtl-optimization/84682
6406 * lra-constraints.c (process_address_1): Check is_address flag
6407 for address constraints.
6408 (process_alt_operands): Likewise.
6409 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6410 preprocess_constraints.
6411 * recog.h (preprocess_constraints): Add oploc parameter.
6413 * recog.c (preprocess_constraints): Test address_operand for
6414 CT_ADDRESS constraints.
6416 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6419 * lra-assigns.c (assign_by_spills): Return a flag of reload
6420 assignment failure. Do not process the reload assignment
6421 failures. Do not spill other reload pseudos if they has the same
6423 (lra_assign): Add a return arg. Set up from the result of
6424 assign_by_spills call.
6425 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6426 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6427 usage_insns if it is not NULL.
6428 (spill_hard_reg_in_range): New function.
6429 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6430 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6431 function prototypes.
6432 (lra_assign): Change prototype.
6433 * lra.c (lra): Add code to deal with fails by splitting hard reg
6436 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6439 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6440 Accept complain bool parameter. Only emit errors if it is true.
6441 (arm_parse_cpu_option_name): Likewise.
6442 (arm_target_thumb_only): Adjust callers of the above.
6443 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6444 prototype to take a default true bool parameter.
6445 (arm_parse_arch_option_name): Likewise.
6447 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6448 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6452 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6453 * configure: Regenerate.
6454 * configure.ac ("linker --version-script option"): New.
6455 ("linker soname option"): New.
6457 2018-03-09 Richard Biener <rguenther@suse.de>
6459 PR tree-optimization/84775
6460 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6461 immediate uses of predicate stmts and mark them modified.
6464 PR tree-optimization/84178
6465 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6467 (version_loop_for_if_conversion): Delay update_ssa call.
6468 (tree_if_conversion): Delay update_ssa until after predicate
6471 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6474 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6475 when the function accesses prior frames.
6477 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6480 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6481 gen_llsym, otherwise call maybe_gen_llsym.
6484 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6485 has ',' character inside of it.
6487 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6490 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6491 as clobbering CC_REGNUM.
6493 2018-03-08 Richard Biener <rguenther@suse.de>
6496 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6497 (follow_copies_to_constant): Do not follow SSA names registered
6500 2018-03-08 Richard Biener <rguenther@suse.de>
6502 PR tree-optimization/84178
6503 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6505 (version_loop_for_if_conversion): Delay update_ssa call.
6506 (tree_if_conversion): Delay update_ssa until after predicate
6509 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6511 PR tree-optimization/84178
6512 * tree-if-conv.c (release_bb_predicate): Remove the
6513 the assertion that the stmts have NULL use_ops.
6514 Discard the statements, asserting that they haven't
6515 yet been added to a BB.
6517 2018-03-08 Richard Biener <rguenther@suse.de>
6519 PR tree-optimization/84746
6520 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6521 (phi_translate): Pass in destination ANTIC_OUT set.
6522 (phi_translate_1): Likewise. For a simplified result lookup
6523 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6524 (phi_translate_set): Adjust.
6525 (do_pre_regular_insertion): Likewise.
6526 (do_pre_partial_partial_insertion): Likewise.
6528 2018-03-08 Martin Liska <mliska@suse.cz>
6530 PR gcov-profile/84735
6531 * doc/gcov.texi: Document usage of profile files.
6532 * gcov-io.h: Document changes in the format.
6534 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6538 * dwarf2out.c (struct dw_line_info_table): Update comments for
6540 (FORCE_RESET_NEXT_VIEW): New.
6541 (FORCE_RESETTING_VIEW_P): New.
6542 (RESETTING_VIEW_P): Check for -1 too.
6543 (ZERO_VIEW_P): Likewise.
6544 (new_line_info_table): Force-reset next view.
6545 (dwarf2out_begin_function): Likewise.
6546 (dwarf2out_source_line): Simplify zero_view_p initialization.
6547 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6548 view directly. Omit view when omitting .loc at line 0.
6550 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6552 PR tree-optimization/84740
6553 * tree-switch-conversion.c (process_switch): Call build_constructors
6554 only if info.phi_count is non-zero.
6556 PR tree-optimization/84739
6557 * tree-tailcall.c (find_tail_calls): Check call arguments against
6558 DECL_ARGUMENTS (current_function_decl) rather than
6559 DECL_ARGUMENTS (func) when checking for tail recursion.
6561 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6563 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6564 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6565 Volker Reichelt's entry and add entries for people that perform
6566 GCC fuzzy testing and report numerous bugs.
6568 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6571 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6572 readonly data in sdata, if that is disabled.
6573 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6574 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6575 -mreadonly-in-sdata option.
6577 2018-03-07 Martin Sebor <msebor@redhat.com>
6579 PR tree-optimization/84468
6580 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6581 basic block when looking for nul assignment.
6583 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6586 * except.h (output_function_exception_table): Adjust prototype.
6587 * except.c (output_function_exception_table): Remove FNNAME parameter
6588 and add SECTION parameter. Ouput one part of the table at a time.
6589 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6590 the first part of the exception table and emit unwind directives.
6591 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6592 (i386_pe_seh_cold_init): Likewise.
6593 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6594 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6595 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6596 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6597 * config/i386/winnt.c: Include except.h.
6598 (struct seh_frame_state): Add reg_offset, after_prologue and
6599 in_cold_section fields.
6600 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6601 (i386_pe_seh_cold_init): New function.
6602 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6603 to seh->in_cold_section.
6604 (seh_emit_push): Record the offset of the push.
6605 (seh_emit_save): Record the offet of the save.
6606 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6607 Test seh->after_prologue to disregard the epilogue.
6608 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6609 (i386_pe_end_cold_function): New function.
6611 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6614 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6615 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6618 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6620 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6621 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6624 * multiple_target.c: Include tree-inline.h and intl.h.
6625 (expand_target_clones): Diagnose and fail if node->definition and
6626 !tree_versionable_function_p (node->decl).
6628 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6630 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6632 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6633 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6634 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6636 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6639 * combine.c (try_combine): Use reg_or_subregno instead of handling
6640 just paradoxical SUBREGs and REGs.
6642 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6644 * config/arc/arc.c (arc_finalize_pic): Remove function.
6645 (arc_must_save_register): We use single base PIC register, remove
6646 checks to save/restore the PIC register.
6647 (arc_expand_prologue): Likewise.
6648 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6650 (arc_verify_short): Likewise.
6651 (arc_attr_type): Likewise.
6652 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6653 (walk_stores): Likewise.
6654 (arc_address_cost): Make it static.
6655 (arc_verify_short): Likewise.
6656 (branch_dest): Likewise.
6657 (arc_attr_type): Likewise.
6658 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6659 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6660 (arc_final_prescan_insn): Remove inserting the nops due to
6661 hardware hazards. It is done in reorg step.
6662 (insn_length_variant_t): Remove.
6663 (insn_length_parameters_t): Likewise.
6664 (arc_insn_length_parameters): Likewise.
6665 (arc_get_insn_variants): Likewise.
6666 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6668 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6671 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6674 PR tree-optimization/84687
6675 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6677 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6679 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6681 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6682 Rename to ppc_speculation_barrier.
6683 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6684 __builtin_ppc_speculation_barrier.
6686 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6689 * combine.c (combine_simplify_rtx): Don't try to simplify if
6690 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6693 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6695 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6696 to 32 bytes when compiling for POWER9.
6698 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6701 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6702 regparm >= 3 with no arg reg available also for calls with
6703 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6706 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6708 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6710 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6713 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6715 2018-03-05 Richard Biener <rguenther@suse.de>
6717 PR tree-optimization/84486
6718 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6719 When inserting a __builtin_assume_aligned call set the LHS
6720 SSA name alignment info accordingly.
6722 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6724 PR tree-optimization/84114
6725 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6726 Avoid reassociation of FLOAT_MODE addition.
6728 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6730 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6731 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6732 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6733 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6734 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6735 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6736 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6738 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6739 __builtin_ia32_wbinvd): New builtins.
6740 (SPECIAL_ARGS2): New.
6741 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6742 (SPECIAL_ARGS2): New.
6743 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6744 (ix86_valid_target_attribute_inner_p): Ditto.
6745 (ix86_init_mmx_sse_builtins): Add special_args2.
6746 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6747 TARGET_WBNOINVD_P): New.
6748 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6749 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6750 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6751 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6752 * config/i386/pconfigintrin.h: New file.
6753 * config/i386/wbnoinvdintrin.h: Ditto.
6754 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6755 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6757 2018-03-05 Richard Biener <rguenther@suse.de>
6759 PR tree-optimization/84670
6760 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6762 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6763 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6764 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6765 doesn't grow if all successors (recursively) were visited at least
6768 2018-03-05 Richard Biener <rguenther@suse.de>
6770 PR tree-optimization/84650
6771 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6772 if executed in the loop pipeline.
6774 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6776 * doc/configfiles.texi (Configuration Files): Move info about
6777 conditionalizing $target-protos.h to...
6778 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6779 differs from $target-protos.h.
6781 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6782 Chung-Ju Wu <jasonwucj@gmail.com>
6784 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6785 * config/nds32/nds32-multiple.md (setmemsi): Define.
6786 * config/nds32/nds32-memory-manipulation.c
6787 (nds32_gen_dup_4_byte_to_word_value): New.
6788 (emit_setmem_word_loop): New.
6789 (emit_setmem_byte_loop): New.
6790 (nds32_expand_setmem_loop): New.
6791 (nds32_expand_setmem_loop_v3m): New.
6792 (nds32_expand_setmem_unroll): New.
6793 (nds32_expand_setmem): New.
6795 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6796 Chung-Ju Wu <jasonwucj@gmail.com>
6798 * config/nds32/nds32-memory-manipulation.c
6799 (nds32_emit_load_store): New.
6800 (nds32_emit_post_inc_load_store): New.
6801 (nds32_emit_mem_move): New.
6802 (nds32_emit_mem_move_block): New.
6803 (nds32_expand_movmemsi_loop_unknown_size): New.
6804 (nds32_expand_movmemsi_loop_known_size): New.
6805 (nds32_expand_movmemsi_loop): New.
6806 (nds32_expand_movmemsi_unroll): New.
6807 (nds32_expand_movmemqi): Rename ...
6808 (nds32_expand_movmemsi): ... to this.
6809 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6810 (movmemsi): ... to this.
6811 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6812 (nds32_expand_movmemsi): ... to this.
6814 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6815 Monk Chiang <sh.chiang04@gmail.com>
6816 Chung-Ju Wu <jasonwucj@gmail.com>
6818 * config/nds32/nds32-protos.h
6819 (nds32_expand_load_multiple): New arguments.
6820 (nds32_expand_store_multiple): Ditto.
6821 (nds32_valid_multiple_load_store): Rename ...
6822 (nds32_valid_multiple_load_store_p): ... to this.
6823 * config/nds32/nds32-memory-manipulation.c
6824 (nds32_expand_load_multiple): Refine implementation.
6825 (nds32_expand_store_multiple): Ditto.
6826 * config/nds32/nds32-multiple.md
6827 (load_multiple): Update nds32_expand_load_multiple interface.
6828 (store_multiple): Update nds32_expand_store_multiple interface.
6829 * config/nds32/nds32-predicates.c
6830 (nds32_valid_multiple_load_store): Rename ...
6831 (nds32_valid_multiple_load_store_p): ... to this and refine
6833 * config/nds32/predicates.md
6834 (nds32_load_multiple_and_update_address_operation): New predicate.
6835 (nds32_store_multiple_and_update_address_operation): New predicate.
6837 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6838 Chung-Ju Wu <jasonwucj@gmail.com>
6840 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6841 (combo): New attribute.
6842 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6844 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6846 * config/nds32/nds32.opt: Change -mcmodel= default value.
6848 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6849 Monk Chiang <sh.chiang04@gmail.com>
6850 Chung-Ju Wu <jasonwucj@gmail.com>
6852 * config/nds32/constants.md (unspec_element): New enum.
6853 * config/nds32/constraints.md (Umw): New constraint.
6854 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6855 * config/nds32/nds32-intrinsic.md: Likewise.
6856 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6857 (nds32_valid_smw_lwm_base_p): New.
6858 (nds32_output_smw_single_word): New.
6859 (nds32_output_lmw_single_word): New.
6860 (nds32_expand_unaligned_load): New.
6861 (nds32_expand_unaligned_store): New.
6862 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6863 (nds32_output_smw_single_word): Declare.
6864 (nds32_output_lmw_single_word): Declare.
6865 (nds32_expand_unaligned_load): Declare.
6866 (nds32_expand_unaligned_store): Declare.
6867 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6868 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6869 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6870 NDS32_BUILTIN_UASTORE_DW.
6871 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6874 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6875 Kito Cheng <kito.cheng@gmail.com>
6876 Chung-Ju Wu <jasonwucj@gmail.com>
6878 * config/nds32/nds32-intrinsic.c
6879 (nds32_expand_builtin_null_ftype_reg): Delete.
6880 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6881 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6882 (nds32_read_argument): New.
6883 (nds32_legitimize_target): Ditto.
6884 (nds32_legitimize_argument): Ditto.
6885 (nds32_check_constant_argument): Ditto.
6886 (nds32_expand_unop_builtin): Ditto.
6887 (nds32_expand_unopimm_builtin): Ditto.
6888 (nds32_expand_binop_builtin): Ditto.
6889 (nds32_builtin_decl_impl): Ditto.
6890 (builtin_description): Ditto.
6891 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6892 (nds32_init_builtins_impl): Ditto.
6893 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6894 (nds32_builtin_decl): New.
6895 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6896 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6898 2018-03-02 Jeff Law <law@redhat.com>
6900 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6901 (redundant_insn, fill_simple_delay_slots): Likewise.
6902 (fill_slots_from_thread): Likewise.
6903 * resource.c (mark_referenced_resources): Likewise.
6904 (mark_set_resources, find_dead_or_set_registers): Likewise.
6906 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6908 * substring-locations.h (format_warning_va): Formatting fix for
6910 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6912 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6913 * substring-locations.c: Include intl.h.
6914 (format_warning_va): Turned into small wrapper around
6915 format_warning_n_va, renamed to ...
6916 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6917 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6919 (format_warning_at_substring_n): New function.
6920 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6921 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6922 format_warning_at_substring with just a shorter name instead of
6923 const function pointer.
6924 (fmtwarn_n): New function.
6925 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6926 appropriate, get rid of all the fmtstr temporaries, move conditionals
6927 with G_() wrapped string literals directly into fmtwarn arguments,
6928 cast dir.len to (int), formatting fixes.
6930 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6932 * doc/invoke.texi: Remove "Cilk Plus" references.
6934 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6935 Richard Biener <rguenther@suse.de>
6938 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6939 for error or warning attributes if CALL_FROM_THUNK_P is set.
6942 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6945 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6946 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6949 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6950 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6952 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6953 instead of -1U in last predictors element's probability member.
6955 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6958 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6959 arguments if they are comparable.
6961 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6963 PR tree-optimization/84634
6964 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6965 masks and masked_loop_p with a single loop_masks, making sure it's
6966 null for bb vectorization.
6968 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6970 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6971 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6972 loop->force_vectorize to check whether there is no alias.
6974 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6977 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6979 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6981 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6982 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6983 instead of a loop around prev_real_insn.
6984 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6988 * config/i386/i386.c (ix86_print_operand): Use conditional
6989 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6992 2018-03-02 Richard Biener <rguenther@suse.de>
6994 PR tree-optimization/84427
6995 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6996 (bitmap_set_subtract_values): Rewrite to handle multiple
6999 (prune_clobbered_mems): Likewise.
7000 (phi_translate): Take edge instead of pred/phiblock.
7001 (phi_translate_1): Likewise.
7002 (phi_translate_set): Likewise. Insert all translated
7003 exprs for a value into the set, keeping possibly multiple
7004 expressions per value.
7005 (compute_antic_aux): Adjust for phi_translate changes.
7006 When intersecting union the expressions and prune those
7007 not in the final value set, keeping possibly multiple
7008 expressions per value. Do not use value-insertion
7009 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7010 all expressions. Add verification that the value-sets
7011 only shrink during iteration.
7012 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7013 (do_pre_regular_insertion): Likewise.
7014 (do_pre_partial_partial_insertion): Likewise.
7016 2018-03-02 Richard Biener <rguenther@suse.de>
7019 * config/darwin.c (saved_debug_info_level): New static global.
7020 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7021 (darwin_asm_lto_end): Restore debug info generation settings.
7023 2018-03-01 Martin Liska <mliska@suse.cz>
7026 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7029 2018-03-01 Richard Biener <rguenther@suse.de>
7032 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7033 type annotation with LTO when debug was disabled at compile-time.
7035 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7037 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7039 (mips_final_postscan_insn): Likewise.
7041 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7043 PR rtl-optimization/84528
7044 * alias.c (init_alias_target): Add commentary.
7045 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7046 a unique base value if the frame pointer is not eliminated
7047 to the stack pointer.
7049 2018-03-01 Tom de Vries <tom@codesourcery.com>
7051 PR rtl-optimization/83327
7052 * lra-int.h (hard_regs_spilled_into): Declare.
7053 * lra.c (hard_regs_spilled_into): Define.
7054 (init_reg_info): Init hard_regs_spilled_into.
7055 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7056 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7057 (process_bb_lives): Handle hard_regs_spilled_into.
7058 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7059 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7061 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7063 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7064 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7065 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7066 * config/rs6000/aix72.h: New file.
7068 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7070 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7071 instead of warning_at with conditional singular and plural messages
7075 * stor-layout.c (update_alignment_for_field): For
7076 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7077 && !DECL_PACKED (field), do the alignment update, just use
7078 only desired_align instead of MAX (type_align, desired_align)
7080 (place_field): Don't do known_align < desired_align handling
7081 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7082 is non-NULL, instead do it after rli->prev_field handling and
7083 only if not within a bitfield word. For DECL_PACKED (field)
7084 use type_align of BITS_PER_UNIT.
7086 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7088 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7089 superfluous parentheses and trailing spaces.
7091 2018-02-28 Richard Biener <rguenther@suse.de>
7093 PR tree-optimization/84584
7094 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7095 SCoPs with fake exit edge.
7097 2018-02-28 Martin Liska <mliska@suse.cz>
7100 * timevar.c (timer::print): Fix format to properly print 100%
7103 2018-02-28 Richard Biener <rguenther@suse.de>
7106 * genmatch.c (capture_info::walk_match): Do not mark
7107 captured expressions without operands as expr_p given
7108 they act more like predicates and should be subject to
7109 "lost tail" side-effect preserving.
7111 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7113 PR rtl-optimization/81611
7114 * auto-inc-dec.c (attempt_change): Move dead note from
7115 mem_insn if it's the next use of regno
7116 (find_address): Take address use of reg holding
7117 non-incremented value. Add parm to limit search to the named
7119 (merge_in_block): Attempt to use a mem insn that is the next
7120 use of the original regno.
7122 2018-02-27 Martin Sebor <msebor@redhat.com>
7125 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7126 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7128 2018-02-27 Martin Sebor <msebor@redhat.com>
7130 PR translation/84207
7131 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7132 n argument to unsigned HOST_WIDE_INT.
7133 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7134 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7135 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7136 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7138 2018-02-27 Richard Biener <rguenther@suse.de>
7140 PR tree-optimization/84512
7141 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7142 Do not use the estimate returned from record_stmt_cost for
7143 the scalar iteration cost but sum properly using add_stmt_cost.
7145 2018-02-27 Richard Biener <rguenther@suse.de>
7147 PR tree-optimization/84466
7148 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7149 Adjust last change to less strictly validate use operands.
7151 2018-02-27 Martin Liska <mliska@suse.cz>
7153 PR gcov-profile/84548
7154 * gcov.c (process_file): Allow partial overlap and consider it
7155 also as group functions.
7156 (output_lines): Properly calculate range of lines for a group.
7158 2018-02-27 Martin Liska <mliska@suse.cz>
7160 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7161 'ggc' suffixes. Change first column width.
7162 (timer::print): Fix formatting of the column.
7164 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7166 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7167 preserve inline entry blocks for the sake of debug inline
7168 entry point markers alone.
7169 (remove_unused_locals): Suggest in comments a better place to
7170 force the preservation of inline entry blocks that are
7171 otherwise unused, but do not preserve them.
7173 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7175 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7177 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7180 * config/i386/constraints.md (Bs): Replace
7181 ix86_indirect_branch_register with
7182 TARGET_INDIRECT_BRANCH_REGISTER.
7184 * config/i386/i386.md (indirect_jump): Likewise.
7185 (tablejump): Likewise.
7186 (*sibcall_memory): Likewise.
7187 (*sibcall_value_memory): Likewise.
7188 Peepholes of indirect call and jump via memory: Likewise.
7189 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7190 (*sibcall_value_GOT_32): Likewise.
7191 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7192 (GOT_memory_operand): Likewise.
7193 (call_insn_operand): Likewise.
7194 (sibcall_insn_operand): Likewise.
7195 (GOT32_symbol_operand): Likewise.
7196 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7198 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7200 PR rtl-optimization/83496
7201 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7202 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7203 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7204 redundant insn, if any.
7205 (relax_delay_slots): Likewise.
7206 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7208 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7210 PR tree-optimization/83965
7211 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7212 that grouped statements are part of a reduction chain. Return
7213 true if the statement is not marked as a reduction itself but
7215 (vect_recog_dot_prod_pattern): Don't check whether the statement
7216 is part of a group here.
7217 (vect_recog_sad_pattern): Likewise.
7218 (vect_recog_widen_sum_pattern): Likewise.
7220 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7223 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7225 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7228 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7230 (ix86_output_indirect_function_return): New prototype.
7231 (ix86_split_simple_return_pop_internal): Likewise.
7232 * config/i386/i386.c (indirect_return_via_cx): New.
7233 (indirect_return_via_cx_bnd): Likewise.
7234 (indirect_thunk_name): Handle return va CX_REG.
7235 (output_indirect_thunk_function): Create alias for
7236 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7237 (ix86_output_indirect_jmp): Remove the bool argument.
7238 (ix86_output_indirect_function_return): New function.
7239 (ix86_split_simple_return_pop_internal): Likewise.
7240 * config/i386/i386.md (*indirect_jump): Don't pass false
7241 to ix86_output_indirect_jmp.
7242 (*tablejump_1): Likewise.
7243 (simple_return_pop_internal): Change it to define_insn_and_split.
7244 Call ix86_split_simple_return_pop_internal to split it for
7246 (simple_return_indirect_internal): Call
7247 ix86_output_indirect_function_return instead of
7248 ix86_output_indirect_jmp.
7250 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7253 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7254 memset and value initialization afterwards.
7256 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7258 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7260 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7263 * common/config/aarch64/aarch64-common.c
7264 (aarch_option_optimization_table[]): Switch
7265 off fomit-frame-pointer
7267 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7268 Chung-Ju Wu <jasonwucj@gmail.com>
7270 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7272 (store_multiple): Ditto.
7274 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7276 * config.gcc: Add --with-cpu support for nds32 target.
7277 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7278 * config/nds32/nds32.opt: Add -mcpu= option.
7280 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7282 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7283 isel=yes): Warn for these deprecated options.
7285 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7287 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7288 ISA_2_5_MASKS_EMBEDDED.
7290 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7292 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7293 p->max as pointers rather than using iterative_hash_expr.
7295 2018-02-23 Carl Love <cel@us.ibm.com>
7297 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7298 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7300 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7301 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7302 P8V_BUILTIN_VEC_VUNSIGNED2.
7304 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7307 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7308 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7310 * lra-constraints.c (curr_insn_transform): Set up
7311 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7313 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7314 * lra-eliminations.c (spill_pseudos): Ditto.
7315 (process_insn_for_elimination): Ditto.
7316 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7317 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7318 LRA_NON_CLOBBERED_ALT.
7320 2018-02-22 Martin Sebor <msebor@redhat.com>
7322 PR tree-optimization/84480
7323 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7324 to maybe_diag_stxncpy_trunc. Call it.
7325 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7326 from gimple_fold_builtin_strcpy. Print inlining stack.
7327 (handle_builtin_stxncpy): Print inlining stack.
7328 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7330 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7333 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7334 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7335 and -fcheck-pointer-bounds are used together.
7336 (indirect_thunk_prefix): New enum.
7337 (indirect_thunk_need_prefix): New function.
7338 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7339 "_nt" instead of "_bnd" for NOTRACK prefix.
7340 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7341 (output_indirect_thunk_function): Likewise.
7343 (ix86_code_end): Update output_indirect_thunk_function calls.
7344 (ix86_output_indirect_branch_via_reg): Replace
7345 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7346 (ix86_output_indirect_branch_via_push): Likewise.
7347 (ix86_output_function_return): Likewise.
7348 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7349 incompatible with -fcf-protection=branch and
7350 -fcheck-pointer-bounds.
7352 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7355 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7356 Change gcc_assert call to output_operand_lossage.
7358 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7360 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7362 2018-02-22 DJ Delorie <dj@redhat.com>
7363 Sebastian Perta <sebastian.perta@renesas.com>
7364 Oleg Endo <olegendo@gcc.gnu.org>
7366 * config/rx/rx.c (rx_rtx_costs): New function.
7367 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7369 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7371 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7373 2018-02-22 Martin Liska <mliska@suse.cz>
7376 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7377 Add "native" as a possible value.
7379 2018-02-22 Martin Liska <mliska@suse.cz>
7382 * config/i386/i386.c (ix86_option_override_internal):
7383 Add "native" as a possible value for -march and -mtune.
7385 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7388 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7389 to all type variants.
7391 PR tree-optimization/84503
7392 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7393 width as info->bitpos + info->bitsize - start.
7394 (merged_store_group::merge_overlapping): Simplify width computation.
7395 (check_no_overlap): New function.
7396 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7397 start + width and last_order of the group, fail if check_no_overlap
7399 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7400 to group if check_no_overlap fails.
7402 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7404 * config/rs6000/altivec.md: Delete contraint arguments to
7405 define_expand, define_split, and define_peephole2, and in
7406 define_insn_and_split if always unused.
7407 * config/rs6000/darwin.md: Ditto.
7408 * config/rs6000/dfp.md: Ditto.
7409 * config/rs6000/rs6000.md: Ditto.
7410 * config/rs6000/sync.md: Ditto.
7411 * config/rs6000/vector.md: Ditto.
7412 * config/rs6000/vsx.md: Ditto.
7414 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7416 * config/rs6000/altivec.md: Write output control strings as braced
7417 blocks instead of double-quoted strings.
7418 * config/rs6000/darwin.md: Ditto.
7419 * config/rs6000/rs6000.md: Ditto.
7420 * config/rs6000/vector.md: Ditto.
7421 * config/rs6000/vsx.md: Ditto.
7423 2018-02-21 Jason Merrill <jason@redhat.com>
7425 PR c++/84314 - ICE with templates and fastcall attribute.
7426 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7428 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7430 * ipa-cp.c (determine_versionability): Fix comment typos.
7432 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7435 * ipa-cp.c (determine_versionability): Do not version functions caling
7438 2018-02-21 Martin Liska <mliska@suse.cz>
7441 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7442 Add "native" as a possible value.
7443 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7444 the macro when native cpu detection is available.
7446 2018-02-21 Martin Liska <mliska@suse.cz>
7449 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7450 Add "native" as a possible value.
7451 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7452 when native cpu detection is available.
7454 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7455 Martin Sebor <msebor@redhat.com>
7457 PR tree-optimization/84478
7458 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7460 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7461 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7462 support which is conservatively correct, for 2 only stay conservative
7463 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7464 argument to the 2 argument get_range_strlen, adjust 6 arg
7465 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7467 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7468 (gimple_fold_builtin_strlen): Pass true as last argument to
7471 2018-02-20 Martin Sebor <msebor@redhat.com>
7474 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7475 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7476 (builtin_memref::builtin_memref): Factor out parts into
7477 set_base_and_offset and call it.
7479 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7482 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7483 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7484 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7485 search at the associated MODE_INT.
7487 2018-02-20 Jeff Law <law@redhat.com>
7490 PR tree-optimization/81592
7492 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7493 for range data rather than using global data.
7494 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7495 range data rather than using global data.
7496 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7497 pass it to children as needed.
7498 (struct directive::fmtresult): Similarly.
7499 (struct directive::set_width): Similarly.
7500 (struct directive::set_precision): Similarly.
7501 (format_integer, format_directive, parse_directive): Similarly.
7502 (format_none): Accept unnamed vr_values parameter.
7503 (format_percent, format_floating, format_character): Similarly.
7504 (format_string, format_plain): Similarly.
7505 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7506 the EVRP range analyzer for range data rather than using global data.
7507 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7508 gimple-ssa-evrp-analyze.h
7509 (class sprintf_dom_walker): Add after_dom_children member function.
7510 Add evrp_range_analyzer member.
7511 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7512 range analyzer as needed.
7513 (sprintf_dom_walker::after_dom_children): New member function.
7514 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7516 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7517 (evrp_range_analyzer::pop_to_marker): Likewise.
7519 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7521 PR tree-optimization/84419
7522 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7523 with the required type if its current type is compatible but
7526 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7529 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7530 after vectorization.
7532 2018-02-20 Martin Liska <mliska@suse.cz>
7535 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7536 possible values if we don't have a hint.
7538 2018-02-20 Martin Liska <mliska@suse.cz>
7542 * final.c (shorten_branches): Build align_tab array with one
7544 * opts.c (finish_options): Add alignment option limit check.
7545 (MAX_CODE_ALIGN): Likewise.
7546 (MAX_CODE_ALIGN_VALUE): Likewise.
7547 * doc/invoke.texi: Document maximum allowed option value for
7548 all -falign-* options.
7550 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7553 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7554 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7555 * var-tracking.c (emit_note_insn_var_location): Remove all references
7556 to NOTE_INSN_CALL_ARG_LOCATION.
7557 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7558 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7559 Use copy_rtx_if_shared.
7560 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7561 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7562 (dwarf2out_var_location): Remove handling of
7563 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7565 * final.c (final_scan_insn): Remove all references to
7566 NOTE_INSN_CALL_ARG_LOCATION.
7567 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7568 before dumping final insns.
7569 * except.c (emit_note_eh_region_end): Remove all references to
7570 NOTE_INSN_CALL_ARG_LOCATION.
7571 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7572 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7573 * config/arc/arc.c (hwloop_optimize): Likewise.
7574 * config/arm/arm.c (create_fix_barrier): Likewise.
7575 * config/s390/s390.c (s390_chunkify_start): Likewise.
7576 * config/sh/sh.c (find_barrier): Likewise.
7577 * config/i386/i386.c (rest_of_insert_endbranch,
7578 ix86_seh_fixup_eh_fallthru): Likewise.
7579 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7580 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7581 * config/frv/frv.c (frv_function_prologue): Likewise.
7582 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7584 (note_outside_basic_block_p): Remove all references to
7585 NOTE_INSN_CALL_ARG_LOCATION.
7586 * gengtype.c (adjust_field_rtx_def): Likewise.
7587 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7589 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7590 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7593 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7596 PR tree-optimization/84452
7597 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7598 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7601 2018-02-19 Martin Liska <mliska@suse.cz>
7604 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7606 2018-02-19 Martin Liska <mliska@suse.cz>
7607 Richard Sandiford <richard.sandiford@linaro.org>
7609 PR tree-optimization/82491
7610 * gimple-fold.c (get_base_constructor): Make earlier bail out
7613 2018-02-19 Carl Love <cel@us.ibm.com>
7615 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7616 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7618 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7619 P8V_BUILTIN_VEC_NEG.
7621 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7623 * config/rl78/rl78.md (movdf): New define expand.
7625 2018-02-19 Martin Liska <mliska@suse.cz>
7628 * doc/invoke.texi: Fix typo.
7629 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7631 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7633 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7634 handle rs6000_single_float and rs6000_double_float specially for
7637 2018-02-16 Jeff Law <law@redhat.com>
7639 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7642 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7645 * ipa-inline.c (inline_small_functions): Fix a typo.
7647 2018-02-16 Nathan Sidwell <nathan@acm.org>
7649 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7651 2018-02-16 Carl Love <cel@us.ibm.com>
7653 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7654 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7655 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7656 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7657 expansion to P8V_BUILTIN_VEC_FLOAT2.
7659 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7661 PR rtl-optimization/70023
7662 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7663 src_regno into account.
7665 2018-02-16 Carl Love <cel@us.ibm.com>
7667 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7668 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7669 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7670 * config/rs6000/rs6000.c: Remove case statements for
7671 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7672 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7673 and P9V_BUILTIN_VEC_VINSERT4B.
7674 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7675 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7676 * config/rs6000/vsx.md:
7677 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7680 2018-02-16 Carl Love <cel@us.ibm.com>
7682 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7684 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7686 * config/rs6000/rs6000-c.c: Add the definitions for
7687 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7688 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7689 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7690 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7691 definition for insert4b and define insn *insert3b_internal.
7692 * doc/extend.texi: Add documentation for vec_extract4b.
7694 2018-02-16 Nathan Sidwell <nathan@acm.org>
7696 * doc/extend.texi (Backwards Compatibility): Mention friend
7697 injection. Note for-scope is deprecated.
7698 * doc/invoke.texi (-ffriend-injection): Deprecate.
7700 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7702 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7703 that moved to I2, also allow destinations that are a paradoxical
7704 subreg (instead of a normal reg).
7706 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7709 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7712 2018-02-16 Richard Biener <rguenther@suse.de>
7714 PR tree-optimization/84037
7715 PR tree-optimization/84016
7717 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7718 Adjust vec_construct for the fact we need additional higher latency
7719 128bit inserts for AVX256 and AVX512 vector builds.
7720 (ix86_add_stmt_cost): Scale vector construction cost for
7723 2018-02-16 Richard Biener <rguenther@suse.de>
7725 PR tree-optimization/84417
7726 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7727 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7728 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7730 2018-02-16 Martin Liska <mliska@suse.cz>
7733 * internal-fn.def (ASAN_CHECK): Set proper flags.
7734 (ASAN_MARK): Likewise.
7736 2018-02-16 Julia Koval <julia.koval@intel.com>
7738 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7739 from PTA_CANNONLAKE.
7741 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7744 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7745 Use ++iter rather than iter++ for std::list iterators.
7746 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7747 defer deleting them until all nodes in the forest are processed. Do
7748 free even leaf nodes. Change to_process into auto_vec.
7751 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7752 * vec.h (vec_default_construct): Use memset instead of placement new
7753 if BROKEN_VALUE_INITIALIZATION is defined.
7754 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7755 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7758 PR rtl-optimization/83723
7759 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7760 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7761 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7763 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7765 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7767 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7769 PR rtl-optimization/81443
7770 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7771 from inner REGs to paradoxical SUBREGs.
7773 2018-02-16 Richard Biener <rguenther@suse.de>
7775 PR tree-optimization/84399
7776 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7777 For operands we can analyze at their definition make sure we can
7778 analyze them at each use as well.
7780 2018-02-16 Richard Biener <rguenther@suse.de>
7782 PR tree-optimization/84190
7783 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7784 volatile accesses if the decl isn't volatile.
7786 2018-02-15 Jason Merrill <jason@redhat.com>
7788 PR c++/84314 - ICE with templates and fastcall attribute.
7789 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7790 TYPE_CANONICAL on an existing type.
7792 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7794 PR tree-optimization/84383
7795 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7796 dstoff nor call operand_equal_p if dstbase is NULL.
7798 PR tree-optimization/84334
7799 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7800 also a CONSTANT_CLASS_P, punt.
7802 2018-02-14 Jim Wilson <jimw@sifive.com>
7804 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7805 first SMALL_OPERAND check. New local min_second_step. Move assert
7806 to where locals are set. Add TARGET_RVC support.
7807 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7809 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7811 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7813 2018-02-14 Martin Sebor <msebor@redhat.com>
7815 PR tree-optimization/83698
7816 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7817 arrays constrain the offset range to their bounds.
7818 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7819 (builtin_access::overlap): Avoid setting the size of overlap if it's
7821 (maybe_diag_overlap): Also consider arrays when deciding what values
7822 of offsets to include in diagnostics.
7824 2018-02-14 Martin Sebor <msebor@redhat.com>
7827 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7828 that correspond to the kind of a declaration.
7830 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7833 * config/pa/pa.md: Load address of PIC label using the linkage table
7834 if the label is nonlocal.
7836 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7838 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7839 warning message if user requests -maltivec=be.
7840 * doc/invoke.texi: Document deprecation of -maltivec=be.
7842 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7845 * config/rs6000/rs6000-c.c: Update definitions for
7846 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7847 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7849 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7852 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7853 add _get_ssp intrinsics. Remove argument from
7854 __builtin_ia32_rdssp[d|q].
7855 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7856 * config/i386/i386-builtin.def: Remove argument from
7857 __builtin_ia32_rdssp[d|q].
7858 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7859 ix86_expand_special_args_builtin for _rdssp[d|q].
7860 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7861 Clear register before usage.
7862 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7863 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7865 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7867 PR tree-optimization/84357
7868 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7869 operand 1 of an ARRAY_REF too.
7871 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7874 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7875 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7877 (set_of_reg): New struct.
7878 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7879 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7880 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7882 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7883 Split into bitclr, bitset, bitinvert patterns if appropriate.
7884 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7885 use rx_fuse_in_memory_bitop.
7886 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7887 to named insn, correct maximum insn length.
7889 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7892 * machmode.def: Define a complex mode for PARTIAL_INT.
7893 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7895 * doc/rtl.texi: Document CSPImode.
7896 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7898 (msp430_hard_regno_nregs_with_padding): Likewise.
7900 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7903 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7905 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7907 PR rtl-optimization/84169
7908 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7909 we generated a parallel as new i3 and we split that to new i2 and i3
7910 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7911 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7912 those to i2, not i1. Partially rewrite this scan code.
7914 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7917 * stor-layout.c (place_field): For variable length fields, adjust
7918 offset_align afterwards not just based on the field's alignment,
7919 but also on the size.
7922 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7923 of exps and logs in the use_exp2 case.
7925 2018-02-13 Jeff Law <law@redhat.com>
7927 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7930 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7933 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7937 * common.opt (gas-loc-support, gas-locview-support): New.
7938 (ginline-points, ginternal-reset-location-views): New.
7939 * doc/invoke.texi: Document them. Use @itemx where intended.
7940 (gvariable-location-views): Adjust.
7941 * target.def (reset_location_view): New.
7942 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7943 (TARGET_RESET_LOCATION_VIEW): New.
7944 * doc/tm.texi: Rebuilt.
7945 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7946 (dwarf2out_default_as_locview_support): New.
7947 (output_asm_line_debug_info): Use option variables.
7948 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7949 (output_loc_list): Likewise.
7950 (add_high_low_attributes): Check option variables.
7951 Don't output entry view attribute in strict mode.
7952 (gen_inlined_subroutine_die): Check option variables.
7953 (dwarf2out_inline_entry): Likewise.
7954 (init_sections_and_labels): Likewise.
7955 (dwarf2out_early_finish): Likewise.
7956 (maybe_reset_location_view): New, from...
7957 (dwarf2out_var_location): ... here. Call it.
7958 * debug.h (dwarf2out_default_as_loc_support): Declare.
7959 (dwarf2out_default_as_locview_support): Declare.
7960 * hooks.c (hook_int_rtx_insn_0): New.
7961 * hooks.h (hook_int_rtx_insn_0): Declare.
7962 * toplev.c (process_options): Take -gas-loc-support and
7963 -gas-locview-support from dwarf2out. Enable
7964 -gvariable-location-views by default only with locview
7965 assembler support. Enable -ginternal-reset-location-views by
7966 default only if the target defines the corresponding hook.
7967 Enable -ginline-points by default if location views are
7968 enabled; force it disabled if statement frontiers are
7970 * tree-inline.c (expand_call_inline): Check option variables.
7971 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7973 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7975 PR tree-optimization/84321
7976 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7977 handling. Also check whether the anti-range contains any values
7978 that satisfy the mask; switch to a VR_RANGE if not.
7980 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7983 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7985 2018-02-13 Martin Jambor <mjambor@suse.cz>
7988 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7989 of call statements, also set location of a load to a temporary.
7991 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7993 * config/rl78/rl78.c (add_vector_labels): New function.
7994 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7995 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7996 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7997 which checks that no arguments are passed.
7998 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7999 * doc/extend.texi: Documentation for the new attribute.
8001 2018-02-13 Andreas Schwab <schwab@suse.de>
8003 * config/riscv/linux.h (CPP_SPEC): Define.
8005 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8008 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8009 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8010 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8011 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8012 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8013 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8014 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8015 temporarily for AES and PCLMUL builtins.
8017 PR tree-optimization/84339
8018 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8019 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8023 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8024 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8025 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8027 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8029 * omp-simd-clone.h: New file.
8030 * omp-simd-clone.c: Include omp-simd-clone.h.
8031 (expand_simd_clones): No longer static.
8032 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8033 cgraph.h and omp-simd-clone.h.
8034 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8035 (vect_recog_widen_shift_pattern): Formatting fix.
8036 (vect_pattern_recog_1): Don't check optab for calls.
8039 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8040 operands[2] into a REG before using gen_lowpart on it.
8042 2018-02-12 Jeff Law <law@redhat.com>
8045 * config/sh/sh.c (find_barrier): Consider a sibling call
8048 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8049 successfully back substituting a reg.
8051 2018-02-12 Richard Biener <rguenther@suse.de>
8053 PR tree-optimization/84037
8054 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8055 parameter, move visited init to caller.
8056 (vect_slp_analyze_operations): Separate cost from validity
8057 check, initialize visited once for all instances.
8058 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8060 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8062 (vect_model_promotion_demotion_cost): Likewise.
8063 (vectorizable_bswap): Guard cost modeling with !slp_node
8064 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8066 (vectorizable_call): Likewise.
8067 (vectorizable_conversion): Likewise.
8068 (vectorizable_assignment): Likewise.
8069 (vectorizable_shift): Likewise.
8070 (vectorizable_operation): Likewise.
8071 (vectorizable_store): Likewise.
8072 (vectorizable_load): Likewise.
8073 (vectorizable_condition): Likewise.
8074 (vectorizable_comparison): Likewise.
8076 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8079 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8080 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8081 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8083 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8086 * params.def (inline-min-speedup): Increase from 8 to 15.
8087 (max-inline-insns-auto): Decrease from 40 to 30.
8088 * ipa-split.c (consider_split): Add some buffer for function to
8089 be considered inlining candidate.
8090 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8093 2018-02-12 Richard Biener <rguenther@suse.de>
8095 PR tree-optimization/84037
8096 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8097 matched stmts if we cannot swap the non-matched ones.
8099 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8101 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8102 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8103 _mm_maskz_scalef_round_ss): New intrinsics.
8104 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8105 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8106 __builtin_ia32_scalefss_round): Remove.
8107 (__builtin_ia32_scalefsd_mask_round,
8108 __builtin_ia32_scalefss_mask_round): New intrinsics.
8109 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8110 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8111 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8112 "<round_constraint>")): Changed to ...
8113 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8114 "<round_scalar_constraint>")): ... this.
8115 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8116 %0, %1, %2<round_op3>}"): Changed to ...
8117 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8118 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8119 %2<round_scalar_mask_op3>}"): ... this.
8120 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8122 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8124 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8125 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8126 (_mm_maskz_sqrt_round_ss): New intrinsics.
8127 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8128 (__builtin_ia32_sqrtsd_mask_round)
8129 (__builtin_ia32_sqrtss_mask_round): New builtins.
8130 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8131 (__builtin_ia32_sqrtss_round): Remove.
8132 (__builtin_ia32_sqrtsd_mask_round)
8133 (__builtin_ia32_sqrtss_mask_round): New builtins.
8134 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8135 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8136 ((match_operand:VF_128 1 "vector_operand"
8137 "xBm,<round_constraint>")): Changed to ...
8138 ((match_operand:VF_128 1 "vector_operand"
8139 "xBm,<round_scalar_constraint>")): ... this.
8140 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8141 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8142 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8143 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8144 %<iptr>1<round_scalar_mask_op3>}): ... this.
8145 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8146 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8148 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8151 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8152 Cast vec_cmpeq result to correct type.
8153 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8154 Cast vec_cmpgt result to correct type.
8156 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8158 * final.c (final_scan_insn_1): Renamed from...
8159 (final_scan_insn): ... this. New wrapper, to recover
8160 seen from the outermost call in recursive ones.
8161 * config/sparc/sparc.c (output_return): Drop seen from call.
8162 (output_sibcall): Likewise.
8163 * config/visium/visium.c (output_branch): Likewise.
8165 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8167 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8170 2018-02-10 Alan Modra <amodra@gmail.com>
8173 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8174 Specify LR as an input.
8176 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8179 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8180 remove_member_access_dummy_vars): New functions.
8181 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8182 lower_omp_1, execute_lower_omp): Use them.
8184 PR rtl-optimization/84308
8185 * shrink-wrap.c (spread_components): Release todo vector.
8187 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8189 PR rtl-optimization/57193
8190 * ira-color.c (struct allocno_color_data): Add member
8191 conflict_allocno_hard_prefs.
8192 (update_conflict_allocno_hard_prefs): New.
8193 (bucket_allocno_compare_func): Add a preference based on
8194 conflict_allocno_hard_prefs.
8195 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8196 (color_allocnos): Remove a dead code. Initiate
8197 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8199 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8202 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8203 constraint from =wa to wa. Avoid a subreg on the output operand,
8204 instead use a pseudo and subreg it in a move.
8205 (p9_xxbrd_<mode>): Changed to ...
8206 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8207 (p9_xxbrd_v2df): New expander.
8208 (p9_xxbrw_<mode>): Changed to ...
8209 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8210 (p9_xxbrw_v4sf): New expander.
8212 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8214 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8216 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8219 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8220 multiply in 32-bit mode.
8221 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8222 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8225 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8227 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8228 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8229 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8230 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8232 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8235 * dwarf2out.c (is_trivial_indirect_ref): New function.
8236 (dwarf2out_late_global_decl): Do not generate a location
8237 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8238 and that are not defined in the current unit.
8240 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8242 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8243 instead of a libcall for UNORDERED.
8245 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8248 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8249 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8251 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8254 * config/s390/s390.c (s390_set_current_function): Invoke
8255 s390_indirect_branch_settings also if fndecl didn't change.
8257 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8259 * config/rs6000/rs6000.md (blockage): Set length to zero.
8261 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8263 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8265 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8268 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8269 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8273 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8274 PARALLEL incoming that failed vt_get_decl_and_offset check.
8277 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8278 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8280 (get_variable_section): For decls in named .bss* sections pass true as
8281 second argument to bss_initializer_p.
8283 2018-02-09 Marek Polacek <polacek@redhat.com>
8284 Jakub Jelinek <jakub@redhat.com>
8287 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8288 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8289 Sync some changes from cxx_fold_indirect_ref.
8291 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8293 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8295 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8296 (BLOCK_INLINE_ENTRY_LABEL): New.
8297 (dwarf2out_var_location): Disregard inline entry markers.
8298 (inline_entry_data): New struct.
8299 (inline_entry_data_hasher): New hashtable type.
8300 (inline_entry_data_hasher::hash): New.
8301 (inline_entry_data_hasher::equal): New.
8302 (inline_entry_data_table): New variable.
8303 (add_high_low_attributes): Add DW_AT_entry_pc and
8304 DW_AT_GNU_entry_view attributes if a pending entry is found
8305 in inline_entry_data_table. Add old entry_pc attribute only
8306 if debug nonbinding markers are disabled.
8307 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8308 markers are enabled.
8309 (block_within_block_p, dwarf2out_inline_entry): New.
8310 (dwarf2out_finish): Check that no entries remained in
8311 inline_entry_data_table.
8312 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8313 (final_scan_insn, notice_source_line): Likewise.
8314 (rest_of_clean_state): Skip inline entry markers.
8315 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8317 * gimple.c (gimple_build_debug_inline_entry): New.
8318 * gimple.h (enum gimple_debug_subcode): Add
8319 GIMPLE_DEBUG_INLINE_ENTRY.
8320 (gimple_build_debug_inline_entry): Declare.
8321 (gimple_debug_inline_entry_p): New.
8322 (gimple_debug_nonbind_marker_p): Adjust.
8323 * insn-notes.def (INLINE_ENTRY): New.
8324 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8325 inline entry marker notes.
8326 (print_insn): Likewise.
8327 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8328 (INSN_DEBUG_MARKER_KIND): Likewise.
8329 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8330 * tree-inline.c (expand_call_inline): Build and insert
8331 debug_inline_entry stmt.
8332 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8333 inline entry blocks early, if nonbind markers are enabled.
8334 (dump_scope_block): Dump fragment info.
8335 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8336 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8337 (gimple_build_debug_inline_entry): New.
8338 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8339 Enable/disable inline entry points too.
8340 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8341 (DEBUG_INSN): Describe inline entry markers.
8343 * common.opt (gvariable-location-views): New.
8344 (gvariable-location-views=incompat5): New.
8345 * config.in: Rebuilt.
8346 * configure: Rebuilt.
8347 * configure.ac: Test assembler for view support.
8348 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8349 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8350 * dwarf2out.c (var_loc_view): New typedef.
8351 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8352 (dwarf2out_locviews_in_attribute): New.
8353 (dwarf2out_locviews_in_loclist): New.
8354 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8355 (enum dw_line_info_opcode): Add LI_adv_address.
8356 (struct dw_line_info_table): Add view.
8357 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8358 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8359 (zero_view_p): New variable.
8360 (ZERO_VIEW_P): New macro.
8361 (output_asm_line_debug_info): New.
8362 (struct var_loc_node): Add view.
8363 (add_AT_view_list, AT_loc_list): New.
8364 (add_var_loc_to_decl): Add view param. Test it against last.
8365 (new_loc_list): Add view params. Record them.
8366 (AT_loc_list_ptr): Handle loc and view lists.
8367 (view_list_to_loc_list_val_node): New.
8368 (print_dw_val): Handle dw_val_class_view_list.
8369 (size_of_die): Likewise.
8370 (value_format): Likewise.
8371 (loc_list_has_views): New.
8372 (gen_llsym): Set vl_symbol too.
8373 (maybe_gen_llsym, skip_loc_list_entry): New.
8374 (dwarf2out_maybe_output_loclist_view_pair): New.
8375 (output_loc_list): Output view list or entries too.
8376 (output_view_list_offset): New.
8377 (output_die): Handle dw_val_class_view_list.
8378 (output_dwarf_version): New.
8379 (output_compilation_unit_header): Use it.
8380 (output_skeleton_debug_sections): Likewise.
8381 (output_rnglists, output_line_info): Likewise.
8382 (output_pubnames, output_aranges): Update version comments.
8383 (output_one_line_info_table): Output view numbers in asm comments.
8384 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8385 Call maybe_gen_llsym.
8386 (loc_list_from_tree_1): Adjust.
8387 (add_AT_location_description): Create view list attribute if
8388 needed, check it's absent otherwise.
8389 (convert_cfa_to_fb_loc_list): Adjust.
8390 (maybe_emit_file): Call output_asm_line_debug_info for test.
8391 (dwarf2out_var_location): Reset views as needed. Precompute
8392 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8393 attribute. Set view.
8394 (new_line_info_table): Reset next view.
8395 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8396 (dwarf2out_source_line): Likewise. Output view resets and labels to
8397 the assembler, or select appropriate line info opcodes.
8398 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8399 (optimize_string_length): Catch it. Adjust.
8400 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8401 dw_val_class_view_list, and remove it if no longer needed.
8402 (hash_loc_list): Hash view numbers.
8403 (loc_list_hasher::equal): Compare them.
8404 (optimize_location_lists): Check whether a view list symbol is
8405 needed, and whether the locview attribute is present, and
8406 whether they match. Remove the locview attribute if no longer
8408 (index_location_lists): Call skip_loc_list_entry for test.
8409 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8410 Use output_dwarf_version.
8411 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8412 (struct dw_val_node): Add val_view_list.
8413 * final.c (SEEN_NEXT_VIEW): New.
8414 (set_next_view_needed): New.
8415 (clear_next_view_needed): New.
8416 (maybe_output_next_view): New.
8417 (final_start_function): Rename to...
8418 (final_start_function_1): ... this. Take pointer to FIRST,
8419 add SEEN parameter. Emit param bindings in the initial view.
8420 (final_start_function): Reintroduce SEEN-less interface.
8421 (final): Rename to...
8422 (final_1): ... this. Take SEEN parameter. Output final pending
8423 next view at the end.
8424 (final): Reintroduce seen-less interface.
8425 (final_scan_insn): Output pending next view before switching
8426 sections or ending a block. Mark the next view as needed when
8427 outputting variable locations. Notify debug backend of section
8428 changes, and of location view changes.
8429 (rest_of_handle_final): Adjust.
8430 * toplev.c (process_options): Autodetect value for debug variable
8431 location views option. Warn on incompat5 without -gdwarf-5.
8432 * doc/invoke.texi (gvariable-location-views): New.
8433 (gvariable-location-views=incompat5): New.
8434 (gno-variable-location-views): New.
8436 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8438 PR tree-optimization/84136
8439 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8440 that the result of find_edge is non-NULL.
8442 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8445 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8446 storing integer register in SImode. Fix cost of 256 and 512
8447 byte aligned SSE register store.
8449 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8451 * config/i386/i386.c (ix86_multiplication_cost): Fix
8452 multiplication cost for TARGET_AVX512DQ.
8454 2018-02-08 Marek Polacek <polacek@redhat.com>
8456 PR tree-optimization/84238
8457 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8460 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8462 PR tree-optimization/84265
8463 * tree-vect-stmts.c (vectorizable_store): Don't treat
8464 VMAT_CONTIGUOUS accesses as grouped.
8465 (vectorizable_load): Likewise.
8467 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8469 PR tree-optimization/81635
8470 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8471 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8472 (test_round_for_mask): New functions.
8473 (wide_int_cc_tests): Call test_round_for_mask.
8474 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8475 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8476 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8477 range returned by get_range_info.
8479 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8482 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8483 * symtab.c: Include builtins.h
8484 (symtab_node::output_to_lto_symbol_table_p): Move here
8485 from lto-streamer-out.c:output_symbol_p.
8486 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8487 (output_symbol_p): Move all logic to symtab.c
8488 (produce_symtab): Update.
8490 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8492 * config/s390/s390-opts.h (enum indirect_branch): Define.
8493 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8494 (s390_indirect_branch_via_thunk)
8495 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8496 (enum s390_indirect_branch_type): Define.
8497 * config/s390/s390.c (struct s390_frame_layout, struct
8498 machine_function): Remove.
8499 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8500 (indirect_branch_table_label_no, indirect_branch_table_name):
8502 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8503 (enum s390_indirect_branch_option): Define.
8504 (s390_return_addr_from_memory): New function.
8505 (s390_handle_string_attribute): New function.
8506 (s390_attribute_table): Add new attribute handler.
8507 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8508 (s390_indirect_branch_via_thunk): New function.
8509 (s390_indirect_branch_via_inline_thunk): New function.
8510 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8511 sibling call optimization for non z10 compiles.
8512 (s390_emit_call): Force indirect branch target to be a single
8513 register. Add r1 clobber for non-z10 compiles.
8514 (s390_emit_epilogue): Emit return jump via return_use expander.
8515 (s390_reorg): Handle JUMP_INSNs as execute targets.
8516 (s390_option_override_internal): Perform validity checks for the
8517 new command line options.
8518 (s390_indirect_branch_attrvalue): New function.
8519 (s390_indirect_branch_settings): New function.
8520 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8521 (s390_output_indirect_thunk_function): New function.
8522 (s390_code_end): Implement target hook.
8523 (s390_case_values_threshold): Implement target hook.
8524 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8526 * config/s390/s390.h (struct s390_frame_layout)
8527 (struct machine_function): Move here from s390.c.
8528 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8529 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8530 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8531 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8532 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8533 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8534 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8535 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8536 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8537 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8538 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8539 (mnemonic attribute): Add values which aren't recognized
8541 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8542 pattern for branch conversion. Fix mnemonic attribute.
8543 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8544 indirect branch via thunk if requested.
8545 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8546 ("*indirect_jump"): Disable for branch conversion using out of
8548 ("indirect_jump_via_thunk<mode>_z10")
8549 ("indirect_jump_via_thunk<mode>")
8550 ("indirect_jump_via_inlinethunk<mode>_z10")
8551 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8552 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8553 ("casesi_jump_via_inlinethunk<mode>_z10")
8554 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8555 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8556 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8557 ("*indirect2_jump"): Disable for branch conversion.
8558 ("casesi_jump"): Turn into expander and expand patterns for branch
8560 ("return_use"): New expander.
8561 ("*return"): Emit return via thunk and rename it to ...
8562 ("*return<mode>"): ... this one.
8563 * config/s390/s390.opt: Add new options and and enum for the
8566 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8568 * lra-constraints.c (match_reload): Unconditionally use
8569 gen_lowpart_SUBREG, rather than selecting between that
8570 and equivalent gen_rtx_SUBREG code.
8572 2018-02-08 Richard Biener <rguenther@suse.de>
8574 PR tree-optimization/84233
8575 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8576 changed flag instead of boguously re-using phi_inserted.
8578 2018-02-08 Martin Jambor <mjambor@suse.cz>
8580 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8581 static local variables.
8583 2018-02-08 Richard Biener <rguenther@suse.de>
8585 PR tree-optimization/84278
8586 * tree-vect-stmts.c (vectorizable_store): When looking for
8587 smaller vector types to perform grouped strided loads/stores
8588 make sure the mode is supported by the target.
8589 (vectorizable_load): Likewise.
8591 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8593 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8594 Increase LDP/STP opportunities by adding adjacent callee-saves.
8596 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8598 PR rtl-optimization/84068
8599 PR rtl-optimization/83459
8600 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8602 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8604 PR tree-optimization/84224
8605 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8606 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8609 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8612 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8613 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8614 count, remove one USE.
8616 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8618 * doc/install.texi (Configuration): Document the
8619 --with-long-double-format={ibm,ieee} PowerPC configuration
8623 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8624 Convert from define_expand to be define_insn_and_split. Rework
8625 float/double/_Float128 conversions to QI/HI/SImode to work with
8626 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8627 conversions to QI/HImode types did a store and then a load to
8628 truncate the value. For conversions to VSX registers, don't split
8629 the insn, instead emit the code directly. Use the code iterator
8630 any_fix to combine signed and unsigned conversions.
8631 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8632 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8633 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8634 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8635 (fix_<mode>di2_hw): Likewise.
8636 (fixuns_<mode>di2_hw): Likewise.
8637 (fix_<mode>si2_hw): Likewise.
8638 (fixuns_<mode>si2_hw): Likewise.
8639 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8640 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8641 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8642 fix<uns>_trunc<SFDF:mode>si2_p8.
8643 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8645 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8646 (fix<uns>_<mode>_mem): Likewise.
8647 (fctiw<u>z_<mode>_mem): Likewise.
8648 (fix<uns>_<mode>_mem): Likewise.
8649 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8650 the register allocator from doing a direct move to the GPRs to do
8651 a store, and instead use the ISA 3.0 store byte/half-word from
8652 vector register instruction. For IEEE 128-bit floating point,
8653 also optimize stores of 32-bit ints.
8654 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8656 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8658 * genextract.c (push_pathstr_operand): New function to support
8660 (walk_rtx): Call push_pathstr_operand.
8661 (print_path): Support [a-zA-Z].
8663 2018-02-07 Richard Biener <rguenther@suse.de>
8665 PR tree-optimization/84037
8666 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8667 (cse_and_gimplify_to_preheader): Declare.
8668 (vect_get_place_in_interleaving_chain): Likewise.
8669 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8671 (_loop_vec_info::~_loop_vec_info): Delete it.
8672 (cse_and_gimplify_to_preheader): New function.
8673 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8674 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8675 (vectorizable_load): Likewise. For grouped stores always base
8676 the IV on the first element.
8677 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8678 condition before gimplifying.
8680 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8682 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8683 *DIV_EXPR and *MOD_EXPR.
8685 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8688 * config/i386/i386.c (ix86_option_override_internal): Mask out
8689 the CF_SET bit when checking -fcf-protection.
8691 2018-02-07 Tom de Vries <tom@codesourcery.com>
8694 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8697 2018-02-07 Richard Biener <rguenther@suse.de>
8699 PR tree-optimization/84204
8700 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8703 PR tree-optimization/84205
8704 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8705 special-case isl_ast_op_zdiv_r.
8707 PR tree-optimization/84223
8708 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8709 Only add conditions from within the region.
8710 (gather_bbs::after_dom_children): Adjust.
8712 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8715 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8716 * config/avr/avr.md: Only post-reload split REG-REG moves if
8717 either register is GENERAL_REG_P.
8719 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8721 PR tree-optimization/84235
8722 * tree-ssa-scopedtables.c
8723 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8724 if the subtraction is performed in floating point type where NaNs are
8725 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8726 build 1. Formatting fix.
8728 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8731 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8732 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8733 and skip it regardless of bb boundaries. Use CALL_P macro,
8734 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8735 unnecessarily, formatting fix.
8737 2018-02-06 Michael Collison <michael.collison@arm.com>
8739 * config/arm/thumb2.md:
8740 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8741 (*thumb_mov_notscc): Ditto.
8743 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8746 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8747 unsigned_fix, not "s".
8749 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8751 * configure.ac (gcc_fn_eh_frame_ro): New function.
8752 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8753 correct .eh_frame permissions.
8754 * configure: Regenerate.
8756 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8758 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8761 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8763 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8764 Display warning message for -mno-speculate-indirect-jumps.
8766 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8768 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8770 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8772 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8774 PR tree-optimization/84225
8775 * tree-eh.c (find_trapping_overflow): Only call
8776 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8778 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8781 * config/i386/i386.c: Reimplement the check of possible options
8782 -mibt/-mshstk conbination. Change error messages.
8783 * doc/invoke.texi: Fix a typo: remove extra '='.
8785 2018-02-06 Marek Polacek <polacek@redhat.com>
8787 PR tree-optimization/84228
8788 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8790 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8793 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8794 emitted arch directives.
8795 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8796 __ARM_FEATURE_COPROC before changing architectures.
8798 2018-02-06 Richard Biener <rguenther@suse.de>
8800 * config/i386/i386.c (print_reg): Fix typo.
8801 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8803 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8805 * configure: Regenerate.
8807 2018-02-05 Martin Sebor <msebor@redhat.com>
8809 PR tree-optimization/83369
8810 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8813 2018-02-05 Martin Liska <mliska@suse.cz>
8815 * doc/invoke.texi: Cherry-pick upstream r323995.
8817 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8819 * ira.c (ira_init_register_move_cost): Adjust comment.
8821 2018-02-05 Martin Liska <mliska@suse.cz>
8823 PR gcov-profile/84137
8824 * doc/gcov.texi: Fix typo in documentation.
8826 2018-02-05 Martin Liska <mliska@suse.cz>
8828 PR gcov-profile/83879
8829 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8830 using dlopen functionality.
8832 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8834 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8835 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8836 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8837 _mm_maskz_range_round_ss): New intrinsics.
8838 (__builtin_ia32_rangesd128_round)
8839 (__builtin_ia32_rangess128_round): Remove.
8840 (__builtin_ia32_rangesd128_mask_round,
8841 __builtin_ia32_rangess128_mask_round): New builtins.
8842 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8843 __builtin_ia32_rangess128_round): Remove.
8844 (__builtin_ia32_rangesd128_mask_round,
8845 __builtin_ia32_rangess128_mask_round): New builtins.
8846 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8847 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8848 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8849 "<round_saeonly_constraint>")): Changed to ...
8850 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8851 "<round_saeonly_scalar_constraint>")): ... this.
8852 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8853 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8854 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8855 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8856 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8858 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8860 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8862 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8863 Remove all values except native, 8540 and 8548.
8865 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8867 * config/i386/i386.c (ix86_output_function_return): Pass
8868 INVALID_REGNUM, instead of -1, as invalid register number to
8869 indirect_thunk_name and output_indirect_thunk.
8871 2018-02-02 Julia Koval <julia.koval@intel.com>
8873 * config.gcc: Add -march=icelake.
8874 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8875 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8876 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8877 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8878 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8879 (processor_target_table): Add icelake.
8880 (ix86_option_override_internal): Handle new PTAs.
8881 (get_builtin_code_for_version): Handle icelake.
8882 (M_INTEL_COREI7_ICELAKE): New.
8883 (fold_builtin_cpu): Handle icelake.
8884 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8885 * doc/invoke.texi: Add -march=icelake.
8887 2018-02-02 Julia Koval <julia.koval@intel.com>
8889 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8890 to wide_int_bitmask.
8891 * wide-int-bitmask.h: New.
8893 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8896 * config/i386/i386.md: Replace Pmode with word_mode in
8897 builtin_setjmp_setup and builtin_longjmp to support x32.
8899 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8903 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8905 (rs6000_supported_cpu_names): New static variable.
8906 (linux_cpu_translation_table): Likewise.
8907 (elf_platform) <cpu>: Define new static variable and use it.
8908 Translate kernel AT_PLATFORM name to canonical name if needed.
8909 Error if platform name is unknown.
8911 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8914 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8916 2018-02-01 Jeff Law <law@redhat.com>
8919 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8920 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8921 the scratch if RELEASE_VIA_POP is false.
8922 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8923 If we have to save a temporary register, decrement SIZE appropriately.
8924 Pass new arguments to release_scratch_register_on_entry.
8925 (ix86_adjust_stack_and_probe): Likewise.
8926 (ix86_emit_probe_stack_range): Pass new arguments to
8927 release_scratch_register_on_entry.
8929 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8931 PR rtl-optimization/84157
8932 * combine.c (change_zero_ext): Use REG_P predicate in
8933 front of HARD_REGISTER_P predicate.
8935 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8937 * config/avr/avr.c (avr_option_override): Move disabling of
8938 -fdelete-null-pointer-checks to...
8939 * common/config/avr/avr-common.c (avr_option_optimization_table):
8942 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8944 PR tree-optimization/81635
8945 * tree-data-ref.c (split_constant_offset_1): For types that
8946 wrap on overflow, try to use range info to prove that wrapping
8949 2018-02-01 Renlin Li <renlin.li@arm.com>
8952 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8954 (aarch64_register_move_cost): Likewise.
8955 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8957 (REG_CLASS_NAMES): Likewise.
8958 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8959 TAILCALL_ADDR_REGS. Remove IP registers.
8960 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8962 2018-02-01 Richard Biener <rguenther@suse.de>
8964 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8965 for specifying RPO order and allow NULL for that.
8966 * domwalk.c (dom_walker::dom_walker): Likewise.
8967 (dom_walker::walk): Handle NULL RPO order.
8968 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8970 (rewrite_update_dom_walker): Likewise.
8971 (mark_def_dom_walker): Likewise.
8973 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8975 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8976 (aarch64_maybe_expand_sve_subreg_move): Declare.
8977 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8978 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8980 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8981 that are semantically a reverse operation.
8982 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8983 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8984 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8986 (aarch64_can_change_mode_class): For big-endian, forbid changes
8987 between two SVE modes if they have different element sizes.
8989 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8991 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8992 the TImode handling for big-endian targets.
8994 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8996 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8997 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8999 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9000 Remove BSWAP handing for big-endian targets and use the form of
9001 LD1RQ appropariate for the mode.
9003 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9005 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9006 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9009 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9012 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9013 check for operands that need to go through aarch64_sve_reload_be.
9015 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9017 PR tree-optimization/81661
9018 PR tree-optimization/84117
9019 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9020 * tree-eh.c: Include gimplify.h.
9021 (find_trapping_overflow, replace_trapping_overflow,
9022 rewrite_to_non_trapping_overflow): New functions.
9023 * tree-vect-loop.c: Include tree-eh.h.
9024 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9025 * tree-data-ref.c: Include tree-eh.h.
9026 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9028 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9030 PR rtl-optimization/84123
9031 * combine.c (change_zero_ext): Check if hard register satisfies
9032 can_change_dest_mode before calling gen_lowpart_SUBREG.
9034 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9037 * ira.c (ira_init_register_move_cost): Remove assert.
9039 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9041 PR rtl-optimization/84071
9042 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9043 * doc/tm.texi: Regenerate.
9045 2018-01-31 Richard Biener <rguenther@suse.de>
9047 PR tree-optimization/84132
9048 * tree-data-ref.c (analyze_miv_subscript): Properly
9049 check whether evolution_function_is_affine_multivariate_p
9050 before calling gcd_of_steps_may_divide_p.
9052 2018-01-31 Julia Koval <julia.koval@intel.com>
9055 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9056 * config/i386/i386.md (rdpid_rex64) New.
9057 (rdpid): Make 32bit only.
9059 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9062 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9063 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9065 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9068 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9070 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9072 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9074 PR rtl-optimization/84071
9075 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9076 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9078 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9080 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9081 (arc_attribute_table): Add 'aux' attribute.
9082 (arc_in_small_data_p): Consider aux like variables.
9083 (arc_is_aux_reg_p): New function.
9084 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9085 (arc_get_aux_arg): New function.
9086 (prepare_move_operands): Handle aux-register access.
9087 (arc_handle_aux_attribute): New function.
9088 * doc/extend.texi (ARC Variable attributes): Add subsection.
9090 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9092 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9093 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9094 (arc_attribute_table): Add 'uncached' attribute.
9095 (arc_print_operand): Print '.di' flag for uncached memory
9097 (arc_in_small_data_p): Do not consider for small data the uncached
9099 (arc_is_uncached_mem_p): New function.
9100 * config/arc/predicates.md (compact_store_memory_operand): Check
9101 for uncached memory accesses.
9102 (nonvol_nonimm_operand): Likewise.
9103 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9105 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9108 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9109 falign-loops=): Add Optimization flag.
9111 2018-01-30 Jeff Law <law@redhat.com>
9114 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9115 INT_REGISTERS_SAVED. Check it prior to calling
9116 get_scratch_register_on_entry.
9117 (ix86_adjust_stack_and_probe): Similarly.
9118 (ix86_emit_probe_stack_range): Similarly.
9119 (ix86_expand_prologue): Corresponding changes.
9121 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9124 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9125 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9127 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9130 * lra-constraints.c (curr_insn_transform): Process AND in the
9133 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9135 PR rtl-optimization/83986
9136 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9137 dependence against last_pending_memory_flush in addition to
9140 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9142 PR tree-optimization/81611
9143 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9146 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9149 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9152 2018-01-30 Richard Biener <rguenther@suse.de>
9153 Jakub Jelinek <jakub@redhat.com>
9155 PR tree-optimization/84111
9156 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9157 inner loops added during recursion, as they don't have up-to-date
9160 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9163 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9164 (can_inline_edge_by_limits_p): ... here.
9165 (can_early_inline_edge_p, check_callers,
9166 update_caller_keys, update_callee_keys, recursive_inlining,
9167 add_new_edges_to_heap, speculation_useful_p,
9168 inline_small_functions,
9169 inline_small_functions, flatten_function,
9170 inline_to_all_callers_1): Update.
9172 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9174 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9177 2018-01-30 Richard Biener <rguenther@suse.de>
9179 PR tree-optimization/83008
9180 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9181 invariant and constant vector uses in stmts when they need
9184 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9187 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9188 * configure: Regenerate.
9190 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9192 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9194 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9195 Use gen_rtx_REG rather than gen_lowpart.
9197 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9199 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9200 rather than 0 when creating partial subregs.
9202 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9204 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9207 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9210 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9211 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9212 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9213 flags. This restores the settings used before the 2017-07-24.
9214 Turning off pre increment/decrement/modify allows IVOPTS to
9215 optimize DF/SF loops where the index is an int.
9217 2018-01-29 Richard Biener <rguenther@suse.de>
9218 Kelvin Nilsen <kelvin@gcc.gnu.org>
9221 * tree-vect-stmts.c (vectorizable_call): Don't call
9222 targetm.vectorize_builtin_md_vectorized_function if callee is
9225 2018-01-22 Carl Love <cel@us.ibm.com>
9227 * doc/extend.tex: Fix typo in second arg in
9228 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9230 2018-01-29 Richard Biener <rguenther@suse.de>
9232 PR tree-optimization/84086
9233 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9234 (flush_ssaname_freelist): When SSA names were released reset
9235 the SCEV hash table.
9237 2018-01-29 Richard Biener <rguenther@suse.de>
9239 PR tree-optimization/84057
9240 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9241 removed paths when removing edges.
9243 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9245 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9246 -mfunction-return=@var{choice}.
9248 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9251 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9253 (layout::print_source_line): Likewise.
9254 (test_get_line_width_without_trailing_whitespace): Add test cases.
9256 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9259 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9262 2018-01-26 Jim Wilson <jimw@sifive.com>
9264 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9266 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9269 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9271 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9272 and CMP + SUB-immediate -> SUBS.
9274 2018-01-26 Martin Sebor <msebor@redhat.com>
9276 PR tree-optimization/83896
9277 * tree-ssa-strlen.c (get_string_len): Rename...
9278 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9279 Avoid assuming length is constant.
9280 (handle_char_store): Use HOST_WIDE_INT for string length.
9282 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9285 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9286 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9288 2018-01-26 Richard Biener <rguenther@suse.de>
9290 PR rtl-optimization/84003
9291 * dse.c (record_store): Only record redundant stores when
9292 the earlier store aliases at least all accesses the later one does.
9294 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9296 PR rtl-optimization/83985
9297 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9298 REG_CFA_RESTORE insns.
9299 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9300 REG_CFA_RESTORE insns here.
9303 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9304 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9306 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9308 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9309 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9310 (arc_init): Likewise.
9311 (arc_override_options): Likewise.
9312 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9314 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9316 * config/arc/arc.h (TARGET_DBNZ): Define.
9317 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9318 properly set the tune attribute.
9319 (dbnz): Use TARGET_DBNZ guard.
9320 * config/arc/arc.opt (mtune): Add core3 option.
9322 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9324 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9325 recognize new pic like addresses.
9326 (arc_delegitimize_address): Clean up.
9328 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9330 * config/arc/arc-arches.def: Option mrf16 valid for all
9332 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9333 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9334 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9335 * config/arc/arc-tables.opt: Regenerate.
9336 * config/arc/arc.c (arc_conditional_register_usage): Handle
9337 reduced register file case.
9338 (arc_file_start): Set must have build attributes.
9339 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9341 * config/arc/arc.opt (mrf16): Add new option.
9342 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9343 * config/arc/genmultilib.awk: Handle new mrf16 option.
9344 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9345 * config/arc/t-multilib: Regenerate.
9346 * doc/invoke.texi (ARC Options): Document mrf16 option.
9348 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9350 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9351 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9352 (arc_attribute_table): Add 'secure_call' attribute.
9353 (arc_print_operand): Print secure call operand.
9354 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9356 (arc_is_secure_call_p): New function. * config/arc/arc.md
9357 (call_i): Add support for sjli instruction.
9358 (call_value_i): Likewise.
9359 * config/arc/constraints.md (Csc): New constraint.
9361 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9362 John Eric Martin <John.Martin@emmicro-us.com>
9364 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9365 * config/arc/arc.c (_arc_jli_section): New struct.
9366 (arc_jli_section): New type.
9367 (rc_jli_sections): New static variable.
9368 (arc_handle_jli_attribute): New function.
9369 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9370 (arc_file_end): New function.
9371 (TARGET_ASM_FILE_END): Define.
9372 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9373 (arc_add_jli_section): New function.
9374 (jli_call_scan): Likewise.
9375 (arc_reorg): Call jli_call_scan.
9376 (arc_output_addsi): Remove 'S' from printing asm operand.
9377 (arc_is_jli_call_p): New function.
9378 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9380 (movhi_insn): Likewise.
9381 (movsi_insn): Likewise.
9382 (movsi_set_cc_insn): Likewise.
9383 (loadqi_update): Likewise.
9384 (load_zeroextendqisi_update): Likewise.
9385 (load_signextendqisi_update): Likewise.
9386 (loadhi_update): Likewise.
9387 (load_zeroextendhisi_update): Likewise.
9388 (load_signextendhisi_update): Likewise.
9389 (loadsi_update): Likewise.
9390 (loadsf_update): Likewise.
9391 (movsicc_insn): Likewise.
9392 (bset_insn): Likewise.
9393 (bxor_insn): Likewise.
9394 (bclr_insn): Likewise.
9395 (bmsk_insn): Likewise.
9396 (bicsi3_insn): Likewise.
9397 (cmpsi_cc_c_insn): Likewise.
9398 (movsi_ne): Likewise.
9399 (movsi_cond_exec): Likewise.
9400 (clrsbsi2): Likewise.
9411 (call_i): Remove 'S' asm letter, add jli instruction.
9412 (call_value_i): Likewise.
9413 * config/arc/arc.op (mjli-always): New option.
9414 * config/arc/constraints.md (Cji): New constraint.
9415 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9417 (subsf3_fpx): Likewise.
9418 (mulsf3_fpx): Likewise.
9419 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9421 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9423 * doc/invoke.texi (ARC): Document mjli-always option.
9425 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9427 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9428 avoid addition with 0 and use incw and decw where possible.
9430 2018-01-26 Richard Biener <rguenther@suse.de>
9432 PR tree-optimization/81082
9433 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9434 association if it requires casting to unsigned.
9435 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9436 from fold_plusminus_mult_expr to catch important cases late when
9437 range info is available.
9439 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9441 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9442 * configure.ac (hidden_linkonce): New test.
9443 * configure: Regenerate.
9444 * config.in: Regenerate.
9446 2018-01-26 Julia Koval <julia.koval@intel.com>
9448 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9449 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9450 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9451 _mm_mask_bitshuffle_epi64_mask): Fix type.
9452 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9453 USI_FTYPE_V4DI_V4DI_USI): Remove.
9454 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9455 __builtin_ia32_vpshufbitqmb256_mask,
9456 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9457 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9458 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9460 2018-01-26 Alan Modra <amodra@gmail.com>
9463 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9464 UNSPEC_VBPERMQ. Sort other unspecs.
9466 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9468 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9470 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9473 * predict.c (drop_profile): Do not push/pop cfun; update also
9475 (handle_missing_profiles): Fix logic looking for zero profiles.
9477 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9480 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9481 on functions with #pragma omp declare simd or functions with simd
9483 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9484 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9485 Remove trailing \n from warning_at calls.
9487 2018-01-25 Tom de Vries <tom@codesourcery.com>
9490 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9491 for neutered workers.
9493 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9496 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9497 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9499 2018-01-24 Jeff Law <law@redhat.com>
9502 * i386.c (get_probe_interval): Move to earlier point.
9503 (ix86_compute_frame_layout): If -fstack-clash-protection and
9504 the frame is larger than the probe interval, then use pushes
9505 to save registers rather than reg->mem moves.
9506 (ix86_expand_prologue): Remove conditional for int_registers_saved
9509 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9512 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9513 min/max for never referenced object.
9515 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9518 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9520 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9521 attributes from DECL_ATTRIBUTES (decl) without affecting
9522 DECL_ATTRIBUTES (current_function_decl).
9523 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9524 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9526 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9528 PR tree-optimization/83979
9529 * fold-const.c (fold_comparison): Use constant_boolean_node
9530 instead of boolean_{true,false}_node.
9532 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9534 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9537 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9539 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9540 Simplify the clause that sets the length attribute.
9541 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9542 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9543 clause that sets the length attribute.
9544 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9546 2018-01-24 Tom de Vries <tom@codesourcery.com>
9549 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9550 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9551 Add strict parameter.
9552 (prevent_branch_around_nothing): Insert dummy insn between branch to
9553 label and label with no ptx insn inbetween.
9554 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9556 2018-01-24 Tom de Vries <tom@codesourcery.com>
9559 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9560 for neutered threads in warp.
9561 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9563 2018-01-24 Richard Biener <rguenther@suse.de>
9565 PR tree-optimization/83176
9566 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9569 2018-01-24 Richard Biener <rguenther@suse.de>
9571 PR tree-optimization/82819
9572 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9573 code generating pluses that are no-ops in the target precision.
9575 2018-01-24 Richard Biener <rguenther@suse.de>
9578 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9580 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9582 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9583 to merge probabilities.
9584 * predict.c (probably_never_executed): Also mark as cold functions
9585 with global 0 profile and guessed local profile.
9586 * profile-count.c (profile_probability::combine_with_count): New
9588 * profile-count.h (profile_probability::operator*,
9589 profile_probability::operator*=, profile_probability::operator/,
9590 profile_probability::operator/=): Reduce precision to adjusted
9591 and set value to guessed on contradictory divisions.
9592 (profile_probability::combine_with_freq): Remove.
9593 (profile_probability::combine_wiht_count): Declare.
9594 (profile_count::force_nonzero):: Set to adjusted.
9595 (profile_count::probability_in):: Set quality to adjusted.
9596 * tree-ssa-tail-merge.c (replace_block_by): Use
9599 2018-01-23 Andrew Waterman <andrew@sifive.com>
9600 Jim Wilson <jimw@sifive.com>
9602 * config/riscv/riscv.c (riscv_stack_boundary): New.
9603 (riscv_option_override): Set riscv_stack_boundary. Handle
9604 riscv_preferred_stack_boundary_arg.
9605 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9606 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9607 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9608 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9609 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9610 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9612 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9615 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9616 of struct ix86_frame.
9617 (ix86_expand_epilogue): Likewise. Add a local variable for
9618 the reg_save_offset field in struct ix86_frame.
9620 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9622 PR tree-optimization/82604
9623 * tree-loop-distribution.c (enum partition_kind): New enum item
9624 PKIND_PARTIAL_MEMSET.
9625 (partition_builtin_p): Support above new enum item.
9626 (generate_code_for_partition): Ditto.
9627 (compute_access_range): Differentiate cases that equality can be
9628 proven at all loops, the innermost loops or no loops.
9629 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9630 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9631 (finalize_partitions, distribute_loop): Don't fuse partition of
9632 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9633 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9636 2018-01-23 Martin Liska <mliska@suse.cz>
9638 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9639 order to ignore the predictor.
9640 (PRED_POLYMORPHIC_CALL): Likewise.
9641 (PRED_RECURSIVE_CALL): Likewise.
9643 2018-01-23 Martin Liska <mliska@suse.cz>
9645 * tree-profile.c (tree_profiling): Print function header to
9646 aware reader which function we are working on.
9647 * value-prof.c (gimple_find_values_to_profile): Do not print
9648 not interesting value histograms.
9650 2018-01-23 Martin Liska <mliska@suse.cz>
9652 * profile-count.h (enum profile_quality): Add
9653 profile_uninitialized as the first value. Do not number values
9654 as they are zero based.
9655 (profile_count::verify): Update sanity check.
9656 (profile_probability::verify): Likewise.
9658 2018-01-23 Nathan Sidwell <nathan@acm.org>
9660 * doc/invoke.texi (ffor-scope): Deprecate.
9662 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9664 PR tree-optimization/83510
9665 * domwalk.c (set_all_edges_as_executable): New function.
9666 (dom_walker::dom_walker): Convert bool param
9667 "skip_unreachable_blocks" to enum reachability. Move setup of
9668 edge flags to set_all_edges_as_executable and only do it when
9669 reachability is REACHABLE_BLOCKS.
9670 * domwalk.h (enum dom_walker::reachability): New enum.
9671 (dom_walker::dom_walker): Convert bool param
9672 "skip_unreachable_blocks" to enum reachability.
9673 (set_all_edges_as_executable): New decl.
9674 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9675 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9677 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9678 but converting true to REACHABLE_BLOCKS.
9679 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9681 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9682 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9683 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9685 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9686 if check_all_array_refs will be called.
9688 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9690 * tree.c (selftest::test_location_wrappers): Add more test
9693 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9695 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9696 (selftest::test_bit_in_range): Likewise.
9698 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9701 * doc/sourcebuild.texi (vect_float): Say that the selector
9702 only describes the situation when -funsafe-math-optimizations is on.
9703 (vect_float_strict): Document.
9705 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9707 PR tree-optimization/83965
9708 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9709 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9710 instead of checking only for a reduction.
9711 (vect_recog_widen_sum_pattern): Likewise.
9713 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9715 * predict.c (probably_never_executed): Only use precise profile info.
9716 (compute_function_frequency): Skip after inlining hack since we now
9717 have quality checking.
9719 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9721 * profile-count.h (profile_probability::very_unlikely,
9722 profile_probability::unlikely, profile_probability::even): Set
9723 precision to guessed.
9725 2018-01-23 Richard Biener <rguenther@suse.de>
9727 PR tree-optimization/83963
9728 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9729 Properly terminate dominator walk when crossing the exit edge not
9730 when visiting its source block.
9732 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9735 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9736 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9738 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9740 PR tree-optimization/83957
9741 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9742 semicolon after for body surrounded by braces.
9744 PR tree-optimization/83081
9745 * profile-count.h (profile_probability::split): New method.
9746 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9747 Use profile_probability::split.
9748 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9749 when splitting a single conditional jump into 2.
9751 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9753 PR tree-optimization/69452
9754 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9757 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9759 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9760 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9761 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9763 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9765 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9766 * config/rl78/rl78.md (movdi): New define_expand.
9767 * config/rl78/rl78.c (rl78_split_movdi): New function.
9769 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9772 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9774 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9775 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9776 128-bit to produce an UNSPEC move to get the double word with the
9777 signbit and then a shift directly to do signbit.
9778 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9779 implementation with a new version that just does either a direct
9780 move or a regular move. Move memory interface to separate insns.
9781 Move insns so they are next to the expander.
9782 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9783 with signbit move. Split big and little endian case.
9784 (signbit<mode>2_dm_mem_le): Likewise.
9785 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9786 (signbit<mode>2_dm2): Likewise.
9788 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9790 * config/rl78/rl78.md (anddi3): New define_expand.
9792 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9794 * config/rl78/rl78.md (umindi3): New define_expand.
9796 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9798 * config/rl78/rl78.md (smindi3): New define_expand.
9800 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9802 * config/rl78/rl78.md (smaxdi3): New define_expand.
9804 2018-01-22 Carl Love <cel@us.ibm.com>
9806 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9807 LVX_V1TI): Add macro expansion.
9808 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9809 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9810 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9811 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9812 Change check to determine if the instruction is a byte reversing
9813 entry. Fix typo in comment.
9814 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9815 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9816 Add def_builtin calls for new builtins.
9817 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9818 Add define_insn expansion.
9820 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9822 * config/rl78/rl78.md (umaxdi3): New define_expand.
9824 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9826 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9827 for non-QImode registers.
9829 2018-01-22 Richard Biener <rguenther@suse.de>
9831 PR tree-optimization/83963
9832 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9833 including the loop exit block.
9834 (scop_detection::merge_sese): Likewise.
9835 (scop_detection::add_scop): Do it here instead.
9837 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9839 * doc/sourcebuild.texi (arm_softfloat): Document.
9841 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9844 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9845 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9846 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9848 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9849 David Edelsohn <dje.gcc@gmail.com>
9852 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9853 Change "crset eq" to "crset 2".
9854 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9855 (*call_indirect_aix<mode>_nospec): Likewise.
9856 (*call_value_indirect_aix<mode>_nospec): Likewise.
9857 (*call_indirect_elfv2<mode>_nospec): Likewise.
9858 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9859 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9860 change assembly output from . to $.
9861 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9862 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9863 (*tablejump<mode>_internal1_nospec): Likewise.
9865 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9868 * config/sh/sh_optimize_sett_clrt.cc:
9869 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9871 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9873 PR tree-optimization/83940
9874 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9875 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9876 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9877 use it to pass back the definition type.
9878 (vect_check_store_rhs): Likewise rhs_dt_out.
9879 (vect_build_gather_load_calls): Add a mask_dt argument and use
9880 it instead of a call to vect_is_simple_use.
9881 (vectorizable_store): Update calls to vect_check_load_store_mask
9882 and vect_check_store_rhs. Use the dt returned by the latter instead
9883 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9884 instead of calls to vect_is_simple_use. Pass the scalar rather
9885 than the vector operand to vect_is_simple_use when handling
9886 second and subsequent copies of an rhs value.
9887 (vectorizable_load): Update calls to vect_check_load_store_mask
9888 and vect_build_gather_load_calls. Use the cached mask_dt and
9889 gs_info.offset_dt instead of calls to vect_is_simple_use.
9891 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9894 * tree-emutls.c: Include gimplify.h.
9895 (lower_emutls_2): New function.
9896 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9897 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9898 it before further processing.
9901 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9902 UINTVAL (trueop1) instead of INTVAL (op1).
9904 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9908 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9909 INCOMING_FRAME_SP_OFFSET if not defined.
9910 (scan_trace): Add ENTRY argument. If true and
9911 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9912 emit a note to adjust the CFA offset.
9913 (create_cfi_notes): Adjust scan_trace callers.
9914 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9915 INCOMING_FRAME_SP_OFFSET in the CIE.
9916 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9917 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9919 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9920 * doc/tm.texi: Regenerated.
9922 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9924 PR rtl-optimization/83147
9925 * lra-constraints.c (remove_inheritance_pseudos): Use
9926 lra_substitute_pseudo_within_insn.
9928 2018-01-19 Tom de Vries <tom@codesourcery.com>
9929 Cesar Philippidis <cesar@codesourcery.com>
9932 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9934 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9937 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9938 spaces for function labels.
9940 2018-01-19 Martin Liska <mliska@suse.cz>
9942 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9943 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9944 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9945 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9946 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9947 (PRED_CONST_RETURN): Change from 69 to 65.
9948 (PRED_NULL_RETURN): Change from 91 to 71.
9949 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9950 (PRED_LOOP_GUARD): Change from 66 to 73.
9952 2018-01-19 Martin Liska <mliska@suse.cz>
9954 * predict.c (predict_insn_def): Add new assert.
9955 (struct branch_predictor): Change type to signed integer.
9956 (test_prediction_value_range): Amend test to cover
9958 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9959 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9960 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9961 (PRED_LOOP_IV_COMPARE): Likewise.
9962 * predict.h (PROB_UNINITIALIZED): Define new constant.
9964 2018-01-19 Martin Liska <mliska@suse.cz>
9966 * predict.c (dump_prediction): Add new format for
9967 analyze_brprob.py script which is enabled with -details
9969 * profile-count.h (precise_p): New function.
9971 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9973 PR tree-optimization/83922
9974 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9975 there are no statements that need masking.
9976 (vect_active_double_reduction_p): New function.
9977 (vect_analyze_loop_operations): Use it when handling phis that
9978 are not in the loop header.
9980 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9982 PR tree-optimization/83914
9983 * tree-vect-loop.c (vectorizable_induction): Don't convert
9984 init_expr or apply the peeling adjustment for inductions
9985 that are nested within the vectorized loop.
9987 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9989 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9992 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9996 * function.h (gimplify_parameters): Add gimple_seq * argument.
9997 * function.c: Include gimple.h and options.h.
9998 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9999 for the added local temporaries if needed.
10000 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10001 if there are any parameter cleanups, wrap whole body into a
10002 try/finally with the cleanups.
10004 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10007 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10008 Use GET_MODE_CLASS for scalar floating point.
10010 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10014 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10015 Fix call of call_cgraph_insertion_hooks.
10017 2018-01-18 Martin Sebor <msebor@redhat.com>
10019 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10021 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10024 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10027 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10030 * common.opt: (-ffile-prefix-map): New option.
10031 * opts.c (common_handle_option): Defer it.
10032 * opts-global.c (handle_common_deferred_options): Handle it.
10033 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10034 * file-prefix-map.h: New file.
10035 (remap_debug_filename, add_debug_prefix_map): ...here.
10036 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10037 * final.c (debug_prefix_map, add_debug_prefix_map
10038 remap_debug_filename): Move to...
10039 * file-prefix-map.c: New file.
10040 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10041 generalize, get rid of alloca(), use strrchr() instead of strchr().
10042 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10043 Implement in terms of add_prefix_map().
10044 (remap_macro_filename, remap_debug_filename): Implement in term of
10046 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10047 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10048 * dbxout.c: Include file-prefix-map.h.
10049 * varasm.c: Likewise.
10050 * vmsdbgout.c: Likewise.
10051 * xcoffout.c: Likewise.
10052 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10053 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10054 * doc/invoke.texi (-ffile-prefix-map): Document.
10055 (-fdebug-prefix-map): Update description.
10057 2018-01-18 Martin Liska <mliska@suse.cz>
10059 * config/i386/i386.c (indirect_thunk_name): Document that also
10061 (output_indirect_thunk): Document why both instructions
10062 (pause and lfence) are generated.
10064 2018-01-18 Richard Biener <rguenther@suse.de>
10066 PR tree-optimization/83887
10067 * graphite-scop-detection.c
10068 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10069 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10070 (scop_detection::merge_sese): Re-implement with a flood-fill
10071 algorithm that properly finds a SESE region if it exists.
10073 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10076 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10077 pointer_diff optimizations use view_convert instead of convert.
10079 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10081 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10082 Generate different code for -mno-speculate-indirect-jumps.
10083 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10084 (*call_indirect_aix<mode>): Disable for
10085 -mno-speculate-indirect-jumps.
10086 (*call_indirect_aix<mode>_nospec): New define_insn.
10087 (*call_value_indirect_aix<mode>): Disable for
10088 -mno-speculate-indirect-jumps.
10089 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10090 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10091 -mno-speculate-indirect-jumps.
10092 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10094 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10096 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10097 long double type, set the flags for noting the default long double
10098 type, even if we don't pass or return a long double type.
10100 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10103 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10106 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10108 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10109 support for merge[hl].
10110 (fold_mergehl_helper): New helper function.
10111 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10112 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10113 (altivec_vmrglw_direct): Add xxmrglw insn.
10115 2018-01-17 Andrew Waterman <andrew@sifive.com>
10117 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10118 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10120 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10123 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10124 call the lto_location_cache before reading the
10125 DECL_SOURCE_LOCATION of the types.
10127 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10128 Richard Sandiford <richard.sandiford@linaro.org>
10130 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10131 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10132 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10133 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10134 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10136 * config/aarch64/constraints.md (aarch64_movti_operand):
10138 * config/aarch64/predicates.md (Uti): Add new constraint.
10140 2018-01-17 Carl Love <cel@us.ibm.com>
10142 * config/rs6000/vsx.md (define_expand xl_len_r,
10143 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10144 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10146 (define_expand, define_insn): Move the shift left from the
10147 define_insn to the define_expand for lxvl and stxvl instructions.
10148 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10149 and XL_LEN_R definitions to PURE.
10151 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10153 * config/i386/i386.c (indirect_thunk_name): Declare regno
10154 as unsigned int. Compare regno with INVALID_REGNUM.
10155 (output_indirect_thunk): Ditto.
10156 (output_indirect_thunk_function): Ditto.
10157 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10158 in the call to output_indirect_thunk_function.
10160 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10162 PR middle-end/83884
10163 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10164 rather than the size of inner_type to determine the stack slot size
10165 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10167 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10170 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10173 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10175 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10176 endian Linux systems to optionally enable multilibs for selecting
10177 the long double type if the user configured an explicit type.
10178 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10179 have no long double multilibs if not defined.
10180 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10181 warn if the user used -mabi={ieee,ibm}longdouble and we built
10182 multilibs for long double.
10183 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10184 appropriate multilib option.
10185 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10187 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10188 for building long double multilibs.
10189 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10191 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10193 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10196 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10198 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10201 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10204 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10207 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10209 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10210 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10212 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10214 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10215 different rtl trees depending on TARGET_64BIT.
10216 (rs6000_gen_lvx): Likewise.
10218 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10220 * config/visium/visium.md (nop): Tweak comment.
10221 (hazard_nop): Likewise.
10223 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10225 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10226 -mspeculate-indirect-jumps.
10227 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10228 for -mno-speculate-indirect-jumps.
10229 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10230 (*call_value_indirect_elfv2<mode>): Disable for
10231 -mno-speculate-indirect-jumps.
10232 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10233 (indirect_jump): Emit different RTL for
10234 -mno-speculate-indirect-jumps.
10235 (*indirect_jump<mode>): Disable for
10236 -mno-speculate-indirect-jumps.
10237 (*indirect_jump<mode>_nospec): New define_insn.
10238 (tablejump): Emit different RTL for
10239 -mno-speculate-indirect-jumps.
10240 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10241 (tablejumpsi_nospec): New define_expand.
10242 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10243 (tablejumpdi_nospec): New define_expand.
10244 (*tablejump<mode>_internal1): Disable for
10245 -mno-speculate-indirect-jumps.
10246 (*tablejump<mode>_internal1_nospec): New define_insn.
10247 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10250 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10252 * caller-save.c (insert_save): Drop unnecessary parameter. All
10255 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10256 Richard Biener <rguenth@suse.de>
10259 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10260 return early, inline manually is_gimple_sizepos. Make sure if we
10261 call gimplify_expr we don't end up with a gimple constant.
10262 * tree.c (variably_modified_type_p): Don't return true for
10263 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10264 * gimplify.h (is_gimple_sizepos): Remove.
10266 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10268 PR tree-optimization/83857
10269 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10270 vectorizable_live_operation for pure SLP statements.
10271 (vectorizable_live_operation): Handle PHIs.
10273 2018-01-16 Richard Biener <rguenther@suse.de>
10275 PR tree-optimization/83867
10276 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10277 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10279 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10282 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10283 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10284 If off is not INTEGER_CST, issue a may not be aligned warning
10285 rather than isn't aligned. Use isn%'t rather than isn't.
10286 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10288 <case MULT_EXPR>: Improve the case when bottom and one of the
10289 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10290 operand, in that case check if the other operand is multiple of
10291 bottom divided by the INTEGER_CST operand.
10293 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10296 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10297 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10298 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10299 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10300 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10301 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10302 (pa_function_arg_size): New function.
10304 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10306 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10307 in a separate statement.
10309 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10311 PR tree-optimization/83847
10312 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10313 group gathers and scatters.
10315 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10317 PR rtl-optimization/86620
10318 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10320 PR rtl-optimization/83213
10321 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10322 to last if both are JUMP_INSNs.
10324 PR tree-optimization/83843
10325 * gimple-ssa-store-merging.c
10326 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10327 store_immediate_info for bswap/nop orig_stores.
10329 2018-01-15 Andrew Waterman <andrew@sifive.com>
10331 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10333 <UDIV>: Increase cost if !TARGET_DIV.
10335 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10337 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10338 (define_attr "cr_logical_3op"): New.
10339 (cceq_ior_compare): Adjust.
10340 (cceq_ior_compare_complement): Adjust.
10341 (*cceq_rev_compare): Adjust.
10342 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10343 (is_cracked_insn): Adjust.
10344 (insn_must_be_first_in_group): Adjust.
10345 * config/rs6000/40x.md: Adjust.
10346 * config/rs6000/440.md: Adjust.
10347 * config/rs6000/476.md: Adjust.
10348 * config/rs6000/601.md: Adjust.
10349 * config/rs6000/603.md: Adjust.
10350 * config/rs6000/6xx.md: Adjust.
10351 * config/rs6000/7450.md: Adjust.
10352 * config/rs6000/7xx.md: Adjust.
10353 * config/rs6000/8540.md: Adjust.
10354 * config/rs6000/cell.md: Adjust.
10355 * config/rs6000/e300c2c3.md: Adjust.
10356 * config/rs6000/e500mc.md: Adjust.
10357 * config/rs6000/e500mc64.md: Adjust.
10358 * config/rs6000/e5500.md: Adjust.
10359 * config/rs6000/e6500.md: Adjust.
10360 * config/rs6000/mpc.md: Adjust.
10361 * config/rs6000/power4.md: Adjust.
10362 * config/rs6000/power5.md: Adjust.
10363 * config/rs6000/power6.md: Adjust.
10364 * config/rs6000/power7.md: Adjust.
10365 * config/rs6000/power8.md: Adjust.
10366 * config/rs6000/power9.md: Adjust.
10367 * config/rs6000/rs64.md: Adjust.
10368 * config/rs6000/titan.md: Adjust.
10370 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10372 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10373 ix86_indirect_branch_register logic.
10375 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10377 * config/i386/constraints.md (Bs): Update
10378 ix86_indirect_branch_register check. Don't check
10379 ix86_indirect_branch_register with GOT_memory_operand.
10381 * config/i386/predicates.md (GOT_memory_operand): Don't check
10382 ix86_indirect_branch_register here.
10383 (GOT32_symbol_operand): Likewise.
10385 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10387 * config/i386/predicates.md (constant_call_address_operand):
10388 Rewrite ix86_indirect_branch_register logic.
10389 (sibcall_insn_operand): Likewise.
10391 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10393 * config/i386/constraints.md (Bs): Replace
10394 ix86_indirect_branch_thunk_register with
10395 ix86_indirect_branch_register.
10397 * config/i386/i386.md (indirect_jump): Likewise.
10398 (tablejump): Likewise.
10399 (*sibcall_memory): Likewise.
10400 (*sibcall_value_memory): Likewise.
10401 Peepholes of indirect call and jump via memory: Likewise.
10402 * config/i386/i386.opt: Likewise.
10403 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10404 (GOT_memory_operand): Likewise.
10405 (call_insn_operand): Likewise.
10406 (sibcall_insn_operand): Likewise.
10407 (GOT32_symbol_operand): Likewise.
10409 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10411 PR middle-end/83837
10412 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10413 type rather than type addr's type points to.
10414 (expand_omp_atomic_mutex): Likewise.
10415 (expand_omp_atomic): Likewise.
10417 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10420 * config/i386/i386.c (output_indirect_thunk_function): Use
10421 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10422 for __x86_return_thunk.
10424 2018-01-15 Richard Biener <rguenther@suse.de>
10426 PR middle-end/83850
10427 * expmed.c (extract_bit_field_1): Fix typo.
10429 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10432 * config/arm/iterators.md (VF): New mode iterator.
10433 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10434 Remove integer-related logic from pattern.
10435 (neon_vabd<mode>_3): Likewise.
10437 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10439 PR middle-end/82694
10440 * common.opt (fstrict-overflow): No longer an alias.
10441 (fwrapv-pointer): New option.
10442 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10443 also for pointer types based on flag_wrapv_pointer.
10444 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10445 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10446 opts->x_flag_wrapv got set.
10447 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10448 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10449 POINTER_TYPE_OVERFLOW_UNDEFINED.
10450 * match.pd: Likewise in address comparison pattern.
10451 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10453 2018-01-15 Richard Biener <rguenther@suse.de>
10456 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10457 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10458 Reset type names to their identifier if their TYPE_DECL doesn't
10459 have linkage (and thus is used for ODR and devirt).
10460 (save_debug_info_for_decl): Remove.
10461 (save_debug_info_for_type): Likewise.
10462 (add_tree_to_fld_list): Adjust.
10463 * tree-pretty-print.c (dump_generic_node): Make dumping of
10464 type names more robust.
10466 2018-01-15 Richard Biener <rguenther@suse.de>
10468 * BASE-VER: Bump to 8.0.1.
10470 2018-01-14 Martin Sebor <msebor@redhat.com>
10473 * builtins.c (check_access): Avoid warning when the no-warning bit
10476 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10478 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10479 * ira-color (allocno_hard_regs_compare): Likewise.
10481 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10484 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10485 Use .pushsection/.popsection.
10487 2018-01-14 Martin Sebor <msebor@redhat.com>
10490 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10492 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10494 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10495 entry from extra_headers.
10496 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10497 extra_headers, make the list bitwise identical to the i?86-*-* one.
10499 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10501 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10502 -mcmodel=large with -mindirect-branch=thunk,
10503 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10504 -mfunction-return=thunk-extern.
10505 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10506 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10507 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10509 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10511 * config/i386/i386.c (print_reg): Print the name of the full
10512 integer register without '%'.
10513 (ix86_print_operand): Handle 'V'.
10514 * doc/extend.texi: Document 'V' modifier.
10516 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10518 * config/i386/constraints.md (Bs): Disallow memory operand for
10519 -mindirect-branch-register.
10521 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10522 (GOT_memory_operand): Likewise.
10523 (call_insn_operand): Likewise.
10524 (sibcall_insn_operand): Likewise.
10525 (GOT32_symbol_operand): Likewise.
10526 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10527 for -mindirect-branch-register.
10528 (tablejump): Likewise.
10529 (*sibcall_memory): Likewise.
10530 (*sibcall_value_memory): Likewise.
10531 Disallow peepholes of indirect call and jump via memory for
10532 -mindirect-branch-register.
10533 (*call_pop): Replace m with Bw.
10534 (*call_value_pop): Likewise.
10535 (*sibcall_pop_memory): Replace m with Bs.
10536 * config/i386/i386.opt (mindirect-branch-register): New option.
10537 * doc/invoke.texi: Document -mindirect-branch-register option.
10539 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10541 * config/i386/i386-protos.h (ix86_output_function_return): New.
10542 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10543 set function_return_type.
10544 (indirect_thunk_name): Add ret_p to indicate thunk for function
10546 (output_indirect_thunk_function): Pass false to
10547 indirect_thunk_name.
10548 (ix86_output_indirect_branch_via_reg): Likewise.
10549 (ix86_output_indirect_branch_via_push): Likewise.
10550 (output_indirect_thunk_function): Create alias for function
10551 return thunk if regno < 0.
10552 (ix86_output_function_return): New function.
10553 (ix86_handle_fndecl_attribute): Handle function_return.
10554 (ix86_attribute_table): Add function_return.
10555 * config/i386/i386.h (machine_function): Add
10556 function_return_type.
10557 * config/i386/i386.md (simple_return_internal): Use
10558 ix86_output_function_return.
10559 (simple_return_internal_long): Likewise.
10560 * config/i386/i386.opt (mfunction-return=): New option.
10561 (indirect_branch): Mention -mfunction-return=.
10562 * doc/extend.texi: Document function_return function attribute.
10563 * doc/invoke.texi: Document -mfunction-return= option.
10565 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10567 * config/i386/i386-opts.h (indirect_branch): New.
10568 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10569 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10570 with local indirect jump when converting indirect call and jump.
10571 (ix86_set_indirect_branch_type): New.
10572 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10573 (indirectlabelno): New.
10574 (indirect_thunk_needed): Likewise.
10575 (indirect_thunk_bnd_needed): Likewise.
10576 (indirect_thunks_used): Likewise.
10577 (indirect_thunks_bnd_used): Likewise.
10578 (INDIRECT_LABEL): Likewise.
10579 (indirect_thunk_name): Likewise.
10580 (output_indirect_thunk): Likewise.
10581 (output_indirect_thunk_function): Likewise.
10582 (ix86_output_indirect_branch_via_reg): Likewise.
10583 (ix86_output_indirect_branch_via_push): Likewise.
10584 (ix86_output_indirect_branch): Likewise.
10585 (ix86_output_indirect_jmp): Likewise.
10586 (ix86_code_end): Call output_indirect_thunk_function if needed.
10587 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10589 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10590 (ix86_attribute_table): Add indirect_branch.
10591 * config/i386/i386.h (machine_function): Add indirect_branch_type
10592 and has_local_indirect_jump.
10593 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10595 (tablejump): Likewise.
10596 (*indirect_jump): Use ix86_output_indirect_jmp.
10597 (*tablejump_1): Likewise.
10598 (simple_return_indirect_internal): Likewise.
10599 * config/i386/i386.opt (mindirect-branch=): New option.
10600 (indirect_branch): New.
10603 (thunk-inline): Likewise.
10604 (thunk-extern): Likewise.
10605 * doc/extend.texi: Document indirect_branch function attribute.
10606 * doc/invoke.texi: Document -mindirect-branch= option.
10608 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10611 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10613 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10615 * ipa-inline.c (want_inline_small_function_p): Return false if
10616 inlining has already failed with CIF_FINAL_ERROR.
10617 (update_caller_keys): Call want_inline_small_function_p before
10619 (update_callee_keys): Likewise.
10621 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10623 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10625 (rs6000_quadword_masked_address_p): Likewise.
10626 (quad_aligned_load_p): Likewise.
10627 (quad_aligned_store_p): Likewise.
10628 (const_load_sequence_p): Add comment to describe the outer-most loop.
10629 (mimic_memory_attributes_and_flags): New function.
10630 (rs6000_gen_stvx): Likewise.
10631 (replace_swapped_aligned_store): Likewise.
10632 (rs6000_gen_lvx): Likewise.
10633 (replace_swapped_aligned_load): Likewise.
10634 (replace_swapped_load_constant): Capitalize argument name in
10635 comment describing this function.
10636 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10637 and stores that access quad-word aligned addresses and replace
10638 with stvx or lvx instructions when appropriate.
10639 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10640 New function prototype.
10641 (rs6000_quadword_masked_address_p): Likewise.
10642 (rs6000_gen_lvx): Likewise.
10643 (rs6000_gen_stvx): Likewise.
10644 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10645 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10646 when memory address is aligned.
10647 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10648 this split to select lvx instruction when memory address is aligned.
10649 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10650 instruction when memory address is aligned.
10651 (*vsx_le_perm_load_v16qi): Likewise.
10652 (four unnamed splitters): Modify to select the stvx instruction
10653 when memory is aligned.
10655 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10657 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10658 which appears in the queue multiple times.
10660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10661 Alan Hayward <alan.hayward@arm.com>
10662 David Sherwood <david.sherwood@arm.com>
10664 * tree-vectorizer.h (vec_lower_bound): New structure.
10665 (_loop_vec_info): Add check_nonzero and lower_bounds.
10666 (LOOP_VINFO_CHECK_NONZERO): New macro.
10667 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10668 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10669 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10670 fields. Make seg_len the distance travelled, not including the
10672 (dr_direction_indicator): Declare.
10673 (dr_zero_step_indicator): Likewise.
10674 (dr_known_forward_stride_p): Likewise.
10675 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10677 (runtime_alias_check_p): Allow runtime alias checks with
10679 (operator ==): Compare access_size and align.
10680 (prune_runtime_alias_test_list): Rework for new distinction between
10681 the access_size and seg_len.
10682 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10684 (get_segment_min_max): New function.
10685 (create_intersect_range_checks): Use it.
10686 (dr_step_indicator): New function.
10687 (dr_direction_indicator): Likewise.
10688 (dr_zero_step_indicator): Likewise.
10689 (dr_known_forward_stride_p): Likewise.
10690 * tree-loop-distribution.c (data_ref_segment_size): Return
10691 DR_STEP * (niters - 1).
10692 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10694 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10695 (vect_preserves_scalar_order_p): New function, split out from...
10696 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10697 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10698 (vect_vfa_access_size): New function.
10699 (vect_vfa_align): Likewise.
10700 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10701 (dump_lower_bound): New function.
10702 (vect_check_lower_bound): Likewise.
10703 (vect_small_gap_p): Likewise.
10704 (vectorizable_with_step_bound_p): Likewise.
10705 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10706 depencies if the vectorization factor is 1. Convert the checks
10707 for nonzero steps into checks on the bounds of DR_STEP. Try using
10708 a bunds check for variable steps if the minimum required step is
10709 relatively small. Update calls to the dr_with_seg_len
10710 constructor and to vect_compile_time_alias.
10711 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10713 (vect_loop_versioning): Call it.
10714 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10716 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10719 Alan Hayward <alan.hayward@arm.com>
10720 David Sherwood <david.sherwood@arm.com>
10722 * doc/sourcebuild.texi (vect_scatter_store): Document.
10723 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10725 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10727 * genopinit.c (main): Add supports_vec_scatter_store and
10728 supports_vec_scatter_store_cached to target_optabs.
10729 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10730 IFN_MASK_SCATTER_STORE.
10731 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10733 * internal-fn.h (internal_store_fn_p): Declare.
10734 (internal_fn_stored_value_index): Likewise.
10735 * internal-fn.c (scatter_store_direct): New macro.
10736 (expand_scatter_store_optab_fn): New function.
10737 (direct_scatter_store_optab_supported_p): New macro.
10738 (internal_store_fn_p): New function.
10739 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10740 IFN_MASK_SCATTER_STORE.
10741 (internal_fn_mask_index): Likewise.
10742 (internal_fn_stored_value_index): New function.
10743 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10744 for scatter stores.
10745 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10746 * optabs-query.c (supports_vec_scatter_store_p): New function.
10747 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10748 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10749 true for scatter stores.
10750 (vect_gather_scatter_fn_p): Handle scatter stores too.
10751 (vect_check_gather_scatter): Consider using scatter stores if
10752 supports_vec_scatter_store_p.
10753 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10754 scatter stores too.
10755 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10756 internal_fn_stored_value_index.
10757 (check_load_store_masking): Handle scatter stores too.
10758 (vect_get_store_rhs): Make public.
10759 (vectorizable_call): Use internal_store_fn_p.
10760 (vectorizable_store): Handle scatter store internal functions.
10761 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10762 when deciding whether the end of the group has been reached.
10763 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10764 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10765 (mask_scatter_store<mode>): New insns.
10767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10768 Alan Hayward <alan.hayward@arm.com>
10769 David Sherwood <david.sherwood@arm.com>
10771 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10772 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10773 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10775 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10776 Use vect_truncate_gather_scatter_offset if we can't treat the
10777 operation as a normal gather load or scatter store.
10778 (get_group_load_store_type): Take the gather_scatter_info
10779 as argument. Try using a gather load or scatter store for
10780 single-element groups.
10781 (get_load_store_type): Update calls to get_group_load_store_type
10782 and vect_use_strided_gather_scatters_p.
10784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10785 Alan Hayward <alan.hayward@arm.com>
10786 David Sherwood <david.sherwood@arm.com>
10788 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10789 optional tree argument.
10790 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10792 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10793 but continue to use the current value as a fallback.
10794 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10795 to compare the updates.
10796 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10797 (get_load_store_type): Use it when handling a strided access.
10798 (vect_get_strided_load_store_ops): New function.
10799 (vect_get_data_ptr_increment): Likewise.
10800 (vectorizable_load): Handle strided gather loads. Always pass
10801 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * doc/md.texi (gather_load@var{m}): Document.
10808 (mask_gather_load@var{m}): Likewise.
10809 * genopinit.c (main): Add supports_vec_gather_load and
10810 supports_vec_gather_load_cached to target_optabs.
10811 * optabs-tree.c (init_tree_optimization_optabs): Use
10812 ggc_cleared_alloc to allocate target_optabs.
10813 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10814 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10816 * internal-fn.h (internal_load_fn_p): Declare.
10817 (internal_gather_scatter_fn_p): Likewise.
10818 (internal_fn_mask_index): Likewise.
10819 (internal_gather_scatter_fn_supported_p): Likewise.
10820 * internal-fn.c (gather_load_direct): New macro.
10821 (expand_gather_load_optab_fn): New function.
10822 (direct_gather_load_optab_supported_p): New macro.
10823 (direct_internal_fn_optab): New function.
10824 (internal_load_fn_p): Likewise.
10825 (internal_gather_scatter_fn_p): Likewise.
10826 (internal_fn_mask_index): Likewise.
10827 (internal_gather_scatter_fn_supported_p): Likewise.
10828 * optabs-query.c (supports_at_least_one_mode_p): New function.
10829 (supports_vec_gather_load_p): Likewise.
10830 * optabs-query.h (supports_vec_gather_load_p): Declare.
10831 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10832 and memory_type field.
10833 (NUM_PATTERNS): Bump to 15.
10834 * tree-vect-data-refs.c: Include internal-fn.h.
10835 (vect_gather_scatter_fn_p): New function.
10836 (vect_describe_gather_scatter_call): Likewise.
10837 (vect_check_gather_scatter): Try using internal functions for
10838 gather loads. Recognize existing calls to a gather load function.
10839 (vect_analyze_data_refs): Consider using gather loads if
10840 supports_vec_gather_load_p.
10841 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10842 (vect_get_gather_scatter_offset_type): Likewise.
10843 (vect_convert_mask_for_vectype): Likewise.
10844 (vect_add_conversion_to_patterm): Likewise.
10845 (vect_try_gather_scatter_pattern): Likewise.
10846 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10847 (vect_vect_recog_func_ptrs): Add it.
10848 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10849 internal_fn_mask_index and internal_gather_scatter_fn_p.
10850 (check_load_store_masking): Take the gather_scatter_info as an
10851 argument and handle gather loads.
10852 (vect_get_gather_scatter_ops): New function.
10853 (vectorizable_call): Check internal_load_fn_p.
10854 (vectorizable_load): Likewise. Handle gather load internal
10856 (vectorizable_store): Update call to check_load_store_masking.
10857 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10858 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10859 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10860 (aarch64_gather_scale_operand_d): New predicates.
10861 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10862 (mask_gather_load<mode>): New insns.
10864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10865 Alan Hayward <alan.hayward@arm.com>
10866 David Sherwood <david.sherwood@arm.com>
10868 * optabs.def (fold_left_plus_optab): New optab.
10869 * doc/md.texi (fold_left_plus_@var{m}): Document.
10870 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10871 * internal-fn.c (fold_left_direct): Define.
10872 (expand_fold_left_optab_fn): Likewise.
10873 (direct_fold_left_optab_supported_p): Likewise.
10874 * fold-const-call.c (fold_const_fold_left): New function.
10875 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10876 * tree-parloops.c (valid_reduction_p): New function.
10877 (gather_scalar_reductions): Use it.
10878 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10879 (vect_finish_replace_stmt): Declare.
10880 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10881 (needs_fold_left_reduction_p): New function, split out from...
10882 (vect_is_simple_reduction): ...here. Accept reductions that
10883 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10884 (vect_force_simple_reduction): Also store the reduction type in
10885 the assignment's STMT_VINFO_REDUC_TYPE.
10886 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10887 (merge_with_identity): New function.
10888 (vect_expand_fold_left): Likewise.
10889 (vectorize_fold_left_reduction): Likewise.
10890 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10891 scalar phi in place for it. Check for target support and reject
10892 cases that would reassociate the operation. Defer the transform
10893 phase to vectorize_fold_left_reduction.
10894 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10895 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10896 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10900 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10901 call to ifc_temp_var.
10903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10904 Alan Hayward <alan.hayward@arm.com>
10905 David Sherwood <david.sherwood@arm.com>
10907 * target.def (legitimize_address_displacement): Take the original
10908 offset as a poly_int.
10909 * targhooks.h (default_legitimize_address_displacement): Update
10911 * targhooks.c (default_legitimize_address_displacement): Likewise.
10912 * doc/tm.texi: Regenerate.
10913 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10914 as an argument, moving assert of ad->disp == ad->disp_term to...
10915 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10916 Try calling targetm.legitimize_address_displacement before expanding
10917 the address rather than afterwards, and adjust for the new interface.
10918 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10919 Match the new hook interface. Handle SVE addresses.
10920 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10921 new hook interface.
10923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10925 * Makefile.in (OBJS): Add early-remat.o.
10926 * target.def (select_early_remat_modes): New hook.
10927 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10928 * doc/tm.texi: Regenerate.
10929 * targhooks.h (default_select_early_remat_modes): Declare.
10930 * targhooks.c (default_select_early_remat_modes): New function.
10931 * timevar.def (TV_EARLY_REMAT): New timevar.
10932 * passes.def (pass_early_remat): New pass.
10933 * tree-pass.h (make_pass_early_remat): Declare.
10934 * early-remat.c: New file.
10935 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10937 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10940 Alan Hayward <alan.hayward@arm.com>
10941 David Sherwood <david.sherwood@arm.com>
10943 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10944 vfm1 with a bound_epilog parameter.
10945 (vect_do_peeling): Update calls accordingly, and move the prologue
10946 call earlier in the function. Treat the base bound_epilog as 0 for
10947 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10948 this base when peeling for gaps.
10949 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10950 with fully-masked loops.
10951 (vect_estimate_min_profitable_iters): Handle the single peeled
10952 iteration in that case.
10954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10955 Alan Hayward <alan.hayward@arm.com>
10956 David Sherwood <david.sherwood@arm.com>
10958 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10959 single-element interleaving even if the size is not a power of 2.
10960 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10961 accesses for single-element interleaving if the group size is
10964 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10965 Alan Hayward <alan.hayward@arm.com>
10966 David Sherwood <david.sherwood@arm.com>
10968 * doc/md.texi (fold_extract_last_@var{m}): Document.
10969 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10970 * optabs.def (fold_extract_last_optab): New optab.
10971 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10972 * internal-fn.c (fold_extract_direct): New macro.
10973 (expand_fold_extract_optab_fn): Likewise.
10974 (direct_fold_extract_optab_supported_p): Likewise.
10975 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10976 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10977 EXTRACT_LAST_REDUCTION.
10978 (get_initial_def_for_reduction): Do not create an initial vector
10979 for EXTRACT_LAST_REDUCTION reductions.
10980 (vectorizable_reduction): Leave the scalar phi in place for
10981 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10982 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10983 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10984 transform phase to vectorizable_condition.
10985 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10987 (vect_finish_stmt_generation): ...here.
10988 (vect_finish_replace_stmt): New function.
10989 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10990 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10992 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10995 Alan Hayward <alan.hayward@arm.com>
10996 David Sherwood <david.sherwood@arm.com>
10998 * doc/md.texi (extract_last_@var{m}): Document.
10999 * optabs.def (extract_last_optab): New optab.
11000 * internal-fn.def (EXTRACT_LAST): New internal function.
11001 * internal-fn.c (cond_unary_direct): New macro.
11002 (expand_cond_unary_optab_fn): Likewise.
11003 (direct_cond_unary_optab_supported_p): Likewise.
11004 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11005 loops using EXTRACT_LAST.
11006 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11007 (extract_last_<mode>): ...this optab.
11008 (vec_extract<mode><Vel>): Update accordingly.
11010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11011 Alan Hayward <alan.hayward@arm.com>
11012 David Sherwood <david.sherwood@arm.com>
11014 * target.def (empty_mask_is_expensive): New hook.
11015 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11016 * doc/tm.texi: Regenerate.
11017 * targhooks.h (default_empty_mask_is_expensive): Declare.
11018 * targhooks.c (default_empty_mask_is_expensive): New function.
11019 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11020 if the target says that empty masks are expensive.
11021 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11023 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11026 Alan Hayward <alan.hayward@arm.com>
11027 David Sherwood <david.sherwood@arm.com>
11029 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11030 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11031 (vect_use_loop_mask_for_alignment_p): New function.
11032 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11033 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11034 niters_skip argument. Make sure that the first niters_skip elements
11035 of the first iteration are inactive.
11036 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11037 Update call to vect_set_loop_masks_directly.
11038 (get_misalign_in_elems): New function, split out from...
11039 (vect_gen_prolog_loop_niters): ...here.
11040 (vect_update_init_of_dr): Take a code argument that specifies whether
11041 the adjustment should be added or subtracted.
11042 (vect_update_init_of_drs): Likewise.
11043 (vect_prepare_for_masked_peels): New function.
11044 (vect_do_peeling): Skip prologue peeling if we're using a mask
11045 instead. Update call to vect_update_inits_of_drs.
11046 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11048 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11049 alignment. Do not include the number of peeled iterations in
11050 the minimum threshold in that case.
11051 (vectorizable_induction): Adjust the start value down by
11052 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11053 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11054 Take the number of skipped iterations into account when calculating
11056 * tree-vect-stmts.c (vect_gen_while_not): New function.
11058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11059 Alan Hayward <alan.hayward@arm.com>
11060 David Sherwood <david.sherwood@arm.com>
11062 * doc/sourcebuild.texi (vect_fully_masked): Document.
11063 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11064 default value to 0.
11065 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11067 (vect_analyze_loop_2): ...here. Don't check the vectorization
11068 factor against the number of loop iterations if the loop is
11071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11072 Alan Hayward <alan.hayward@arm.com>
11073 David Sherwood <david.sherwood@arm.com>
11075 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11076 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11077 (dump_groups): Update accordingly.
11078 (iv_use::mem_type): New member variable.
11079 (address_p): New function.
11080 (record_use): Add a mem_type argument and initialize the new
11082 (record_group_use): Add a mem_type argument. Use address_p.
11083 Remove obsolete null checks of base_object. Update call to record_use.
11084 (find_interesting_uses_op): Update call to record_group_use.
11085 (find_interesting_uses_cond): Likewise.
11086 (find_interesting_uses_address): Likewise.
11087 (get_mem_type_for_internal_fn): New function.
11088 (find_address_like_use): Likewise.
11089 (find_interesting_uses_stmt): Try find_address_like_use before
11090 calling find_interesting_uses_op.
11091 (addr_offset_valid_p): Use the iv mem_type field as the type
11092 of the addressed memory.
11093 (add_autoinc_candidates): Likewise.
11094 (get_address_cost): Likewise.
11095 (split_small_address_groups_p): Use address_p.
11096 (split_address_groups): Likewise.
11097 (add_iv_candidate_for_use): Likewise.
11098 (autoinc_possible_for_pair): Likewise.
11099 (rewrite_groups): Likewise.
11100 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11101 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11102 (get_alias_ptr_type_for_ptr_address): New function.
11103 (rewrite_use_address): Rewrite address uses in calls that were
11104 identified by find_address_like_use.
11106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11107 Alan Hayward <alan.hayward@arm.com>
11108 David Sherwood <david.sherwood@arm.com>
11110 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11112 * gimple-expr.h (is_gimple_addressable: Likewise.
11113 * gimple-expr.c (is_gimple_address): Likewise.
11114 * internal-fn.c (expand_call_mem_ref): New function.
11115 (expand_mask_load_optab_fn): Use it.
11116 (expand_mask_store_optab_fn): Likewise.
11118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11119 Alan Hayward <alan.hayward@arm.com>
11120 David Sherwood <david.sherwood@arm.com>
11122 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11123 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11124 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11125 (cond_umax@var{mode}): Document.
11126 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11127 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11128 (cond_umin_optab, cond_umax_optab): New optabs.
11129 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11130 (COND_IOR, COND_XOR): New internal functions.
11131 * internal-fn.h (get_conditional_internal_fn): Declare.
11132 * internal-fn.c (cond_binary_direct): New macro.
11133 (expand_cond_binary_optab_fn): Likewise.
11134 (direct_cond_binary_optab_supported_p): Likewise.
11135 (get_conditional_internal_fn): New function.
11136 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11137 Cope with reduction statements that are vectorized as calls rather
11139 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11140 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11141 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11142 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11143 (UNSPEC_COND_EOR): New unspecs.
11144 (optab): Add mappings for them.
11145 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11146 (sve_int_op, sve_fp_op): New int attributes.
11148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11149 Alan Hayward <alan.hayward@arm.com>
11150 David Sherwood <david.sherwood@arm.com>
11152 * optabs.def (while_ult_optab): New optab.
11153 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11154 * internal-fn.def (WHILE_ULT): New internal function.
11155 * internal-fn.h (direct_internal_fn_supported_p): New override
11156 that takes two types as argument.
11157 * internal-fn.c (while_direct): New macro.
11158 (expand_while_optab_fn): New function.
11159 (convert_optab_supported_p): Likewise.
11160 (direct_while_optab_supported_p): New macro.
11161 * wide-int.h (wi::udiv_ceil): New function.
11162 * tree-vectorizer.h (rgroup_masks): New structure.
11163 (vec_loop_masks): New typedef.
11164 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11165 and fully_masked_p.
11166 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11167 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11168 (vect_max_vf): New function.
11169 (slpeel_make_loop_iterate_ntimes): Delete.
11170 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11171 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11172 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11173 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11174 internal-fn.h, stor-layout.h and optabs-query.h.
11175 (vect_set_loop_mask): New function.
11176 (add_preheader_seq): Likewise.
11177 (add_header_seq): Likewise.
11178 (interleave_supported_p): Likewise.
11179 (vect_maybe_permute_loop_masks): Likewise.
11180 (vect_set_loop_masks_directly): Likewise.
11181 (vect_set_loop_condition_masked): Likewise.
11182 (vect_set_loop_condition_unmasked): New function, split out from
11183 slpeel_make_loop_iterate_ntimes.
11184 (slpeel_make_loop_iterate_ntimes): Rename to..
11185 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11186 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11187 (vect_do_peeling): Update call accordingly.
11188 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11190 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11191 mask_compare_type, can_fully_mask_p and fully_masked_p.
11192 (release_vec_loop_masks): New function.
11193 (_loop_vec_info): Use it to free the loop masks.
11194 (can_produce_all_loop_masks_p): New function.
11195 (vect_get_max_nscalars_per_iter): Likewise.
11196 (vect_verify_full_masking): Likewise.
11197 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11198 retries, and free the mask rgroups before retrying. Check loop-wide
11199 reasons for disallowing fully-masked loops. Make the final decision
11200 about whether use a fully-masked loop or not.
11201 (vect_estimate_min_profitable_iters): Do not assume that peeling
11202 for the number of iterations will be needed for fully-masked loops.
11203 (vectorizable_reduction): Disable fully-masked loops.
11204 (vectorizable_live_operation): Likewise.
11205 (vect_halve_mask_nunits): New function.
11206 (vect_double_mask_nunits): Likewise.
11207 (vect_record_loop_mask): Likewise.
11208 (vect_get_loop_mask): Likewise.
11209 (vect_transform_loop): Handle the case in which the final loop
11210 iteration might handle a partial vector. Call vect_set_loop_condition
11211 instead of slpeel_make_loop_iterate_ntimes.
11212 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11213 (check_load_store_masking): New function.
11214 (prepare_load_store_mask): Likewise.
11215 (vectorizable_store): Handle fully-masked loops.
11216 (vectorizable_load): Likewise.
11217 (supportable_widening_operation): Use vect_halve_mask_nunits for
11219 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11220 (vect_gen_while): New function.
11221 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11222 (aarch64_uqdec<mode>): New insn.
11224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11225 Alan Hayward <alan.hayward@arm.com>
11226 David Sherwood <david.sherwood@arm.com>
11228 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11229 (reduc_xor_scal_optab): New optabs.
11230 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11231 (reduc_xor_scal_@var{m}): Document.
11232 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11233 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11234 internal functions.
11235 * fold-const-call.c (fold_const_call): Handle them.
11236 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11237 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11238 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11239 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11240 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11241 (UNSPEC_XORV): New unspecs.
11242 (optab): Add entries for them.
11243 (BITWISEV): New int iterator.
11244 (bit_reduc_op): New int attributes.
11246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11247 Alan Hayward <alan.hayward@arm.com>
11248 David Sherwood <david.sherwood@arm.com>
11250 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11251 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11252 * optabs.def (vec_shl_insert_optab): New optab.
11253 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11254 (duplicate_and_interleave): Likewise.
11255 * tree-vect-loop.c: Include internal-fn.h.
11256 (neutral_op_for_slp_reduction): New function, split out from
11257 get_initial_defs_for_reduction.
11258 (get_initial_def_for_reduction): Handle option 2 for variable-length
11259 vectors by loading the neutral value into a vector and then shifting
11260 the initial value into element 0.
11261 (get_initial_defs_for_reduction): Replace the code argument with
11262 the neutral value calculated by neutral_op_for_slp_reduction.
11263 Use gimple_build_vector for constant-length vectors.
11264 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11265 but the first group_size elements have a neutral value.
11266 Use duplicate_and_interleave otherwise.
11267 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11268 Update call to get_initial_defs_for_reduction. Handle SLP
11269 reductions for variable-length vectors by creating one vector
11270 result for each scalar result, with the elements associated
11271 with other scalar results stubbed out with the neutral value.
11272 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11273 Require IFN_VEC_SHL_INSERT for double reductions on
11274 variable-length vectors, or SLP reductions that have
11275 a neutral value. Require can_duplicate_and_interleave_p
11276 support for variable-length unchained SLP reductions if there
11277 is no neutral value, such as for MIN/MAX reductions. Also require
11278 the number of vector elements to be a multiple of the number of
11279 SLP statements when doing variable-length unchained SLP reductions.
11280 Update call to vect_create_epilog_for_reduction.
11281 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11282 and remove initial values.
11283 (duplicate_and_interleave): Make public.
11284 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11285 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11288 Alan Hayward <alan.hayward@arm.com>
11289 David Sherwood <david.sherwood@arm.com>
11291 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11292 (can_duplicate_and_interleave_p): New function.
11293 (vect_get_and_check_slp_defs): Take the vector of statements
11294 rather than just the current one. Remove excess parentheses.
11295 Restriction rejectinon of vect_constant_def and vect_external_def
11296 for variable-length vectors to boolean types, or types for which
11297 can_duplicate_and_interleave_p is false.
11298 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11299 (duplicate_and_interleave): New function.
11300 (vect_get_constant_vectors): Use gimple_build_vector for
11301 constant-length vectors and suitable variable-length constant
11302 vectors. Use duplicate_and_interleave for other variable-length
11303 vectors. Don't defer the update when inserting new statements.
11305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11306 Alan Hayward <alan.hayward@arm.com>
11307 David Sherwood <david.sherwood@arm.com>
11309 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11310 min_profitable_iters doesn't go negative.
11312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11313 Alan Hayward <alan.hayward@arm.com>
11314 David Sherwood <david.sherwood@arm.com>
11316 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11317 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11318 * optabs.def (vec_mask_load_lanes_optab): New optab.
11319 (vec_mask_store_lanes_optab): Likewise.
11320 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11321 (MASK_STORE_LANES): Likewise.
11322 * internal-fn.c (mask_load_lanes_direct): New macro.
11323 (mask_store_lanes_direct): Likewise.
11324 (expand_mask_load_optab_fn): Handle masked operations.
11325 (expand_mask_load_lanes_optab_fn): New macro.
11326 (expand_mask_store_optab_fn): Handle masked operations.
11327 (expand_mask_store_lanes_optab_fn): New macro.
11328 (direct_mask_load_lanes_optab_supported_p): Likewise.
11329 (direct_mask_store_lanes_optab_supported_p): Likewise.
11330 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11332 (vect_load_lanes_supported): Likewise.
11333 * tree-vect-data-refs.c (strip_conversion): New function.
11334 (can_group_stmts_p): Likewise.
11335 (vect_analyze_data_ref_accesses): Use it instead of checking
11336 for a pair of assignments.
11337 (vect_store_lanes_supported): Take a masked_p parameter.
11338 (vect_load_lanes_supported): Likewise.
11339 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11340 vect_store_lanes_supported and vect_load_lanes_supported.
11341 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11342 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11343 parameter. Don't allow gaps for masked accesses.
11344 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11345 and vect_load_lanes_supported.
11346 (get_load_store_type): Take a masked_p parameter and update
11347 call to get_group_load_store_type.
11348 (vectorizable_store): Update call to get_load_store_type.
11349 Handle IFN_MASK_STORE_LANES.
11350 (vectorizable_load): Update call to get_load_store_type.
11351 Handle IFN_MASK_LOAD_LANES.
11353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11354 Alan Hayward <alan.hayward@arm.com>
11355 David Sherwood <david.sherwood@arm.com>
11357 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11359 * config/aarch64/aarch64-protos.h
11360 (aarch64_sve_struct_memory_operand_p): Declare.
11361 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11362 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11363 (VPRED, vpred): Handle SVE structure modes.
11364 * config/aarch64/constraints.md (Utx): New constraint.
11365 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11366 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11367 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11368 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11369 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11370 structure modes. Split into pieces after RA.
11371 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11372 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11374 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11375 SVE structure modes.
11376 (aarch64_classify_address): Likewise.
11377 (sizetochar): Move earlier in file.
11378 (aarch64_print_operand): Handle SVE register lists.
11379 (aarch64_array_mode): New function.
11380 (aarch64_sve_struct_memory_operand_p): Likewise.
11381 (TARGET_ARRAY_MODE): Redefine.
11383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11384 Alan Hayward <alan.hayward@arm.com>
11385 David Sherwood <david.sherwood@arm.com>
11387 * target.def (array_mode): New target hook.
11388 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11389 * doc/tm.texi: Regenerate.
11390 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11391 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11392 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11393 targetm.array_mode.
11394 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11398 Alan Hayward <alan.hayward@arm.com>
11399 David Sherwood <david.sherwood@arm.com>
11401 * fold-const.c (fold_binary_loc): Check the argument types
11402 rather than the result type when testing for a vector operation.
11404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11406 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11407 * doc/tm.texi: Regenerate.
11409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11410 Alan Hayward <alan.hayward@arm.com>
11411 David Sherwood <david.sherwood@arm.com>
11413 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11414 (sve): Document new AArch64 extension.
11415 * doc/md.texi (w): Extend the description of the AArch64
11416 constraint to include SVE vectors.
11417 (Upl, Upa): Document new AArch64 predicate constraints.
11418 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11420 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11421 (msve-vector-bits=): New option.
11422 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11423 SVE when these are disabled.
11424 (sve): New extension.
11425 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11426 modes. Adjust their number of units based on aarch64_sve_vg.
11427 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11428 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11429 aarch64_addr_query_type.
11430 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11431 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11432 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11433 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11434 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11435 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11436 (aarch64_simd_imm_zero_p): Delete.
11437 (aarch64_check_zero_based_sve_index_immediate): Declare.
11438 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11439 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11440 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11441 (aarch64_sve_float_mul_immediate_p): Likewise.
11442 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11443 rather than an rtx.
11444 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11445 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11446 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11447 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11448 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11449 (aarch64_regmode_natural_size): Likewise.
11450 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11451 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11453 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11454 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11455 for VG and the SVE predicate registers.
11456 (V_ALIASES): Add a "z"-prefixed alias.
11457 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11458 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11459 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11460 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11461 (REG_CLASS_NAMES): Add entries for them.
11462 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11463 and the predicate registers.
11464 (aarch64_sve_vg): Declare.
11465 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11466 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11467 (REGMODE_NATURAL_SIZE): Define.
11468 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11470 * config/aarch64/aarch64.c: Include cfgrtl.h.
11471 (simd_immediate_info): Add a constructor for series vectors,
11472 and an associated step field.
11473 (aarch64_sve_vg): New variable.
11474 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11475 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11476 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11477 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11478 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11479 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11480 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11481 (aarch64_get_mask_mode): New functions.
11482 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11483 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11484 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11485 predicate modes and predicate registers. Explicitly restrict
11486 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11487 to store a vector mode if it is recognized by
11488 aarch64_classify_vector_mode.
11489 (aarch64_regmode_natural_size): New function.
11490 (aarch64_hard_regno_caller_save_mode): Return the original mode
11492 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11493 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11494 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11495 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11497 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11498 does not overlap dest if the function is frame-related. Handle
11500 (aarch64_split_add_offset): New function.
11501 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11502 them aarch64_add_offset.
11503 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11504 and update call to aarch64_sub_sp.
11505 (aarch64_add_cfa_expression): New function.
11506 (aarch64_expand_prologue): Pass extra temporary registers to the
11507 functions above. Handle the case in which we need to emit new
11508 DW_CFA_expressions for registers that were originally saved
11509 relative to the stack pointer, but now have to be expressed
11510 relative to the frame pointer.
11511 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11513 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11514 IP0 and IP1 values for SVE frames.
11515 (aarch64_expand_vec_series): New function.
11516 (aarch64_expand_sve_widened_duplicate): Likewise.
11517 (aarch64_expand_sve_const_vector): Likewise.
11518 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11519 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11520 into the register, rather than emitting a SET directly.
11521 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11522 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11523 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11524 (offset_9bit_signed_scaled_p): New functions.
11525 (aarch64_replicate_bitmask_imm): New function.
11526 (aarch64_bitmask_imm): Use it.
11527 (aarch64_cannot_force_const_mem): Reject expressions involving
11528 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11529 (aarch64_classify_index): Handle SVE indices, by requiring
11530 a plain register index with a scale that matches the element size.
11531 (aarch64_classify_address): Handle SVE addresses. Assert that
11532 the mode of the address is VOIDmode or an integer mode.
11533 Update call to aarch64_classify_symbol.
11534 (aarch64_classify_symbolic_expression): Update call to
11535 aarch64_classify_symbol.
11536 (aarch64_const_vec_all_in_range_p): New function.
11537 (aarch64_print_vector_float_operand): Likewise.
11538 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11539 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11540 and the FP immediates 1.0 and 0.5.
11541 (aarch64_print_address_internal): Handle SVE addresses.
11542 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11543 (aarch64_regno_regclass): Handle predicate registers.
11544 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11546 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11547 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11548 (aarch64_convert_sve_vector_bits): New function.
11549 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11550 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11551 rather than an rtx.
11552 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11553 Handle SVE vector and predicate modes. Accept VL-based constants
11554 that need only one temporary register, and VL offsets that require
11555 no temporary registers.
11556 (aarch64_conditional_register_usage): Mark the predicate registers
11557 as fixed if SVE isn't available.
11558 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11559 Return true for SVE vector and predicate modes.
11560 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11561 rather than an unsigned int. Handle SVE modes.
11562 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11564 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11566 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11567 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11568 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11569 (aarch64_sve_float_mul_immediate_p): New functions.
11570 (aarch64_sve_valid_immediate): New function.
11571 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11572 Explicitly reject structure modes. Check for INDEX constants.
11573 Handle PTRUE and PFALSE constants.
11574 (aarch64_check_zero_based_sve_index_immediate): New function.
11575 (aarch64_simd_imm_zero_p): Delete.
11576 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11577 vector modes. Accept constants in the range of CNT[BHWD].
11578 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11579 ask for an Advanced SIMD mode.
11580 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11581 (aarch64_simd_vector_alignment): Handle SVE predicates.
11582 (aarch64_vectorize_preferred_vector_alignment): New function.
11583 (aarch64_simd_vector_alignment_reachable): Use it instead of
11585 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11586 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11588 (MAX_VECT_LEN): Delete.
11589 (expand_vec_perm_d): Add a vec_flags field.
11590 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11591 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11592 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11594 (aarch64_evpc_rev): Rename to...
11595 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11596 (aarch64_evpc_rev_global): New function.
11597 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11598 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11600 (aarch64_evpc_sve_tbl): New function.
11601 (aarch64_expand_vec_perm_const_1): Update after rename of
11602 aarch64_evpc_rev. Handle SVE permutes too, trying
11603 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11604 than aarch64_evpc_tbl.
11605 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11606 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11607 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11608 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11609 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11610 (aarch64_expand_sve_vcond): New functions.
11611 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11612 of aarch64_vector_mode_p.
11613 (aarch64_dwarf_poly_indeterminate_value): New function.
11614 (aarch64_compute_pressure_classes): Likewise.
11615 (aarch64_can_change_mode_class): Likewise.
11616 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11617 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11618 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11619 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11620 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11621 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11622 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11623 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11625 (Dn, Dl, Dr): Accept const as well as const_vector.
11626 (Dz): Likewise. Compare against CONST0_RTX.
11627 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11628 of "vector" where appropriate.
11629 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11630 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11631 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11632 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11633 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11634 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11635 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11636 (v_int_equiv): Extend to SVE modes.
11637 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11639 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11640 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11641 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11642 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11643 (SVE_COND_FP_CMP): New int iterators.
11644 (perm_hilo): Handle the new unpack unspecs.
11645 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11647 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11648 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11649 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11650 (aarch64_equality_operator, aarch64_constant_vector_operand)
11651 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11652 (aarch64_sve_nonimmediate_operand): Likewise.
11653 (aarch64_sve_general_operand): Likewise.
11654 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11655 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11656 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11657 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11658 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11659 (aarch64_sve_float_arith_immediate): Likewise.
11660 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11661 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11662 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11663 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11664 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11665 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11666 (aarch64_sve_float_arith_operand): Likewise.
11667 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11668 (aarch64_sve_float_mul_operand): Likewise.
11669 (aarch64_sve_vec_perm_operand): Likewise.
11670 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11671 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11672 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11673 as well as const_vector.
11674 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11675 in file. Use CONST0_RTX and CONSTM1_RTX.
11676 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11677 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11678 Use aarch64_simd_imm_zero.
11679 * config/aarch64/aarch64-sve.md: New file.
11680 * config/aarch64/aarch64.md: Include it.
11681 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11682 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11683 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11684 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11685 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11686 (sve): New attribute.
11687 (enabled): Disable instructions with the sve attribute unless
11689 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11690 aarch64_expand_mov_immediate.
11691 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11692 CNT[BHSD] immediates.
11693 (movti): Split CONST_POLY_INT moves into two halves.
11694 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11695 Split additions that need a temporary here if the destination
11696 is the stack pointer.
11697 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11698 (*add<mode>3_poly_1): New instruction.
11699 (set_clobber_cc): New expander.
11701 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11703 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11704 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11705 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11706 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11707 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11708 Change innermode from fixed_mode_size to machine_mode.
11709 (simplify_subreg): Update call accordingly. Handle a constant-sized
11710 subreg of a variable-length CONST_VECTOR.
11712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11713 Alan Hayward <alan.hayward@arm.com>
11714 David Sherwood <david.sherwood@arm.com>
11716 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11717 (add_offset_to_base): New function, split out from...
11718 (create_mem_ref): ...here. When handling a scale other than 1,
11719 check first whether the address is valid without the offset.
11720 Add it into the base if so, leaving the index and scale as-is.
11722 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11725 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11726 fold_for_warn before checking if arg2 is INTEGER_CST.
11728 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11730 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11731 (store_multiple_operation): Delete.
11732 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11733 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11734 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11735 guarded by TARGET_STRING.
11736 (rs6000_output_load_multiple): Delete.
11737 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11738 OPTION_MASK_STRING / TARGET_STRING handling.
11739 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11740 (const rs6000_opt_masks) <"string">: Change mask to 0.
11741 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11742 (MASK_STRING): Delete.
11743 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11745 (load_multiple): Delete.
11752 (store_multiple): Delete.
11759 (movmemsi_8reg): Delete.
11760 (corresponding unnamed define_insn): Delete.
11761 (movmemsi_6reg): Delete.
11762 (corresponding unnamed define_insn): Delete.
11763 (movmemsi_4reg): Delete.
11764 (corresponding unnamed define_insn): Delete.
11765 (movmemsi_2reg): Delete.
11766 (corresponding unnamed define_insn): Delete.
11767 (movmemsi_1reg): Delete.
11768 (corresponding unnamed define_insn): Delete.
11769 * config/rs6000/rs6000.opt (mno-string): New.
11770 (mstring): Replace by deprecation warning stub.
11771 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11773 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11775 * regrename.c (regrename_do_replace): If replacing the same
11776 reg multiple times, try to reuse last created gen_raw_REG.
11779 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11780 main to workaround a bug in GDB.
11782 2018-01-12 Tom de Vries <tom@codesourcery.com>
11785 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11787 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11789 PR rtl-optimization/80481
11790 * ira-color.c (get_cap_member): New function.
11791 (allocnos_conflict_by_live_ranges_p): Use it.
11792 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11793 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11795 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11798 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11799 (*saddl_se_1): Ditto.
11800 (*ssubsi_1): Ditto.
11801 (*ssubl_se_1): Ditto.
11803 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11805 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11806 rather than wi::to_widest for DR_INITs.
11807 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11808 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11809 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11811 (vect_analyze_group_access_1): Note that here.
11813 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11815 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11816 polynomial type sizes.
11818 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11820 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11821 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11822 (gimple_add_tmp_var): Likewise.
11824 2018-01-12 Martin Liska <mliska@suse.cz>
11826 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11827 (gimple_alloc_sizes): Likewise.
11828 (dump_gimple_statistics): Use PRIu64 in printf format.
11829 * gimple.h: Change uint64_t to int.
11831 2018-01-12 Martin Liska <mliska@suse.cz>
11833 * tree-core.h: Use uint64_t instead of int.
11834 * tree.c (tree_node_counts): Likewise.
11835 (tree_node_sizes): Likewise.
11836 (dump_tree_statistics): Use PRIu64 in printf format.
11838 2018-01-12 Martin Liska <mliska@suse.cz>
11840 * Makefile.in: As qsort_chk is implemented in vec.c, add
11841 vec.o to linkage of gencfn-macros.
11842 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11843 passing the info to record_node_allocation_statistics.
11844 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11846 * ggc-common.c (struct ggc_usage): Add operator== and use
11847 it in operator< and compare function.
11848 * mem-stats.h (struct mem_usage): Likewise.
11849 * vec.c (struct vec_usage): Remove operator< and compare
11850 function. Can be simply inherited.
11852 2018-01-12 Martin Jambor <mjambor@suse.cz>
11855 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11856 * tree-ssa-math-opts.c: Include domwalk.h.
11857 (convert_mult_to_fma_1): New function.
11858 (fma_transformation_info): New type.
11859 (fma_deferring_state): Likewise.
11860 (cancel_fma_deferring): New function.
11861 (result_of_phi): Likewise.
11862 (last_fma_candidate_feeds_initial_phi): Likewise.
11863 (convert_mult_to_fma): Added deferring logic, split actual
11864 transformation to convert_mult_to_fma_1.
11865 (math_opts_dom_walker): New type.
11866 (math_opts_dom_walker::after_dom_children): New method, body moved
11867 here from pass_optimize_widening_mul::execute, added deferring logic
11869 (pass_optimize_widening_mul::execute): Moved most of code to
11870 math_opts_dom_walker::after_dom_children.
11871 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11872 * config/i386/i386.c (ix86_option_override_internal): Added
11873 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11875 2018-01-12 Richard Biener <rguenther@suse.de>
11878 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11879 inline instance vars.
11881 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11884 * config/rx/rx.c (rx_is_restricted_memory_address):
11885 Handle SUBREG case.
11887 2018-01-12 Richard Biener <rguenther@suse.de>
11889 PR tree-optimization/80846
11890 * target.def (split_reduction): New target hook.
11891 * targhooks.c (default_split_reduction): New function.
11892 * targhooks.h (default_split_reduction): Declare.
11893 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11894 target requests first reduce vectors by combining low and high
11896 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11897 (get_vectype_for_scalar_type_and_size): Export.
11898 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11899 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11900 * doc/tm.texi: Regenerate.
11901 * config/i386/i386.c (ix86_split_reduction): Implement
11902 TARGET_VECTORIZE_SPLIT_REDUCTION.
11904 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11907 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11908 in PIC mode except for TARGET_VXWORKS_RTP.
11909 * config/sparc/sparc.c: Include cfgrtl.h.
11910 (TARGET_INIT_PIC_REG): Define.
11911 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11912 (sparc_pic_register_p): New predicate.
11913 (sparc_legitimate_address_p): Use it.
11914 (sparc_legitimize_pic_address): Likewise.
11915 (sparc_delegitimize_address): Likewise.
11916 (sparc_mode_dependent_address_p): Likewise.
11917 (gen_load_pcrel_sym): Remove 4th parameter.
11918 (load_got_register): Adjust call to above. Remove obsolete stuff.
11919 (sparc_expand_prologue): Do not call load_got_register here.
11920 (sparc_flat_expand_prologue): Likewise.
11921 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11922 (sparc_use_pseudo_pic_reg): New function.
11923 (sparc_init_pic_reg): Likewise.
11924 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11925 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11927 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11929 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11930 Add item for branch_cost.
11932 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11934 PR rtl-optimization/83565
11935 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11936 not extend the result to a larger mode for rotate operations.
11937 (num_sign_bit_copies1): Likewise.
11939 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11942 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11944 Use values-Xc.o for -pedantic.
11945 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11947 2018-01-12 Martin Liska <mliska@suse.cz>
11950 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11952 (possible_polymorphic_call_targets): Use it.
11953 (ipa_devirt): Likewise.
11955 2018-01-12 Martin Liska <mliska@suse.cz>
11957 * profile-count.h (enum profile_quality): Use 0 as invalid
11958 enum value of profile_quality.
11960 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11962 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11963 -mext-string options.
11965 2018-01-12 Richard Biener <rguenther@suse.de>
11967 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11968 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11969 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11971 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11973 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11975 * configure.ac (--with-long-double-format): Add support for the
11976 configuration option to change the default long double format on
11978 * config.gcc (powerpc*-linux*-*): Likewise.
11979 * configure: Regenerate.
11980 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11981 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11982 used without modification.
11984 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11986 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11987 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11988 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11989 MISC_BUILTIN_SPEC_BARRIER.
11990 (rs6000_init_builtins): Likewise.
11991 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11993 (speculation_barrier): New define_insn.
11994 * doc/extend.texi: Document __builtin_speculation_barrier.
11996 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11999 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12000 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12001 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12003 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12004 integral modes instead of "ss" and "sd".
12005 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12006 vectors with 32-bit and 64-bit elements.
12007 (vecdupssescalarmodesuffix): New mode attribute.
12008 (vec_dup<mode>): Use it.
12010 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12013 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12014 frame if argument is passed on stack.
12016 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12019 * ree.c (combine_reaching_defs): Optimize also
12020 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12021 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12023 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12025 PR middle-end/83189
12026 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12028 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12030 PR middle-end/83718
12031 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12032 after they are computed.
12034 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12036 PR tree-optimization/83695
12037 * gimple-loop-linterchange.cc
12038 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12039 reset cached scev information after interchange.
12040 (pass_linterchange::execute): Remove call to scev_reset_htab.
12042 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12044 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12045 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12046 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12047 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12048 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12049 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12050 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12051 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12052 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12053 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12054 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12055 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12056 (V_lane_reg): Likewise.
12057 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12059 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12060 (vfmal_lane_low<mode>_intrinsic,
12061 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12062 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12063 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12064 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12065 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12066 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12068 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12070 * config/arm/arm-cpus.in (fp16fml): New feature.
12071 (ALL_SIMD): Add fp16fml.
12072 (armv8.2-a): Add fp16fml as an option.
12073 (armv8.3-a): Likewise.
12074 (armv8.4-a): Add fp16fml as part of fp16.
12075 * config/arm/arm.h (TARGET_FP16FML): Define.
12076 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12078 * config/arm/arm-modes.def (V2HF): Define.
12079 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12080 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12081 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12082 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12083 vfmsl_low, vfmsl_high): New set of builtins.
12084 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12085 (vfml_op): New code attribute.
12086 (VFMLHALVES): New int iterator.
12087 (VFML, VFMLSEL): New mode attributes.
12088 (V_reg): Define mapping for V2HF.
12089 (V_hi, V_lo): New mode attributes.
12090 (VF_constraint): Likewise.
12091 (vfml_half, vfml_half_selector): New int attributes.
12092 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12094 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12095 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12097 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12098 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12099 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12100 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12102 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12103 Document new effective target and option set.
12105 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12107 * config/arm/arm-cpus.in (armv8_4): New feature.
12108 (ARMv8_4a): New fgroup.
12109 (armv8.4-a): New arch.
12110 * config/arm/arm-tables.opt: Regenerate.
12111 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12112 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12113 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12114 Add matching rules for -march=armv8.4-a and extensions.
12115 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12117 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12120 * config/rx/rx.md (BW): New mode attribute.
12121 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12123 2018-01-11 Richard Biener <rguenther@suse.de>
12125 PR tree-optimization/83435
12126 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12127 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12128 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12130 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12131 Alan Hayward <alan.hayward@arm.com>
12132 David Sherwood <david.sherwood@arm.com>
12134 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12136 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12137 (aarch64_print_address_internal): Use it to check for a zero offset.
12139 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12140 Alan Hayward <alan.hayward@arm.com>
12141 David Sherwood <david.sherwood@arm.com>
12143 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12144 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12145 Return a poly_int64 rather than a HOST_WIDE_INT.
12146 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12147 rather than a HOST_WIDE_INT.
12148 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12149 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12150 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12151 final_offset from HOST_WIDE_INT to poly_int64.
12152 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12153 to_constant when getting the number of units in an Advanced SIMD
12155 (aarch64_builtin_vectorized_function): Check for a constant number
12157 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12159 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12160 attribute instead of GET_MODE_NUNITS.
12161 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12162 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12163 GET_MODE_SIZE for fixed-size registers.
12164 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12165 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12166 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12167 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12168 (aarch64_print_operand, aarch64_print_address_internal)
12169 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12170 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12171 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12172 Handle polynomial GET_MODE_SIZE.
12173 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12174 wider than SImode without modification.
12175 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12176 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12177 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12178 passing and returning SVE modes.
12179 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12180 rather than GEN_INT.
12181 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12182 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12183 (aarch64_allocate_and_probe_stack_space): Likewise.
12184 (aarch64_layout_frame): Cope with polynomial offsets.
12185 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12186 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12187 polynomial offsets.
12188 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12189 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12190 poly_int64 rather than a HOST_WIDE_INT.
12191 (aarch64_get_separate_components, aarch64_process_components)
12192 (aarch64_expand_prologue, aarch64_expand_epilogue)
12193 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12194 (aarch64_anchor_offset): New function, split out from...
12195 (aarch64_legitimize_address): ...here.
12196 (aarch64_builtin_vectorization_cost): Handle polynomial
12197 TYPE_VECTOR_SUBPARTS.
12198 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12200 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12201 number of elements from the PARALLEL rather than the mode.
12202 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12203 rather than GET_MODE_BITSIZE.
12204 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12205 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12206 (aarch64_expand_vec_perm_const_1): Handle polynomial
12207 d->perm.length () and d->perm elements.
12208 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12209 Apply to_constant to d->perm elements.
12210 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12211 polynomial CONST_VECTOR_NUNITS.
12212 (aarch64_move_pointer): Take amount as a poly_int64 rather
12214 (aarch64_progress_pointer): Avoid temporary variable.
12215 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12216 the mode attribute instead of GET_MODE.
12218 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12219 Alan Hayward <alan.hayward@arm.com>
12220 David Sherwood <david.sherwood@arm.com>
12222 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12223 x exists before using it.
12224 (aarch64_add_constant_internal): Rename to...
12225 (aarch64_add_offset_1): ...this. Replace regnum with separate
12226 src and dest rtxes. Handle the case in which they're different,
12227 including when the offset is zero. Replace scratchreg with an rtx.
12228 Use 2 additions if there is no spare register into which we can
12229 move a 16-bit constant.
12230 (aarch64_add_constant): Delete.
12231 (aarch64_add_offset): Replace reg with separate src and dest
12232 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12233 Use aarch64_add_offset_1.
12234 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12235 an rtx rather than an int. Take the delta as a poly_int64
12236 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12237 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12238 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12239 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12240 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12241 and aarch64_add_sp.
12242 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12243 aarch64_add_constant.
12245 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12247 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12248 Use scalar_float_mode.
12250 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12252 * config/aarch64/aarch64-simd.md
12253 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12254 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12255 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12256 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12257 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12258 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12259 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12260 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12261 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12262 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12264 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12267 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12268 targ_options->x_arm_arch_string is non NULL.
12270 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12272 * config/aarch64/aarch64.h
12273 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12275 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12278 * expmed.c (emit_store_flag_force): Swap if const op0
12279 and change VOIDmode to mode of op0.
12281 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12283 PR rtl-optimization/83761
12284 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12285 than bytes to mode_for_size.
12287 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12289 PR middle-end/83189
12290 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12291 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12294 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12296 PR middle-end/83575
12297 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12298 when in layout mode.
12299 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12300 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12303 2018-01-10 Michael Collison <michael.collison@arm.com>
12305 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12306 * config/aarch64/aarch64-option-extension.def: Add
12307 AARCH64_OPT_EXTENSION of 'fp16fml'.
12308 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12309 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12310 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12311 * config/aarch64/constraints.md (Ui7): New constraint.
12312 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12313 (VFMLA_SEL_W): Ditto.
12316 (VFMLA16_LOW): New int iterator.
12317 (VFMLA16_HIGH): Ditto.
12318 (UNSPEC_FMLAL): New unspec.
12319 (UNSPEC_FMLSL): Ditto.
12320 (UNSPEC_FMLAL2): Ditto.
12321 (UNSPEC_FMLSL2): Ditto.
12322 (f16mac): New code attribute.
12323 * config/aarch64/aarch64-simd-builtins.def
12324 (aarch64_fmlal_lowv2sf): Ditto.
12325 (aarch64_fmlsl_lowv2sf): Ditto.
12326 (aarch64_fmlalq_lowv4sf): Ditto.
12327 (aarch64_fmlslq_lowv4sf): Ditto.
12328 (aarch64_fmlal_highv2sf): Ditto.
12329 (aarch64_fmlsl_highv2sf): Ditto.
12330 (aarch64_fmlalq_highv4sf): Ditto.
12331 (aarch64_fmlslq_highv4sf): Ditto.
12332 (aarch64_fmlal_lane_lowv2sf): Ditto.
12333 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12334 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12335 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12336 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12337 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12338 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12339 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12340 (aarch64_fmlal_lane_highv2sf): Ditto.
12341 (aarch64_fmlsl_lane_highv2sf): Ditto.
12342 (aarch64_fmlal_laneq_highv2sf): Ditto.
12343 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12344 (aarch64_fmlalq_lane_highv4sf): Ditto.
12345 (aarch64_fmlsl_lane_highv4sf): Ditto.
12346 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12347 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12348 * config/aarch64/aarch64-simd.md:
12349 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12350 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12351 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12352 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12353 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12354 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12355 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12356 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12357 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12358 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12359 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12360 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12361 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12362 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12363 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12364 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12365 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12366 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12367 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12368 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12369 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12370 (vfmlsl_low_u32): Ditto.
12371 (vfmlalq_low_u32): Ditto.
12372 (vfmlslq_low_u32): Ditto.
12373 (vfmlal_high_u32): Ditto.
12374 (vfmlsl_high_u32): Ditto.
12375 (vfmlalq_high_u32): Ditto.
12376 (vfmlslq_high_u32): Ditto.
12377 (vfmlal_lane_low_u32): Ditto.
12378 (vfmlsl_lane_low_u32): Ditto.
12379 (vfmlal_laneq_low_u32): Ditto.
12380 (vfmlsl_laneq_low_u32): Ditto.
12381 (vfmlalq_lane_low_u32): Ditto.
12382 (vfmlslq_lane_low_u32): Ditto.
12383 (vfmlalq_laneq_low_u32): Ditto.
12384 (vfmlslq_laneq_low_u32): Ditto.
12385 (vfmlal_lane_high_u32): Ditto.
12386 (vfmlsl_lane_high_u32): Ditto.
12387 (vfmlal_laneq_high_u32): Ditto.
12388 (vfmlsl_laneq_high_u32): Ditto.
12389 (vfmlalq_lane_high_u32): Ditto.
12390 (vfmlslq_lane_high_u32): Ditto.
12391 (vfmlalq_laneq_high_u32): Ditto.
12392 (vfmlslq_laneq_high_u32): Ditto.
12393 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12394 (AARCH64_FL_FOR_ARCH8_4): New.
12395 (AARCH64_ISA_F16FML): New ISA flag.
12396 (TARGET_F16FML): New feature flag for fp16fml.
12397 (doc/invoke.texi): Document new fp16fml option.
12399 2018-01-10 Michael Collison <michael.collison@arm.com>
12401 * config/aarch64/aarch64-builtins.c:
12402 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12403 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12404 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12405 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12406 (AARCH64_ISA_SHA3): New ISA flag.
12407 (TARGET_SHA3): New feature flag for sha3.
12408 * config/aarch64/iterators.md (sha512_op): New int attribute.
12409 (CRYPTO_SHA512): New int iterator.
12410 (UNSPEC_SHA512H): New unspec.
12411 (UNSPEC_SHA512H2): Ditto.
12412 (UNSPEC_SHA512SU0): Ditto.
12413 (UNSPEC_SHA512SU1): Ditto.
12414 * config/aarch64/aarch64-simd-builtins.def
12415 (aarch64_crypto_sha512hqv2di): New builtin.
12416 (aarch64_crypto_sha512h2qv2di): Ditto.
12417 (aarch64_crypto_sha512su0qv2di): Ditto.
12418 (aarch64_crypto_sha512su1qv2di): Ditto.
12419 (aarch64_eor3qv8hi): Ditto.
12420 (aarch64_rax1qv2di): Ditto.
12421 (aarch64_xarqv2di): Ditto.
12422 (aarch64_bcaxqv8hi): Ditto.
12423 * config/aarch64/aarch64-simd.md:
12424 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12425 (aarch64_crypto_sha512su0qv2di): Ditto.
12426 (aarch64_crypto_sha512su1qv2di): Ditto.
12427 (aarch64_eor3qv8hi): Ditto.
12428 (aarch64_rax1qv2di): Ditto.
12429 (aarch64_xarqv2di): Ditto.
12430 (aarch64_bcaxqv8hi): Ditto.
12431 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12432 (vsha512h2q_u64): Ditto.
12433 (vsha512su0q_u64): Ditto.
12434 (vsha512su1q_u64): Ditto.
12435 (veor3q_u16): Ditto.
12436 (vrax1q_u64): Ditto.
12437 (vxarq_u64): Ditto.
12438 (vbcaxq_u16): Ditto.
12439 * config/arm/types.md (crypto_sha512): New type attribute.
12440 (crypto_sha3): Ditto.
12441 (doc/invoke.texi): Document new sha3 option.
12443 2018-01-10 Michael Collison <michael.collison@arm.com>
12445 * config/aarch64/aarch64-builtins.c:
12446 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12447 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12448 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12449 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12450 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12451 (AARCH64_ISA_SM4): New ISA flag.
12452 (TARGET_SM4): New feature flag for sm4.
12453 * config/aarch64/aarch64-simd-builtins.def
12454 (aarch64_sm3ss1qv4si): Ditto.
12455 (aarch64_sm3tt1aq4si): Ditto.
12456 (aarch64_sm3tt1bq4si): Ditto.
12457 (aarch64_sm3tt2aq4si): Ditto.
12458 (aarch64_sm3tt2bq4si): Ditto.
12459 (aarch64_sm3partw1qv4si): Ditto.
12460 (aarch64_sm3partw2qv4si): Ditto.
12461 (aarch64_sm4eqv4si): Ditto.
12462 (aarch64_sm4ekeyqv4si): Ditto.
12463 * config/aarch64/aarch64-simd.md:
12464 (aarch64_sm3ss1qv4si): Ditto.
12465 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12466 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12467 (aarch64_sm4eqv4si): Ditto.
12468 (aarch64_sm4ekeyqv4si): Ditto.
12469 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12470 (sm3part_op): Ditto.
12471 (CRYPTO_SM3TT): Ditto.
12472 (CRYPTO_SM3PART): Ditto.
12473 (UNSPEC_SM3SS1): New unspec.
12474 (UNSPEC_SM3TT1A): Ditto.
12475 (UNSPEC_SM3TT1B): Ditto.
12476 (UNSPEC_SM3TT2A): Ditto.
12477 (UNSPEC_SM3TT2B): Ditto.
12478 (UNSPEC_SM3PARTW1): Ditto.
12479 (UNSPEC_SM3PARTW2): Ditto.
12480 (UNSPEC_SM4E): Ditto.
12481 (UNSPEC_SM4EKEY): Ditto.
12482 * config/aarch64/constraints.md (Ui2): New constraint.
12483 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12484 * config/arm/types.md (crypto_sm3): New type attribute.
12485 (crypto_sm4): Ditto.
12486 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12487 (vsm3tt1aq_u32): Ditto.
12488 (vsm3tt1bq_u32): Ditto.
12489 (vsm3tt2aq_u32): Ditto.
12490 (vsm3tt2bq_u32): Ditto.
12491 (vsm3partw1q_u32): Ditto.
12492 (vsm3partw2q_u32): Ditto.
12493 (vsm4eq_u32): Ditto.
12494 (vsm4ekeyq_u32): Ditto.
12495 (doc/invoke.texi): Document new sm4 option.
12497 2018-01-10 Michael Collison <michael.collison@arm.com>
12499 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12500 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12501 (AARCH64_FL_FOR_ARCH8_4): New.
12502 (AARCH64_FL_V8_4): New flag.
12503 (doc/invoke.texi): Document new armv8.4-a option.
12505 2018-01-10 Michael Collison <michael.collison@arm.com>
12507 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12508 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12509 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12510 * config/aarch64/aarch64-option-extension.def: Add
12511 AARCH64_OPT_EXTENSION of 'sha2'.
12512 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12513 (crypto): Disable sha2 and aes if crypto disabled.
12514 (crypto): Enable aes and sha2 if enabled.
12515 (simd): Disable sha2 and aes if simd disabled.
12516 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12518 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12519 (TARGET_SHA2): New feature flag for sha2.
12520 (TARGET_AES): New feature flag for aes.
12521 * config/aarch64/aarch64-simd.md:
12522 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12523 conditional on TARGET_AES.
12524 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12525 (aarch64_crypto_sha1hsi): Make pattern conditional
12527 (aarch64_crypto_sha1hv4si): Ditto.
12528 (aarch64_be_crypto_sha1hv4si): Ditto.
12529 (aarch64_crypto_sha1su1v4si): Ditto.
12530 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12531 (aarch64_crypto_sha1su0v4si): Ditto.
12532 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12533 (aarch64_crypto_sha256su0v4si): Ditto.
12534 (aarch64_crypto_sha256su1v4si): Ditto.
12535 (doc/invoke.texi): Document new aes and sha2 options.
12537 2018-01-10 Martin Sebor <msebor@redhat.com>
12539 PR tree-optimization/83781
12540 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12543 2018-01-11 Martin Sebor <msebor@gmail.com>
12544 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12546 PR tree-optimization/83501
12547 PR tree-optimization/81703
12549 * tree-ssa-strlen.c (get_string_cst): Rename...
12550 (get_string_len): ...to this. Handle global constants.
12551 (handle_char_store): Adjust.
12553 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12554 Jim Wilson <jimw@sifive.com>
12556 * config/riscv/riscv-protos.h (riscv_output_return): New.
12557 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12558 (riscv_attribute_table, riscv_output_return),
12559 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12560 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12561 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12563 (riscv_expand_prologue): Add early return for naked function.
12564 (riscv_expand_epilogue): Likewise.
12565 (riscv_function_ok_for_sibcall): Return false for naked function.
12566 (riscv_set_current_function): New.
12567 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12568 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12569 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12570 * doc/extend.texi (RISC-V Function Attributes): New.
12572 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12574 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12575 check for 128-bit long double before checking TCmode.
12576 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12577 128-bit long doubles before checking TFmode or TCmode.
12578 (FLOAT128_IBM_P): Likewise.
12580 2018-01-10 Martin Sebor <msebor@redhat.com>
12582 PR tree-optimization/83671
12583 * builtins.c (c_strlen): Unconditionally return zero for the empty
12585 Use -Warray-bounds for warnings.
12586 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12587 for non-constant array indices with COMPONENT_REF, arrays of
12588 arrays, and pointers to arrays.
12589 (gimple_fold_builtin_strlen): Determine and set length range for
12590 non-constant character arrays.
12592 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12594 PR middle-end/81897
12595 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12598 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12600 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12602 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12605 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12606 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12607 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12608 indexed_or_indirect_operand predicate.
12609 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12610 (*vsx_le_perm_load_v8hi): Likewise.
12611 (*vsx_le_perm_load_v16qi): Likewise.
12612 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12613 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12614 (*vsx_le_perm_store_v8hi): Likewise.
12615 (*vsx_le_perm_store_v16qi): Likewise.
12616 (eight unnamed splitters): Likewise.
12618 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12620 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12621 * config/rs6000/emmintrin.h: Likewise.
12622 * config/rs6000/mmintrin.h: Likewise.
12623 * config/rs6000/xmmintrin.h: Likewise.
12625 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12628 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12630 * tree.c (tree_nop_conversion): Return true for location wrapper
12632 (maybe_wrap_with_location): New function.
12633 (selftest::check_strip_nops): New function.
12634 (selftest::test_location_wrappers): New function.
12635 (selftest::tree_c_tests): Call it.
12636 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12637 (maybe_wrap_with_location): New decl.
12638 (EXPR_LOCATION_WRAPPER_P): New macro.
12639 (location_wrapper_p): New inline function.
12640 (tree_strip_any_location_wrapper): New inline function.
12642 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12645 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12646 stack_realign_offset for the largest alignment of stack slot
12648 (ix86_find_max_used_stack_alignment): New function.
12649 (ix86_finalize_stack_frame_flags): Use it. Set
12650 max_used_stack_alignment if we don't realign stack.
12651 * config/i386/i386.h (machine_function): Add
12652 max_used_stack_alignment.
12654 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12656 * config/arm/arm.opt (-mbranch-cost): New option.
12657 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12660 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12663 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12664 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12666 2018-01-10 Richard Biener <rguenther@suse.de>
12669 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12670 early out so it also covers the case where we have a non-NULL
12673 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12675 PR tree-optimization/83753
12676 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12677 for non-strided grouped accesses if the number of elements is 1.
12679 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12682 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12683 * i386.h (TARGET_USE_GATHER): Define.
12684 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12686 2018-01-10 Martin Liska <mliska@suse.cz>
12689 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12690 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12692 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12693 CLEANUP_NO_PARTITIONING is not set.
12695 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12697 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12698 for vectors, as a partial revert of r254296.
12699 * rtl.h (const_vec_p): Delete.
12700 (const_vec_duplicate_p): Don't test for vector CONSTs.
12701 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12702 * expmed.c (make_tree): Likewise.
12705 * common.md (E, F): Use CONSTANT_P instead of checking for
12707 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12708 checking for CONST_VECTOR.
12710 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12712 PR middle-end/83575
12713 * predict.c (force_edge_cold): Handle in more sane way edges
12714 with no prediction.
12716 2018-01-09 Carl Love <cel@us.ibm.com>
12718 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12720 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12721 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12722 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12723 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12724 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12725 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12726 * config/rs6000/rs6000-protos.h: Add extern defition for
12727 rs6000_generate_float2_double_code.
12728 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12730 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12731 (float2_v2df): Add define_expand.
12733 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12736 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12737 op_mode in the force_to_mode call.
12739 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12741 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12742 instead of checking each element individually.
12743 (aarch64_evpc_uzp): Likewise.
12744 (aarch64_evpc_zip): Likewise.
12745 (aarch64_evpc_ext): Likewise.
12746 (aarch64_evpc_rev): Likewise.
12747 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12748 instead of checking each element individually. Return true without
12750 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12751 whether all selected elements come from the same input, instead of
12752 checking each element individually. Remove calls to gen_rtx_REG,
12753 start_sequence and end_sequence and instead assert that no rtl is
12756 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12758 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12759 order of HIGH and CONST checks.
12761 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12763 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12764 if the destination isn't an SSA_NAME.
12766 2018-01-09 Richard Biener <rguenther@suse.de>
12768 PR tree-optimization/83668
12769 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12771 (canonicalize_loop_form): ... here, renamed from ...
12772 (canonicalize_loop_closed_ssa_form): ... this and amended to
12773 swap successor edges for loop exit blocks to make us use
12774 the RPO order we need for initial schedule generation.
12776 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12778 PR tree-optimization/64811
12779 * match.pd: When optimizing comparisons with Inf, avoid
12780 introducing or losing exceptions from comparisons with NaN.
12782 2018-01-09 Martin Liska <mliska@suse.cz>
12785 * asan.c (shadow_mem_size): Add gcc_assert.
12787 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12789 Don't save registers in main().
12792 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12793 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12794 * config/avr/avr.c (avr_set_current_function): Don't error if
12795 naked, OS_task or OS_main are specified at the same time.
12796 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12798 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12800 * common/config/avr/avr-common.c (avr_option_optimization_table):
12801 Switch on -mmain-is-OS_task for optimizing compilations.
12803 2018-01-09 Richard Biener <rguenther@suse.de>
12805 PR tree-optimization/83572
12806 * graphite.c: Include cfganal.h.
12807 (graphite_transform_loops): Connect infinite loops to exit
12808 and remove fake edges at the end.
12810 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12812 * ipa-inline.c (edge_badness): Revert accidental checkin.
12814 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12817 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12818 symbols; not inline clones.
12820 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12823 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12824 hard registers. Formatting fixes.
12826 PR preprocessor/83722
12827 * gcc.c (try_generate_repro): Pass
12828 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12829 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12832 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12833 Kito Cheng <kito.cheng@gmail.com>
12835 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12836 (riscv_leaf_function_p): Delete.
12837 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12839 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12841 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12843 (do_ifelse): New function.
12844 (do_isel): New function.
12845 (do_sub3): New function.
12846 (do_add3): New function.
12847 (do_load_mask_compare): New function.
12848 (do_overlap_load_compare): New function.
12849 (expand_compare_loop): New function.
12850 (expand_block_compare): Call expand_compare_loop() when appropriate.
12851 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12852 option description.
12853 (-mblock-compare-inline-loop-limit): New option.
12855 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12858 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12859 Reverse order of second and third operands in first alternative.
12860 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12861 of first and second elements in UNSPEC_VPERMR vector.
12862 (altivec_expand_vec_perm_le): Likewise.
12864 2018-01-08 Jeff Law <law@redhat.com>
12866 PR rtl-optimizatin/81308
12867 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12868 (process_switch): If group_case_labels makes a change, then set
12870 (pass_convert_switch::execute): If a switch is converted, then
12871 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12873 PR rtl-optimization/81308
12874 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12877 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12879 PR target/83663 - Revert r255946
12880 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12881 generation for cases where splatting a value is not useful.
12882 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12883 across a vec_duplicate and a paradoxical subreg forming a vector
12884 mode to a vec_concat.
12886 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12888 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12889 -march=armv8.3-a variants.
12890 * config/arm/t-multilib: Likewise.
12891 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12893 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12895 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12897 (cceq_ior_compare_complement): Give it a name so I can use it, and
12898 change boolean_or_operator predicate to boolean_operator so it can
12899 be used to generate a crand.
12900 (eqne): New code iterator.
12901 (bd/bd_neg): New code_attrs.
12902 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12903 a single define_insn.
12904 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12905 decrement (bdnzt/bdnzf/bdzt/bdzf).
12906 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12907 with the new names of the branch decrement patterns, and added the
12908 names of the branch decrement conditional patterns.
12910 2018-01-08 Richard Biener <rguenther@suse.de>
12912 PR tree-optimization/83563
12913 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12916 2018-01-08 Richard Biener <rguenther@suse.de>
12918 PR middle-end/83713
12919 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12921 2018-01-08 Richard Biener <rguenther@suse.de>
12923 PR tree-optimization/83685
12924 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12925 references to abnormals.
12927 2018-01-08 Richard Biener <rguenther@suse.de>
12930 * dwarf2out.c (output_indirect_strings): Handle empty
12931 skeleton_debug_str_hash.
12932 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12934 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12936 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12937 (emit_store_direct): Likewise.
12938 (arc_trampoline_adjust_address): Likewise.
12939 (arc_asm_trampoline_template): New function.
12940 (arc_initialize_trampoline): Use asm_trampoline_template.
12941 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12942 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12943 * config/arc/arc.md (flush_icache): Delete pattern.
12945 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12947 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12948 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12951 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12954 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12955 by not USED_FOR_TARGET.
12956 (make_pass_resolve_sw_modes): Likewise.
12958 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12960 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12963 2018-01-08 Richard Biener <rguenther@suse.de>
12965 PR middle-end/83580
12966 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12968 2018-01-08 Richard Biener <rguenther@suse.de>
12970 PR middle-end/83517
12971 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12973 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12975 PR middle-end/81897
12976 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12977 basic blocks with a small number of successors.
12978 (convert_control_dep_chain_into_preds): Improve handling of
12980 (dump_predicates): Split apart into...
12981 (dump_pred_chain): ...here...
12982 (dump_pred_info): ...and here.
12983 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12984 (can_chain_union_be_invalidated_p): Improve check for invalidation
12986 (uninit_uses_cannot_happen): Avoid unnecessary if
12987 convert_control_dep_chain_into_preds yielded nothing.
12989 2018-01-06 Martin Sebor <msebor@redhat.com>
12991 PR tree-optimization/83640
12992 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12993 subtracting negative offset from size.
12994 (builtin_access::overlap): Adjust offset bounds of the access to fall
12995 within the size of the object if possible.
12997 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12999 PR rtl-optimization/83699
13000 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13001 extract_bit_field_as_subreg to cases in which the extracted
13002 value is also a vector.
13004 * lra-constraints.c (process_alt_operands): Test for the equivalence
13005 substitutions when detecting a possible reload cycle.
13007 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13010 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13011 by default if flag_selective_schedling{,2}. Formatting fixes.
13013 PR rtl-optimization/83682
13014 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13015 if it has non-VECTOR_MODE element mode.
13016 (vec_duplicate_p): Likewise.
13018 PR middle-end/83694
13019 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13020 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13022 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13025 * config/i386/i386-builtin.def
13026 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13027 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13028 Require also OPTION_MASK_ISA_AVX512F in addition to
13029 OPTION_MASK_ISA_GFNI.
13030 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13031 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13032 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13033 to OPTION_MASK_ISA_GFNI.
13034 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13035 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13036 OPTION_MASK_ISA_AVX512BW.
13037 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13038 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13039 addition to OPTION_MASK_ISA_GFNI.
13040 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13041 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13042 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13043 to OPTION_MASK_ISA_GFNI.
13044 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13045 a requirement for all ISAs rather than any of them with a few
13047 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13049 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13050 bitmasks to be enabled with 3 exceptions, instead of requiring any
13051 enabled ISA with lots of exceptions.
13052 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13053 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13054 Change avx512bw in isa attribute to avx512f.
13055 * config/i386/sgxintrin.h: Add license boilerplate.
13056 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13057 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13058 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13059 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13061 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13062 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13063 temporarily sse2 rather than sse if not enabled already.
13066 * config/i386/sse.md (VI248_VLBW): Rename to ...
13067 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13068 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13069 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13070 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13071 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13072 mode iterator instead of VI248_VLBW.
13074 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13076 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13077 (record_modified): Skip clobbers; add debug output.
13078 (param_change_prob): Use sreal frequencies.
13080 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13082 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13083 punt for user-aligned variables.
13085 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13087 * tree-chrec.c (chrec_contains_symbols): Return true for
13090 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13093 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13094 of (x|y) == x for BICS pattern.
13096 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13098 PR tree-optimization/83605
13099 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13100 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13103 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13105 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13106 * config/epiphany/rtems.h: New file.
13108 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13109 Uros Bizjak <ubizjak@gmail.com>
13112 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13113 QIreg_operand instead of register_operand predicate.
13114 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13115 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13116 comments instead of -fmitigate[-_]rop.
13118 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13121 * cgraphunit.c (symbol_table::compile): Switch to text_section
13122 before calling assembly_start debug hook.
13123 * run-rtl-passes.c (run_rtl_passes): Likewise.
13126 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13128 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13129 range_int_cst_p rather than !symbolic_range_p before calling
13130 extract_range_from_multiplicative_op_1.
13132 2018-01-04 Jeff Law <law@redhat.com>
13134 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13135 redundant test in assertion.
13137 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13139 * doc/rtl.texi: Document machine_mode wrapper classes.
13141 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13143 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13144 using tree_to_uhwi.
13146 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13148 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13149 the VEC_PERM_EXPR fold to fail.
13151 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13154 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13155 to switched_sections.
13157 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13160 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13161 test for d.testing.
13163 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13166 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13167 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13169 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13172 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13173 is BLKmode and bitpos not zero or mode change is needed.
13175 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13178 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13181 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13184 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13185 instead of MULT rtx. Update all corresponding splitters.
13186 (*saddl_se): Ditto.
13187 (*ssub<modesuffix>): Ditto.
13188 (*ssubl_se): Ditto.
13189 (*cmp_sadd_di): Update split patterns.
13190 (*cmp_sadd_si): Ditto.
13191 (*cmp_sadd_sidi): Ditto.
13192 (*cmp_ssub_di): Ditto.
13193 (*cmp_ssub_si): Ditto.
13194 (*cmp_ssub_sidi): Ditto.
13195 * config/alpha/predicates.md (const23_operand): New predicate.
13196 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13197 Look for ASHIFT, not MULT inner operand.
13198 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13200 2018-01-04 Martin Liska <mliska@suse.cz>
13202 PR gcov-profile/83669
13203 * gcov.c (output_intermediate_file): Add version to intermediate
13205 * doc/gcov.texi: Document new field 'version' in intermediate
13206 file format. Fix location of '-k' option of gcov command.
13208 2018-01-04 Martin Liska <mliska@suse.cz>
13211 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13213 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13215 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13217 2018-01-03 Martin Sebor <msebor@redhat.com>
13219 PR tree-optimization/83655
13220 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13221 checking calls with invalid arguments.
13223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13225 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13226 (vectorizable_mask_load_store): Delete.
13227 (vectorizable_call): Return false for masked loads and stores.
13228 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13229 instead of gimple_assign_rhs1.
13230 (vectorizable_load): Handle IFN_MASK_LOAD.
13231 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13235 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13237 (vectorizable_mask_load_store): ...here.
13238 (vectorizable_load): ...and here.
13240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13242 * tree-vect-stmts.c (vect_build_all_ones_mask)
13243 (vect_build_zero_merge_argument): New functions, split out from...
13244 (vectorizable_load): ...here.
13246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13248 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13250 (vectorizable_mask_load_store): ...here.
13251 (vectorizable_store): ...and here.
13253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13255 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13257 (vectorizable_mask_load_store): ...here.
13259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13261 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13262 (vect_model_store_cost): Take a vec_load_store_type instead of a
13264 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13265 (vect_model_store_cost): Take a vec_load_store_type instead of a
13267 (vectorizable_mask_load_store): Update accordingly.
13268 (vectorizable_store): Likewise.
13269 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13273 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13274 IFN_MASK_LOAD calls here rather than...
13275 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13278 Alan Hayward <alan.hayward@arm.com>
13279 David Sherwood <david.sherwood@arm.com>
13281 * expmed.c (extract_bit_field_1): For vector extracts,
13282 fall back to extract_bit_field_as_subreg if vec_extract
13285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13286 Alan Hayward <alan.hayward@arm.com>
13287 David Sherwood <david.sherwood@arm.com>
13289 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13290 they are variable or constant sized.
13291 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13292 slots for constant-sized data.
13294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13295 Alan Hayward <alan.hayward@arm.com>
13296 David Sherwood <david.sherwood@arm.com>
13298 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13299 handling COND_EXPRs with boolean comparisons, try to find a better
13300 basis for the mask type than the boolean itself.
13302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13304 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13305 is calculated and how it can be overridden.
13306 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13307 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13309 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13313 Alan Hayward <alan.hayward@arm.com>
13314 David Sherwood <david.sherwood@arm.com>
13316 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13317 Remove the mode argument.
13318 (aarch64_simd_valid_immediate): Remove the mode and inverse
13320 * config/aarch64/iterators.md (bitsize): New iterator.
13321 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13322 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13323 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13324 aarch64_simd_valid_immediate.
13325 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13326 (aarch64_reg_or_bic_imm): Likewise.
13327 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13328 with an insn_type enum and msl with a modifier_type enum.
13329 Replace element_width with a scalar_mode. Change the shift
13330 to unsigned int. Add constructors for scalar_float_mode and
13331 scalar_int_mode elements.
13332 (aarch64_vect_float_const_representable_p): Delete.
13333 (aarch64_can_const_movi_rtx_p)
13334 (aarch64_simd_scalar_immediate_valid_for_move)
13335 (aarch64_simd_make_constant): Update call to
13336 aarch64_simd_valid_immediate.
13337 (aarch64_advsimd_valid_immediate_hs): New function.
13338 (aarch64_advsimd_valid_immediate): Likewise.
13339 (aarch64_simd_valid_immediate): Remove mode and inverse
13340 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13341 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13342 and aarch64_float_const_representable_p on the result.
13343 (aarch64_output_simd_mov_immediate): Remove mode argument.
13344 Update call to aarch64_simd_valid_immediate and use of
13345 simd_immediate_info.
13346 (aarch64_output_scalar_simd_mov_immediate): Update call
13349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13350 Alan Hayward <alan.hayward@arm.com>
13351 David Sherwood <david.sherwood@arm.com>
13353 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13354 (mode_nunits): Likewise CONST_MODE_NUNITS.
13355 * machmode.def (ADJUST_NUNITS): Document.
13356 * genmodes.c (mode_data::need_nunits_adj): New field.
13357 (blank_mode): Update accordingly.
13358 (adj_nunits): New variable.
13359 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13361 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13362 listed in adj_nunits.
13363 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13364 listed in adj_nunits. Don't emit case statements for such modes.
13365 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13366 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13367 nothing if adj_nunits is nonnull.
13368 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13369 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13370 (emit_mode_fbit): Update use of print_maybe_const_decl.
13371 (emit_move_size): Likewise. Treat the array as non-const
13373 (emit_mode_adjustments): Handle adj_nunits.
13375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13377 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13378 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13379 (VECTOR_MODES): Use it.
13380 (make_vector_modes): Take the prefix as an argument.
13382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13383 Alan Hayward <alan.hayward@arm.com>
13384 David Sherwood <david.sherwood@arm.com>
13386 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13387 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13388 for MODE_VECTOR_BOOL.
13389 * machmode.def (VECTOR_BOOL_MODE): Document.
13390 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13391 (make_vector_bool_mode): New function.
13392 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13394 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13395 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13397 * stor-layout.c (int_mode_for_mode): Likewise.
13398 * tree.c (build_vector_type_for_mode): Likewise.
13399 * varasm.c (output_constant_pool_2): Likewise.
13400 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13401 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13402 for MODE_VECTOR_BOOL.
13403 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13404 of mode class checks.
13405 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13406 instead of a list of mode class checks.
13407 (expand_vector_scalar_condition): Likewise.
13408 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13411 Alan Hayward <alan.hayward@arm.com>
13412 David Sherwood <david.sherwood@arm.com>
13414 * machmode.h (mode_size): Change from unsigned short to
13416 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13417 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13418 or if measurement_type is not polynomial.
13419 (fixed_size_mode::includes_p): Check for constant-sized modes.
13420 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13421 return a poly_uint16 rather than an unsigned short.
13422 (emit_mode_size): Change the type of mode_size from unsigned short
13423 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13424 (emit_mode_adjustments): Cope with polynomial vector sizes.
13425 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13427 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13429 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13430 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13431 * caller-save.c (setup_save_areas): Likewise.
13432 (replace_reg_with_saved_mem): Likewise.
13433 * calls.c (emit_library_call_value_1): Likewise.
13434 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13435 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13436 (gen_lowpart_for_combine): Likewise.
13437 * convert.c (convert_to_integer_1): Likewise.
13438 * cse.c (equiv_constant, cse_insn): Likewise.
13439 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13440 (cselib_subst_to_values): Likewise.
13441 * dce.c (word_dce_process_block): Likewise.
13442 * df-problems.c (df_word_lr_mark_ref): Likewise.
13443 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13444 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13445 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13446 (rtl_for_decl_location): Likewise.
13447 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13448 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13449 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13450 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13451 (expand_expr_real_1): Likewise.
13452 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13453 (pad_below): Likewise.
13454 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13455 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13456 * ira.c (get_subreg_tracking_sizes): Likewise.
13457 * ira-build.c (ira_create_allocno_objects): Likewise.
13458 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13459 (ira_sort_regnos_for_alter_reg): Likewise.
13460 * ira-costs.c (record_operand_costs): Likewise.
13461 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13462 (resolve_simple_move): Likewise.
13463 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13464 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13465 (lra_constraints): Likewise.
13466 (CONST_POOL_OK_P): Reject variable-sized modes.
13467 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13468 (add_pseudo_to_slot, lra_spill): Likewise.
13469 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13470 * optabs-query.c (get_best_extraction_insn): Likewise.
13471 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13472 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13473 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13474 * recog.c (offsettable_address_addr_space_p): Likewise.
13475 * regcprop.c (maybe_mode_change): Likewise.
13476 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13477 * regrename.c (build_def_use): Likewise.
13478 * regstat.c (dump_reg_info): Likewise.
13479 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13480 (find_reloads, find_reloads_subreg_address): Likewise.
13481 * reload1.c (eliminate_regs_1): Likewise.
13482 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13483 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13484 (simplify_binary_operation_1, simplify_subreg): Likewise.
13485 * targhooks.c (default_function_arg_padding): Likewise.
13486 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13487 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13488 (verify_gimple_assign_ternary): Likewise.
13489 * tree-inline.c (estimate_move_cost): Likewise.
13490 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13491 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13492 (get_address_cost_ainc): Likewise.
13493 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13494 (vect_supportable_dr_alignment): Likewise.
13495 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13496 (vectorizable_reduction): Likewise.
13497 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13498 (vectorizable_operation, vectorizable_load): Likewise.
13499 * tree.c (build_same_sized_truth_vector_type): Likewise.
13500 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13501 * var-tracking.c (emit_note_insn_var_location): Likewise.
13502 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13503 (ADDR_VEC_ALIGN): Likewise.
13505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13506 Alan Hayward <alan.hayward@arm.com>
13507 David Sherwood <david.sherwood@arm.com>
13509 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13511 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13512 or if measurement_type is polynomial.
13513 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13514 * combine.c (make_extraction): Likewise.
13515 * dse.c (find_shift_sequence): Likewise.
13516 * dwarf2out.c (mem_loc_descriptor): Likewise.
13517 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13518 (extract_bit_field, extract_low_bits): Likewise.
13519 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13520 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13521 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13522 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13523 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13524 * reload.c (find_reloads): Likewise.
13525 * reload1.c (alter_reg): Likewise.
13526 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13527 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13528 * tree-if-conv.c (predicate_mem_writes): Likewise.
13529 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13530 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13531 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13532 * valtrack.c (dead_debug_insert_temp): Likewise.
13533 * varasm.c (mergeable_constant_section): Likewise.
13534 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13537 Alan Hayward <alan.hayward@arm.com>
13538 David Sherwood <david.sherwood@arm.com>
13540 * expr.c (expand_assignment): Cope with polynomial mode sizes
13541 when assigning to a CONCAT.
13543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13544 Alan Hayward <alan.hayward@arm.com>
13545 David Sherwood <david.sherwood@arm.com>
13547 * machmode.h (mode_precision): Change from unsigned short to
13549 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13551 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13552 or if measurement_type is not polynomial.
13553 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13554 in which the mode is already known to be a scalar_int_mode.
13555 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13556 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13558 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13559 for GET_MODE_PRECISION.
13560 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13561 for GET_MODE_PRECISION.
13562 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13564 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13565 (expand_field_assignment, make_extraction): Likewise.
13566 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13567 (get_last_value): Likewise.
13568 * convert.c (convert_to_integer_1): Likewise.
13569 * cse.c (cse_insn): Likewise.
13570 * expr.c (expand_expr_real_1): Likewise.
13571 * lra-constraints.c (simplify_operand_subreg): Likewise.
13572 * optabs-query.c (can_atomic_load_p): Likewise.
13573 * optabs.c (expand_atomic_load): Likewise.
13574 (expand_atomic_store): Likewise.
13575 * ree.c (combine_reaching_defs): Likewise.
13576 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13577 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13578 * tree.h (type_has_mode_precision_p): Likewise.
13579 * ubsan.c (instrument_si_overflow): Likewise.
13581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13582 Alan Hayward <alan.hayward@arm.com>
13583 David Sherwood <david.sherwood@arm.com>
13585 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13586 polynomial numbers of units.
13587 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13588 (valid_vector_subparts_p): New function.
13589 (build_vector_type): Remove temporary shim and take the number
13590 of units as a poly_uint64 rather than an int.
13591 (build_opaque_vector_type): Take the number of units as a
13592 poly_uint64 rather than an int.
13593 * tree.c (build_vector_from_ctor): Handle polynomial
13594 TYPE_VECTOR_SUBPARTS.
13595 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13596 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13597 (build_vector_from_val): If the number of units is variable,
13598 use build_vec_duplicate_cst for constant operands and
13599 VEC_DUPLICATE_EXPR otherwise.
13600 (make_vector_type): Remove temporary is_constant ().
13601 (build_vector_type, build_opaque_vector_type): Take the number of
13602 units as a poly_uint64 rather than an int.
13603 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13605 * cfgexpand.c (expand_debug_expr): Likewise.
13606 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13607 (store_constructor, expand_expr_real_1): Likewise.
13608 (const_scalar_mask_from_tree): Likewise.
13609 * fold-const-call.c (fold_const_reduction): Likewise.
13610 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13611 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13612 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13613 (fold_relational_const): Likewise.
13614 (native_interpret_vector): Likewise. Change the size from an
13615 int to an unsigned int.
13616 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13617 TYPE_VECTOR_SUBPARTS.
13618 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13619 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13620 duplicating a non-constant operand into a variable-length vector.
13621 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13622 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13623 * ipa-icf.c (sem_variable::equals): Likewise.
13624 * match.pd: Likewise.
13625 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13626 * print-tree.c (print_node): Likewise.
13627 * stor-layout.c (layout_type): Likewise.
13628 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13629 * tree-cfg.c (verify_gimple_comparison): Likewise.
13630 (verify_gimple_assign_binary): Likewise.
13631 (verify_gimple_assign_ternary): Likewise.
13632 (verify_gimple_assign_single): Likewise.
13633 * tree-pretty-print.c (dump_generic_node): Likewise.
13634 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13635 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13636 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13637 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13638 (vect_shift_permute_load_chain): Likewise.
13639 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13640 (expand_vector_condition, optimize_vector_constructor): Likewise.
13641 (lower_vec_perm, get_compute_type): Likewise.
13642 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13643 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13644 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13645 (vect_recog_mask_conversion_pattern): Likewise.
13646 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13647 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13648 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13649 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13650 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13651 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13652 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13653 (supportable_widening_operation): Likewise.
13654 (supportable_narrowing_operation): Likewise.
13655 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13657 * varasm.c (output_constant): Likewise.
13659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13660 Alan Hayward <alan.hayward@arm.com>
13661 David Sherwood <david.sherwood@arm.com>
13663 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13664 so that both the length == 3 and length != 3 cases set up their
13665 own permute vectors. Add comments explaining why we know the
13666 number of elements is constant.
13667 (vect_permute_load_chain): Likewise.
13669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13670 Alan Hayward <alan.hayward@arm.com>
13671 David Sherwood <david.sherwood@arm.com>
13673 * machmode.h (mode_nunits): Change from unsigned char to
13675 (ONLY_FIXED_SIZE_MODES): New macro.
13676 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13677 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13678 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13680 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13681 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13682 or if measurement_type is not polynomial.
13683 * genmodes.c (ZERO_COEFFS): New macro.
13684 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13686 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13687 Use ZERO_COEFFS when emitting initializers.
13688 * data-streamer.h (bp_pack_poly_value): New function.
13689 (bp_unpack_poly_value): Likewise.
13690 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13691 for GET_MODE_NUNITS.
13692 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13693 for GET_MODE_NUNITS.
13694 * tree.c (make_vector_type): Remove temporary shim and make
13695 the real function take the number of units as a poly_uint64
13696 rather than an int.
13697 (build_vector_type_for_mode): Handle polynomial nunits.
13698 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13699 * emit-rtl.c (const_vec_series_p_1): Likewise.
13700 (gen_rtx_CONST_VECTOR): Likewise.
13701 * fold-const.c (test_vec_duplicate_folding): Likewise.
13702 * genrecog.c (validate_pattern): Likewise.
13703 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13704 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13705 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13706 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13707 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13708 * rtlanal.c (subreg_get_info): Likewise.
13709 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13710 (vect_grouped_load_supported): Likewise.
13711 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13712 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13713 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13714 (simplify_const_unary_operation, simplify_binary_operation_1)
13715 (simplify_const_binary_operation, simplify_ternary_operation)
13716 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13717 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13718 instead of CONST_VECTOR_NUNITS.
13719 * varasm.c (output_constant_pool_2): Likewise.
13720 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13721 explicit-encoded elements in the XVEC for variable-length vectors.
13723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13725 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13728 Alan Hayward <alan.hayward@arm.com>
13729 David Sherwood <david.sherwood@arm.com>
13731 * coretypes.h (fixed_size_mode): Declare.
13732 (fixed_size_mode_pod): New typedef.
13733 * builtins.h (target_builtins::x_apply_args_mode)
13734 (target_builtins::x_apply_result_mode): Change type to
13735 fixed_size_mode_pod.
13736 * builtins.c (apply_args_size, apply_result_size, result_vector)
13737 (expand_builtin_apply_args_1, expand_builtin_apply)
13738 (expand_builtin_return): Update accordingly.
13740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13742 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13743 * cselib.c (cselib_hash_rtx): Likewise.
13744 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13745 CONST_VECTOR encoding.
13747 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13748 Jeff Law <law@redhat.com>
13751 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13752 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13753 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13754 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13757 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13758 explicitly probe *sp in a noreturn function if there were any callee
13759 register saves or frame pointer is needed.
13761 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13764 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13765 BLKmode for ternary, binary or unary expressions.
13768 * var-tracking.c (delete_vta_debug_insn): New inline function.
13769 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13770 insns from get_insns () to NULL instead of each bb separately.
13771 Use delete_vta_debug_insn. No longer static.
13772 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13773 delete_vta_debug_insns callers.
13774 * rtl.h (delete_vta_debug_insns): Declare.
13775 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13776 instead of variable_tracking_main.
13778 2018-01-03 Martin Sebor <msebor@redhat.com>
13780 PR tree-optimization/83603
13781 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13782 arguments past the endof the argument list in functions declared
13783 without a prototype.
13784 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13785 Avoid checking when arguments are null.
13787 2018-01-03 Martin Sebor <msebor@redhat.com>
13790 * doc/extend.texi (attribute const): Fix a typo.
13791 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13792 issuing -Wsuggest-attribute for void functions.
13794 2018-01-03 Martin Sebor <msebor@redhat.com>
13796 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13797 offset_int::from instead of wide_int::to_shwi.
13798 (maybe_diag_overlap): Remove assertion.
13799 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13800 * gimple-ssa-sprintf.c (format_directive): Same.
13801 (parse_directive): Same.
13802 (sprintf_dom_walker::compute_format_length): Same.
13803 (try_substitute_return_value): Same.
13805 2018-01-03 Jeff Law <law@redhat.com>
13807 PR middle-end/83654
13808 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13809 non-constant residual for zero at runtime and avoid probing in
13810 that case. Reorganize code for trailing problem to mirror handling
13813 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13815 PR tree-optimization/83501
13816 * tree-ssa-strlen.c (get_string_cst): New.
13817 (handle_char_store): Call get_string_cst.
13819 2018-01-03 Martin Liska <mliska@suse.cz>
13821 PR tree-optimization/83593
13822 * tree-ssa-strlen.c: Include tree-cfg.h.
13823 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13824 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13825 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13827 (strlen_dom_walker::before_dom_children): Call
13828 gimple_purge_dead_eh_edges. Dump tranformation with details
13830 (strlen_dom_walker::before_dom_children): Update call by adding
13831 new argument cleanup_eh.
13832 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13834 2018-01-03 Martin Liska <mliska@suse.cz>
13837 * cif-code.def (VARIADIC_THUNK): New enum value.
13838 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13841 2018-01-03 Jan Beulich <jbeulich@suse.com>
13843 * sse.md (mov<mode>_internal): Tighten condition for when to use
13844 vmovdqu<ssescalarsize> for TI and OI modes.
13846 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13848 Update copyright years.
13850 2018-01-03 Martin Liska <mliska@suse.cz>
13853 * ipa-visibility.c (function_and_variable_visibility): Skip
13854 functions with noipa attribure.
13856 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13858 * gcc.c (process_command): Update copyright notice dates.
13859 * gcov-dump.c (print_version): Ditto.
13860 * gcov.c (print_version): Ditto.
13861 * gcov-tool.c (print_version): Ditto.
13862 * gengtype.c (create_file): Ditto.
13863 * doc/cpp.texi: Bump @copying's copyright year.
13864 * doc/cppinternals.texi: Ditto.
13865 * doc/gcc.texi: Ditto.
13866 * doc/gccint.texi: Ditto.
13867 * doc/gcov.texi: Ditto.
13868 * doc/install.texi: Ditto.
13869 * doc/invoke.texi: Ditto.
13871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13873 * vector-builder.h (vector_builder::m_full_nelts): Change from
13874 unsigned int to poly_uint64.
13875 (vector_builder::full_nelts): Update prototype accordingly.
13876 (vector_builder::new_vector): Likewise.
13877 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13878 (vector_builder::operator ==): Likewise.
13879 (vector_builder::finalize): Likewise.
13880 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13881 Take the number of elements as a poly_uint64 rather than an
13883 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13884 from unsigned int to poly_uint64.
13885 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13886 (vec_perm_indices::new_vector): Likewise.
13887 (vec_perm_indices::length): Likewise.
13888 (vec_perm_indices::nelts_per_input): Likewise.
13889 (vec_perm_indices::input_nelts): Likewise.
13890 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13891 number of elements per input as a poly_uint64 rather than an
13892 unsigned int. Use the original encoding for variable-length
13893 vectors, rather than clamping each individual element.
13894 For the second and subsequent elements in each pattern,
13895 clamp the step and base before clamping their sum.
13896 (vec_perm_indices::series_p): Handle polynomial element counts.
13897 (vec_perm_indices::all_in_range_p): Likewise.
13898 (vec_perm_indices_to_tree): Likewise.
13899 (vec_perm_indices_to_rtx): Likewise.
13900 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13901 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13902 (tree_vector_builder::new_binary_operation): Handle polynomial
13903 element counts. Return false if we need to know the number
13904 of elements at compile time.
13905 * fold-const.c (fold_vec_perm): Punt if the number of elements
13906 isn't known at compile time.
13908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13910 * vec-perm-indices.h (vec_perm_builder): Change element type
13911 from HOST_WIDE_INT to poly_int64.
13912 (vec_perm_indices::element_type): Update accordingly.
13913 (vec_perm_indices::clamp): Handle polynomial element_types.
13914 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13915 (vec_perm_indices::all_in_range_p): Likewise.
13916 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13918 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13919 polynomial vec_perm_indices element types.
13920 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13921 * fold-const.c (fold_vec_perm): Likewise.
13922 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13923 * tree-vect-generic.c (lower_vec_perm): Likewise.
13924 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13925 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13926 element type to HOST_WIDE_INT.
13928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13929 Alan Hayward <alan.hayward@arm.com>
13930 David Sherwood <david.sherwood@arm.com>
13932 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13933 rather than an int. Use plus_constant.
13934 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13935 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13938 Alan Hayward <alan.hayward@arm.com>
13939 David Sherwood <david.sherwood@arm.com>
13941 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13942 a HOST_WIDE_INT to a poly_int64.
13944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13945 Alan Hayward <alan.hayward@arm.com>
13946 David Sherwood <david.sherwood@arm.com>
13948 * calls.c (load_register_parameters): Cope with polynomial
13949 mode sizes. Require a constant size for BLKmode parameters
13950 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13951 forces a parameter to be padded at the lsb end in order to
13952 fill a complete number of words, require the parameter size
13953 to be ordered wrt UNITS_PER_WORD.
13955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13956 Alan Hayward <alan.hayward@arm.com>
13957 David Sherwood <david.sherwood@arm.com>
13959 * reload1.c (spill_stack_slot_width): Change element type
13960 from unsigned int to poly_uint64_pod.
13961 (alter_reg): Treat mode sizes as polynomial.
13963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13964 Alan Hayward <alan.hayward@arm.com>
13965 David Sherwood <david.sherwood@arm.com>
13967 * reload.c (complex_word_subreg_p): New function.
13968 (reload_inner_reg_of_subreg, push_reload): Use it.
13970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13971 Alan Hayward <alan.hayward@arm.com>
13972 David Sherwood <david.sherwood@arm.com>
13974 * lra-constraints.c (process_alt_operands): Reject matched
13975 operands whose sizes aren't ordered.
13976 (match_reload): Refer to this check here.
13978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13979 Alan Hayward <alan.hayward@arm.com>
13980 David Sherwood <david.sherwood@arm.com>
13982 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13983 that the mode size is in the set {1, 2, 4, 8, 16}.
13985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13986 Alan Hayward <alan.hayward@arm.com>
13987 David Sherwood <david.sherwood@arm.com>
13989 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13990 Use plus_constant instead of gen_rtx_PLUS.
13992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13993 Alan Hayward <alan.hayward@arm.com>
13994 David Sherwood <david.sherwood@arm.com>
13996 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13997 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13998 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13999 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14000 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14001 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14002 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14003 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14004 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14005 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14007 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14008 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14009 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14010 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14011 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14012 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14013 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14014 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14015 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14016 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14018 * expr.c (emit_move_resolve_push): Treat the input and result
14019 of PUSH_ROUNDING as a poly_int64.
14020 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14021 (emit_push_insn): Likewise.
14022 * lra-eliminations.c (mark_not_eliminable): Likewise.
14023 * recog.c (push_operand): Likewise.
14024 * reload1.c (elimination_effects): Likewise.
14025 * rtlanal.c (nonzero_bits1): Likewise.
14026 * calls.c (store_one_arg): Likewise. Require the padding to be
14027 known at compile time.
14029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14030 Alan Hayward <alan.hayward@arm.com>
14031 David Sherwood <david.sherwood@arm.com>
14033 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14034 Use plus_constant instead of gen_rtx_PLUS.
14036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14037 Alan Hayward <alan.hayward@arm.com>
14038 David Sherwood <david.sherwood@arm.com>
14040 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14041 rather than an int.
14043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14044 Alan Hayward <alan.hayward@arm.com>
14045 David Sherwood <david.sherwood@arm.com>
14047 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14048 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14049 via stack temporaries. Treat the mode size as polynomial too.
14051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14052 Alan Hayward <alan.hayward@arm.com>
14053 David Sherwood <david.sherwood@arm.com>
14055 * expr.c (expand_expr_real_2): When handling conversions involving
14056 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14057 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14058 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14067 Alan Hayward <alan.hayward@arm.com>
14068 David Sherwood <david.sherwood@arm.com>
14070 * combine.c (can_change_dest_mode): Handle polynomial
14071 REGMODE_NATURAL_SIZE.
14072 * expmed.c (store_bit_field_1): Likewise.
14073 * expr.c (store_constructor): Likewise.
14074 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14075 and polynomial REGMODE_NATURAL_SIZE.
14076 (gen_lowpart_common): Likewise.
14077 * reginfo.c (record_subregs_of_mode): Likewise.
14078 * rtlanal.c (read_modify_subreg_p): Likewise.
14080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14081 Alan Hayward <alan.hayward@arm.com>
14082 David Sherwood <david.sherwood@arm.com>
14084 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14085 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
14093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14094 Alan Hayward <alan.hayward@arm.com>
14095 David Sherwood <david.sherwood@arm.com>
14097 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14098 in a POINTER_PLUS_EXPR.
14100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14101 Alan Hayward <alan.hayward@arm.com>
14102 David Sherwood <david.sherwood@arm.com>
14104 * omp-simd-clone.c (simd_clone_subparts): New function.
14105 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14106 (ipa_simd_modify_function_body): Likewise.
14108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14109 Alan Hayward <alan.hayward@arm.com>
14110 David Sherwood <david.sherwood@arm.com>
14112 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14113 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14114 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14115 (expand_vector_condition, vector_element): Likewise.
14116 (subparts_gt): New function.
14117 (get_compute_type): Use subparts_gt.
14118 (count_type_subparts): Delete.
14119 (expand_vector_operations_1): Use subparts_gt instead of
14120 count_type_subparts.
14122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14123 Alan Hayward <alan.hayward@arm.com>
14124 David Sherwood <david.sherwood@arm.com>
14126 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14127 (vect_compile_time_alias): ...this new function. Do the calculation
14128 on poly_ints rather than trees.
14129 (vect_prune_runtime_alias_test_list): Update call accordingly.
14131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14132 Alan Hayward <alan.hayward@arm.com>
14133 David Sherwood <david.sherwood@arm.com>
14135 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14137 (vect_schedule_slp_instance): Likewise.
14139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14140 Alan Hayward <alan.hayward@arm.com>
14141 David Sherwood <david.sherwood@arm.com>
14143 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14144 constant and extern definitions for variable-length vectors.
14145 (vect_get_constant_vectors): Note that the number of units
14146 is known to be constant.
14148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14149 Alan Hayward <alan.hayward@arm.com>
14150 David Sherwood <david.sherwood@arm.com>
14152 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14153 of units as polynomial. Choose between WIDE and NARROW based
14156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14157 Alan Hayward <alan.hayward@arm.com>
14158 David Sherwood <david.sherwood@arm.com>
14160 * tree-vect-stmts.c (simd_clone_subparts): New function.
14161 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14164 Alan Hayward <alan.hayward@arm.com>
14165 David Sherwood <david.sherwood@arm.com>
14167 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14168 vectors as polynomial. Use build_index_vector for
14169 IFN_GOMP_SIMD_LANE.
14171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14172 Alan Hayward <alan.hayward@arm.com>
14173 David Sherwood <david.sherwood@arm.com>
14175 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14176 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14177 for variable-length vectors.
14178 (vectorizable_mask_load_store): Treat the number of units as
14179 polynomial, asserting that it is constant if the condition has
14180 already been enforced.
14181 (vectorizable_store, vectorizable_load): Likewise.
14183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14184 Alan Hayward <alan.hayward@arm.com>
14185 David Sherwood <david.sherwood@arm.com>
14187 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14188 of units as polynomial. Punt if we can't tell at compile time
14189 which vector contains the final result.
14191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14192 Alan Hayward <alan.hayward@arm.com>
14193 David Sherwood <david.sherwood@arm.com>
14195 * tree-vect-loop.c (vectorizable_induction): Treat the number
14196 of units as polynomial. Punt on SLP inductions. Use an integer
14197 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14198 cast of such a series for variable-length floating-point
14201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14202 Alan Hayward <alan.hayward@arm.com>
14203 David Sherwood <david.sherwood@arm.com>
14205 * tree.h (build_index_vector): Declare.
14206 * tree.c (build_index_vector): New function.
14207 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14208 of units as polynomial, forcibly converting it to a constant if
14209 vectorizable_reduction has already enforced the condition.
14210 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14211 to create a {1,2,3,...} vector.
14212 (vectorizable_reduction): Treat the number of units as polynomial.
14213 Choose vectype_in based on the largest scalar element size rather
14214 than the smallest number of units. Enforce the restrictions
14217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14218 Alan Hayward <alan.hayward@arm.com>
14219 David Sherwood <david.sherwood@arm.com>
14221 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14222 number of units as polynomial.
14224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14225 Alan Hayward <alan.hayward@arm.com>
14226 David Sherwood <david.sherwood@arm.com>
14228 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14229 * target.def (autovectorize_vector_sizes): Return the vector sizes
14230 by pointer, using vector_sizes rather than a bitmask.
14231 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14232 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14233 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14235 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14236 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14237 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14238 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14239 * omp-general.c (omp_max_vf): Likewise.
14240 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14241 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14242 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14243 * tree-vect-slp.c (vect_slp_bb): Likewise.
14244 * doc/tm.texi: Regenerate.
14245 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14247 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14248 the vector size as a poly_uint64 rather than an unsigned int.
14249 (current_vector_size): Change from an unsigned int to a poly_uint64.
14250 (get_vectype_for_scalar_type): Update accordingly.
14251 * tree.h (build_truth_vector_type): Take the size and number of
14252 units as a poly_uint64 rather than an unsigned int.
14253 (build_vector_type): Add a temporary overload that takes
14254 the number of units as a poly_uint64 rather than an unsigned int.
14255 * tree.c (make_vector_type): Likewise.
14256 (build_truth_vector_type): Take the number of units as a poly_uint64
14257 rather than an unsigned int.
14259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14260 Alan Hayward <alan.hayward@arm.com>
14261 David Sherwood <david.sherwood@arm.com>
14263 * target.def (get_mask_mode): Take the number of units and length
14264 as poly_uint64s rather than unsigned ints.
14265 * targhooks.h (default_get_mask_mode): Update accordingly.
14266 * targhooks.c (default_get_mask_mode): Likewise.
14267 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14268 * doc/tm.texi: Regenerate.
14270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14271 Alan Hayward <alan.hayward@arm.com>
14272 David Sherwood <david.sherwood@arm.com>
14274 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14275 * omp-general.c (omp_max_vf): Likewise.
14276 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14277 (expand_omp_simd): Handle polynomial safelen.
14278 * omp-low.c (omplow_simd_context): Add a default constructor.
14279 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14280 (lower_rec_simd_input_clauses): Update accordingly.
14281 (lower_rec_input_clauses): Likewise.
14283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14284 Alan Hayward <alan.hayward@arm.com>
14285 David Sherwood <david.sherwood@arm.com>
14287 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14288 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14289 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14290 (vect_analyze_slp_cost): Likewise.
14291 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14292 (vect_model_load_cost): Likewise.
14294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14295 Alan Hayward <alan.hayward@arm.com>
14296 David Sherwood <david.sherwood@arm.com>
14298 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14299 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14300 from an unsigned int * to a poly_uint64_pod *.
14301 (calculate_unrolling_factor): New function.
14302 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14305 Alan Hayward <alan.hayward@arm.com>
14306 David Sherwood <david.sherwood@arm.com>
14308 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14309 from an unsigned int to a poly_uint64.
14310 (_loop_vec_info::slp_unrolling_factor): Likewise.
14311 (_loop_vec_info::vectorization_factor): Change from an int
14313 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14314 (vect_get_num_vectors): New function.
14315 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14316 (vect_get_num_copies): Use vect_get_num_vectors.
14317 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14318 to an unsigned int *.
14319 (vect_analyze_data_refs): Change min_vf from an int * to a
14321 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14322 than an unsigned HOST_WIDE_INT.
14323 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14324 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14325 to an unsigned int *.
14326 (vect_analyze_data_ref_dependences): Likewise.
14327 (vect_compute_data_ref_alignment): Handle polynomial vf.
14328 (vect_enhance_data_refs_alignment): Likewise.
14329 (vect_prune_runtime_alias_test_list): Likewise.
14330 (vect_shift_permute_load_chain): Likewise.
14331 (vect_supportable_dr_alignment): Likewise.
14332 (dependence_distance_ge_vf): Take the vectorization factor as a
14333 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14334 (vect_analyze_data_refs): Change min_vf from an int * to a
14336 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14337 vfm1 as a poly_uint64 rather than an int. Make the same change
14338 for the returned bound_scalar.
14339 (vect_gen_vector_loop_niters): Handle polynomial vf.
14340 (vect_do_peeling): Likewise. Update call to
14341 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14342 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14344 * tree-vect-loop.c (vect_determine_vectorization_factor)
14345 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14346 (vect_get_known_peeling_cost): Likewise.
14347 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14348 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14349 (vect_transform_loop): Likewise. Use the lowest possible VF when
14350 updating the upper bounds of the loop.
14351 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14352 rather than an int.
14353 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14354 polynomial unroll factors.
14355 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14356 (vect_make_slp_decision): Likewise.
14357 (vect_supported_load_permutation_p): Likewise, and polynomial
14359 (vect_analyze_slp_cost): Handle polynomial vf.
14360 (vect_slp_analyze_node_operations): Likewise.
14361 (vect_slp_analyze_bb_1): Likewise.
14362 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14363 than an unsigned HOST_WIDE_INT.
14364 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14365 (vectorizable_load): Handle polynomial vf.
14366 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14368 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14371 Alan Hayward <alan.hayward@arm.com>
14372 David Sherwood <david.sherwood@arm.com>
14374 * match.pd: Handle bit operations involving three constants
14375 and try to fold one pair.
14377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14379 * tree-vect-loop-manip.c: Include gimple-fold.h.
14380 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14381 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14382 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14383 Add a path that uses a step of VF instead of 1, but disable it
14385 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14386 and niters_no_overflow parameters. Update calls to
14387 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14388 Create a new SSA name if the latter choses to use a ste other
14389 than zero, and return it via niters_vector_mult_vf_var.
14390 * tree-vect-loop.c (vect_transform_loop): Update calls to
14391 vect_do_peeling, vect_gen_vector_loop_niters and
14392 slpeel_make_loop_iterate_ntimes.
14393 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14394 (vect_gen_vector_loop_niters): Update declarations after above changes.
14396 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14398 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14399 128-bit round to integer instructions.
14400 (ceil<mode>2): Likewise.
14401 (btrunc<mode>2): Likewise.
14402 (round<mode>2): Likewise.
14404 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14406 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14407 unaligned VSX load/store on P8/P9.
14408 (expand_block_clear): Allow the use of unaligned VSX
14409 load/store on P8/P9.
14411 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14413 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14415 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14416 swap associated with both a load and a store.
14418 2018-01-02 Andrew Waterman <andrew@sifive.com>
14420 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14421 * config/riscv/riscv.md (clear_cache): Use it.
14423 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14425 * web.c: Remove out-of-date comment.
14427 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14429 * expr.c (fixup_args_size_notes): Check that any existing
14430 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14431 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14432 (emit_single_push_insn): ...here.
14434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14436 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14437 (const_vector_encoded_nelts): New function.
14438 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14439 (const_vector_int_elt, const_vector_elt): Declare.
14440 * emit-rtl.c (const_vector_int_elt_1): New function.
14441 (const_vector_elt): Likewise.
14442 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14443 of CONST_VECTOR_ELT.
14445 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14447 * expr.c: Include rtx-vector-builder.h.
14448 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14449 directly on the tree encoding.
14450 (const_vector_from_tree): Likewise.
14451 * optabs.c: Include rtx-vector-builder.h.
14452 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14453 sequence of "u" values.
14454 * vec-perm-indices.c: Include rtx-vector-builder.h.
14455 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14456 directly on the vec_perm_indices encoding.
14458 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14460 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14461 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14462 * rtx-vector-builder.h: New file.
14463 * rtx-vector-builder.c: Likewise.
14464 * rtl.h (rtx_def::u2): Add a const_vector field.
14465 (CONST_VECTOR_NPATTERNS): New macro.
14466 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14467 (CONST_VECTOR_DUPLICATE_P): Likewise.
14468 (CONST_VECTOR_STEPPED_P): Likewise.
14469 (CONST_VECTOR_ENCODED_ELT): Likewise.
14470 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14471 (unwrap_const_vec_duplicate): Likewise.
14472 (const_vec_series_p): Check for a non-duplicated vector encoding.
14473 Say that the function only returns true for integer vectors.
14474 * emit-rtl.c: Include rtx-vector-builder.h.
14475 (gen_const_vec_duplicate_1): Delete.
14476 (gen_const_vector): Call gen_const_vec_duplicate instead of
14477 gen_const_vec_duplicate_1.
14478 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14479 (gen_const_vec_duplicate): Use rtx_vector_builder.
14480 (gen_const_vec_series): Likewise.
14481 (gen_rtx_CONST_VECTOR): Likewise.
14482 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14483 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14484 Build a new vector rather than modifying a CONST_VECTOR in-place.
14485 (handle_special_swappables): Update call accordingly.
14486 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14487 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14488 Build a new vector rather than modifying a CONST_VECTOR in-place.
14489 (handle_special_swappables): Update call accordingly.
14491 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14493 * simplify-rtx.c (simplify_const_binary_operation): Use
14494 CONST_VECTOR_ELT instead of XVECEXP.
14496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14498 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14499 the selector elements to be different from the data elements
14500 if the selector is a VECTOR_CST.
14501 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14502 ssizetype for the selector.
14504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14506 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14507 before testing each element individually.
14508 * tree-vect-generic.c (lower_vec_perm): Likewise.
14510 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14512 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14513 * selftest-run-tests.c (selftest::run_tests): Call it.
14514 * vector-builder.h (vector_builder::operator ==): New function.
14515 (vector_builder::operator !=): Likewise.
14516 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14517 (vec_perm_indices::all_from_input_p): New function.
14518 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14519 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14520 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14521 instead of reading the VECTOR_CST directly. Detect whether both
14522 vector inputs are the same before constructing the vec_perm_indices,
14523 and update the number of inputs argument accordingly. Use the
14524 utility functions added above. Only construct sel2 if we need to.
14526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14528 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14529 the broadcast of the low byte.
14530 (expand_mult_highpart): Use an explicit encoding for the permutes.
14531 * optabs-query.c (can_mult_highpart_p): Likewise.
14532 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14533 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14534 (vectorizable_bswap): Likewise.
14535 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14536 explicit encoding for the power-of-2 permutes.
14537 (vect_permute_store_chain): Likewise.
14538 (vect_grouped_load_supported): Likewise.
14539 (vect_permute_load_chain): Likewise.
14541 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14543 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14544 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14545 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14546 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14547 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14548 (vect_gen_perm_mask_any): Likewise.
14550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14552 * int-vector-builder.h: New file.
14553 * vec-perm-indices.h: Include int-vector-builder.h.
14554 (vec_perm_indices): Redefine as an int_vector_builder.
14555 (auto_vec_perm_indices): Delete.
14556 (vec_perm_builder): Redefine as a stand-alone class.
14557 (vec_perm_indices::vec_perm_indices): New function.
14558 (vec_perm_indices::clamp): Likewise.
14559 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14560 (vec_perm_indices::new_vector): New function.
14561 (vec_perm_indices::new_expanded_vector): Update for new
14562 vec_perm_indices class.
14563 (vec_perm_indices::rotate_inputs): New function.
14564 (vec_perm_indices::all_in_range_p): Operate directly on the
14565 encoded form, without computing elided elements.
14566 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14567 encoding. Update for new vec_perm_indices class.
14568 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14569 the given vec_perm_builder.
14570 (expand_vec_perm_var): Update vec_perm_builder constructor.
14571 (expand_mult_highpart): Use vec_perm_builder instead of
14572 auto_vec_perm_indices.
14573 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14574 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14575 or double series encoding as appropriate.
14576 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14577 vec_perm_indices instead of auto_vec_perm_indices.
14578 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14579 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14580 (vect_permute_store_chain): Likewise.
14581 (vect_grouped_load_supported): Likewise.
14582 (vect_permute_load_chain): Likewise.
14583 (vect_shift_permute_load_chain): Likewise.
14584 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14585 (vect_transform_slp_perm_load): Likewise.
14586 (vect_schedule_slp_instance): Likewise.
14587 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14588 (vectorizable_mask_load_store): Likewise.
14589 (vectorizable_bswap): Likewise.
14590 (vectorizable_store): Likewise.
14591 (vectorizable_load): Likewise.
14592 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14593 vec_perm_indices instead of auto_vec_perm_indices. Use
14594 tree_to_vec_perm_builder to read the vector from a tree.
14595 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14596 vec_perm_builder instead of a vec_perm_indices.
14597 (have_whole_vector_shift): Use vec_perm_builder and
14598 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14599 truncation to calc_vec_perm_mask_for_shift.
14600 (vect_create_epilog_for_reduction): Likewise.
14601 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14602 from auto_vec_perm_indices to vec_perm_indices.
14603 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14604 instead of changing individual elements.
14605 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14606 the vector in d.perm.
14607 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14608 from auto_vec_perm_indices to vec_perm_indices.
14609 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14610 instead of changing individual elements.
14611 (arm_vectorize_vec_perm_const): Use new_vector to install
14612 the vector in d.perm.
14613 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14614 Update vec_perm_builder constructor.
14615 (rs6000_expand_interleave): Likewise.
14616 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14617 (rs6000_expand_interleave): Likewise.
14619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14621 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14622 to qimode could truncate the indices.
14623 * optabs.c (expand_vec_perm_var): Likewise.
14625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14627 * Makefile.in (OBJS): Add vec-perm-indices.o.
14628 * vec-perm-indices.h: New file.
14629 * vec-perm-indices.c: Likewise.
14630 * target.h (vec_perm_indices): Replace with a forward class
14632 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14633 * optabs.h: Include vec-perm-indices.h.
14634 (expand_vec_perm): Delete.
14635 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14636 (expand_vec_perm_const): Declare.
14637 * target.def (vec_perm_const_ok): Replace with...
14638 (vec_perm_const): ...this new hook.
14639 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14640 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14641 * doc/tm.texi: Regenerate.
14642 * optabs.def (vec_perm_const): Delete.
14643 * doc/md.texi (vec_perm_const): Likewise.
14644 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14645 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14646 expand_vec_perm for constant permutation vectors. Assert that
14647 the mode of variable permutation vectors is the integer equivalent
14648 of the mode that is being permuted.
14649 * optabs-query.h (selector_fits_mode_p): Declare.
14650 * optabs-query.c: Include vec-perm-indices.h.
14651 (selector_fits_mode_p): New function.
14652 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14653 is defined, instead of checking whether the vec_perm_const_optab
14654 exists. Use targetm.vectorize.vec_perm_const instead of
14655 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14656 fit in the vector mode before using a variable permute.
14657 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14658 vec_perm_indices instead of an rtx.
14659 (expand_vec_perm): Replace with...
14660 (expand_vec_perm_const): ...this new function. Take the selector
14661 as a vec_perm_indices rather than an rtx. Also take the mode of
14662 the selector. Update call to shift_amt_for_vec_perm_mask.
14663 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14664 Use vec_perm_indices::new_expanded_vector to expand the original
14665 selector into bytes. Check whether the indices fit in the vector
14666 mode before using a variable permute.
14667 (expand_vec_perm_var): Make global.
14668 (expand_mult_highpart): Use expand_vec_perm_const.
14669 * fold-const.c: Includes vec-perm-indices.h.
14670 * tree-ssa-forwprop.c: Likewise.
14671 * tree-vect-data-refs.c: Likewise.
14672 * tree-vect-generic.c: Likewise.
14673 * tree-vect-loop.c: Likewise.
14674 * tree-vect-slp.c: Likewise.
14675 * tree-vect-stmts.c: Likewise.
14676 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14678 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14679 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14680 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14681 (aarch64_vectorize_vec_perm_const): ...this new function.
14682 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14683 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14684 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14685 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14686 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14687 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14688 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14690 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14691 check for NEON modes.
14692 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14693 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14694 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14695 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14697 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14698 the old VEC_PERM_CONST conditions.
14699 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14700 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14701 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14702 (ia64_vectorize_vec_perm_const_ok): Merge into...
14703 (ia64_vectorize_vec_perm_const): ...this new function.
14704 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14705 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14706 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14707 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14708 * config/mips/mips.c (mips_expand_vec_perm_const)
14709 (mips_vectorize_vec_perm_const_ok): Merge into...
14710 (mips_vectorize_vec_perm_const): ...this new function.
14711 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14712 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14713 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14714 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14715 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14716 (rs6000_expand_vec_perm_const): Delete.
14717 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14720 (altivec_expand_vec_perm_const_le): Take each operand individually.
14721 Operate on constant selectors rather than rtxes.
14722 (altivec_expand_vec_perm_const): Likewise. Update call to
14723 altivec_expand_vec_perm_const_le.
14724 (rs6000_expand_vec_perm_const): Delete.
14725 (rs6000_vectorize_vec_perm_const_ok): Delete.
14726 (rs6000_vectorize_vec_perm_const): New function.
14727 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14728 an element count and rtx array.
14729 (rs6000_expand_extract_even): Update call accordingly.
14730 (rs6000_expand_interleave): Likewise.
14731 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14732 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14733 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14734 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14735 (rs6000_expand_vec_perm_const): Delete.
14736 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14737 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14738 (altivec_expand_vec_perm_const_le): Take each operand individually.
14739 Operate on constant selectors rather than rtxes.
14740 (altivec_expand_vec_perm_const): Likewise. Update call to
14741 altivec_expand_vec_perm_const_le.
14742 (rs6000_expand_vec_perm_const): Delete.
14743 (rs6000_vectorize_vec_perm_const_ok): Delete.
14744 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14745 reference to the SPE evmerge intructions.
14746 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14747 an element count and rtx array.
14748 (rs6000_expand_extract_even): Update call accordingly.
14749 (rs6000_expand_interleave): Likewise.
14750 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14751 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14753 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14755 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14757 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14758 vector mode and that that mode matches the mode of the data
14760 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14761 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14762 directly using expand_vec_perm_1 when forcing selectors into
14764 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14766 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14768 * optabs-query.h (can_vec_perm_p): Delete.
14769 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14770 * optabs-query.c (can_vec_perm_p): Split into...
14771 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14772 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14773 particular selector is valid.
14774 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14775 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14776 (vect_grouped_load_supported): Likewise.
14777 (vect_shift_permute_load_chain): Likewise.
14778 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14779 (vect_transform_slp_perm_load): Likewise.
14780 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14781 (vectorizable_bswap): Likewise.
14782 (vect_gen_perm_mask_checked): Likewise.
14783 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14784 implementations of variable permutation vectors into account
14785 when deciding which selector to use.
14786 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14787 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14788 with a false third argument.
14789 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14790 to test whether the constant selector is valid and can_vec_perm_var_p
14791 to test whether a variable selector is valid.
14793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14795 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14796 * optabs-query.c (can_vec_perm_p): Likewise.
14797 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14798 instead of vec_perm_indices.
14799 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14800 (vect_gen_perm_mask_checked): Likewise,
14801 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14802 (vect_gen_perm_mask_checked): Likewise,
14804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14806 * optabs-query.h (qimode_for_vec_perm): Declare.
14807 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14808 (qimode_for_vec_perm): ...this new function.
14809 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14811 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14813 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14814 does not have a conditional at the top.
14816 2018-01-02 Richard Biener <rguenther@suse.de>
14818 * ipa-inline.c (big_speedup_p): Fix expression.
14820 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14823 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14826 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14830 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14831 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14832 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14833 cond_taken_branch_cost 3->4.
14835 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14837 PR tree-optimization/83581
14838 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14839 TODO_cleanup_cfg if any changes have been made.
14841 PR middle-end/83608
14842 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14843 convert_modes if target mode has the right side, but different mode
14846 PR middle-end/83609
14847 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14848 last argument when extracting from CONCAT. If either from_real or
14849 from_imag is NULL, use expansion through memory. If result is not
14850 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14851 the parts directly to inner mode, if even that fails, use expansion
14854 PR middle-end/83623
14855 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14856 check for bswap in mode rather than HImode and use that in expand_unop
14859 Copyright (C) 2018 Free Software Foundation, Inc.
14861 Copying and distribution of this file, with or without modification,
14862 are permitted in any medium without royalty provided the copyright
14863 notice and this notice are preserved.