| blob | 2bdf916284061436a9484b11bc81502e319224d8 |
1 2018-05-31 Martin Sebor <msebor@redhat.com>
3 PR c/82063
4 * calls.c (alloc_max_size): Correct a logic error/typo.
5 Treat excessive arguments as infinite. Warn for invalid arguments.
6 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10 PR target/85829
11 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12 and movx for Haswell.
14 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
15 Cesar Philippidis <cesar@codesourcery.com>
17 PR middle-end/85879
18 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
19 when emitting error on private/firstprivate reductions.
20 * omp-low.c (lower_omp_target): Avoid reference-type processing
21 on pointers for firstprivate clause.
23 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
25 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
26 (st1x2): Likewise.
27 (st1x3): Likewise.
28 * config/aarch64/aarch64-simd.md
29 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
30 (aarch64_ld1_x3_<mode>): Likewise
31 (aarch64_st1x2<VALLDIF:mode>): Likewise
32 (aarch64_st1_x2_<mode>): Likewise
33 (aarch64_st1x3<VALLDIF:mode>): Likewise
34 (aarch64_st1_x3_<mode>): Likewise
35 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
36 (vld1_s8_x3): Likewise.
37 (vld1_u16_x3): Likewise.
38 (vld1_s16_x3): Likewise.
39 (vld1_u32_x3): Likewise.
40 (vld1_s32_x3): Likewise.
41 (vld1_u64_x3): Likewise.
42 (vld1_s64_x3): Likewise.
43 (vld1_f16_x3): Likewise.
44 (vld1_f32_x3): Likewise.
45 (vld1_f64_x3): Likewise.
46 (vld1_p8_x3): Likewise.
47 (vld1_p16_x3): Likewise.
48 (vld1_p64_x3): Likewise.
49 (vld1q_u8_x3): Likewise.
50 (vld1q_s8_x3): Likewise.
51 (vld1q_u16_x3): Likewise.
52 (vld1q_s16_x3): Likewise.
53 (vld1q_u32_x3): Likewise.
54 (vld1q_s32_x3): Likewise.
55 (vld1q_u64_x3): Likewise.
56 (vld1q_s64_x3): Likewise.
57 (vld1q_f16_x3): Likewise.
58 (vld1q_f32_x3): Likewise.
59 (vld1q_f64_x3): Likewise.
60 (vld1q_p8_x3): Likewise.
61 (vld1q_p16_x3): Likewise.
62 (vld1q_p64_x3): Likewise.
63 (vst1_s64_x2): Likewise.
64 (vst1_u64_x2): Likewise.
65 (vst1_f64_x2): Likewise.
66 (vst1_s8_x2): Likewise.
67 (vst1_p8_x2): Likewise.
68 (vst1_s16_x2): Likewise.
69 (vst1_p16_x2): Likewise.
70 (vst1_s32_x2): Likewise.
71 (vst1_u8_x2): Likewise.
72 (vst1_u16_x2): Likewise.
73 (vst1_u32_x2): Likewise.
74 (vst1_f16_x2): Likewise.
75 (vst1_f32_x2): Likewise.
76 (vst1_p64_x2): Likewise.
77 (vst1q_s8_x2): Likewise.
78 (vst1q_p8_x2): Likewise.
79 (vst1q_s16_x2): Likewise.
80 (vst1q_p16_x2): Likewise.
81 (vst1q_s32_x2): Likewise.
82 (vst1q_s64_x2): Likewise.
83 (vst1q_u8_x2): Likewise.
84 (vst1q_u16_x2): Likewise.
85 (vst1q_u32_x2): Likewise.
86 (vst1q_u64_x2): Likewise.
87 (vst1q_f16_x2): Likewise.
88 (vst1q_f32_x2): Likewise.
89 (vst1q_f64_x2): Likewise.
90 (vst1q_p64_x2): Likewise.
91 (vst1_s64_x3): Likewise.
92 (vst1_u64_x3): Likewise.
93 (vst1_f64_x3): Likewise.
94 (vst1_s8_x3): Likewise.
95 (vst1_p8_x3): Likewise.
96 (vst1_s16_x3): Likewise.
97 (vst1_p16_x3): Likewise.
98 (vst1_s32_x3): Likewise.
99 (vst1_u8_x3): Likewise.
100 (vst1_u16_x3): Likewise.
101 (vst1_u32_x3): Likewise.
102 (vst1_f16_x3): Likewise.
103 (vst1_f32_x3): Likewise.
104 (vst1_p64_x3): Likewise.
105 (vst1q_s8_x3): Likewise.
106 (vst1q_p8_x3): Likewise.
107 (vst1q_s16_x3): Likewise.
108 (vst1q_p16_x3): Likewise.
109 (vst1q_s32_x3): Likewise.
110 (vst1q_s64_x3): Likewise.
111 (vst1q_u8_x3): Likewise.
112 (vst1q_u16_x3): Likewise.
113 (vst1q_u32_x3): Likewise.
114 (vst1q_u64_x3): Likewise.
115 (vst1q_f16_x3): Likewise.
116 (vst1q_f32_x3): Likewise.
117 (vst1q_f64_x3): Likewise.
118 (vst1q_p64_x3): Likewise.
120 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
122 * config/msp430/msp430.c (msp430_output_labelref): Prepend
123 user_label_prefix to name.
125 * tree-core.h: Update comment about the format of NAME string
126 passed to handler in attribute_spec.
128 * config/msp430/msp430.md: Remove erroneous subreg expression from
129 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
130 zero_extend{q,h}isi2.
132 2018-05-30 Borislav Petkov <bp@suse.de>
134 * doc/extend.texi: Document some architecture specific
135 constraints and sort entries.
137 2018-05-30 Martin Sebor <msebor@redhat.com>
139 PR middle-end/85369
140 * builtins.c (expand_builtin_stpcpy_1): New function.
141 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
142 only if the former succeeds.
144 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
146 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
147 in saphira.
149 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
151 * doc/invoke.texi (-flinker-output): Document
153 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
155 * passes.c (ipa_write_summaries): Only modify statements if body
156 is in memory.
157 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
158 incrementally linking.
159 (ipa_passes): Likewise.
160 * lto-cgraph.c (lto_output_node): When incrementally linking do not
161 pass down resolution info.
162 * common.opt (flag_incremental_link): Update info.
163 * gcc.c (plugin specs): Turn flinker-output=* to
164 -plugin-opt=-linker-output-known
165 * toplev.c (compile_file): Also cut compilation when doing incremental
166 link.
167 * flag-types. (enum lto_partition_model): Add
168 LTO_LINKER_OUTPUT_NOLTOREL.
169 (invoke.texi): Add -flinker-output docs.
170 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
171 link same way as WPA; do not stream in dead initializers.
173 * dwarf2out.c (dwarf2out_die_ref_for_decl,
174 darf2out_register_external_decl): Support incremental link.
176 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
178 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
180 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
182 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
183 it down to simple_object_copy_lto_debug_sections.
184 (run_gcc): Determine incremental LTO link time and configure
185 lto1 into non-wpa mode, disable renaming of debug sections.
187 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
189 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
190 descriptions of various incorrectly documented functions.
192 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
194 Revert:
195 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
196 address check not strict.
198 2018-05-30 Richard Biener <rguenther@suse.de>
200 PR tree-optimization/85964
201 * tracer.c (better_p): Drop initialized count check, we only
202 call the function with initialized counts now.
203 (find_best_successor): Do find a best edge if one
204 has uninitialized count.
205 (find_best_predecessor): Likewise. Do BB frequency check only
206 if count is initialized.
208 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
210 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
211 (aarch64_ldrstr_offset_compare): New.
212 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
213 load/store orderings.
214 (aarch64_gen_adjusted_ldpstp): Likewise.
216 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
218 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
219 Check for subset of GENERAL_REGS and FP_REGS.
220 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
221 r=w alternative.
223 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
225 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
226 and wi::to_poly_offset. Add the current offset and then check
227 whether the sum fits, rather than using an unchecked addition of
228 a checked term. Check for a shwi rather than a uhwi.
229 * expr.c (get_bit_range): Use tree_to_poly_uint64.
230 (store_constructor): Use poly_int_tree_p.
231 (expand_expr_real_1): Likewise.
232 * function.c (assign_temp): Likewise.
233 * fold-const.c (const_binop): Use poly_int_tree_p and
234 wi::to_poly_offset.
235 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
236 division.
237 * ipa-icf-gimple.c (func_checker::compare_operand): Use
238 to_poly_offset for MEM offsets.
239 * ipa-icf.c (sem_variable::equals): Likewise.
240 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
241 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
242 wi::to_poly_offset for BIT_FIELD_REF offsets.
243 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
244 wi::to_poly_offset.
245 * var-tracking.c (emit_note_insn_var_location): Use
246 tree_to_poly_uint64.
248 2018-05-29 Jim Wilson <jimw@sifive.com>
250 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
252 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
254 PR target/85950
255 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
256 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
257 sequence.
258 (sse4_1_round<mode>2): Use nonimmediate_operand
259 for operand 1 predicate.
261 2018-05-29 Martin Sebor <msebor@redhat.com>
262 Richard Biener <rguenther@suse.de>
264 PR testsuite/85888
265 * calls.c (get_size_range): Call determine_value_range instead
266 of get_value_range..
267 * tree-vrp.h (determine_value_range): Declared new function.
268 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
270 2018-05-29 Richard Biener <rguenther@suse.de>
272 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
273 sure to use non-pattern stmts for get_earlier_stmt arguments.
274 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
275 called on pattern stmts.
276 (get_later_stmt): Likewise.
278 2018-05-29 Martin Liska <mliska@suse.cz>
280 PR gcov-profile/85759
281 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
282 env variables.
284 2018-05-29 Jakub Jelinek <jakub@redhat.com>
286 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
287 VEC_UNPACK_*_EXPR.
288 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
289 VEC_PACK_*_EXPR.
291 PR target/85918
292 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
293 VEC_PACK_FLOAT_EXPR): New tree codes.
294 * tree-pretty-print.c (op_code_prio): Handle
295 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
296 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
297 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
298 * tree-inline.c (estimate_operator_cost): Likewise.
299 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
300 * fold-const.c (const_binop): Likewise.
301 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
302 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
303 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
304 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
305 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
306 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
307 * expr.c (expand_expr_real_2): Likewise.
308 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
309 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
310 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
311 optabs.
312 * optabs.c (expand_widen_pattern_expr): For
313 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
314 sign from result type rather than operand's type.
315 (expand_binop_directly): For vec_packu_float_optab and
316 vec_packs_float_optab allow result type to be different from operand's
317 type.
318 * optabs-tree.c (optab_for_tree_code): Handle
319 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
320 VEC_PACK_FLOAT_EXPR. Formatting fixes.
321 * tree-vect-generic.c (expand_vector_operations_1): Handle
322 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
323 VEC_PACK_FLOAT_EXPR.
324 * tree-vect-stmts.c (supportable_widening_operation): Handle
325 FIX_TRUNC_EXPR.
326 (supportable_narrowing_operation): Handle FLOAT_EXPR.
327 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
328 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
329 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
330 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
331 mode attributes.
332 (vec_pack<floatprefix>_float_<mode>): New expander.
333 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
334 attributes.
335 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
336 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
337 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
338 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
339 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
340 Document.
341 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
342 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
343 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
344 VEC_PACK_FLOAT_EXPR): Document.
346 2018-05-29 Richard Biener <rguenther@suse.de>
348 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
349 member.
350 (stmt_vec_info_vec): Make pointer.
351 (init_stmt_vec_info_vec): Remove.
352 (free_stmt_vec_info_vec): Likewise.
353 (set_stmt_vec_info_vec): New function.
354 (free_stmt_vec_infos): Likewise.
355 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
356 (set_vinfo_for_stmt): Likewise.
357 (get_earlier_stmt): Likewise.
358 (get_later_stmt): Likewise.
359 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
360 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
361 (vec_info::~vec_info): Free stmt_vec_infos.
362 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
363 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
364 (pass_slp_vectorize::execute): Likewise.
365 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
366 (free_stmt_vec_info_vec): Likewise.
367 (set_stmt_vec_info_vec): New function.
368 (free_stmt_vec_infos): Likewise.
369 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
370 the global stmt_vec_info_vec.
371 * tree-parloops.c (gather_scalar_reductions): Use
372 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
373 vector.
375 2018-05-29 Richard Biener <rguenther@suse.de>
377 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
379 2018-05-29 Martin Liska <mliska@suse.cz>
380 David Malcolm <dmalcolm@redhat.com>
382 * vec.c (test_reverse): New.
383 (vec_c_tests): Add new test.
384 * vec.h (vl_ptr>::reverse): New function.
386 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
388 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
390 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
391 and later.
393 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
395 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
397 2018-05-28 Richard Biener <rguenther@suse.de>
399 PR tree-optimization/85933
400 * tree-vect-data-refs.c (vect_record_base_alignments): Only
401 look at stmts marked as vectorizable.
403 2018-05-28 Richard Biener <rguenther@suse.de>
405 PR tree-optimization/85934
406 * tree-vect-generic.c (expand_vector_operations_1): Hoist
407 vector boolean check before scalar optimization.
409 2018-05-28 Jakub Jelinek <jakub@redhat.com>
411 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
412 for armv5te.
414 2018-05-28 Mark Wielaard <mark@klomp.org>
416 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
417 if it is an expression containing a minus sign.
419 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
421 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
423 2018-05-27 Paul Koning <ni1d@arrl.net>
425 * config/pdp11/pdp11.md (truncsihi2): Remove.
427 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
428 Chung-Ju Wu <jasonwucj@gmail.com>
430 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
431 implementation.
432 (unaligned_store_dw): Ditto.
433 * config/nds32/nds32-memory-manipulation.c
434 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
435 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
436 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
437 (emit_setmem_word_loop): Rename to ...
438 (emit_setmem_doubleword_loop): ... this.
439 (nds32_gen_dup_4_byte_to_word_value): New function.
440 (nds32_gen_dup_8_byte_to_double_word_value): New function.
441 (nds32_expand_setmem_loop): Refine implementation.
442 (nds32_expand_setmem_loop_v3m): Ditto.
443 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
444 pattern.
446 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
448 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
450 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
452 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
453 (nds32_init_machine_status): Initialize machine->attr_naked_p and
454 machine->attr_no_prologue_p.
455 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
456 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
457 (nds32_expand_epilogue): Consider attr_naked_p.
458 (nds32_expand_epilogue_v3pop): Likewise.
459 (nds32_can_use_return_insn): Likewise.
460 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
461 attr_no_prologue_p fields.
462 * config/nds32/nds32.opt (mret-in-naked-func): New option.
464 2018-05-27 Jakub Jelinek <jakub@redhat.com>
466 PR target/85918
467 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
468 attributes.
469 * config/i386/sse.md
470 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
471 Rename to ...
472 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
473 ... this.
474 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
475 Rename to ...
476 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
477 ... this.
478 (*<floatsuffix>floatv2div2sf2): Rename to ...
479 (*float<floatunssuffix>v2div2sf2): ... this.
480 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
481 (float<floatunssuffix>v2div2sf2_mask): ... this.
482 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
483 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
484 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
485 to ...
486 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
487 ... this.
488 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
489 Rename to ...
490 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
491 ... this.
492 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
493 Rename to ...
494 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
495 ... this.
496 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
497 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
498 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
499 gen_ufix_truncv8dfv8si2.
500 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
501 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
502 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
503 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
504 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
505 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
506 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
507 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
509 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
511 PR target/85900
512 PR target/85345
513 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
515 2018-05-25 Jim Wilson <jimw@sifive.com>
517 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
518 * config/riscv/riscv.c (struct machine_function): Add
519 interrupt_handler_p and attribute_checked_p fields.
520 (riscv_attribute_table): Add interrupt.
521 (riscv_interrupt_type_p): New.
522 (riscv_save_reg_p): Save extra regs for interrupt handler.
523 (riscv_use_save_libcall): Return false for interrupt handler.
524 (riscv_first_stack_step): Add forward declaration.
525 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
526 for interrupt handler with large frame. Use it for saved reg list.
527 (riscv_expand_prologue): Move flag_stack_usage_info support to
528 eliminate duplication.
529 (riscv_expand_epilogue): Generate mret for interrupt handler.
530 (riscv_epilogue_uses): New.
531 (riscv_can_use_return_insn): Return false for interrupt handler.
532 (riscv_function_ok_for_sibcall): Likewise.
533 (riscv_set_current_function): Add interrupt handler support.
534 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
535 * config/riscv/riscv.md (UNSPECV_MRET): New.
536 (GP_REGNUM): New.
537 (riscv_frflags, riscv_fsflags): Use tab after opcode.
538 (riscv_mret): New.
539 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
541 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
543 PR tree-optimization/85712
544 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
545 this candidate has already been replaced in-situ by a copy.
547 2018-05-25 Jason Merrill <jason@redhat.com>
549 PR c++/80485 - inline function non-zero address.
550 * symtab.c (nonzero_address): Check DECL_COMDAT.
552 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
554 PR target/83628
555 * config/alpha/alpha.md (ashlsi3): New insn pattern.
556 (*ashlsi_se): Rename from *ashldi_se. Define as sign
557 extension of SImode operation. Use const123_operand predicate.
558 (*saddsi_1): Remove.
559 (*saddl_se_1): Ditto.
560 (*ssubsi_1): Ditto.
561 (*ssubl_se_1): Ditto.
562 * config/alpha/predicates.md (const123_operand): New predicate.
563 * config/alpha/constraints.md (P): Use IN_RANGE.
565 2018-05-25 Richard Biener <rguenther@suse.de>
567 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
568 defaulted to true.
569 (ref_maybe_used_by_stmt_p): Likewise.
570 (stmt_may_clobber_ref_p): Likewise.
571 (stmt_may_clobber_ref_p_1): Likewise.
572 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
573 and pass it along.
574 (ref_maybe_used_by_stmt_p): Likewise.
575 (stmt_may_clobber_ref_p): Likewise.
576 (stmt_may_clobber_ref_p_1): Likewise.
577 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
578 the alias oracle to disambiguate DRs with stmts DR analysis
579 couldn't handle.
580 (vect_analyze_data_refs): Do not give up on not analyzable
581 DRs for BB vectorization. Remove code truncating the dataref
582 vector.
584 2018-05-25 Jakub Jelinek <jakub@redhat.com>
586 PR target/85832
587 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
588 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
589 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
591 2018-05-25 Richard Biener <rguenther@suse.de>
593 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
594 function, combining stmt data ref gathering and fatal analysis
595 parts.
596 (vect_analyze_data_refs): Remove now redudnant code and simplify.
597 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
598 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
599 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
600 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
602 2018-05-25 Bin Cheng <bin.cheng@arm.com>
604 PR tree-optimization/85720
605 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
606 SCC if all partitions are builtins.
607 (version_loop_by_alias_check): New parameter. Generate cancelable
608 runtime alias check if all partitions are builtins.
609 (distribute_loop): Update call to above function.
611 2018-05-25 Bin Cheng <bin.cheng@arm.com>
613 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
614 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
615 (parm_default_def_partition_arg): Ditto.
616 (set_parm_default_def_partition): Ditto.
617 (get_parm_default_def_partitions): Ditto and make it static.
618 (get_undefined_value_partitions): Ditto and make it static.
619 (remove_ssa_form): Refactor call to init_var_map here.
620 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
621 computation for loop region.
622 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
623 (register_default_def): Delete.
624 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
625 (parm_default_def_partition_arg): Ditto.
626 (set_parm_default_def_partition): Ditto.
627 (get_parm_default_def_partitions): Ditto and make it static.
628 (get_undefined_value_partitions): Ditto and make it static.
629 (coalesce_with_default, coalesce_with_default): Update comment.
630 (create_coalesce_list_for_region): New func factored out from
631 create_outofssa_var_map.
632 (populate_coalesce_list_for_outofssa): New func factored out from
633 create_outofssa_var_map and coalesce_ssa_name.
634 (create_outofssa_var_map): Delete.
635 (coalesce_ssa_name): Refactor to support live range computation.
636 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
637 (get_parm_default_def_partitions): Delete.
638 (get_undefined_value_partitions): Ditto.
639 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
640 computation for loop region.
641 (new_tree_live_info, loe_visit_block): Ditto.
642 (live_worklist, set_var_live_on_entry): Ditto.
643 (calculate_live_on_exit, verify_live_on_entry): Ditto.
644 * tree-ssa-live.h (struct _var_map): New fields.
645 (init_var_map): Change decl.
646 (region_contains_p): New.
648 2018-05-25 Bin Cheng <bin.cheng@arm.com>
650 * tree-ssa-live.h (live_merge_and_clear): Delete.
652 2018-05-25 Richard Biener <rguenther@suse.de>
654 PR c++/85912
655 * tree-dump.c (dequeue_and_dump): Remove access to removed
656 operand 2 of a SWITCH_EXPR.
658 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
660 * doc/sourcebuild.texi (vect_double_cond_arith): Include
661 multiplication and division.
662 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
663 (cond_udiv@var{m}, cond_umod@var{m}): Document.
664 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
665 (cond_udiv_optab, cond_umod_optab): New optabs.
666 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
667 (IFN_COND_RDIV): New internal functions.
668 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
669 TRUNC_MOD_EXPR and RDIV_EXPR.
670 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
671 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
672 New unspecs.
673 (SVE_INT_BINARY): Include mult.
674 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
675 (optab, sve_int_op): Handle mult.
676 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
677 UNSPEC_COND_DIV.
678 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
679 for SVE_INT_BINARY_SD.
681 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
683 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
684 (optab, sve_int_op): Handle div and udiv.
685 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
686 for SVE_INT_BINARY_SD.
687 (*<optab><mode>3): New insn for the same.
689 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
691 * tree-vect-patterns.c: Include predict.h.
692 (vect_recog_divmod_pattern): Restrict check for division support
693 to when optimizing for size.
695 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
697 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
698 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
699 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
700 (gimple_match_op::set_op): Likewise.
701 (gimple_resimplify4): Declare.
702 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
703 (expr::gen_transform): Likewise.
704 (decision_tree::gen): Generate a simplification routine for 4 operands.
705 * gimple-match-head.c (gimple_simplify): Add an overload for
706 4 operands. In the top-level function, handle up to 4 call
707 arguments and call gimple_resimplify4.
708 (gimple_resimplify4): New function.
709 (build_call_internal): Pass a fourth operand.
710 (maybe_push_to_seq): Likewise.
711 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
712 Fold VEC_COND_EXPRs of an operation and a default value into
713 an IFN_COND_* function if possible.
714 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
715 New unspecs.
716 (SVE_COND_FP_BINARY): Include them.
717 (optab, sve_fp_op): Handle them.
718 (SVE_INT_BINARY_REV): New code iterator.
719 (SVE_COND_FP_BINARY_REV): New int iterator.
720 (commutative): New int attribute.
721 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
722 Declare.
723 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
724 function.
725 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
726 (*cond_<optab><mode>): New patterns for reversed operands.
728 2018-05-25 Richard Biener <rguenther@suse.de>
730 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
731 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
732 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
733 (STMT_VINFO_GROUPED_ACCESS): Adjust.
734 * tree-vect-data-refs.c (everywhere): Adjust users.
735 * tree-vect-loop.c (everywhere): Likewise.
736 * tree-vect-slp.c (everywhere): Likewise.
737 * tree-vect-stmts.c (everywhere): Likewise.
738 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
740 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
742 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
743 Rename to...
744 (gcc_cv_as_section_exclude): ... this.
745 Try Solaris as #exclude syntax.
746 * configure: Regenerate.
747 * config.in: Regenerate.
748 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
749 SECTION_EXCLUDE.
750 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
751 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
753 * varasm.c (default_elf_asm_named_section): Don't check if
754 HAVE_GAS_SECTION_EXCLUDE is defined.
756 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
758 * doc/md.texi: Update the documentation of the cond_* optabs
759 to mention the new final operand. Fix GET_MODE_NUNITS call.
760 Describe the scalar case too.
761 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
762 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
763 instead of 2.
764 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
765 (get_conditional_internal_fn): Update comment.
766 * tree-vect-loop.c (vectorizable_reduction): Pass the original
767 accumulator value as a final argument to conditional functions.
768 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
769 a define_expand and add an "else" operand. Assert for now that
770 the else operand is equal to operand 2. Use SVE_INT_BINARY and
771 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
772 (*cond_<optab><mode>): New patterns.
773 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
774 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
775 (UNSPEC_COND_EOR): Delete.
776 (optab): Remove associated mappings.
777 (SVE_INT_BINARY): New code iterator.
778 (sve_int_op): Remove int attribute and add "minus" to the code
779 attribute.
780 (SVE_COND_INT_OP): Delete.
781 (SVE_COND_FP_OP): Rename to...
782 (SVE_COND_FP_BINARY): ...this.
784 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
786 * optabs.c (can_reuse_operands_p): New function.
787 (maybe_legitimize_operands): Try to reuse the results for
788 earlier operands.
790 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
792 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
793 Add {q} suffix to insn mnemonic.
795 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
797 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
798 (msp430_warn_func_return): New.
800 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
802 * fold-const.c (tree_nonzero_bits): New function.
803 * fold-const.h (tree_nonzero_bits): Likewise.
804 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
805 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
807 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
809 PR target/85900
810 PR target/85345
811 * varasm.c (assemble_alias): Check ifunc_resolver only on
812 FUNCTION_DECL.
814 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
816 PR target/85903
817 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
818 when memory input operand is handled.
820 2018-05-24 Luis Machado <luis.machado@linaro.org>
822 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
823 global.
824 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
826 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
828 * match.pd: Delay FMA folds until after vectorization.
830 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
832 PR target/83009
833 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
834 address check not strict.
836 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
838 * gimple-match.h (gimple_match_op): New class.
839 (mprts_hook): Replace parameters with a gimple_match_op *.
840 (maybe_build_generic_op): Likewise.
841 (gimple_simplified_result_is_gimple_val): Replace parameters with
842 a const gimple_match_op *.
843 (gimple_simplify): Replace code_helper * and tree * parameters with
844 a gimple_match_op * parameter.
845 (gimple_resimplify1): Replace code_helper *, tree and tree *
846 parameters with a gimple_match_op * parameter.
847 (gimple_resimplify2): Likewise.
848 (gimple_resimplify3): Likewise.
849 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
850 parameters with a gimple_match_op * parameter.
851 * gimple-match-head.c (gimple_simplify): Change prototypes of
852 auto-generated functions to take a gimple_match_op * instead of
853 separate code_helper * and tree * parameters. Make the same
854 change in the top-level overload and update calls to the
855 gimple_resimplify routines. Update calls to the auto-generated
856 functions and to maybe_push_res_to_seq in the publicly-facing
857 operation-specific gimple_simplify overloads.
858 (gimple_match_op::MAX_NUM_OPS): Define.
859 (gimple_resimplify1): Replace rcode and ops with a single res_op
860 parameter. Update call to gimple_simplify.
861 (gimple_resimplify2): Likewise.
862 (gimple_resimplify3): Likewise.
863 (mprts_hook): Replace parameters with a gimple_match_op *.
864 (maybe_build_generic_op): Likewise.
865 (build_call_internal): Replace type, nargs and ops with
866 a gimple_match_op *.
867 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
868 with a single gimple_match_op *. Update calls to mprts_hook,
869 build_call_internal and gimple_simplified_result_is_gimple_val.
870 Factor out code that is common to the tree_code and combined_fn cases.
871 * genmatch.c (expr::gen_transform): Replace tem_code and
872 tem_ops with a gimple_match_op called tem_op. Update calls
873 to the gimple_resimplify functions and maybe_push_res_to_seq.
874 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
875 res_ops. Update call to the gimple_resimplify functions.
876 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
877 (decision_tree::gen): Make the functions take a gimple_match_op *
878 called res_op instead of separate res_code and res_ops parameters.
879 Update call accordingly.
880 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
881 and ops with a single res_op parameter. Update calls to
882 maybe_build_generic_op and maybe_push_res_to_seq.
883 (fold_stmt_1): Update calls to gimple_simplify and
884 replace_stmt_with_simplification.
885 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
886 and gimple_simplified_result_is_gimple_val.
887 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
888 gimple_simplify.
889 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
890 with a gimple_match_op *.
891 (vn_nary_build_or_lookup): Likewise. Update call to
892 vn_nary_build_or_lookup_1.
893 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
894 gimple_match_op *. Update calls to the gimple_resimplify routines
895 and to gimple_simplified_result_is_gimple_val.
896 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
897 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
898 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
899 (visit_nary_op): Likewise.
900 (visit_reference_op_load): Likewise.
902 2018-05-23 Luis Machado <luis.machado@linaro.org>
904 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
905 modifier for printing the step amount.
907 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
909 PR target/78849
910 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
911 types.
913 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
915 * doc/sourcebuild.texi (Endianness): New subsubsection.
917 2018-05-23 Luis Machado <luis.machado@linaro.org>
919 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
920 <prefetch_dynamic_strides>: New const bool field.
921 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
922 prefetch_dynamic_strides.
923 (exynosm1_prefetch_tune): Likewise.
924 (thunderxt88_prefetch_tune): Likewise.
925 (thunderx_prefetch_tune): Likewise.
926 (thunderx2t99_prefetch_tune): Likewise.
927 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
928 false.
929 (aarch64_override_options_internal): Update to set
930 PARAM_PREFETCH_DYNAMIC_STRIDES.
931 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
932 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
933 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
934 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
935 prefetch-dynamic-strides setting.
937 2018-05-23 Luis Machado <luis.machado@linaro.org>
939 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
940 <minimum_stride>: New const int field.
941 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
942 minimum_stride field defaulting to -1.
943 (exynosm1_prefetch_tune): Likewise.
944 (thunderxt88_prefetch_tune): Likewise.
945 (thunderx_prefetch_tune): Likewise.
946 (thunderx2t99_prefetch_tune): Likewise.
947 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
948 <default_opt_level>: Set to 3.
949 (aarch64_override_options_internal): Update to set
950 PARAM_PREFETCH_MINIMUM_STRIDE.
951 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
952 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
953 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
954 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
955 stride is constant and is below the minimum stride threshold.
957 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
959 * config/arm/arm-cpus.in (mode26): Delete.
960 (armv4): Delete mode26 reference.
961 * config/arm/arm.c (arm_configure_build_target): Delete use of
962 isa_bit_mode26.
964 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
966 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
967 New insn pattern.
968 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
969 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
970 for non-SSE modes.
971 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
972 (floatunsdidf2): Ditto.
974 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
976 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
977 (fixuns_trunc<mode>si2_avx512f): Ditto.
978 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
979 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
980 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
982 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
984 PR rtl-optimization/79985
985 * df-scan.c (df_insn_refs_collect): Remove special case for
986 global registers and asm statements.
988 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
990 * extend.texi (Global Register Variables): Rewrite the bullet list.
991 Note that the register is available for allocation. Note that access
992 via inline asm must use constraints. Add note about async-signal
993 handlers. Remove paragraph about automagic register selection.
995 2018-05-23 Richard Biener <rguenther@suse.de>
997 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
998 of fixed offset from memset VN.
1000 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1002 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1003 first_interp field.
1004 (alloc_cand_and_find_basis): Initialize first_interp field.
1005 (slsr_process_mul): Modify first_interp field.
1006 (slsr_process_add): Likewise.
1007 (slsr_process_cast): Modify first_interp field for each new
1008 interpretation.
1009 (slsr_process_copy): Likewise.
1010 (dump_candidate): Dump first_interp field.
1011 (replace_mult_candidate): Process all interpretations, not just
1012 subsequent ones.
1013 (replace_rhs_if_not_dup): Likewise.
1014 (replace_one_candidate): Likewise.
1016 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1018 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1019 Add new boolean.
1020 (aarch64_needs_frame_chain): New function.
1021 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1023 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1025 PR target/84882
1026 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1027 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1028 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1029 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1030 as true for strict-align.
1031 (aarch64_can_inline_p): Perform checks even when callee has no
1032 attributes to check for strict alignment.
1033 * doc/extend.texi (AArch64 Function Attributes): Document
1034 no-strict-align.
1035 * doc/invoke.texi: (AArch64 Options): Likewise.
1037 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1039 PR tree-optimization/85853
1040 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1041 the handling of the root of the node to...
1042 (vect_slp_analyze_node_operations_1): ...this new function,
1043 and run the whole thing with the child nodes' def types
1044 set according to their SLP node's def type.
1046 2018-05-23 Richard Biener <rguenther@suse.de>
1048 PR middle-end/85874
1049 * tree-data-ref.c (create_runtime_alias_checks): Defer
1050 and ignore overflow warnings.
1052 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1054 PR tree-optimization/85822
1055 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1056 constants.
1058 2018-05-23 Richard Biener <rguenther@suse.de>
1060 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1061 memset constants via native_interpret_expr.
1063 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1065 PR target/85345
1066 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1067 attribute.
1068 (cgraph_node::create_alias): Likewise.
1069 (cgraph_node::get_availability): Check ifunc_resolver instead
1070 of looking up ifunc attribute.
1071 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1072 * varasm.c (do_assemble_alias): Likewise.
1073 (assemble_alias): Likewise.
1074 (default_binds_local_p_3): Likewise.
1075 * cgraph.h (cgraph_node): Add ifunc_resolver.
1076 (cgraph_node::only_called_directly_or_aliased_p): Return false
1077 for IFUNC resolver.
1078 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1079 attribute.
1080 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1081 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1082 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1083 instead of looking up ifunc attribute.
1085 2018-05-22 Luis Machado <luis.machado@linaro.org>
1087 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1089 2018-05-22 Martin Sebor <msebor@redhat.com>
1091 PR middle-end/85359
1092 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1093 only when expasion succeeds.
1094 (expand_builtin_strcmp): Same.
1095 (expand_builtin_strncmp): Same.
1097 2018-05-22 Martin Sebor <msebor@redhat.com>
1099 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1101 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1102 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1104 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1105 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1106 code.
1107 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1108 Add check for legitimate_address.
1109 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1110 (aarch64_swap_ldrstr_operands): New.
1111 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1112 Define prototype.
1114 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1115 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1117 * config/aarch64/aarch64.md: New patterns to generate stp
1118 and ldp.
1119 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1120 single words and double words.
1121 (load_pair_sw, load_pair_dw): Likewise.
1122 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1123 Delete.
1124 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1125 Delete.
1126 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1127 for different mode ldpstp and add peephole for merged zero stores.
1128 Likewise for loads.
1129 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1130 Add size check.
1131 (aarch64_gen_store_pair): Rename calls to match new patterns.
1132 (aarch64_gen_load_pair): Rename calls to match new patterns.
1133 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1134 (load_pair<DREG:mode><DREG2:mode>): ... This.
1135 (store_pair<mode>): Rename to...
1136 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1137 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1138 New mode iterators.
1139 (V_INT_EQUIV): Handle SImode.
1140 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1141 New predicate.
1143 2018-05-22 Martin Sebor <msebor@redhat.com>
1145 PR c/85623
1146 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1147 or ajust the presumed bound on an operation to avoid unnecessary
1148 warnings.
1150 2018-05-22 Martin Sebor <msebor@redhat.com>
1152 PR tree-optimization/85826
1153 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1154 assuming that a DECL necesarily has a constant size.
1156 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1158 PR middle-end/85862
1159 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1161 2018-05-22 Richard Biener <rguenther@suse.de>
1163 PR tree-optimization/85834
1164 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1165 non-constant and non-zero memset arguments.
1167 2018-05-22 Martin Liska <mliska@suse.cz>
1169 PR ipa/85607
1170 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1172 2018-05-22 Richard Biener <rguenther@suse.de>
1174 PR tree-optimization/85863
1175 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1176 comparisons when vectype is specified.
1177 (vectorizable_condition): Do not specify vectype for
1178 vect_is_simple_cond when SLP vectorizing.
1180 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1182 PR target/85657
1183 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1184 define __ibm128 as long double.
1185 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1186 as a distinct type when IEEE 128-bit support is enabled.
1187 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1188 128-bit types to use the correct functions.
1189 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1190 convert between 128-bit floating point types that have different
1191 modes but the same representation, instead of using gen_lowpart to
1192 makean alias.
1193 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1194 KFmode.
1195 (IFKF_reg): New attributes to give the register constraints for
1196 IFmode and KFmode.
1197 (extend<mode>tf2_internal): New insns to mark an explicit
1198 conversion between 128-bit floating point types that have a
1199 different mode but share the same representation.
1201 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1203 PR tree-optimization/85814
1204 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1205 a null return from get_strinfo when unsharing the next
1206 strinfo in the chain.
1208 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1210 PR gcc/84923
1211 * varasm.c (weak_finish): Clean up weak_decls.
1213 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1215 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1216 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1217 UNSPEC_UADALP values.
1218 * config/aarch64/iterators.md (ABAL): New int iterator.
1219 (ABDL2): Likewise.
1220 (ADALP): Likewise.
1221 (sur): Add mappings for the above.
1222 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1223 New define_insn.
1224 (aarch64_<sur>abal<mode>_4): Likewise.
1225 (aarch64_<sur>adalp<mode>_3): Likewise.
1226 (<sur>sadv16qi): New define_expand.
1228 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1230 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1231 (*movdf_internal): Ditto.
1232 (*rcpsf2_sse): Ditto.
1233 (*rsqrtsf2_sse): Ditto.
1234 (*sqrt<mode>2_sse): Ditto.
1236 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1238 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1239 eor3q<mode>4.
1240 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1241 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1242 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1243 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1244 vbcaxq_s64): New.
1245 * config/aarch64/arm_neon.h: Likewise.
1246 * config/aarch64/iterators.md (VQ_I): New.
1248 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1250 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1251 arc*-*-linux*.
1252 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1253 MULTILIB_DIRNAMES
1255 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1257 * config/nds32/constraints.md (S): New constraint.
1258 * config/nds32/nds32.md (call_internal): Use constraint S.
1259 (call_value_internal): Likewise.
1260 (sibcall_internal): Likewise.
1261 (sibcall_value_internal): Likewise.
1263 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1264 Chung-Ju Wu <jasonwucj@gmail.com>
1266 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1267 into consideration.
1269 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1270 Chung-Ju Wu <jasonwucj@gmail.com>
1272 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1273 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1274 (nds32_rtx_costs_impl): Simplify.
1275 (nds32_address_cost_impl): Simplify.
1276 (nds32_init_rtx_costs): New function.
1277 (nds32_rtx_costs_speed_prefer): Likewise.
1278 (nds32_rtx_costs_size_prefer): Likewise.
1279 (nds32_address_cost_speed_prefer): Likewise.
1280 (nds32_address_cost_speed_fwprop): Likewise.
1281 (nds32_address_cost_size_prefer): Likewise.
1282 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1283 * config/nds32/nds32.c (nds32_option_override): Use
1284 nds32_init_rtx_costs function.
1286 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1289 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1290 (TARGET_PIPELINE_N8): Likewise.
1291 (TARGET_PIPELINE_N10): Likewise.
1292 (TARGET_PIPELINE_N13): Likewise.
1293 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1295 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1297 * config/nds32/nds32-fpu.md: Update copyright year.
1299 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1301 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1303 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1305 * config/nds32/nds32.c
1306 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1307 * config/nds32/nds32.opt (minline-asm-r15): New option.
1309 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1311 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1312 MASK_HW_ABS.
1313 * config/nds32/nds32.md (abssi2): New pattern.
1315 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1317 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1318 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1319 Merge insn pattern from sse_cvtsi2ss<round_name> and
1320 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1321 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1322 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1323 using SWI48 mode iterator.
1324 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1325 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1326 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1327 pattern from sse_cvttss2si<round_saeonly_name>
1328 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1329 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1330 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1331 using SWI48 mode iterator.
1332 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1333 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1334 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1335 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1336 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1337 using SWI48 mode iterator.
1338 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1339 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1340 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1341 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1342 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1343 SWI48 mode iterator.
1344 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1345 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1346 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1347 pattern from sse_cvttsd2si<round_saeonly_name>
1348 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1350 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1352 * config/nds32/nds32-md-auxiliary.c
1353 (nds32_valid_smw_lwm_base_p): Refine.
1354 (nds32_output_smw_single_word): Refine.
1355 (nds32_output_smw_double_word): New.
1356 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1358 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1360 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1361 (nds32_output_stack_pop): Refine.
1362 (nds32_expand_unaligned_load): Refine.
1363 (nds32_expand_unaligned_store): Refine.
1365 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1366 Chung-Ju Wu <jasonwucj@gmail.com>
1368 * config/nds32/constants.md: Add TP_REGNUM constant.
1369 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1370 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1371 UNSPEC_ADD32.
1372 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1373 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1374 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1375 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1376 and PIC code generation.
1377 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1378 code generation.
1379 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1380 optimization.
1381 * config/nds32/nds32.md: Support TLS and PIC.
1382 * config/nds32/nds32.c: Support TLS and PIC.
1383 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1384 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1385 predicate.
1387 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1389 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1390 mode with E_ prefix.
1392 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1393 Chung-Ju Wu <jasonwucj@gmail.com>
1395 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1396 * config/nds32/nds32-md-auxiliary.c
1397 (symbolic_reference_mentioned_p): New.
1398 (nds32_legitimize_ict_address): New.
1399 (nds32_expand_ict_move): New.
1400 (nds32_indirect_call_referenced_p): New.
1401 (nds32_symbol_binds_local_p): Delete.
1402 (nds32_long_call_p): Modify.
1403 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1404 * config/nds32/nds32-protos.h
1405 (symbolic_reference_mentioned_p): Declare.
1406 (nds32_legitimize_ict_address): Declare.
1407 (nds32_expand_ict_move): Declare.
1408 (nds32_indirect_call_referenced_p): Declare.
1409 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1410 (nds32_relax_group): Use nds32_ict_const_p as condition.
1411 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1412 (nds32_asm_file_start): Output ict_model directive in asm code.
1413 (nds32_legitimate_address_p): Consider indirect call.
1414 (nds32_print_operand): Consider indirect call.
1415 (nds32_print_operand_address): Consider indirect call.
1416 (nds32_insert_attributes): Handle "indirect_call" attribute.
1417 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1418 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1419 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1420 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1421 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1422 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1423 (TARGET_ICT_MODEL_SMALL): Define.
1424 (TARGET_ICT_MODEL_LARGE): Define.
1425 * config/nds32/nds32.md (movsi): Consider ict model.
1426 (call, call_value): Consider ict model.
1427 (sibcall, sibcall_value): Consider ict model.
1428 * config/nds32/nds32.opt (mict-model): New option.
1429 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1430 model.
1432 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1433 Monk Chiang <sh.chiang04@gmail.com>
1434 Jim Wilson <jimw@sifive.com>
1436 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1437 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1438 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1439 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1440 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1441 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1442 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1443 compute save_libcall_adjustment properly.
1444 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1445 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1446 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1447 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1448 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1449 (ABI_SPEC): Handle mabi=ilp32e.
1450 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1451 (RVE): Add RVE mask.
1452 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1453 <-march>: Add rv32e as an example.
1455 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1457 PR c++/82899
1458 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1459 (intra_create_variable_infos): Handle C++ constructors.
1461 2018-05-18 Martin Liska <mliska@suse.cz>
1463 * passes.def: Remove a redundant pass.
1465 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1467 PR bootstrap/85838
1468 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1470 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1472 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1473 (ARMv4): Update.
1474 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1475 (ARMv6m): Update.
1476 (armv2, armv2a, armv3, armv3m): Delete architectures.
1477 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1478 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1479 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1480 Delete cpus.
1481 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1482 (*mulsidi3adddi): Likewise.
1483 (mulsidi3): Likewise.
1484 (*mulsidi3_nov6): Likewise.
1485 (umulsidi3): Likewise.
1486 (umulsidi3_nov6): Likewise.
1487 (umaddsidi4): Likewise.
1488 (*umulsidi3adddi): Likewise.
1489 (smulsi3_highpart): Likewise.
1490 (*smulsi3_highpart_nov6): Likewise.
1491 (umulsi3_highpart): Likewise.
1492 (*umulsi3_highpart_nov6): Likewise.
1493 * config/arm/arm.h (arm_arch3m): Delete.
1494 * config/arm/arm.c (arm_arch3m): Delete.
1495 (arm_option_override_internal): Update armv3-related comment.
1496 (arm_configure_build_target): Delete use of isa_bit_mode32.
1497 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1498 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1499 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1500 (mulsa3): Likewise.
1501 (mulusa3): Likewise.
1502 * config/arm/arm-protos.h (arm_arch3m): Delete.
1503 * config/arm/arm-tables.opt: Regenerate.
1504 * config/arm/arm-tune.md: Likewise.
1505 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1506 deleted architectures.
1508 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1510 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1511 (armv5t, armv5te): New features.
1512 (ARMv5, ARMv5e): Delete fgroups.
1513 (ARMv5t, ARMv5te): Adjust for above changes.
1514 (ARMv6m): Likewise.
1515 (armv5, armv5e): Delete arches.
1516 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1517 arm_arch5.
1518 (*call_reg_arm): Likewise.
1519 (*call_value_reg_armv5): Likewise.
1520 (*call_value_reg_arm): Likewise.
1521 (*call_symbol): Likewise.
1522 (*call_value_symbol): Likewise.
1523 (*sibcall_insn): Likewise.
1524 (*sibcall_value_insn): Likewise.
1525 (clzsi2): Likewise.
1526 (prefetch): Likewise.
1527 (define_split and define_peephole2 dependent on arm_arch5):
1528 Likewise.
1529 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1530 arm_arch5e.
1531 (TARGET_ARM_QBIT): Likewise.
1532 (TARGET_DSP_MULTIPLY): Likewise.
1533 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1534 (arm_arch5, arm_arch5e): Delete.
1535 (arm_arch5t, arm_arch5te): Declare.
1536 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1537 (arm_arch5t): Declare.
1538 (arm_option_reconfigure_globals): Update for the above.
1539 (arm_options_perform_arch_sanity_checks): Update comment, replace
1540 use of arm_arch5 with arm_arch5t.
1541 (use_return_insn): Likewise.
1542 (arm_emit_call_insn): Likewise.
1543 (output_return_instruction): Likewise.
1544 (arm_final_prescan_insn): Likewise.
1545 (arm_coproc_builtin_available): Likewise.
1546 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1547 arm_arch5e with arm_arch5t and arm_arch5te.
1548 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1549 (arm_arch5t, arm_arch5te): Declare.
1550 * config/arm/arm-tables.opt: Regenerate.
1551 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1552 * config/arm/t-multilib: Likewise.
1553 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1554 instead of arm_arch5.
1555 (*call_reg_thumb1): Likewise.
1556 (*call_value_reg_thumb1_v5): Likewise.
1557 (*call_value_reg_thumb1): Likewise.
1558 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1559 unreachable path.
1560 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1562 2018-05-18 Martin Liska <mliska@suse.cz>
1564 PR gcov-profile/84846
1565 * doc/gcov.texi: Document -t option of gcov tool.
1567 2018-05-18 Martin Liska <mliska@suse.cz>
1569 PR gcov-profile/84846
1570 * gcov.c (print_usage): Add new -t option.
1571 (process_args): Handle the option.
1572 (generate_results): Use stdout as output when requested by
1573 the option.
1575 2018-05-18 Martin Liska <mliska@suse.cz>
1577 PR gcov-profile/84846
1578 * coverage.c (coverage_init): Write PWD to .gcno file.
1579 * doc/gcov.texi: Document how working directory is printed.
1580 * gcov-dump.c (dump_gcov_file): Print PWD.
1581 * gcov.c (output_intermediate_file): Likewise.
1582 (read_graph_file): Read PWD string.
1583 (output_lines): Print PWD.
1585 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1587 PR middle-end/85817
1588 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1589 for retval and return false if all args to phi are zero.
1591 2018-05-18 Richard Biener <rguenther@suse.de>
1593 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1594 method.
1595 (evrp_dom_walker::before_dom_children): Call it.
1597 2018-05-18 Richard Biener <rguenther@suse.de>
1599 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1600 results when processing array refs with variable index.
1602 2018-05-18 Toon Moene <toon@moene.org>
1604 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1605 directly after that of -floop-interchange. Indicate that both
1606 options are enabled by default when specifying -O3.
1608 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1610 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1611 iterator. Delete separate integer-mode vec_set<mode> expander.
1612 (aarch64_simd_vec_setv2di): Delete.
1613 (vec_setv2di): Delete.
1614 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1615 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1616 the "w, r" alternative.
1618 2018-05-18 Martin Liska <mliska@suse.cz>
1620 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1621 * tree-pass.h (make_pass_lower_switch_O0): New function.
1622 * tree-switch-conversion.c (node_has_low_bound): Remove.
1623 (node_has_high_bound): Likewise.
1624 (node_is_bounded): Likewise.
1625 (class pass_lower_switch): Make it a template type and create
1626 two instances.
1627 (pass_lower_switch::execute): Add template argument.
1628 (make_pass_lower_switch): New function.
1629 (make_pass_lower_switch_O0): New function.
1630 (do_jump_if_equal): Remove.
1631 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1632 all the hard work to tree optimization passes.
1634 2018-05-18 Martin Liska <mliska@suse.cz>
1636 * dbgcnt.c (limit_low): Renamed from limit.
1637 (limit_high): New variable.
1638 (dbg_cnt_is_enabled): Check for upper limit.
1639 (dbg_cnt): Adjust dumping.
1640 (dbg_cnt_set_limit_by_index): Add new argument for high
1641 value.
1642 (dbg_cnt_set_limit_by_name): Likewise.
1643 (dbg_cnt_process_single_pair): Parse new format.
1644 (dbg_cnt_process_opt): Use strtok.
1645 (dbg_cnt_list_all_counters): Remove 'value' and add
1646 'limit_high'.
1647 * doc/invoke.texi: Document changes.
1649 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1651 * doc/sourcebuild.texi (scalar_all_fma): Document.
1652 * tree.def (FMA_EXPR): Delete.
1653 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1654 * internal-fn.c (ternary_direct): New macro.
1655 (expand_ternary_optab_fn): Likewise.
1656 (direct_ternary_optab_supported_p): Likewise.
1657 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1658 * builtins.c (fold_builtin_fma): Delete.
1659 (fold_builtin_3): Don't call it.
1660 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1661 * expr.c (expand_expr_real_2): Likewise.
1662 * fold-const.c (operand_equal_p): Likewise.
1663 (fold_ternary_loc): Likewise.
1664 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1665 * gimple.c (DEFTREECODE): Likewise.
1666 * gimplify.c (gimplify_expr): Likewise.
1667 * optabs-tree.c (optab_for_tree_code): Likewise.
1668 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1669 * tree-eh.c (operation_could_trap_p): Likewise.
1670 (stmt_could_throw_1_p): Likewise.
1671 * tree-inline.c (estimate_operator_cost): Likewise.
1672 * tree-pretty-print.c (dump_generic_node): Likewise.
1673 (op_code_prio): Likewise.
1674 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1675 * tree-ssa-operands.c (get_expr_operands): Likewise.
1676 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1677 * fold-const-call.h (fold_fma): Delete.
1678 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1679 CFN_FNMA and CFN_FNMS.
1680 (fold_fma): Delete.
1681 * genmatch.c (combined_fn): New enum.
1682 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1683 (commutative_op): New function.
1684 (commutate): Use it. Handle more than 2 operands.
1685 (dt_operand::gen_gimple_expr): Use commutative_op.
1686 (parser::parse_expr): Allow :c to be used with non-binary
1687 operators if the commutative operand is known.
1688 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1689 CFN_FMS, CFN_FNMA and CFN_FNMS.
1690 (backprop::process_assign_use): Remove FMA_EXPR handling.
1691 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1692 (gen_hsa_fma): New function.
1693 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1694 IFN_FNMA and IFN_FNMS.
1695 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1696 * gimple-fold.h (follow_all_ssa_edges): Declare.
1697 * gimple-fold.c (follow_all_ssa_edges): New function.
1698 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1699 gimple_build interface and use follow_all_ssa_edges to fold the result.
1700 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1701 instead of checking for optabs directly.
1702 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1703 rather than FMA_EXPRs.
1704 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1705 call to IFN_FMA instead of an FMA_EXPR.
1707 2018-05-17 Jim Wilson <jimw@sifive.com>
1709 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1710 sign extended promoted subreg, and the range does not have the sign bit
1711 set, then do a sign extend.
1713 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1714 test, check for sign extended subreg and/or constant operands, and
1715 do a sign extend in that case.
1717 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1719 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1720 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1721 Add untyped.
1722 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1723 Change logics_shift_reg to logics_shift_imm.
1724 (thunderx2t99_fp_loadpair_basic): Delete.
1725 (thunderx2t99_fp_storepair_basic): Delete.
1726 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1727 (thunderx2t99_asimd_polynomial): Delete.
1728 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1729 and neon_fp_mul_d_scalar_q.
1730 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1731 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1732 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1733 (thunderx2t99_asimd_lut): Add missing tbl types.
1734 (thunderx2t99_asimd_ext): Delete.
1735 (thunderx2t99_asimd_load1_1_mult): Delete.
1736 (thunderx2t99_asimd_load1_2_mult): Delete.
1737 (thunderx2t99_asimd_load1_ldp): New.
1738 (thunderx2t99_asimd_load1): New.
1739 (thunderx2t99_asimd_load2): Add missing *load2* types.
1740 (thunderx2t99_asimd_load3): New.
1741 (thunderx2t99_asimd_load4): New.
1742 (thunderx2t99_asimd_store1_1_mult): Delete.
1743 (thunderx2t99_asimd_store1_2_mult): Delete.
1744 (thunderx2t99_asimd_store2_mult): Delete.
1745 (thunderx2t99_asimd_store2_onelane): Delete.
1746 (thunderx2t99_asimd_store_stp): New.
1747 (thunderx2t99_asimd_store1): New.
1748 (thunderx2t99_asimd_store2): New.
1749 (thunderx2t99_asimd_store3): New.
1750 (thunderx2t99_asimd_store4): New.
1752 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1754 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1755 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1757 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1758 Segher Boessenkool <segher@kernel.crashing.org>
1760 PR target/85698
1761 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1762 operand.
1764 2018-05-17 Richard Biener <rguenther@suse.de>
1766 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1767 for pruning loop and prune defs feeding only already visited PHIs.
1769 2018-05-17 Richard Biener <rguenther@suse.de>
1771 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1773 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1774 Richard Biener <rguenther@suse.de>
1776 PR tree-optimization/85793
1777 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1778 for VMAT_ELEMENTWISE.
1780 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1782 * internal-fn.h (lookup_internal_fn): Declare
1783 * internal-fn.c (lookup_internal_fn): New function.
1784 * gimple.c (gimple_build_call_from_tree): Handle calls to
1785 internal functions.
1786 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1787 internal function names.
1788 * tree-pretty-print.c (dump_generic_node): Likewise.
1789 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1791 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1793 * gimple-fold.h (gimple_build): Make the function forms take
1794 combined_fn rather than built_in_function.
1795 (gimple_simplify): Likewise.
1796 * gimple-match-head.c (gimple_simplify): Likewise.
1797 * gimple-fold.c (gimple_build): Likewise.
1798 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1799 rather than gimple_build_call_internal.
1800 (get_initial_defs_for_reduction): Likewise.
1801 (vect_create_epilog_for_reduction): Likewise.
1802 (vectorizable_live_operation): Likewise.
1804 2018-05-17 Martin Liska <mliska@suse.cz>
1806 * gimple-ssa-sprintf.c (format_directive): Do not use
1807 space in between 'G_' and '('.
1809 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1811 PR target/85323
1812 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1813 even if the mask is not all ones.
1815 PR target/85323
1816 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1817 vector.
1818 (ix86_gimple_fold_builtin): Likewise.
1820 PR target/85323
1821 * config/i386/i386.c: Include tree-vector-builder.h.
1822 (ix86_vector_shift_count): New function.
1823 (ix86_fold_builtin): Fold shift builtins by scalar count.
1824 (ix86_gimple_fold_builtin): Likewise.
1826 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1827 _mm512_setzero): New intrinsics.
1829 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1830 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1832 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1833 code generation for cases where splatting a value is not useful.
1834 * simplify-rtx.c (simplify_ternary_operation): Simplify
1835 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1836 mode to a vec_concat.
1838 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1840 * config.gcc: Support "goldmont-plus".
1841 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1842 "goldmont-plus".
1843 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1844 PROCESSOR_GOLDMONT_PLUS.
1845 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1846 (processor_target_table): Add "goldmont-plus".
1847 (PTA_GOLDMONT_PLUS): Define.
1848 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1849 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1850 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1851 (fold_builtin_cpu): Add "goldmont-plus".
1852 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1853 (ix86_option_override_internal): Add "goldmont-plus".
1854 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1855 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1856 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1857 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1859 2018-05-17 Richard Biener <rguenther@suse.de>
1861 PR tree-optimization/85757
1862 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1863 remove defs that only feed that PHI from further processing.
1865 2018-05-16 Jim Wilson <jimw@sifive.com>
1867 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1868 asterisk to name.
1869 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1871 2018-05-16 Mark Wielaard <mark@klomp.org>
1873 * dwarf2out.c (count_index_strings): New function.
1874 (output_indirect_strings): Call count_index_strings and generate
1875 header for dwarf_version >= 5.
1877 2018-05-16 Mark Wielaard <mark@klomp.org>
1879 * dwarf2out.c (dwarf_FORM): New function.
1880 (set_indirect_string): Use dwarf_FORM.
1881 (reset_indirect_string): Likewise.
1882 (size_of_die): Likewise.
1883 (value_format): Likewise.
1884 (output_die): Likewise.
1885 (add_skeleton_AT_string): Likewise.
1886 (output_macinfo_op): Likewise.
1887 (index_string): Likewise.
1888 (output_index_string_offset): Likewise.
1889 (output_index_string): Likewise.
1890 (count_index_strings): Likewise.
1892 2018-05-16 Carl Love <cel@us.ibm.com>
1894 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1895 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1897 2018-05-16 Martin Jambor <mjambor@suse.cz>
1899 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1900 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1902 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1904 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1905 (fnma<mode>4): Likewise.
1906 (fms<mode>4): Likewise.
1907 (fnms<mode>4): Likewise.
1908 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1909 (aarch64_fnma<mode>4): Likewise.
1910 (aarch64_fms<mode>4): Likewise.
1911 (aarch64_fnms<mode>4): Likewise.
1912 (aarch64_fnmadd<mode>4): Likewise.
1914 2018-05-16 Jason Merrill <jason@redhat.com>
1916 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1918 2018-05-16 Richard Biener <rguenther@suse.de>
1920 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1921 (dump_stmt_cost): Declare.
1922 (add_stmt_cost): Dump cost we add.
1923 (add_stmt_costs): New function.
1924 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1925 No longer exported.
1926 (vect_analyze_stmt): Adjust prototype.
1927 (vectorizable_condition): Likewise.
1928 (vectorizable_live_operation): Likewise.
1929 (vectorizable_reduction): Likewise.
1930 (vectorizable_induction): Likewise.
1931 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1932 cost vector to pass to vectorizable_ and record afterwards.
1933 (vect_model_reduction_cost): Take cost vector argument and adjust.
1934 (vect_model_induction_cost): Likewise.
1935 (vectorizable_reduction): Likewise.
1936 (vectorizable_induction): Likewise.
1937 (vectorizable_live_operation): Likewise.
1938 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1939 SLP_TREE_NUMBER_OF_VEC_STMTS.
1940 (vect_analyze_slp_cost_1): Remove.
1941 (vect_analyze_slp_cost): Likewise.
1942 (vect_slp_analyze_node_operations): Take visited args and
1943 a target cost vector. Avoid processing already visited stmt sets.
1944 (vect_slp_analyze_operations): Use a local cost vector to gather
1945 costs and register those of non-discarded instances.
1946 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1947 (vect_schedule_slp_instance): Remove copying of
1948 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1949 zero.
1950 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1951 adding cost. Record cost entry location.
1952 (vect_prologue_cost_for_slp_op): Function to compute cost of
1953 a constant or invariant generated for SLP vect in the prologue,
1954 split out from vect_analyze_slp_cost_1.
1955 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1956 (vect_model_promotion_demotion_cost): Likewise.
1957 (vect_model_store_cost): Likewise, make static.
1958 (vect_model_load_cost): Likewise.
1959 (vectorizable_bswap): Add cost vector arg and adjust.
1960 (vectorizable_call): Likewise.
1961 (vectorizable_simd_clone_call): Likewise.
1962 (vectorizable_conversion): Likewise.
1963 (vectorizable_assignment): Likewise.
1964 (vectorizable_shift): Likewise.
1965 (vectorizable_operation): Likewise.
1966 (vectorizable_store): Likewise.
1967 (vectorizable_load): Likewise.
1968 (vectorizable_condition): Likewise.
1969 (vectorizable_comparison): Likewise.
1970 (can_vectorize_live_stmts): Likewise.
1971 (vect_analyze_stmt): Likewise.
1972 (vect_transform_stmt): Adjust calls to vectorizable_*.
1973 * tree-vectorizer.c: Include gimple-pretty-print.h.
1974 (dump_stmt_cost): New function.
1976 2018-05-16 Richard Biener <rguenther@suse.de>
1978 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1979 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1980 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1981 (check_name): New callback.
1982 (dse_classify_store): Track cycles via a visited bitmap of PHI
1983 defs and simplify handling of in-loop and across loop dead stores
1984 and properly fail for loop-variant refs. Handle byte-tracking with
1985 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1986 limiting the walk.
1988 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1990 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1991 (vect_get_mask_type_for_stmt): Likewise.
1992 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1993 split out from...
1994 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1995 to determine the statement's vector type and the vector type that
1996 should be used for calculating nunits. Deal with cases in which
1997 the type has to be deferred.
1998 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1999 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2000 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2001 (vect_determine_vf_for_stmt): New functions, split out from...
2002 (vect_determine_vectorization_factor): ...here.
2003 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2004 (vect_get_mask_type_for_stmt): New functions, split out from
2005 vect_determine_vectorization_factor.
2007 2018-05-16 Richard Biener <rguenther@suse.de>
2009 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2010 verify the [VEC_]COND_EXPR embedded comparison.
2012 2018-05-15 Martin Sebor <msebor@redhat.com>
2014 PR tree-optimization/85753
2015 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2016 RECORD_TYPE in addition to ARRAY_TYPE.
2018 2018-05-15 Martin Sebor <msebor@redhat.com>
2020 PR middle-end/85643
2021 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2023 2018-05-15 Richard Biener <rguenther@suse.de>
2025 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2026 add by_clobber_p one. Change algorithm to collect all defs
2027 representing uses we need to walk and try reducing them to
2028 a single one before failing.
2029 (dse_dom_walker::dse_optimize_stmt): Adjust.
2031 2018-05-13 Mark Wielaard <mark@klomp.org>
2033 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2034 (size_of_loc_descr): Likewise.
2035 (output_loc_operands): Likewise.
2036 (output_loc_operands_raw): Likewise.
2037 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2038 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2039 (hash_loc_operands): Likewise.
2040 (compare_loc_operands): Likewise.
2042 2018-05-14 Mark Wielaard <mark@klomp.org>
2044 * dwarf2out.c (count_index_addrs): New function.
2045 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2047 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2049 PR tree-optimization/83648
2050 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2051 return value as malloc candidate.
2053 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2055 PR ipa/85734
2056 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2057 as true in call to suggest_attribute.
2059 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2061 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2062 -mreadonly-in-sdata.
2064 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2066 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2067 New pattern.
2068 (aarch64_crypto_aesd_fused): Likewise.
2070 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2072 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2073 (movsi_aarch64): Likewise.
2074 (load_pairsi): Likewise.
2075 (load_pairdi): Likewise.
2076 (store_pairsi): Likewise.
2077 (store_pairdi): Likewise.
2078 (load_pairsf): Likewise.
2079 (load_pairdf): Likewise.
2080 (store_pairsf): Likewise.
2081 (store_pairdf): Likewise.
2082 (zero_extend): Likewise.
2083 (trunc): Swap alternatives.
2084 (fcvt_target): Add '?' to prefer w over r.
2086 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2088 PR target/85756
2089 * config/i386/i386.md: Disallow non-commutative arithmetics in
2090 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2091 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2092 in the peephole2 before it.
2094 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2096 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2097 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2098 (ix86_handle_option): Handle -mcldemote.
2099 * config.gcc: New header.
2100 * config/i386/cldemoteintrin.h: New file.
2101 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2102 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2103 -mcldemote.
2104 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2105 OPTION_MASK_ISA_CLDEMOTE.
2106 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2107 (ix86_valid_target_attribute_inner_p): Ditto.
2108 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2109 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2110 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2111 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2112 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2113 (cldemote): New.
2114 * config/i386/i386.opt: Add -mcldemote.
2115 * config/i386/x86intrin.h: New header.
2116 * doc/invoke.texi: Add -mcldemote.
2118 2018-05-14 Richard Biener <rguenther@suse.de>
2120 * doc/match-and-simplify.texi: Adjust :s documentation.
2122 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2124 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2125 intended memcpy size.
2126 (REORDER_45): Likewise.
2128 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2130 * sort.cc: New file.
2131 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2132 * vec.c (qsort_chk): Use gcc_qsort.
2133 * Makefile.in (OBJS-libcommon): Add sort.o.
2134 (build/sort.o): New target. Use it...
2135 (BUILD_RTL): ... here, and...
2136 (build/gencfn-macros): ... here, and...
2137 (build/genmatch): ... here.
2139 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2140 Chung-Ju Wu <jasonwucj@gmail.com>
2142 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2143 * config/nds32/nds32-graywolf.md: New file.
2144 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2145 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2146 pipeline.
2147 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2148 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2149 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2150 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2151 * config/nds32/pipelines.md: Include n15 settings.
2153 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2154 Chung-Ju Wu <jasonwucj@gmail.com>
2156 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2157 * config/nds32/nds32-n13.md: New file.
2158 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2159 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2160 pipeline.
2161 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2162 * config/nds32/nds32.md (pipeline_model): Add n13.
2163 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2164 * config/nds32/pipelines.md: Include n13 settings.
2166 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2167 Chung-Ju Wu <jasonwucj@gmail.com>
2169 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2170 * config/nds32/nds32-n10.md: New file.
2171 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2172 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2173 pipeline.
2174 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2175 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2176 * config/nds32/nds32.md (pipeline_model): Add n10.
2177 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2178 * config/nds32/pipelines.md: Include n10 settings.
2180 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2181 Kito Cheng <kito.cheng@gmail.com>
2182 Chung-Ju Wu <jasonwucj@gmail.com>
2184 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2185 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2186 Add enum values for DSP extension instructions.
2187 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2188 New constraints.
2189 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2190 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2191 New code iterators.
2192 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2193 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2194 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2195 * config/nds32/nds32-intrinsic.md: Likewise.
2196 * config/nds32/nds32_intrinsic.h: Likewise.
2197 * config/nds32/nds32-md-auxiliary.c: Likewise.
2198 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2199 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2200 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2201 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2202 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2203 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2204 TYPE_DMAC in switch statement.
2205 * config/nds32/nds32.c: New checking and implementation for DSP
2206 extension instructions.
2207 * config/nds32/nds32.h: Likewise.
2208 * config/nds32/nds32.md: Likewise.
2209 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2210 * config/nds32/predicates.md: Implement new predicates for DSP
2211 extension.
2213 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2215 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2216 Reformat alternatives and attributes so it is easier to identify
2217 which constraints/attributes go with which instruction.
2218 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2219 (mov<mode>_softfloat32, FMOVE64): Likewise.
2220 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2221 (mov<mode>_softfloat64, FMOVE64): Likewise.
2223 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2225 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2226 subsection.
2227 (Basic PowerPC Built-in Functions): The new name of the
2228 subsection previously known as "PowerPC Built-in Functions".
2229 (Basic PowerPC Built-in Functions Available on all Configurations):
2230 New subsubsection.
2231 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2232 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2233 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2234 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2236 2018-05-11 Martin Jambor <mjambor@suse.cz>
2238 PR ipa/85655
2239 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2240 single const.
2242 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2244 PR target/85733
2245 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2247 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2249 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2250 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2251 (ix86_handle_option): Handle -mwaitpkg.
2252 * config.gcc: New header.
2253 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2254 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2255 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2256 function type.
2257 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2258 OPTION_MASK_ISA_WAITPKG.
2259 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2260 (ix86_option_override_internal): Add PTA_WAITPKG.
2261 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2262 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2263 IX86_BUILTIN_TPAUSE.
2264 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2265 __builtin_ia32_umwait and __builtin_ia32_tpause.
2266 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2267 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2268 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2269 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2270 UNSPECV_TPAUSE): New.
2271 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2272 * config/i386/i386.opt: Add -mwaitpkg.
2273 * config/i386/waitpkgintrin.h: New file.
2274 * config/i386/x86intrin.h: New header.
2275 * doc/invoke.texi: Add -mwaitpkg.
2277 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2279 PR target/85606
2280 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2281 equivalent.
2282 (cortex-m0): Use armv6s-m isa.
2283 (cortex-m0plus): Likewise.
2284 (cortex-m1): Likewise.
2285 (cortex-m0.small-multiply): Likewise.
2286 (cortex-m0plus.small-multiply): Likewise.
2287 (cortex-m1.small-multiply): Likewise.
2289 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2290 Jakub Jelinek <jakub@redhat.com>
2292 PR tree-optimization/85692
2293 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2294 source permute as well.
2296 2018-05-11 Martin Liska <mliska@suse.cz>
2298 PR sanitizer/85556
2299 * doc/extend.texi: Document LLVM style format for no_sanitize
2300 attribute.
2302 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2304 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2305 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2306 (mode_supports_vsx_dform_quad): Likewise.
2307 (mode_supports_vmx_dform): Move these functions to be next to the
2308 other mode_supports functions.
2309 (mode_supports_dq_form): Likewise.
2310 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2311 mode_supports_dq_form.
2312 (reg_offset_addressing_ok_p): Likewise.
2313 (offsettable_ok_by_alignment): Likewise.
2314 (rs6000_legitimate_offset_address_p): Likewise.
2315 (legitimate_lo_sum_address_p): Likewise.
2316 (rs6000_legitimize_address): Likewise.
2317 (rs6000_legitimize_reload_address): Likewise.
2318 (rs6000_secondary_reload_inner): Likewise.
2319 (rs6000_preferred_reload_class): Likewise.
2320 (rs6000_output_move_128bit): Likewise.
2322 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2324 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2325 Generate SImode target register for null target.
2326 <case IX86_BUILTIN_XGETBV>: Ditto.
2327 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2328 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2330 2018-05-10 Carl Love <cel@us.ibm.com>
2332 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2333 dcbtt and dcbtstt if operands[2] is 0.
2335 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2337 PR target/85693
2338 * config/i386/sse.md (usadv64qi): New expander.
2340 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2342 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2343 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2344 -maltivec=be support.
2345 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2346 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2347 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2348 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2349 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2350 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2351 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2352 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2353 altivec_vsumsws): Adjust.
2354 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2355 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2356 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2357 support.
2358 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2359 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2360 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2361 (altivec_lve<VI_char>x): Delete expand.
2362 (*altivec_lve<VI_char>x_internal): Rename to...
2363 (altivec_lve<VI_char>x): ... this.
2364 (altivec_lvxl_<mode>): Delete expand.
2365 (*altivec_lvxl_<mode>_internal): Rename to ...
2366 (altivec_lvxl_<mode>): ... this.
2367 (altivec_stvxl_<mode>): Delete expand.
2368 (*altivec_stvxl_<mode>_internal): Rename to ...
2369 (altivec_stvxl_<mode>): ... this.
2370 (altivec_stve<VI_char>x): Delete expand.
2371 (*altivec_stve<VI_char>x_internal): Rename to ...
2372 (altivec_stve<VI_char>x): ... this.
2373 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2374 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2375 reduc_plus_scal_<mode>): Adjust.
2376 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2377 comment.
2378 (rs6000_cpu_cpp_builtins): Adjust.
2379 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2380 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2381 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2382 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2383 -maltivec=be support.
2384 (rs6000_split_vec_extract_var): Adjust.
2385 (rs6000_split_v4si_init): Adjust.
2386 (swap_selector_for_mode): Delete.
2387 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2388 altivec_expand_stvex_be): Delete.
2389 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2390 -maltivec=be support.
2391 (rs6000_gimple_fold_builtin): Ditto.
2392 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2393 Adjust.
2394 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2395 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2396 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2397 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2398 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2399 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2400 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2401 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2402 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2403 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2404 anonymous split): Adjust.
2405 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2406 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2408 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2410 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2411 when --with-gxx-include-dir is also specified.
2412 * configure: Regenerate.
2414 2018-05-09 Jim Wilson <jimw@sifive.com>
2416 PR target/84797
2417 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2418 * config/riscv/t-withmultilib: New.
2419 * config/riscv/withmultilib.h: New.
2420 * doc/install.texi: Document RISC-V --with-multilib-list support.
2422 2018-05-09 Richard Biener <rguenther@suse.de>
2424 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2425 vector.
2426 (vect_bb_vectorization_profitable_p): Adjust. Compute
2427 actual scalar cost using the cost vector and the add_stmt_cost
2428 machinery.
2430 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2432 PR rtl-optimization/85645
2433 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2434 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2436 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2438 PR rtl-optimization/85645
2439 * shrink-wrap.c (spread_components): Return a boolean saying if
2440 anything was changed.
2441 (try_shrink_wrapping_separate): Iterate spread_components until
2442 nothing changes anymore.
2444 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2446 PR rtl-optimization/85645
2447 * regrename.c (build_def_use): Also kill the chains that include the
2448 destination of a REG_CFA_REGISTER note.
2450 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2452 PR rtl-optimization/85645
2453 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2454 insn that has a REG_CFA_REGISTER note.
2456 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2458 * cfgexpand.c (expand_clobber): New function.
2459 (expand_gimple_stmt_1): Use it.
2460 * tree-vect-stmts.c (vect_clobber_variable): New function,
2461 split out from...
2462 (vectorizable_simd_clone_call): ...here.
2463 (vectorizable_store): Emit a clobber either side of an
2464 IFN_STORE_LANES sequence.
2465 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2467 2018-05-09 Tom de Vries <tom@codesourcery.com>
2469 PR target/85626
2470 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2471 (define_insn "trap_if_false"): Add exit after trap.
2473 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2475 PR rtl-optimization/85638
2476 * bb-reorder.c: Include common/common-target.h.
2477 (create_forwarder_block): New function extracted from...
2478 (fix_up_crossing_landing_pad): ...here. Rename into...
2479 (dw2_fix_up_crossing_landing_pad): ...this.
2480 (sjlj_fix_up_crossing_landing_pad): New function.
2481 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2482 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2483 from both partitions and exit the loop after one iteration.
2485 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2487 Revert:
2488 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2489 subsection.
2490 (Basic PowerPC Built-in Functions): The new name of the
2491 subsection previously known as "PowerPC Built-in Functions".
2492 (Basic PowerPC Built-in Functions Available on all Configurations):
2493 New subsubsection.
2494 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2495 subsubsection.
2496 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2497 subsubsection.
2498 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2499 subsubsection.
2500 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2501 subsubsection.
2503 2018-05-08 Jim Wilson <jimw@sifive.com>
2505 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2506 (LD_EMUL_SUFFIX): New.
2507 (LINK_SPEC): Use it.
2509 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2511 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2512 subsection.
2513 (Basic PowerPC Built-in Functions): The new name of the
2514 subsection previously known as "PowerPC Built-in Functions".
2515 (Basic PowerPC Built-in Functions Available on all Configurations):
2516 New subsubsection.
2517 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2518 subsubsection.
2519 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2520 subsubsection.
2521 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2522 subsubsection.
2523 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2524 subsubsection.
2526 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2528 PR target/85683
2529 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2530 after cmpelim optimization.
2532 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2534 * config.gcc: Support "goldmont".
2535 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2537 PROCESSOR_GOLDMONT.
2538 * config/i386/i386.c (m_GOLDMONT): Define.
2539 (processor_target_table): Add "goldmont".
2540 (PTA_GOLDMONT): Define.
2541 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2542 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2543 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2544 (fold_builtin_cpu): Add "goldmont".
2545 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2546 (ix86_option_override_internal): Add "goldmont".
2547 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2548 (processor_type): Add PROCESSOR_GOLDMONT.
2549 * config/i386/i386.md: Add CPU "glm".
2550 * config/i386/glm.md: New file.
2551 * config/i386/x86-tune.def: Add m_GOLDMONT.
2552 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2554 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2556 PR target/85572
2557 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2558 E_V4DImode.
2559 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2560 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2561 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2563 PR target/85317
2564 * config/i386/i386.c (ix86_fold_builtin): Handle
2565 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2567 PR target/85480
2568 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2569 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2571 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2573 PR target/85658
2574 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2575 (check_arch): Likewise.
2576 (check_fpu): Return the result rather than printing it.
2577 (end arch): Fix operator precedence.
2578 (end cpu): Likewise.
2579 (END): Print the result from check_fpu.
2581 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2582 Alan Hayward <alan.hayward@arm.com>
2583 David Sherwood <david.sherwood@arm.com>
2585 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2586 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2587 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2588 (*fcmuo<mode>_and): New patterns.
2590 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2592 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2593 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2594 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2595 (cmp_op, sve_imm_con): New code attributes.
2596 (SVE_COND_INT_CMP, imm_con): Delete.
2597 (cmp_op): Remove above unspecs from int attribute.
2598 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2599 to...
2600 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2601 comparison-specific unspecs.
2602 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2603 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2604 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2605 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2606 (*vec_fcm<cmp_op><mode>): Rename to...
2607 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2608 (*vec_fcmuo<mode>): Rename to...
2609 (*fcmuo<mode>): ...this and adjust likewise.
2610 (*pred_fcm<cmp_op><mode>): New pattern.
2611 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2612 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2613 functions.
2614 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2615 and UNORDERED.
2616 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2617 (aarch64_emit_sve_predicated_cond): New function.
2618 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2619 (aarch64_emit_unspec_cond_or): Replace with...
2620 (aarch64_emit_sve_or_conds): ...this new function. Use
2621 aarch64_emit_sve_ptrue_op for the individual comparisons and
2622 aarch64_emit_binop to OR them together.
2623 (aarch64_emit_inverted_unspec_cond): Replace with...
2624 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2625 aarch64_emit_sve_ptrue_op for the comparison and
2626 aarch64_emit_unop to invert the result.
2627 (aarch64_expand_sve_vec_cmp_float): Update after the above
2628 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2630 2018-05-07 Nathan Sidwell <nathan@acm.org>
2632 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2633 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2634 (Backwards Compatibility): Likewise.
2636 2018-05-07 Luis Machado <luis.machado@linaro.org>
2638 PR bootstrap/85681
2639 Revert:
2640 2018-05-07 Luis Machado <luis.machado@linaro.org>
2642 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2643 <prefetch_dynamic_strides>: New const bool field.
2644 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2645 prefetch_dynamic_strides.
2646 (exynosm1_prefetch_tune): Likewise.
2647 (thunderxt88_prefetch_tune): Likewise.
2648 (thunderx_prefetch_tune): Likewise.
2649 (thunderx2t99_prefetch_tune): Likewise.
2650 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2651 (aarch64_override_options_internal): Update to set
2652 PARAM_PREFETCH_DYNAMIC_STRIDES.
2653 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2654 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2655 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2656 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2657 prefetch-dynamic-strides setting.
2659 2018-05-07 Luis Machado <luis.machado@linaro.org>
2661 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2662 <minimum_stride>: New const int field.
2663 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2664 minimum_stride field.
2665 (exynosm1_prefetch_tune): Likewise.
2666 (thunderxt88_prefetch_tune): Likewise.
2667 (thunderx_prefetch_tune): Likewise.
2668 (thunderx2t99_prefetch_tune): Likewise.
2669 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2670 (aarch64_override_options_internal): Update to set
2671 PARAM_PREFETCH_MINIMUM_STRIDE.
2672 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2673 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2674 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2675 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2676 stride is constant and is below the minimum stride threshold.
2678 2018-05-07 Luis Machado <luis.machado@linaro.org>
2680 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2681 to 512.
2683 2018-05-07 Luis Machado <luis.machado@linaro.org>
2685 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2686 <prefetch_dynamic_strides>: New const bool field.
2687 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2688 prefetch_dynamic_strides.
2689 (exynosm1_prefetch_tune): Likewise.
2690 (thunderxt88_prefetch_tune): Likewise.
2691 (thunderx_prefetch_tune): Likewise.
2692 (thunderx2t99_prefetch_tune): Likewise.
2693 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2694 (aarch64_override_options_internal): Update to set
2695 PARAM_PREFETCH_DYNAMIC_STRIDES.
2696 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2697 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2698 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2699 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2700 prefetch-dynamic-strides setting.
2702 2018-05-07 Luis Machado <luis.machado@linaro.org>
2704 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2705 <minimum_stride>: New const int field.
2706 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2707 minimum_stride field.
2708 (exynosm1_prefetch_tune): Likewise.
2709 (thunderxt88_prefetch_tune): Likewise.
2710 (thunderx_prefetch_tune): Likewise.
2711 (thunderx2t99_prefetch_tune): Likewise.
2712 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2713 (aarch64_override_options_internal): Update to set
2714 PARAM_PREFETCH_MINIMUM_STRIDE.
2715 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2716 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2717 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2718 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2719 stride is constant and is below the minimum stride threshold.
2721 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2723 PR c++/85659
2724 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2725 the type is addressable. Don't force op into register if it has
2726 BLKmode.
2728 2018-05-05 Roland McGrath <mcgrathr@google.com>
2730 PR other/77609
2731 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2732 any section for which we don't know a specific type it should have,
2733 regardless of name. Previously this was done only for the exact
2734 names ".init_array", ".fini_array", and ".preinit_array".
2735 (default_elf_asm_named_section): Add comment about
2736 relationship with default_section_type_flags and SECTION_NOTYPE.
2737 (get_section): Don't consider it a type conflict if one side has
2738 SECTION_NOTYPE and the other doesn't, as long as neither has the
2739 SECTION_BSS et al used in the default_section_type_flags logic.
2741 2018-05-05 Tom de Vries <tom@codesourcery.com>
2743 PR target/85653
2744 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2745 (workaround_barsyncs): New function.
2746 (nvptx_reorg): Use workaround_barsyncs.
2747 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2748 (define_expand "nvptx_membar_cta"): New define_expand.
2749 (define_insn "*nvptx_membar_cta"): New insn.
2751 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2753 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2754 To improve optimization opportunities.
2755 * builtin-types.def: The new needed builtin types for the above.
2757 2018-05-04 Richard Biener <rguenther@suse.de>
2759 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2760 * gimple-ssa-store-merging.c
2761 (imm_store_chain_info::output_merged_store): Remove redundant create,
2762 release split_store vector contents on failure.
2763 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2764 scalar stmt vector on cache hit.
2766 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2768 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2769 Xilinx FP support.
2770 * config.gcc (powerpc-xilinx-eabi*): Remove.
2771 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2772 support.
2773 (fusion_addis_mem_combo_load): Ditto.
2774 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2775 FP support.
2776 (rs6000_cpu_cpp_builtins): Ditto.
2777 * config/rs6000/rs6000-linux.c
2778 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2779 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2780 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2781 support.
2782 (rs6000_setup_reg_addr_masks): Ditto.
2783 (rs6000_init_hard_regno_mode_ok): Ditto.
2784 (rs6000_option_override_internal): Ditto.
2785 (legitimate_lo_sum_address_p): Ditto.
2786 (rs6000_legitimize_address): Ditto.
2787 (rs6000_legitimize_reload_address): Ditto.
2788 (rs6000_legitimate_address_p): Ditto.
2789 (abi_v4_pass_in_fpr): Ditto.
2790 (setup_incoming_varargs): Ditto.
2791 (rs6000_gimplify_va_arg): Ditto.
2792 (rs6000_split_multireg_move): Ditto.
2793 (rs6000_savres_strategy): Ditto.
2794 (rs6000_emit_prologue_components): Ditto.
2795 (rs6000_emit_epilogue_components): Ditto.
2796 (rs6000_emit_prologue): Ditto.
2797 (rs6000_emit_epilogue): Ditto.
2798 (rs6000_elf_file_end): Ditto.
2799 (rs6000_function_value): Ditto.
2800 (rs6000_libcall_value): Ditto.
2801 * config/rs6000/rs6000.h: Ditto.
2802 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2803 (TARGET_MINMAX): ... this. New.
2804 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2805 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2806 (*movsi_internal1_single): Delete.
2807 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2808 mfpu=, mxilinx-fpu): Delete.
2809 * config/rs6000/singlefp.h: Delete.
2810 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2811 * config/rs6000/t-rs6000: Ditto.
2812 * config/rs6000/t-xilinx: Delete.
2813 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2814 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2815 (VStype_simple): Delete.
2816 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2817 * config/rs6000/xfpu.h: Delete.
2818 * config/rs6000/xfpu.md: Delete.
2819 * config/rs6000/xilinx.h: Delete.
2820 * config/rs6000/xilinx.opt: Delete.
2821 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2822 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2824 2018-05-04 Tom de Vries <tom@codesourcery.com>
2826 PR libgomp/85639
2827 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2828 if ignore == 0.
2830 2018-05-04 Richard Biener <rguenther@suse.de>
2832 PR middle-end/85627
2833 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2834 (expand_complex_div_wide): Likewise.
2835 (expand_complex_operations_1): Likewise.
2836 (expand_complex_libcall): Preserve EH info of the original stmt.
2837 (tree_lower_complex): Handle removed blocks.
2838 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2839 on complex multiplication and division libcall builtins.
2841 2018-05-04 Richard Biener <rguenther@suse.de>
2843 PR middle-end/85574
2844 * fold-const.c (negate_expr_p): Restrict negation of operand
2845 zero of a division to when we know that can happen without
2846 overflow.
2847 (fold_negate_expr_1): Likewise.
2849 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2851 PR libstdc++/85466
2852 * real.h (real_nextafter): Declare.
2853 * real.c (real_nextafter): New function.
2854 * fold-const-call.c (fold_const_nextafter): New function.
2855 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2856 CASE_CFN_NEXTTOWARD.
2857 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2858 even when arg1_mode is different from arg0_mode.
2860 2018-05-03 Nathan Sidwell <nathan@acm.org>
2862 * doc/extend.texi (Deprecated Features): Remove
2863 -ffriend-injection.
2864 (Backwards Compatibility): Likewise.
2865 * doc/invoke.texi (C++ Language Options): Likewise.
2866 (C++ Dialect Options): Likewise.
2868 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2870 PR target/85530
2871 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2872 _mm512_mask_mullox_epi64): New intrinsics.
2874 2018-05-03 Tom de Vries <tom@codesourcery.com>
2876 PR testsuite/85106
2877 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2878 dump files): Add offload-tree.
2880 2018-05-03 Richard Biener <rguenther@suse.de>
2882 PR tree-optimization/85615
2883 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2884 to loops not nested in BBs loop father to avoid creating multi-entry
2885 loops.
2887 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2889 PR tree-optimization/70291
2890 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2891 arguments. Change return type to tree. Emit libcall as a new
2892 statement rather than replacing existing one when inplace_p is true.
2893 (expand_complex_multiplication_components): New function.
2894 (expand_complex_multiplication): Expand floating-point complex
2895 multiplication using the above.
2896 (expand_complex_division): Rename inner_type parameter to type.
2897 Update expand_complex_libcall call-site.
2898 (expand_complex_operations_1): Update expand_complex_multiplication
2899 and expand_complex_division call-sites.
2901 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2903 PR target/85582
2904 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2905 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2906 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2907 the highest significant bit of the shift count mask is clear. In
2908 check whether and[sq]i3 is needed verify that all significant bits
2909 of the shift count other than the highest are set.
2911 2018-05-02 Tom de Vries <tom@codesourcery.com>
2913 PR libgomp/82428
2914 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2915 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2916 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2917 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2918 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2919 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2920 __builtin_goacc_parlevel_size.
2922 2018-05-02 Richard Biener <rguenther@suse.de>
2924 PR tree-optimization/85597
2925 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2926 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2927 directly.
2929 2018-05-02 Tom de Vries <tom@codesourcery.com>
2931 PR testsuite/85106
2932 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2933 dump files): Add ltrans-tree.
2935 2018-05-02 Tom de Vries <tom@codesourcery.com>
2937 PR testsuite/85106
2938 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2939 dump files): Add wpa-ipa.
2941 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2943 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2944 powerpc*-*-linux*paired* target.
2945 * config/rs6000/750cl.h: Delete.
2946 * config/rs6000/paired.h: Delete.
2947 * config/rs6000/paired.md: Delete.
2948 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2949 float support.
2950 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2951 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2952 comment. Remove paired float support.
2953 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2954 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2955 VECTOR_PAIRED.
2956 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2957 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2958 declarations.
2959 * config/rs6000/rs6000.c: Remove paired float support.
2960 (paired_expand_vector_init, paired_expand_vector_move,
2961 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2962 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2963 paired_expand_builtin, paired_expand_predicate_builtin,
2964 paired_init_builtins): Delete.
2965 * config/rs6000/rs6000.h: Remove paired float support.
2966 * config/rs6000/rs6000.md: Remove paired float support.
2967 (move_from_CR_ov_bit): Delete.
2968 * config/rs6000/rs6000.opt (mpaired): Delete.
2969 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2970 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2972 2018-05-02 Richard Biener <rguenther@suse.de>
2974 PR middle-end/85567
2975 * gimplify.c (gimplify_save_expr): When in SSA form allow
2976 SAVE_EXPRs to compute to SSA vars.
2978 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2980 PR target/85582
2981 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2982 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2983 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2984 clobber operands[2], instead use a new pseudo. Formatting fixes.
2986 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2988 PR tree-optimization/85586
2989 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2990 exit early for statements in the same group if the accesses are
2991 not strided.
2993 2018-05-02 Tom de Vries <tom@codesourcery.com>
2995 PR lto/85451
2996 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2997 error message.
2999 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3001 PR tree-optimization/85143
3002 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3004 2018-05-01 Tom de Vries <tom@codesourcery.com>
3006 PR lto/85451
3007 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3008 not found" error message.
3010 2018-05-01 Tom de Vries <tom@codesourcery.com>
3012 PR other/83786
3013 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3014 * vec.c (test_ordered_remove_if): New function.
3015 (vec_c_tests): Call test_ordered_remove_if.
3016 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3017 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3018 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3019 VEC_ORDERED_REMOVE_IF.
3021 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3023 PR tree-optimization/82665
3024 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3025 pointer subtraction where arguments come from a memchr call.
3027 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3029 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3030 --push-state --as-needed and --pop-state instead of --as-needed and
3031 --no-as-needed if ld supports it.
3032 * configure: Regenerated.
3034 PR web/85578
3035 * doc/install.texi2html: Replace _002d with - and _002a with * in
3036 generated html files using sed.
3038 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3040 PR c++/85523
3041 * gcc-rich-location.c (blank_line_before_p): New function.
3042 (use_new_line): New function.
3043 (gcc_rich_location::add_fixit_insert_formatted): New function.
3044 * gcc-rich-location.h
3045 (gcc_rich_location::add_fixit_insert_formatted): New function.
3047 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3049 * selftest.c (assert_streq): Rename "expected" and "actual" to
3050 "val1" and "val2". Extend NULL-handling to cover both inputs
3051 symmetrically, while still requiring both to be non-NULL for a pass.
3052 * selftest.h (assert_streq): Rename "expected" and "actual" to
3053 "val1" and "val2".
3054 (ASSERT_EQ): Likewise.
3055 (ASSERT_EQ_AT): Likewise.
3056 (ASSERT_KNOWN_EQ): Likewise.
3057 (ASSERT_KNOWN_EQ_AT): Likewise.
3058 (ASSERT_NE): Likewise.
3059 (ASSERT_MAYBE_NE): Likewise.
3060 (ASSERT_MAYBE_NE_AT): Likewise.
3061 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3062 the assertion to pass.
3063 (ASSERT_STREQ_AT): Likewise.
3065 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3067 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3068 interaction with -pie.
3070 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3072 * selftest.h: Fix alphabetization of per-source-file selftest
3073 declarations.
3075 2018-04-30 Jason Merrill <jason@redhat.com>
3077 PR c++/61982 - dead stores to destroyed objects.
3078 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3079 of clobber.
3081 2018-04-30 Jason Merrill <jason@redhat.com>
3083 * tree.c (build_clobber): New.
3084 * tree.h: Declare it.
3085 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3087 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3089 * diagnostic-show-locus.c (layout::layout): Update for
3090 location_get_source_line returning a char_span.
3091 (struct char_span): Move to input.h.
3092 (struct correction): Update for fields in char_span becoming
3093 private.
3094 (struct source_line): Update for location_get_source_line
3095 returning a char_span.
3096 (layout::print_line): Likewise.
3097 * edit-context.c (edited_file::print_content): Likewise.
3098 (edited_file::print_diff_hunk): Likewise.
3099 (edited_file::print_run_of_changed_lines): Likewise.
3100 (edited_file::get_num_lines): Likewise.
3101 (edited_line::edited_line): Likewise.
3102 * final.c (asm_show_source): Likewise.
3103 * input.c (location_get_source_line): Convert return type
3104 from const char * to char_span, losing the final "line_len"
3105 param.
3106 (dump_location_info): Update for the above.
3107 (get_substring_ranges_for_loc): Likewise. Use a char_span
3108 when handling the literal within the line.
3109 (test_reading_source_line): Update for location_get_source_line
3110 returning a char_span.
3111 * input.h (class char_span): Move here from
3112 diagnostic-show-locus.c, converting from a struct to a class.
3113 Make data members private.
3114 (char_span::operator bool): New.
3115 (char_span::length): New.
3116 (char_span::get_buffer): New.
3117 (char_span::operator[]): New.
3118 (char_span::subspan): Make const.
3119 (char_span::xstrdup): New.
3120 (location_get_source_line): Convert return type from const char *
3121 to char_span, losing the final "line_size" param.
3123 2018-04-30 Jan Hubicka <jh@suse.cz>
3125 * lto-wrapper.c (ltrans_priorities): New static var.
3126 (cmp_priority): New.
3127 (run_gcc): Read priorities and if doing parallel build order
3128 the Makefile by them.
3130 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3132 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3134 2018-04-30 Richard Biener <rguenther@suse.de>
3136 * tree-cfg.c (verify_address): Remove base argument, add
3137 flag whether to check TREE_ADDRESSABLE and do that.
3138 (verify_expr): Remove.
3139 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3140 (verify_gimple_assign_single): Likewise.
3141 (verify_gimple_switch): Likewise.
3142 (verify_expr_location_1): Dereference tp once. Add (disabled)
3143 piece from verify_expr.
3144 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3146 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3148 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3150 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3152 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3153 (small_data_pattern): Likewise.
3154 (arc_rewrite_small_data): Likewise.
3155 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3156 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3157 (get_symbol_alignment): New function.
3158 (legitimate_small_data_address_p): Likewise.
3159 (legitimate_scaled_address): Update, call
3160 legitimate_small_data_address_p.
3161 (output_sdata): New static variable.
3162 (arc_print_operand): Update how we handle small data operands.
3163 (arc_print_operand_address): Likewise.
3164 (arc_legitimate_address_p): Update, use
3165 legitimate_small_data_address_p.
3166 (arc_rewrite_small_data_p): Remove.
3167 (arc_rewrite_small_data_1): Likewise.
3168 (arc_rewrite_small_data): Likewise.
3169 (small_data_pattern): Likewise.
3170 (compact_sda_memory_operand): Update to use
3171 legitimate_small_data_address_p and get_symbol_alignment.
3172 (prepare_move_operands): Don't rewite sdata pattern.
3173 (prepare_extend_operands): Remove.
3174 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3175 pattern.
3176 (zero_extendqisi2): Likewise.
3177 (zero_extendhisi2): Likewise.
3178 (extendqihi2): Likewise.
3179 (extendqisi2): Likewise.
3180 (extendhisi2): Likewise.
3181 (addsi3): Likewise.
3182 (subsi3): Likewise.
3183 (andsi3): Likewise.
3184 * config/arc/constraints.md (Usd): Change it to memory constraint.
3186 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3188 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3189 as source of std instructions.
3190 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3191 allow 6-bit constants as source for store instructions.
3192 (movdi_insn): Update instruction pattern to allow 6-bit constants
3193 as source for store instructions.
3195 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3197 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3199 2018-04-30 Nathan Sidwell <nathan@acm.org>
3200 Sandra Loosemore <sandra@codesourcery.com>
3202 * dumpfile.c (dump_open): Allow '-' for stdout.
3203 * doc/invoke.texi (Developer Options): Document dump filename
3204 determination early. Document stdin/stdout selection.
3206 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3208 Microblaze Target: PIC data text relative
3210 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3211 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3212 Add declaration.
3213 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3214 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3215 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3216 New addressing mode for data-text relative position indepenedent code.
3217 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3218 'ADDRESS_SYMBOLIC_TXT_REL'.
3219 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3220 (microblaze_legitimate_pic_operand): Exclude function calls from
3221 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3222 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3223 addresses cases.
3224 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3225 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3226 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3227 for 'address + offset'.
3228 (microblaze_expand_prologue): Add new function prologue call for
3229 'r20' assignation.
3230 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3231 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3232 table in case of TARGET_PIC_DATA_TEXT_REL.
3233 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3234 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3235 Add new macros 'UNSPEC_TEXT',
3236 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3237 + exclude function calls from 'UNSPEC_PLT' in case of data text
3238 relative mode.
3239 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3240 new target hook for generating address diff vector tables in case of
3241 flag_pic.
3242 * doc/tm.texi : Regenerate.
3243 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3244 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3245 of addr diff vector generation.
3246 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3247 target hook definition.
3248 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3249 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3250 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3251 option for data text relative.
3253 2018-04-30 Richard Biener <rguenther@suse.de>
3255 * tree-chrec.h (evolution_function_is_constant_p): Remove
3256 redundant check.
3257 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3259 2018-04-30 Richard Biener <rguenther@suse.de>
3261 PR bootstrap/85571
3262 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3264 2018-04-30 Richard Biener <rguenther@suse.de>
3266 PR tree-optimization/28364
3267 PR tree-optimization/85275
3268 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3269 copying first exit test.
3271 2018-04-28 Mark Wielaard <mark@klomp.org>
3273 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3274 dwarf_version >= 5.
3275 (dwarf_AT): Handle DW_AT_addr_base.
3276 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3278 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3280 PR target/84431
3281 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3282 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3283 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3284 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3286 2018-04-28 Richard Biener <rguenther@suse.de>
3288 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3289 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3290 to reflect use. Only add interesting stmts.
3292 2018-04-27 Martin Jambor <mjambor@suse.cz>
3294 PR ipa/85549
3295 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3296 the jump function allows for passing through aggregate values.
3298 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3300 * input.h (in_system_header_at): Convert from macro to inline
3301 function.
3302 (from_macro_expansion_at): Likewise.
3303 (from_macro_definition_at): Likewise.
3305 2018-04-27 Jeff Law <law@redhat.com>
3307 * config.gcc: Mark tile* targets as deprecated/obsolete.
3309 2018-04-27 Richard Biener <rguenther@suse.de>
3311 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3312 fix for ILP32.
3314 2018-04-27 Richard Biener <rguenther@suse.de>
3316 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3318 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3320 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3321 with Yd constraint. Set "preferred_for_speed" attribute from
3322 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3323 with Yd constraint.
3324 (*movdi_internal): Ditto.
3325 (movti_interunit splitters): Remove
3326 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3327 (movdi_interunit splitters): Ditto.
3328 * config/i386/constraints.md (Ye): Remove.
3329 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3331 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3333 PR target/85512
3334 * config/aarch64/constraints.md (Usg): Limit to 31.
3335 (Usj): Limit to 63.
3337 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3339 PR tree-optimization/85529
3340 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3341 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3342 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3343 zero extension or masking of the MSB bit.
3344 (optimize_range_tests): Add FIRST_BB argument, pass it through
3345 to optimize_range_tests_var_bound.
3346 (maybe_optimize_range_tests, reassociate_bb): Adjust
3347 optimize_range_tests callers.
3349 2018-04-26 Richard Biener <rguenther@suse.de>
3350 Jakub Jelinek <jakub@redhat.com>
3352 * cgraph.h (symbol_table): Just declare debug method here.
3353 * symtab.c (symbol_table::debug): Define.
3355 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3357 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3359 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3361 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3362 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3363 (*movdi_internal): Substitute Yi and Yj constraint with x
3364 and Ym and Yn constraint with y constraint. Update "isa"
3365 attribute and set "preferred_for_speed" attribute from
3366 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3367 (*movsi_internal): Ditto.
3368 (*movdf_internal): Ditto.
3369 (*movsf_internal): Ditto.
3370 (*zero_extendsidi2): Ditto.
3371 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3372 (sse2_loadld): Ditto.
3373 (*vec_extract<ssevecmodelower>_0): Ditto.
3374 (*vec_extractv4si_0_zext_sse4): Ditto.
3375 (vec_concatv2di): Ditto.
3376 (*vec_dup<mode>): Ditto.
3377 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3378 * config/i386/constraints.md (Yi): Remove.
3379 (Yj): Remove.
3380 (Ym): Remove.
3381 (Yn): Remove.
3383 2018-04-26 Nathan Sidwell <nathan@acm.org>
3385 * dumpfile.c (dump_open): New.
3386 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3387 (dump_finish): Detect stdio/stderr by value not name.
3389 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3391 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3393 2018-04-26 Tom de Vries <tom@codesourcery.com>
3395 PR target/84952
3396 * config/nvptx/nvptx.c (verify_neutering_jumps)
3397 (verify_neutering_labels): New function
3398 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3400 2018-04-26 Tom de Vries <tom@codesourcery.com>
3402 PR target/84025
3403 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3404 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3405 need neutering.
3407 2018-04-26 Richard Biener <rguenther@suse.de>
3408 Tom de Vries <tom@codesourcery.com>
3410 PR lto/85422
3411 * lto-streamer-out.c (output_function): Fixup loops if required to match
3412 discovery done in the reader.
3414 2018-04-26 Richard Biener <rguenther@suse.de>
3416 PR tree-optimization/85116
3417 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3418 have a loop exit from the single latch predecessor. Remove
3419 case of header with just condition.
3420 (ch_base::copy_headers): Exclude infinite loops from any
3421 processing.
3422 (pass_ch::execute): Record exits.
3424 2018-04-26 Richard Biener <rguenther@suse.de>
3426 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3427 prologue cost vector and pass it to vect_get_load_cost.
3428 (vect_get_peeling_costs_all_drs): Likewise.
3429 (vect_peeling_hash_get_lowest_cost): Likewise.
3430 (vect_enhance_data_refs_alignment): Likewise.
3432 2018-04-26 Richard Biener <rguenther@suse.de>
3434 PR middle-end/85450
3435 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3436 checking of integer<->pointer conversions.
3437 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3438 sign-/zero-extending pointer types.
3439 (expand_omp_for_static_chunk): Likewise.
3441 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3442 Jean Lee <xiaoyur347@gmail.com>
3444 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3445 (TARGET_ASAN_SHADOW_OFFSET): Define.
3446 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3447 true for -fsanitize=address.
3449 2018-04-25 Mark Wielaard <mark@klomp.org>
3451 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3452 shorter ones.
3454 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3456 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3457 than "alu", remove explicit "memory" and "imm_disp" attributes.
3458 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3460 PR middle-end/85414
3461 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3462 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3463 gen_lowpart_no_emit.
3465 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3467 PR target/85473
3468 * config/i386/i386.c (ix86_expand_builtin): Change memory
3469 operand to XI, extend p0 to Pmode.
3470 * config/i386/i386.md: Change unspec volatile and operand
3471 1 mode to XI, change operand 0 mode to P.
3473 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3475 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3476 GET_MODE_MASK before any checking.
3477 (nds32_can_use_bset_p): Likewise.
3478 (nds32_can_use_btgl_p): Likewise.
3480 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3482 * config/nds32/nds32-doubleword.md: New define_split pattern for
3483 illegal register number.
3485 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3487 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3489 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3491 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3493 2018-04-25 Richard Biener <rguenther@suse.de>
3495 * lto-streamer.h (LTO_major_version): Bump to 8.
3497 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3499 * BASE-VER: Set to 9.0.0.
3501 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3503 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3504 in __abskf2 and __powikf2.
3506 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3508 PR target/85512
3509 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3510 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3511 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3512 Use the above on operand 2. Reindent.
3513 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3515 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3517 PR target/85485
3518 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3519 handle OPT_mcet.
3520 * config/i386/i386.opt (mcet): Removed.
3521 * doc/install.texi: Remove -mcet documentation.
3522 * doc/invoke.texi: Likewise.
3524 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3526 PR target/85485
3527 * doc/install.texi: Remove -mcet from bootstrap-cet.
3529 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3531 PR target/85511
3532 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3533 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3534 if TARGET_64BIT.
3536 PR target/85503
3537 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3538 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3539 containing a CONST_VECTOR.
3541 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3543 * doc/install.texi: Update newlib dependency for nvptx.
3545 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3547 PR target/85508
3548 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3549 instead of INTVAL when shifting x left.
3551 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3553 PR tree-optimization/85478
3554 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3555 vect_grouped_store_supported for single element vectors.
3557 2018-04-24 Richard Biener <rguenther@suse.de>
3559 PR target/85491
3560 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3561 load cost increase to the case of non-constant step.
3563 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3565 PR target/84828
3566 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3567 destination if any_malformed_asm.
3569 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3571 PR middle-end/85496
3572 * expr.c (store_field): In the bitfield case, if the value comes from
3573 a function call and is returned in registers by means of a PARALLEL,
3574 do not change the mode of the temporary unless BLKmode and VOIDmode.
3576 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3578 PR rtl-optimization/85423
3579 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3580 dependencies to debug insns when the previous insn is non-debug.
3582 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3584 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3585 enums into a single definition.
3586 (fls): Fix predicates and printing.
3587 (seti): Likewise.
3589 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3591 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3592 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3593 and short u6 immediate.
3594 (check_if_valid_sleep_operand): Remove.
3595 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3597 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3599 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3600 flag_always_save_lp condition.
3601 * config/nds32/nds32.opt (malways-save-lp): New option.
3603 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3605 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3606 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3607 * config/nds32/nds32.h
3608 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3609 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3611 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3613 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3614 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3616 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3617 Chung-Ju Wu <jasonwucj@gmail.com>
3619 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3620 Declare.
3621 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3622 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3624 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3626 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3628 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3630 * config/nds32/nds32-protos.h (nds32_data_alignment,
3631 nds32_local_alignment): Declare.
3632 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3633 nds32_local_alignment): New functions.
3634 (TARGET_CONSTANT_ALIGNMENT): Define.
3635 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3637 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3639 * config/nds32/nds32.c
3640 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3641 (TARGET_MODES_TIEABLE_P): Likewise.
3643 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3645 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3646 level Ofast and Og.
3648 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3649 Chung-Ju Wu <jasonwucj@gmail.com>
3651 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3652 for unaligned access.
3653 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3654 unaligned access.
3655 * config/nds32/nds32-intrinsic.md: Likewise.
3656 * config/nds32/nds32_intrinsic.h: Likewise.
3657 * config/nds32/nds32.h (nds32_builtins): Likewise.
3658 * config/nds32/nds32.opt (munaligned-access): New option.
3659 * config/nds32/nds32.c (nds32_asm_file_start): Display
3660 flag_unaligned_access status.
3662 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3664 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3665 -mno-relax is present.
3666 * config/riscv/linux.h (LINK_SPEC): Ditto.
3668 2018-04-20 Martin Sebor <msebor@redhat.com>
3670 PR c/85365
3671 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3672 for null pointers.
3673 (gimple_fold_builtin_stxcpy_chk): Same.
3674 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3676 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3678 PR target/85456
3679 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3680 __powikf2 when long double is IEEE 128-bit.
3682 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3684 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3685 step to make sure stack always aligned.
3687 2018-04-20 Carl Love <cel@us.ibm.com>
3689 PR target/83402
3690 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3691 size check for arg0.
3693 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3694 Tom de Vries <tom@codesourcery.com>
3696 PR target/85445
3697 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3698 Emit insns for calls too.
3699 (nvptx_find_par): Always look for worker-level predecessor insn.
3700 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3701 calls.
3702 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3703 (nvptx_process_pars): Propagate frames for calls.
3705 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3707 PR target/85469
3708 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3709 Removed.
3710 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3711 (ix86_handle_option): Don't handle OPT_mibt.
3712 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3713 __SHSTK__.
3714 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3715 has_ibt and ibt.
3716 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3717 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3718 (ix86_target_macros): Define __CET__ with flag_cf_protection
3719 for -fcf-protection.
3720 * config/i386/i386.c (isa2_opts): Remove -mibt.
3721 * config/i386/i386.h (TARGET_IBT): Removed.
3722 (TARGET_IBT_P): Likewise.
3723 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3724 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3725 * config/i386/i386.opt (mcet): Update help message.
3726 (mshstk): Likewise.
3727 (mibt): Removed.
3728 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3729 -mcet as an alias for -mshstk.
3731 2018-04-20 Richard Biener <rguenther@suse.de>
3733 PR middle-end/85475
3734 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3735 complexity by forcing a single use of the multiply operand.
3737 2018-04-20 Martin Jambor <mjambor@suse.cz>
3739 ipa/85449
3740 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3741 recursion dependency to only apply to non-clones.
3743 2018-04-20 Martin Jambor <mjambor@suse.cz>
3745 ipa/85447
3746 * ipa-cp.c (create_specialized_node): Check that clones of
3747 self-recursive edges exist during IPA-CP.
3749 2018-04-19 Toon Moene <toon@moene.org>
3751 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3752 by -O3.
3754 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3756 PR tree-optimization/85467
3757 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3758 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3759 VECTOR_CST element to type.
3761 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3763 PR target/85397
3764 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3765 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3766 (builtin_longjmp): Likewise.
3767 (save_stack_nonlocal): New pattern.
3768 (restore_stack_nonlocal): Likewise.
3770 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3772 PR target/85404
3773 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3774 Replace ASM_OUTPUT_LABEL with fprintf.
3776 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3778 PR target/85417
3779 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3780 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3781 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3782 define __IBT__ and __SHSTK__ for -fcf-protection.
3783 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3784 TARGET_IBT.
3785 (ix86_trampoline_init): Likewise.
3786 (x86_output_mi_thunk): Likewise.
3787 (ix86_notrack_prefixed_insn_p): Likewise.
3788 (ix86_option_override_internal): Don't disallow -fcf-protection.
3789 * config/i386/i386.md (rdssp<mode>): Also enable for
3790 -fcf-protection.
3791 (incssp<mode>): Likewise.
3792 (nop_endbr): Likewise.
3793 * config/i386/i386.opt (mcet): Change help message to built-in
3794 functions only.
3795 (mibt): Likewise.
3796 (mshstk): Likewise.
3797 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3798 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3799 enable CET built-in functions.
3801 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3803 * common/config/i386/i386-common.c
3804 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3805 OPTION_MASK_ISA_MOVDIRI_UNSET,
3806 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3807 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3808 * config.gcc (movdirintrin.h): New header.
3809 * config/i386/cpuid.h (bit_MOVDIRI,
3810 bit_MOVDIR64B): New bits.
3811 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3812 and -mmvodir64b.
3813 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3814 (VOID, PVOID, PCVOID)): New function types.
3815 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3816 __builtin_ia32_directstoreu_u64,
3817 __builtin_ia32_movdir64b): New builtins.
3818 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3819 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3820 and -mmovdiri.
3821 (ix86_valid_target_attribute_inner_p): Ditto.
3822 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3823 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3824 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3825 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3826 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3827 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3828 (movdiri<mode>, movdir64b_<mode>): New.
3829 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3830 * config/i386/immintrin.h: Include movdirintrin.h.
3831 * config/i386/movdirintrin.h: New file.
3832 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3834 2018-04-19 Richard Biener <rguenther@suse.de>
3836 PR middle-end/85455
3837 * cfg.c (clear_bb_flags): When loop state says we have
3838 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3840 2018-04-19 Richard Biener <rguenther@suse.de>
3842 PR tree-optimization/84737
3843 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3844 copying restrict info.
3845 (vect_setup_realignment): Use it.
3846 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3847 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3848 the first DR to all generated stores.
3849 (vectorizable_load): Likewise for loads.
3851 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3853 PR tree-optimization/85446
3854 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3855 the integral and pointer types to have the same precision.
3857 * doc/install.texi: Document --disable-cet being the default and
3858 --enable-cet=auto.
3860 2018-04-18 Martin Liska <mliska@suse.cz>
3862 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3863 style.
3865 2018-04-18 Martin Liska <mliska@suse.cz>
3867 Revert
3868 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3870 PR ipa/83983
3871 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3872 arguments if they are comparable.
3874 2018-04-18 Martin Liska <mliska@suse.cz>
3876 Revert
3877 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3879 PR lto/84805
3880 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3881 incomplete types.
3883 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3885 PR target/85388
3886 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3887 ENDBR after calling __morestack.
3889 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3891 PR jit/85384
3892 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3893 by using gcc_base_ver to generate a gcc_driver_version, and use
3894 it when generating GCC_DRIVER_NAME.
3895 * configure: Regenerate.
3897 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3899 PR target/81084
3900 * config.gcc: Obsolete powerpc*-*-*spe*.
3902 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3904 PR debug/84637
3905 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3906 (stabstr_D): Change type of unum from unsigned int to
3907 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3908 type.
3910 2018-04-17 Jim Wilson <jimw@sifive.com>
3912 PR 84856
3913 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3914 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3915 Set arg_pointer_offset after using pretend_args_size.
3917 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3919 PR rtl-optimization/85431
3920 * dse.c (record_store): Ignore zero width stores.
3922 PR sanitizer/85230
3923 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3924 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3925 __builtin_stack_restore rather than after it.
3926 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3927 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3928 argument instead of virtual_dynamic_stack_rtx.
3930 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3932 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3933 New prototype.
3934 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3935 Add note to error message to explain internal mapping of overloaded
3936 built-in function name to non-overloaded built-in function name.
3937 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3938 function.
3940 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3942 PR target/85424
3943 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3944 where the inputs overlap with the output.
3946 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3948 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3949 (=v, v) alternative and explicit "memory" attribute.
3950 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3951 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3952 attributes.
3953 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3954 "sselog1" type instead of "sselog".
3955 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3956 "sselog". Remove explicit "memory" attribute.
3957 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3958 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3959 attributes.
3960 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3961 "sselog1" type instead of "sselog". Remove explicit "memory"
3962 attribute.
3963 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3964 use "sselog1" type instead of "sselog". Remove explicit "memory"
3965 attribute.
3966 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3967 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3968 attributes.
3969 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3970 "sselog1" type instead of "sselog". Remove explicit "memory"
3971 attribute.
3972 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3973 use "sselog1" type instead of "sselog". Remove explicit "memory"
3974 attribute.
3976 PR target/85430
3977 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3979 PR middle-end/85414
3980 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3981 on a SUBREG.
3983 2018-04-17 Martin Jambor <mjambor@suse.cz>
3985 PR ipa/85421
3986 * ipa-cp.c (create_specialized_node): Call
3987 expand_all_artificial_thunks if necessary.
3989 2018-04-17 Martin Liska <mliska@suse.cz>
3991 PR lto/85405
3992 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3993 in message, remote space in between '_G' and '('.
3995 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3997 PR target/85281
3998 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3999 avx512f_vmcmp<mode>3<round_saeonly_name>,
4000 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4001 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4002 avx512f_rndscale<mode><round_saeonly_name>,
4003 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4004 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4005 Use %<iptr>2 instead of %2 for -masm=intel.
4006 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4007 avx512f_vcvttss2usi<round_saeonly_name>,
4008 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4009 -masm=intel.
4010 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4011 avx512f_vcvttsd2usi<round_saeonly_name>,
4012 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4013 Use %q1 instead of %1 for -masm=intel.
4014 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4015 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4016 of %3 for -masm=intel.
4017 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4018 -masm=intel.
4019 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4020 -masm=intel.
4021 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4022 -masm=intel.
4023 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4024 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4025 %g1.
4026 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4027 -masm=intel.
4028 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4029 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4030 %g1 and one with %0 and %1.
4031 (avx512er_vmrcp28<mode><round_saeonly_name>,
4032 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4033 %1 for -masm=intel.
4034 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4035 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4036 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4037 of %0 and %{%4%} for -masm=intel.
4038 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4039 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4040 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4041 order of %0 and %{%5%}%{z%} for -masm=intel.
4043 2018-04-17 Jan Hubicka <jh@suse.cz>
4045 PR lto/85405
4046 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4048 2018-04-17 Martin Liska <mliska@suse.cz>
4050 PR ipa/85329
4051 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4052 for target_clone error message. Make default implementation
4053 clone to be a local declaration.
4054 (separate_attrs): Add new argument and check for an empty
4055 string.
4056 (expand_target_clones): Handle it.
4057 (ipa_target_clone): Make redirection just for target_clones
4058 functions.
4060 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4061 Tom de Vries <tom@codesourcery.com>
4063 PR middle-end/84955
4064 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4065 tiled basic blocks without omp continue statements.
4067 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4069 PR target/83660
4070 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4071 vec_extract expression as having side effects to make sure it gets
4072 a cleanup point.
4074 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4076 PR target/85403
4077 * config/i386/i386.c (get_builtin_code_for_version): Check
4078 error_mark_node.
4080 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4082 PR target/84331
4083 * gcc/config.gcc: Support "skylake".
4084 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4085 PROCESSOR_SKYLAKE.
4086 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4087 (processor_target_table): Add "skylake".
4088 (ix86_option_override_internal): Add "skylake".
4089 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4090 PROCESSOR_CANNONLAKE.
4091 (get_builtin_code_for_version): Fix priority for
4092 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4093 PROCESSOR_SKYLAKE-AVX512.
4094 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4095 (processor_type): Add PROCESSOR_SKYLAKE.
4097 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4098 Jason Merrill <jason@redhat.com>
4100 PR c++/85112
4101 * convert.c (convert_to_integer_1): Use direct recursion for
4102 enumeral types and types with a precision less than the number
4103 of bits in their mode.
4105 2018-04-16 Julia Koval <julia.koval@intel.com>
4107 PR target/84413
4108 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4109 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4111 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4113 PR target/85293
4114 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4115 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4116 and -mno-direct-move.
4118 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4120 PR target/83402
4121 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4122 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4123 Ensure negative shifts result in {0}.
4125 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4127 PR rtl-optimization/79916
4128 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4129 regs (if any) to define how to gnerate SD moves when LRA is in
4130 progress.
4132 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4134 PR rtl-optimization/85393
4135 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4136 * except.c (expand_dw2_landing_pad_for_region): Make static.
4137 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4138 a label and unconditional jump to old_bb, rather than
4139 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4140 basic block.
4142 PR rtl-optimization/85376
4143 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4144 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4145 instead of a specific value.
4147 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4148 Bin Cheng <bin.cheng@arm.com>
4150 PR tree-optimization/82965
4151 PR tree-optimization/83991
4152 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4153 by_profile_only parameter.
4154 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4155 information if the loop was predicted to iterate too many times.
4156 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4158 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4160 PR lto/71991
4161 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4162 always inline.
4164 2018-04-13 Martin Liska <mliska@suse.cz>
4165 Jakub Jelinek <jakub@redhat.com>
4167 PR middle-end/81657
4168 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4169 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4170 * builtins.c (expand_builtin_memory_copy_args): Use
4171 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4172 handle dest_addr == pc_rtx.
4174 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4176 PR target/85291
4177 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4178 asked to not generate direct moves.
4179 (fix_trunc<mode>si2_stfiwx): Similar.
4180 (fix_trunc<mode>si2_internal): Similar.
4182 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4184 PR debug/83157
4185 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4186 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4187 lookup if dest in some wider mode is known to be const0_rtx and
4188 if so, record permanent equivalence for it to be ZERO_EXTEND of
4189 the narrower mode destination.
4191 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4193 * lto-streamer-out.c (output_function): Revert 259346.
4194 * omp-expand.c (expand_oacc_for): Likewise.
4196 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4198 PR rtl-optimization/85354
4199 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4200 * sel-sched.c (sel_global_init): ... here.
4202 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4204 PR target/85238
4205 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4206 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4207 mode for PE-COFF targets.
4208 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4209 (i386_pe_asm_lto_end): Likewise.
4210 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4211 (TARGET_ASM_LTO_END): Likewise.
4212 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4213 (i386_pe_asm_lto_start): New function.
4214 (i386_pe_asm_lto_end): Likewise.
4216 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4217 Richard Biener <rguenther@suse.de>
4219 PR middle-end/84955
4220 * lto-streamer-out.c (output_function): Fix CFG loop state before
4221 streaming out.
4222 * omp-expand.c (expand_oacc_for): Handle calls to internal
4223 functions like regular functions.
4225 2018-04-12 Richard Biener <rguenther@suse.de>
4227 PR lto/85371
4228 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4229 for the early LTO debug to properly generate references to it
4230 during DIE emission. Do not re-use that for the skeleton for
4231 split-dwarf.
4232 (dwarf2out_early_finish): Likewise.
4234 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4236 PR target/85328
4237 * config/i386/sse.md
4238 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4239 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4240 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4241 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4242 and output is a reg, avoid creating invalid lowpart subreg, but
4243 instead split into a 512-bit move. Don't split if not AVX512VL,
4244 input is xmm16+ reg and output is a mem.
4245 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4246 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4247 xmm16+ reg and output is a mem.
4249 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4251 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4252 also for flag_dwarf2_cfi_asm.
4254 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4256 PR rtl-optimization/85342
4257 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4258 a bool scalar var inside of the loop instead. Don't try to update
4259 recog_data.operand after failed apply_change_group.
4261 2018-04-12 Tom de Vries <tom@codesourcery.com>
4263 PR target/85296
4264 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4265 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4266 array with flexible array member as array without given dimension.
4267 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4268 argument for undefined param to true.
4270 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4272 PR target/85321
4273 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4274 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4275 from PowerPC section.
4276 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4277 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4278 help text that is too long.
4279 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4280 help text that is too long.
4281 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4282 help text that is too long.
4284 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4286 * config/alpha/alpha.md (stack_probe_internal): Rename
4287 from "probe_stack". Update all callers.
4289 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4291 PR rtl-optimization/84566
4292 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4293 sched_macro_fuse_insns.
4295 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4297 PR target/84301
4298 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4299 (compute_block_dependences): ... from here.
4301 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4303 PR tree-optimization/85331
4304 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4305 from int to HOST_WIDE_INT.
4307 2018-04-11 Martin Jambor <mjambor@suse.cz>
4309 PR ipa/84149
4310 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4311 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4312 not the same as the source val.
4313 (cgraph_edge_brings_value_p): New parameter.
4314 (gather_edges_for_value): Pass destination value to
4315 cgraph_edge_brings_value_p.
4316 (perhaps_add_new_callers): Likewise.
4317 (get_info_about_necessary_edges): Likewise and exclude values brought
4318 only by self-recursive edges.
4319 (create_specialized_node): Redirect only clones of self-calling edges.
4320 (+self_recursive_pass_through_p): New function.
4321 (find_more_scalar_values_for_callers_subset): Use it.
4322 (find_aggregate_values_for_callers_subset): Likewise.
4323 (known_aggs_to_agg_replacement_list): Removed.
4324 (decide_whether_version_node): Re-calculate known constants for all
4325 remaining context clones.
4327 2018-04-11 Richard Biener <rguenther@suse.de>
4329 PR lto/85339
4330 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4331 from early DWARF output.
4332 (dwarf2out_early_finish): Output line info unconditionally into
4333 early DWARF and add reference to it.
4335 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4337 PR target/85281
4338 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4339 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4340 other than V2DFmode using iptr mode attribute.
4341 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4343 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4345 PR rtl-optimization/84659
4346 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4348 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4350 PR debug/85302
4351 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4352 SIZEP is NULL.
4353 (output_loc_list): Pass address of a dummy size variable even in the
4354 locview handling loop.
4355 (index_location_lists): Add comment on why skip_loc_list_entry can't
4356 call size_of_locs.
4358 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4360 PR target/85261
4361 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4362 into register.
4364 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4366 PR target/85321
4367 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4368 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4369 and -mstring-compare-inline-limit.
4371 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4373 PR target/85287
4374 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4375 for stack clash protection in a register whenever we need it to be in
4376 a register.
4378 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4380 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4381 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4383 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4385 PR target/85321
4386 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4387 the help text.
4388 (mlong-double-): Ditto.
4389 * config/rs6000/sysv4.opt (msdata=): Ditto.
4390 (mtls-size=): Ditto.
4392 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4394 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4395 erroneous entries for
4396 "vector int vec_ldl (int, long int *)", and
4397 "vector unsigned int vec_ldl (int, unsigned long int *)".
4398 Add comments and entries for
4399 "vector bool char vec_ldl (int, bool char *)",
4400 "vector bool short vec_ldl (int, bool short *)",
4401 "vector bool int vec_ldl (int, bool int *)",
4402 "vector bool long long vec_ldl (int, bool long long *)",
4403 "vector pixel vec_ldl (int, pixel *)",
4404 "vector long long vec_ldl (int, long long *)",
4405 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4406 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4407 type tree bool_long_long_type_node and correct definition of
4408 bool_V2DI_type_node to make reference to this new type tree.
4409 (rs6000_mangle_type): Replace erroneous reference to
4410 bool_long_type_node with bool_long_long_type_node.
4411 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4412 comments to emphasize sign distinctions for char and int types and
4413 replace RS6000_BTI_bool_long constant with
4414 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4415 use of RS6000_BTI_pixel.
4416 (bool_long_type_node): Remove this macro definition.
4417 (bool_long_long_type_node): New macro definition
4419 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4421 PR rtl-optimization/85300
4422 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4423 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4424 simplify_unary_operation fails.
4426 2018-04-10 Martin Liska <mliska@suse.cz>
4428 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4429 cgraph_edge and ipa_ref.
4431 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4433 PR target/85177
4434 PR target/85255
4435 * config/i386/sse.md
4436 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4437 computation of the VEC_MERGE selector from mask.
4438 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4439 Fix decoding of the VEC_MERGE selector into mask.
4441 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4443 PR tree-optimization/85286
4444 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4446 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4448 * final.c (final_1): Set insn_last_address as well as
4449 insn_current_address.
4451 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4453 PR target/85173
4454 * explow.c (emit_stack_probe): Call validize_mem on memory location
4455 before passing it to gen_probe_stack. Create address operand and
4456 legitimize it for the probe_stack_address case.
4458 2018-04-09 Jan Hubicka <jh@suse.cz>
4460 PR lto/85078
4461 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4462 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4463 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4464 walk basetypes.
4465 (free_lang_data): Rebuild type inheritance graph.
4467 2018-04-09 Martin Sebor <msebor@redhat.com>
4469 * invoke.texi (-finline-small-functions): Mention other optimization
4470 options.
4471 (-findirect-inlining, -fpartial-inlining): Same.
4472 (-finline-functions-called-once): Same.
4473 (-freorder-blocks-and-partition): Same.
4475 2018-04-09 Jan Hubicka <jh@suse.cz>
4477 PR rtl/84058
4478 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4479 jumps; choose last target that matches the criteria (i.e.
4480 no partition changes for non-crossing jumps).
4481 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4482 support for redirecting crossing jumps to non-crossing.
4484 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4486 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4487 also for naked functions.
4489 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4491 * config/arc/arc.md (add_shift): New pattern.
4492 (add_shift2): Likewise.
4493 (sub_shift): Likewise.
4494 (sub_shift_cmp0_noout): Likewise.
4495 (compare_si_ashiftsi): Likewise.
4496 (xbfu_cmp0_noout): New combine pattern.
4497 (xbfu_cmp0"): Likewise.
4498 (movsi_set_cc_insn): Place the predicable variant first.
4499 (commutative_binary_cmp0_noout): Remove clobber.
4500 (commutative_binary_cmp0): New pattern.
4501 (noncommutative_binary_cmp0): Likewise.
4502 (noncommutative_binary_cmp0_noout): Likewise.
4503 (noncommutative_binary_comparison_result_used): Removed.
4504 (rsub_cmp0): New pattern.
4505 (rsub_cmp0_noout): Likewise.
4506 (extzvsi): Changed, keep only meaningful variants.
4507 (SQH, SEZ): New iterators.
4508 (SQH_postfix): New mode attribute.
4509 (SEZ_prefix): New code attribute.
4510 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4511 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4512 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4513 of numerical value.
4514 (noncommutative_operator): Check the availability of barrel
4515 shifter option.
4517 2018-04-09 Richard Biener <rguenther@suse.de>
4519 PR tree-optimization/85284
4520 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4521 Only use the niter constraining form of simple_iv when the exit
4522 is always executed.
4524 2018-04-09 Tom de Vries <tom@codesourcery.com>
4526 PR target/84041
4527 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4528 (define_expand "*memory_barrier"): New define_expand.
4529 (define_insn "memory_barrier"): New insn.
4531 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4533 PR rtl-optimization/80463
4534 PR rtl-optimization/83972
4535 PR rtl-optimization/83480
4537 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4538 correct producer for the insn.
4539 (tidy_control_flow): Fixup seqnos in case of debug insns.
4541 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4543 PR rtl-optimization/83913
4545 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4546 different sched-times when merging exprs.
4548 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4550 PR rtl-optimization/83962
4552 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4553 tidy_fallthru_edge and tidy_control_flow.
4555 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4557 PR rtl-optimization/83530
4559 * sel-sched.c (force_next_insn): New global variable.
4560 (remove_insn_for_debug): When force_next_insn is true, also leave only
4561 next insn in the ready list.
4562 (sel_sched_region): When the region wasn't scheduled, make another pass
4563 over it with force_next_insn set to 1.
4565 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4567 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4568 into tm_file.
4569 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4570 for interrupt control.
4571 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4572 functions for interrupt control.
4573 * config/nds32/nds32-intrinsic.md: Likewise.
4574 * config/nds32/nds32_intrinsic.h: Likewise.
4575 * config/nds32/nds32.h (nds32_builtins): Likewise.
4577 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4579 * config/nds32/nds32.c (nds32_init_machine_status,
4580 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4581 strict_aligned_p field.
4582 (nds32_expand_to_rtl_hook): New function.
4583 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4584 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4586 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4587 Chung-Ju Wu <jasonwucj@gmail.com>
4589 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4590 * config/nds32/nds32-n7.md: New file.
4591 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4593 pipeline.
4594 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4595 * config/nds32/nds32.md (pipeline_model): Add n7.
4596 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4597 * config/nds32/pipelines.md: Include n7 settings.
4599 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4600 Chung-Ju Wu <jasonwucj@gmail.com>
4602 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4603 * config/nds32/nds32-e8.md: New file.
4604 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4605 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4606 pipeline.
4607 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4608 * config/nds32/nds32.md (pipeline_model): Add e8.
4609 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4610 * config/nds32/pipelines.md: Include e8 settings.
4612 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4613 Chung-Ju Wu <jasonwucj@gmail.com>
4615 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4616 * config/nds32/nds32-n8.md: New file.
4617 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4618 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4619 pipeline.
4620 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4621 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4622 * config/nds32/nds32.md (pipeline_model): Add n8.
4623 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4624 * config/nds32/pipelines.md: Include n8 settings.
4626 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4627 Chung-Ju Wu <jasonwucj@gmail.com>
4629 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4630 * config/nds32/nds32-n9-2r1w.md: New file.
4631 * config/nds32/nds32-n9-3r2w.md: New file.
4632 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4633 nds32_register_ports): New or modify for cpu n9.
4634 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4635 pipeline.
4636 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4637 * config/nds32/nds32-utils.c: New file.
4638 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4639 TARGET_MUL_SLOW): Define.
4640 * config/nds32/nds32.md (pipeline_model): New attribute.
4641 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4642 New options that support cpu n9.
4643 * config/nds32/pipelines.md: Include n9 settings.
4644 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4646 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4648 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4649 information if necessary.
4650 (output_cond_branch_compare_zero): Likewise.
4651 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4652 (nds32_target_alignment): Refine for alignment.
4653 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4654 (FUNCTION_BOUNDARY): Modify.
4655 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4656 align case.
4657 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4659 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4661 * config/nds32/constants.md (unspec_volatile_element): Add values for
4662 TLB operation and data prefetch.
4663 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4664 functions for TLB operation and data prefetch.
4665 * config/nds32/nds32-intrinsic.md: Likewise.
4666 * config/nds32/nds32_intrinsic.h: Likewise.
4667 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4668 (nds32_print_operand): Likewise.
4669 * config/nds32/nds32.h (nds32_builtins): Likewise.
4671 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4672 Andrew Pinski <pinsika@gcc.gnu.org>
4674 PR middle-end/82976
4675 * match.pd: Use constant_boolean_node of correct type instead of
4676 boolean_true_node or boolean_false_node for simplifying
4677 pointer comparisons to zero.
4679 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4681 PR tree-optimization/80021
4682 * tree.c (verify_type_variant): Make error call in verify_variant_match
4683 translatable and remove final full stop.
4685 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4687 * config/nds32/constants.md (unspec_volatile_element): Add
4688 UNSPEC_VOLATILE_EH_RETURN.
4689 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4690 nds32_output_stack_pop): Support dwarf exception handling process.
4691 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4692 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4693 exception handling process.
4694 (nds32_compute_stack_frame): Likewise.
4695 (nds32_return_addr_rtx): Likewise.
4696 (nds32_initial_elimination_offset): Likewise.
4697 (nds32_expand_prologue): Likewise.
4698 (nds32_expand_epilogue): Likewise.
4699 (nds32_dynamic_chain_address): New function.
4700 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4701 exception handling.
4702 (DYNAMIC_CHAIN_ADDRESS): Define.
4703 (EH_RETURN_DATA_REGNO): Define.
4704 (EH_RETURN_STACKADJ_RTX): Define.
4705 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4706 patterns for dwarf exception handling.
4708 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config/nds32/nds32.h: Clean up obsolete macros.
4712 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4714 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4715 Add enum values for particular instructions.
4716 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4717 particular intrinsic functions.
4718 * config/nds32/nds32-intrinsic.md: Likewise.
4719 * config/nds32/nds32_intrinsic.h: Likewise.
4720 * config/nds32/nds32.h (nds32_builtins): Likewise.
4721 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4722 (btst, ave): New patterns for particular instructions.
4724 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4726 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4727 Add enum values for atomic load/store and memory sync.
4728 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4729 and memory sync.
4730 * config/nds32/nds32-intrinsic.md: Likewise.
4731 * config/nds32/nds32_intrinsic.h: Likewise.
4732 * config/nds32/nds32.h (nds32_builtins): Likewise.
4734 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4736 PR tree-optimization/85257
4737 * fold-const.c (native_encode_vector): If not all elts could fit
4738 and off is -1, return 0 rather than offset.
4739 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4740 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4741 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4742 adjust buffer in native_interpret_expr call.
4744 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4746 * config/nds32/constants.md (unspec_volatile_element): Add cache
4747 control enum values.
4748 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4749 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4750 * config/nds32/nds32.c (nds32_cctl_names): New.
4751 (nds32_print_operand): Handle cache control register names.
4752 * config/nds32/nds32.h (nds32_builtins): New enum values.
4753 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4754 macros.
4755 * config/nds32/nds32.md (type): Add mmu.
4756 * config/nds32/pipelines.md (simple_insn): Add mmu.
4758 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4760 * config/nds32/nds32.md (type): Remove call.
4761 * config/nds32/pipelines.md (simple_insn): Likewise.
4763 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4765 * config/nds32/constants.md (unspec_volatile_element): Add
4766 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4767 UNSPEC_VOLATILE_FMFCFG.
4768 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4769 description for fmfcfg and fmfcsr.
4770 (bdesc_1arg): Add fmtcsr.
4771 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4772 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4773 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4774 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4775 unspec_fmfcfg): New patterns.
4776 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4777 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4778 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4779 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4780 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4781 __nds32__fmfcfg): Define.
4783 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4785 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4786 intrinsic register names.
4787 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4788 intrinsic register enum values and macros.
4790 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4792 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4793 for load/store addressing form.
4794 (nds32_print_operand_address): Likewise.
4796 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4798 PR target/85196
4799 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4800 based on LABEL_REF. Remove useless assertion.
4801 (pic_address_needs_scratch): Fix formatting.
4802 (sparc_legitimize_pic_address): Minor tweaks.
4803 (sparc_delegitimize_address): Adjust assertion accordingly.
4804 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4805 into symbolic_operand.
4806 (movsi_high_pic_label_ref): Likewise.
4807 (movsi_lo_sum_pic_label_ref): Likewise.
4808 (movdi_pic_label_ref): Likewise.
4809 (movdi_high_pic_label_ref): Likewise.
4810 (movdi_lo_sum_pic_label_ref): Likewise.
4812 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4814 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4815 custom LIB_SPEC setup.
4817 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4818 Kito Cheng <kito.cheng@gmail.com>
4820 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4821 * config/riscv/freebsd.h: New.
4823 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4825 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4826 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4827 file.
4829 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4830 Kito Cheng <kito.cheng@gmail.com>
4832 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4833 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4834 * config/nds32/nds32-protos.h (nds32_output_call,
4835 nds32_output_return): Declare.
4836 * config/nds32/nds32.md: Refine all the call and return patterns.
4838 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4840 PR debug/85252
4841 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4842 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4844 PR rtl-optimization/84872
4845 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4846 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4847 EDGE_CROSSING edge.
4849 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4851 * expr.c (copy_blkmode_to_reg): Revert 254862.
4852 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4854 2018-04-06 Richard Biener <rguenther@suse.de>
4856 PR middle-end/85244
4857 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4858 after seeing a component reference with an adjacent field. Treat
4859 refs to arrays at struct end of external decls similar to
4860 refs to unconstrained commons.
4862 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4864 PR sanitizer/85213
4865 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4866 look through SAVE_EXPRs with non-side-effects argument. Adjust
4867 recursive calls.
4868 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4869 save_p here.
4871 2018-04-06 Richard Biener <rguenther@suse.de>
4873 PR middle-end/85180
4874 * alias.c (find_base_term): New wrapper around find_base_term
4875 unwinding CSELIB_VAL_PTR changes.
4876 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4877 recursion.
4879 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4881 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4882 instructions.
4883 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4884 constant definitions.
4885 ("nop"): lr 0,0 -> nopr r0
4886 ("nop_lr0", "nop_lr1"): New insn definitions.
4888 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4890 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4891 NDS32_V3PUSH_AVAILABLE_P macro.
4893 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4894 Chung-Ju Wu <jasonwucj@gmail.com>
4896 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4897 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4898 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4899 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4900 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4901 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4902 * config/nds32/constraints.md: New constraints and checking for hard
4903 float configuration.
4904 * config/nds32/iterators.md: New mode iterator and attribute for hard
4905 float configuration.
4906 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4907 patterns.
4908 * config/nds32/nds32-fpu.md: New file.
4909 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4910 deal with hard float code generation.
4911 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4912 ARCH_V3S.
4913 (abi_type, float_reg_number): New enum type.
4914 * config/nds32/nds32-predicates.c: New predicates for hard float.
4915 * config/nds32/nds32-protos.h: Declare functions for hard float.
4916 * config/nds32/nds32.c: Implementation for hard float configuration.
4917 * config/nds32/nds32.h: Definitions for hard float configuration.
4918 * config/nds32/nds32.md: Include hard float machine description and
4919 modify patterns for hard float configuration.
4920 * config/nds32/nds32.opt: New options for hard float configuration.
4921 * config/nds32/predicates.md: New predicates for hard float
4922 configuration.
4924 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4926 * common/config/nds32/nds32-common.c
4927 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4929 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4931 PR middle-end/85195
4932 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4933 CONSTRUCTOR_ELT (ctor, ...)->value.
4935 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4937 PR target/85193
4938 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4940 2018-04-05 Tom de Vries <tom@codesourcery.com>
4942 PR target/85204
4943 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4944 cond jump.
4946 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4947 Kito Cheng <kito.cheng@gmail.com>
4949 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4950 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4951 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4952 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4954 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4955 Kito Cheng <kito.cheng@gmail.com>
4957 * config/nds32/constraints.md (Ufe): New memory constraint.
4958 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4959 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4960 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4961 operands.
4962 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4963 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4965 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4967 * config/nds32/nds32.md: Use optimize_size in the condition for
4968 alu-shift instructions.
4970 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4972 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4974 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4976 * config/nds32/nds32.md (negsi2): Refine pattern.
4978 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4979 Chung-Ju Wu <jasonwucj@gmail.com>
4981 * config/nds32/iterators.md (shift_rotate): New code iterator.
4982 (shift): New code attribute.
4983 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4984 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4985 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4986 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4987 bit-wise operations.
4988 (andsi3, *andsi3): Ditto.
4989 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4990 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4991 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4992 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4993 nds32_ior_operand, nds32_xor_operand): New predicates.
4995 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4997 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4998 (addsi3, subsi3): ... this.
5000 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5004 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5006 * config/nds32/nds32.md: Adjust indention.
5008 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5010 * config/nds32/nds32.md (feature): New attribute.
5012 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5014 * config/nds32/nds32.md (subtype): New attribute.
5016 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5018 PR target/85203
5019 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5020 expansion to perform a bitwise AND of the argument followed by a
5021 boolean negation of the result.
5023 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5025 PR rtl-optimization/84878
5026 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5027 the basic block. Assert the use reference is not artificial and that
5028 it has an associated insn.
5030 2018-04-04 Michael Matz <matz@suse.de>
5032 * builtins.c (compute_objsize): Pass correct operand
5033 to array_at_struct_end_p.
5035 2018-04-04 Richard Biener <rguenther@suse.de>
5037 PR lto/85176
5038 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5039 from contexts for DINFO_LEVEL_TERSE and below.
5041 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5043 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5044 resiter_operand condition.
5045 * config/nds32/nds32.md (*move<mode>): Ditto.
5047 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5048 Monk Chiang <sh.chiang04@gmail.com>
5050 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5052 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5054 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5056 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5057 Kito Cheng <kito.cheng@gmail.com>
5059 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5060 nds32_cond_code_str, output_cond_branch,
5061 output_cond_branch_compare_zero, nds32_expand_cbranch,
5062 nds32_expand_cstore, nds32_expand_movcc,
5063 nds32_output_cbranchsi4_equality_zero,
5064 nds32_output_cbranchsi4_equality_reg,
5065 nds32_output_cbranchsi4_equality_reg_or_const_int,
5066 nds32_output_cbranchsi4_greater_less_zero: New functions.
5067 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5068 nds32_expand_cstore, nds32_expand_movcc,
5069 nds32_output_cbranchsi4_equality_zero,
5070 nds32_output_cbranchsi4_equality_reg,
5071 nds32_output_cbranchsi4_equality_reg_or_const_int,
5072 nds32_output_cbranchsi4_greater_less_zero): Declare.
5073 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5074 nds32_rimm11s_operand): New predicates.
5075 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5076 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5077 patterns.
5079 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5081 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5082 * config/nds32/nds32.md: Ditto.
5083 * config/nds32/pipelines.md: Ditto.
5085 2018-04-04 Richard Biener <rguenther@suse.de>
5087 PR tree-optimization/85168
5088 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5089 propagating abnormals.
5091 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5093 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5095 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5096 Kito Cheng <kito.cheng@gmail.com>
5098 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5099 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5100 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5101 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5102 * config/nds32/nds32.md (sibcall_internal): New.
5103 (sibcall_register): Remove.
5104 (sibcall_immediate): Remove.
5105 (sibcall_value_internal): New.
5106 (sibcall_value_register): Remove.
5107 (sibcall_value_immediate): Remove.
5108 * config/nds32/predicates.md (nds32_general_register_operand): New.
5109 (nds32_call_address_operand): New.
5111 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5113 PR rtl-optimization/85167
5114 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5115 bb_defs if *split_p, instead preinitialize it to NULL.
5117 PR tree-optimization/85156
5118 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5119 evaluating the argument multiple times.
5121 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5123 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5124 than vector.
5125 (_mm_cvtpd_ps): Likewise.
5126 (_mm_cvttpd_epi32): Likewise.
5127 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5128 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5129 vector, pixel, and bool following altivec.h include.
5131 2018-04-03 Martin Sebor <msebor@redhat.com>
5133 * doc/extend.texi (Common Function Attributes): Clarify.
5134 (const attribute): Likewise.
5135 (pure attribute): Likewise.
5137 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5139 PR target/85169
5140 * config/i386/i386.c (ix86_expand_vector_set): Use
5141 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5143 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5145 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5146 instructions when changing rounding bits to preserve precision bits
5147 in the x87 control word.
5149 2018-04-03 Martin Liska <mliska@suse.cz>
5151 PR tree-optimization/82491
5152 * rtl.h (strip_offset_and_add): Replace += suboffset with
5153 poly_uint64 () + suboffset.
5155 2018-03-29 Martin Liska <mliska@suse.cz>
5156 Martin Jambor <mjambor@suse.cz>
5158 PR ipa/84947
5159 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5160 param_type is not an integral or pointer type.
5162 2018-04-03 Richard Biener <rguenther@suse.de>
5164 * sese.h (recompute_all_dominators): Remove.
5166 2018-04-02 Martin Sebor <msebor@redhat.com>
5168 * doc/invoke.texi (-Wrestrict): Fix typos.
5170 2018-04-02 Jim Wilson <jimw@sifive.com>
5172 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5173 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5174 (<optab>di3, <optab>si3_extend): Likewise.
5175 (<optab>si3_mask, <optab>si3_mask_1): New.
5176 (<optab>di3_mask, <optab>di3_mask_1): New.
5177 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5178 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5179 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5181 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5183 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5184 example.
5186 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5188 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5189 (nds32_canonicalize_comparison): New function.
5191 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5192 Kito Cheng <kito.cheng@gmail.com>
5193 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5195 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5196 * config/nds32/constants.md (unspec_volatile_element): Add
5197 UNSPEC_VOLATILE_RELAX_GROUP.
5198 * config/nds32/nds32-relax-opt.c: New file.
5199 * config/nds32/nds32-predicates.c
5200 (nds32_symbol_load_store_p): New function.
5201 * config/nds32/nds32-protos.h
5202 (nds32_symbol_load_store_p): Declare function.
5203 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5204 * config/nds32/nds32.c
5205 (nds32_register_pass): New function to register pass.
5206 (nds32_register_passes): New function to register passes.
5207 * config/nds32/nds32.md (relax_group): New pattern.
5208 * config/nds32/nds32.opt (mrelax-hint): New option.
5209 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5211 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5213 * config/nds32/t-nds32: Modify files dependency.
5215 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5217 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5218 (PROFILE_HOOK): Define its implementation.
5220 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5222 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5223 type and 32-bit size.
5225 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5227 PR middle-end/85090
5228 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5229 (V_128_256): New mode iterator.
5230 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5231 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5232 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5233 of V.
5234 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5235 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5237 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5239 PR target/83315
5240 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5241 NaN inputs correctly.
5243 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5245 PR target/80546
5246 * config/rs6000/vsx.md (??r): New mode attribute.
5247 (*vsx_mov<mode>_64bit): Use it.
5248 (*vsx_mov<mode>_32bit): Likewise.
5250 2018-03-30 Martin Sebor <msebor@redhat.com>
5252 PR tree-optimization/84818
5253 * builtins.c (check_access): Use warning_n.
5255 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5257 PR target/83822
5258 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5259 condition.
5260 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5261 condition.
5263 2018-03-30 Julia Koval <julia.koval@intel.com>
5265 PR target/84413
5266 * x86-tune.def (movx, partial_reg_dependency): Enable for
5267 m_SKYLAKE_AVX512.
5269 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5271 PR inline-asm/84985
5272 * lra-constraints.c (process_alt_operands): Move setting
5273 this_alternative_matches below.
5275 2018-03-29 Martin Liska <mliska@suse.cz>
5277 PR lto/84995.
5278 * doc/invoke.texi: Document how LTO works with debug info.
5279 Describe auto-load support of binutils. Mention 'x86-64'
5280 as valid option value of -march option.
5282 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5284 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5286 PR c/85094
5287 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5288 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5289 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5290 checking.
5292 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5294 PR target/84912
5295 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5296 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5297 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5298 for RS6000_BTM_POWERPC64.
5299 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5300 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5301 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5302 definition.
5303 (DIVDE): Use it.
5304 (DIVDEU): Likewise.
5306 2018-03-28 Carl Love <cel@us.ibm.com>
5308 Revert
5309 2017-09-27 Carl Love <cel@us.ibm.com>
5311 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5312 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5313 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5314 fctiw instruction.
5316 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5318 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5319 instead of __vector bool.
5320 (_mm_max_pu8): Likewise.
5321 (_mm_min_pi16): Likewise.
5323 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5325 PR target/84912
5326 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5327 (DIVWEUO): Likewise.
5328 (DIVDEO): Likewise.
5329 (DIVDEUO): Likewise.
5330 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5331 DIVWEUO and DIVDEUO.
5332 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5333 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5334 (div_extend): Likewise.
5335 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5336 builtin function.
5337 (__builtin_divweuo): Likewise.
5338 (__builtin_divdeo): Likewise.
5339 (__builtin_divdeuo): Likewise.
5341 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5343 PR target/85095
5344 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5345 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5347 PR tree-optimization/82004
5348 * gimple-match-head.c (optimize_pow_to_exp): New function.
5349 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5350 Don't fold to exp if optimize_pow_to_exp is false.
5352 2018-03-28 Martin Liska <mliska@suse.cz>
5354 PR other/84819
5355 * calls.c (initialize_argument_information): Fix trailing space.
5356 * common.opt: Fix typo and provide better explanation for
5357 -fsanitize-coverage option.
5358 * config/i386/i386.opt: Fix typo.
5360 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5361 Martin Liska <mliska@suse.cz>
5363 PR sanitizer/85081
5364 * gimplify.c (asan_poison_variable): Don't do the check for
5365 gimplify_omp_ctxp here.
5366 (gimplify_decl_expr): Do it here.
5367 (gimplify_target_expr): Likewise.
5369 2018-03-28 Martin Liska <mliska@suse.cz>
5371 PR target/84988
5372 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5373 chkp_type_bounds_count if MPX is not enabled.
5375 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5377 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5379 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5381 PR target/84914
5382 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5383 function to create the function decl for complex long double
5384 multiply and divide for -mabi=ieeelongdouble.
5385 (init_float128_ieee): Call it.
5387 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5389 PR target/85044
5390 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5391 -fcf-protection=branch -mibt.
5392 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5394 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5396 PR target/81863
5397 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5399 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5401 PR target/85056
5402 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5403 extern array declarations.
5405 2018-03-27 Richard Biener <rguenther@suse.de>
5407 PR middle-end/84067
5408 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5409 explicit single_use checks.
5411 2018-03-27 Richard Biener <rguenther@suse.de>
5413 PR tree-optimization/85082
5414 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5415 Valueize the VUSE.
5417 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5419 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5420 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5421 Turn on fasynchronous-unwind-tables and funwind-tables.
5423 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5425 PR target/85073
5426 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5427 (*bmi_blsr_<mode>_ccz): Ditto.
5429 2018-03-26 Tom de Vries <tom@codesourcery.com>
5431 PR tree-optimization/85063
5432 * omp-general.c (offloading_function_p): New function. Factor out
5433 of ...
5434 * omp-offload.c (pass_omp_target_link::gate): ... here.
5435 * omp-general.h (offloading_function_p): Declare.
5436 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5437 with attribute omp declare target for offloading functions.
5439 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5441 PR tree-optimization/84005
5442 * tree-data-ref.h (get_base_for_alignment): Declare.
5443 * tree-data-ref.c (get_base_for_alignment_1): New function.
5444 (get_base_for_alignment): Likewise.
5445 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5446 get_base_for_alignment to find a suitable base object, instead
5447 of always using drb->base_address.
5449 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5451 PR inline-asm/85022
5452 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5453 known size by default.
5455 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5457 PR inline-asm/85030
5458 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5459 and non BLKmode operands.
5461 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5463 PR target/85026
5464 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5465 Clean up attributes.
5467 2018-03-23 Richard Biener <rguenther@suse.de>
5469 PR debug/85020
5470 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5471 we are going to emit early debug for LTO.
5473 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5475 PR inline-asm/85034
5476 * function.c (match_asm_constraints_1): Don't optimize if input
5477 doesn't satisfy general_operand predicate for output's mode.
5479 PR inline-asm/85022
5480 * alias.c (write_dependence_p): Don't require for x_canonicalized
5481 non-VOIDmode if x has VOIDmode.
5483 PR sanitizer/85029
5484 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5485 just don't try to optimize it rather than assert it never happens.
5487 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5489 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5490 macro expansions for definition of ST_INTERNAL_<mode> and
5491 LD_INTERNAL_<mode> builtins.
5492 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5493 Remove prototype.
5494 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5495 function.
5496 (altivec_expand_st_builtin): Likewise.
5497 (altivec_expand_builtin): Remove calls to deleted functions.
5498 (rs6000_address_for_altivec): Delete this function.
5499 * config/rs6000/vector.md: Remove expands for
5500 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5502 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5504 PR target/84826
5505 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5506 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5507 re-computing once computed.
5508 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5509 (arm_init_machine_status): Initialize
5510 machine->static_chain_stack_bytes.
5512 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5514 PR target/84760
5515 * doc/extend.texi: Add four new prototypes for vec_ld.
5516 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5517 definitions for more logical presentation.
5518 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5519 entries for V1TI variants of __builtin_altivec_ld builtin.
5520 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5521 handling of V1TI variant of LVX icode pattern.
5522 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5523 (rs6000_gimple_fold_builtin): Likewise.
5524 (altivec_init_builtins): Add code to define
5525 __builtin_altivec_lvx_v1ti function.
5527 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5529 PR inline-asm/84941
5530 * function.c (match_asm_constraints_1): Don't do the optimization
5531 if input isn't a REG, SUBREG, MEM or constant.
5533 2018-03-22 Tom de Vries <tom@codesourcery.com>
5535 PR tree-optimization/84956
5536 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5537 bb_has_abnormal_pred.
5539 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5541 PR sanitizer/85018
5542 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5543 DECL_INITIAL (decl) to decl at the end.
5544 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5545 adjust the comment.
5547 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5549 * doc/extend.texi (__builtin_tgmath): Document when complex
5550 integer types are treated as _Complex _Float64.
5552 2018-03-21 Tom de Vries <tom@codesourcery.com>
5554 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5556 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5558 PR tree-optimization/84960
5559 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5560 if it is ENTRY block, move them into single succ of ENTRY in that case.
5562 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5564 PR tree-optimization/84811
5565 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5566 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5567 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5568 (known_subrange_p): Update accordingly. Cast each value involved
5569 in the size comparison, rather than casting the result of the
5570 subtraction.
5572 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5574 PR tree-optimization/84982
5575 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5576 by flipping the least significant bit rather than all bits from
5577 bitpos to bitpos + bitsize - 1.
5579 2018-03-21 Nathan Sidwell <nathan@acm.org>
5581 * doc/extend.texi (Deprecated Features): Remove mention of
5582 long-deleted deprecations.
5584 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5586 PR jit/84288
5587 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5588 * configure: Regenerate.
5590 2018-03-21 Tom de Vries <tom@codesourcery.com>
5592 PR tree-optimization/83126
5593 * tree-parloops.c (num_phis): New function.
5594 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5596 2018-03-21 Nathan Sidwell <nathan@acm.org>
5598 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5599 mention anon-struct/union members and trailing attributes.
5601 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5603 PR tree-optimization/84969
5604 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5605 builtin memset partitions if they set different rhs values.
5607 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5609 PR rtl-optimization/84989
5610 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5611 VEC_DUPLICATE with scalar result mode.
5613 2018-03-21 Martin Liska <mliska@suse.cz>
5615 PR ipa/84963
5616 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5617 not intended return statement.
5619 2018-03-21 Martin Liska <mliska@suse.cz>
5621 PR target/84988
5622 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5623 (chkp_find_bound_slots_1): Limit number of iterations.
5625 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5627 PR target/84838
5628 * Minor grammar fixes for x86 options.
5630 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5632 PR debug/84875
5633 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5634 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5636 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5638 PR target/83789
5639 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5640 (altivec_lvx_<mode>_1op): Likewise.
5641 (altivec_stvx_<mode>_2op): Likewise.
5642 (altivec_stvx_<mode>_1op): Likewise.
5643 (altivec_lvx_<VM2:mode>): New define_expand.
5644 (altivec_stvx_<VM2:mode>): Likewise.
5645 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5646 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5647 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5648 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5649 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5650 (rs6000_gen_lvx): Likewise.
5651 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5652 (altivec_expand_stv_builtin): Likewise.
5653 (altivec_expand_builtin): Likewise.
5654 * config/rs6000/vector.md: Likewise.
5656 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5658 PR target/82518
5659 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5660 BYTES_BIG_ENDIAN.
5662 2018-03-20 Richard Biener <rguenther@suse.de>
5664 PR target/84986
5665 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5666 sign-conversions as zero, fall back to standard scalar_stmt
5667 cost for the rest.
5669 2018-03-20 Martin Liska <mliska@suse.cz>
5671 PR ipa/84825
5672 * predict.c (rebuild_frequencies): Handle case when we have
5673 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5675 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5677 PR target/84990
5678 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5679 flag_section_anchors.
5680 * varasm.c (use_blocks_for_decl_p): Remove hack for
5681 dw2_force_const_mem.
5683 PR target/84845
5684 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5685 to ...
5686 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5687 be created, use lowpart_subreg of operands[0] rather than operands[0]
5688 itself.
5689 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5690 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5691 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5692 and n constraint instead of aarch64_shift_imm_di and Usd.
5693 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5694 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5696 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5698 PR target/82989
5699 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5700 to favor GPR over NEON registers.
5701 (<shift>di3_neon): Likewise.
5703 2018-03-20 Tom de Vries <tom@codesourcery.com>
5705 PR target/84952
5706 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5707 (nvptx_process_pars): Emit bar.sync asap and alap.
5709 2018-03-20 Tom de Vries <tom@codesourcery.com>
5711 PR target/84954
5712 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5713 seen_label if seen_label is already set.
5715 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5717 PR target/84945
5718 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5719 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5720 Use 1U instead of 1. Formatting fixes.
5722 PR c/84953
5723 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5724 instead of TREE_TYPE (s1) for the return value.
5726 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5728 PR tree-optimization/84946
5729 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5730 bitsize + bitsize in poly_uint64 rather than poly_int64.
5732 PR sanitizer/78651
5733 * dwarf2asm.c: Include fold-const.c.
5734 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5735 of decl rather than decl itself.
5737 PR rtl-optimization/84643
5738 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5740 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5742 PR sanitizer/78651
5743 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5744 calling assemble_variable.
5746 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5748 PR target/81647
5749 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5750 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5752 2018-03-19 Jim Wilson <jimw@sifive.com>
5754 PR bootstrap/84856
5755 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5756 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5757 (riscv_first_stack_step): Likewise.
5758 (riscv_option_override): Use STACK_BOUNDARY instead of
5759 MIN_STACK_BOUNDARY.
5760 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5761 MIN_STACK_BOUNDARY.
5762 (BIGGEST_ALIGNMENT): Set to 128.
5763 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5764 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5765 STACK_BOUNDARY.
5767 2018-03-19 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/84933
5770 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5771 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5773 2018-03-19 Richard Biener <rguenther@suse.de>
5775 PR tree-optimization/84859
5776 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5777 (cond_if_else_store_replacement): Perform sinking operation on
5778 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5779 Generalize what a BB with a single eligible store is.
5781 2018-03-19 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/84929
5784 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5785 chrec_is_positive against non-chrec arg.
5787 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5789 PR target/84711
5790 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5792 2018-03-18 Martin Liska <mliska@suse.cz>
5794 PR rtl-optimization/84635
5795 * regrename.c (build_def_use): Use matches_mode only when
5796 matches >= 0.
5798 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5800 PR tree-optimization/84913
5801 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5802 vectorize chains of COND_EXPRs.
5804 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5806 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5808 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5810 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5812 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5814 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5816 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5817 Kito Cheng <kito.cheng@gmail.com>
5819 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5820 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5821 (nds32_adjust_reg_alloc_order): New function.
5822 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5824 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5826 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5827 nds32_print_operand, nds32_print_operand_address): Use
5828 HOST_WIDE_INT_PRINT_DEC instead.
5830 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5832 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5834 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5836 PR target/84902
5837 * config/i386/i386.c (initial_ix86_tune_features,
5838 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5839 unsigned long long.
5840 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5841 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5842 rather than 1u << ix86_tune. Formatting fix.
5843 (ix86_option_override_internal): Change ix86_arch_mask from
5844 unsigned int to unsigned HOST_WIDE_INT, initialize to
5845 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5846 (ix86_function_specific_restore): Likewise.
5848 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5850 PR target/84899
5851 * postreload.c (reload_combine_recognize_pattern): Perform
5852 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5853 truncate_int_for_mode the result for the destination's mode.
5855 PR c/84909
5856 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5857 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5858 Likewise.
5859 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5860 Likewise.
5862 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5864 PR target/84876
5865 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5866 regno_allocno_class_array and sorted_pseudos.
5867 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5868 insns where regno is used.
5870 2018-03-16 Martin Liska <mliska@suse.cz>
5872 PR ipa/84833
5873 * multiple_target.c (create_dispatcher_calls): Redirect
5874 reference in the symbol table.
5876 2018-03-16 Martin Liska <mliska@suse.cz>
5878 PR ipa/84722
5879 * multiple_target.c (create_dispatcher_calls): Redirect also
5880 an alias.
5882 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5884 PR c++/79937
5885 PR c++/82410
5886 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5887 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5888 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5890 2018-03-16 Julia Koval <julia.koval@intel.com>
5892 * doc/invoke.texi (Skylake Server): Add CLWB.
5893 Cannonlake): Remove CLWB.
5895 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5897 PR tree-optimization/84841
5898 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5899 1 << 3.
5900 (FLOAT_ONE_CONST_TYPE): Define.
5901 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5902 (sort_by_operand_rank): Put entries with higher constant_type last
5903 rather than first to match comments.
5905 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5907 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5908 split predicate.
5910 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5912 PR c++/79085
5913 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5914 check and use address of target always.
5916 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5918 PR target/84574
5919 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5920 (indirect_thunk_bnd_needed): Likewise.
5921 (indirect_thunks_used): Likewise.
5922 (indirect_thunks_bnd_used): Likewise.
5923 (indirect_return_needed): New.
5924 (indirect_return_bnd_needed): Likewise.
5925 (output_indirect_thunk_function): Add a bool argument for
5926 function return.
5927 (output_indirect_thunk_function): Don't generate alias for
5928 function return thunk.
5929 (ix86_code_end): Call output_indirect_thunk_function to generate
5930 function return thunks.
5931 (ix86_output_function_return): Set indirect_return_bnd_needed
5932 and indirect_return_needed instead of indirect_thunk_bnd_needed
5933 and indirect_thunk_needed.
5935 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5937 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5938 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5939 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5941 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5942 Paul Hua <paul.hua.gm@gmail.com>
5944 PR c/84852
5945 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5947 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5949 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5950 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5951 resp. SFmode cases.
5953 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5955 PR target/84711
5956 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5957 instead of GET_MODE_SIZE when comparing Units.
5959 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5961 PR target/68256
5962 * varasm.c (hash_section): Return an unchangeble hash value
5963 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5964 Return !aarch64_can_use_per_function_literal_pools_p ().
5966 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5968 PR target/84860
5969 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5970 rather than address of cmode as last argument to prepare_cmp_insn.
5972 2018-03-15 Julia Koval <julia.koval@intel.com>
5974 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5975 F_AVX512VNNI, F_AVX512BITALG): New.
5977 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5979 PR target/83451
5980 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5981 insn for floating-point loads and stores.
5983 2018-03-14 Carl Love <cel@us.ibm.com>
5985 * config/rs6000/rs6000-c.c: Add macro definitions for
5986 ALTIVEC_BUILTIN_VEC_PERMXOR.
5987 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5988 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5989 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5990 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5991 UNSPEC_VPERMXOR.
5992 * config/doc/extend.texi: Add prototypes for vec_permxor.
5994 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5996 PR c/84852
5997 * diagnostic-show-locus.c (class layout_point): Convert m_line
5998 from int to linenum_type.
5999 (line_span::comparator): Use linenum "compare" function when
6000 comparing line numbers.
6001 (test_line_span): New function.
6002 (layout_range::contains_point): Convert param "row" from int to
6003 linenum_type.
6004 (layout_range::intersects_line_p): Likewise.
6005 (layout::will_show_line_p): Likewise.
6006 (layout::print_source_line): Likewise.
6007 (layout::should_print_annotation_line_p): Likewise.
6008 (layout::print_annotation_line): Likewise.
6009 (layout::print_leading_fixits): Likewise.
6010 (layout::annotation_line_showed_range_p): Likewise.
6011 (struct line_corrections): Likewise for field m_row.
6012 (line_corrections::line_corrections): Likewise for param "row".
6013 (layout::print_trailing_fixits): Likewise.
6014 (layout::get_state_at_point): Likewise.
6015 (layout::get_x_bound_for_row): Likewise.
6016 (layout::print_line): Likewise.
6017 (diagnostic_show_locus): Likewise for locals "last_line" and
6018 "row".
6019 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6020 * input.c (selftest::test_linenum_comparisons): New function.
6021 (selftest::input_c_tests): Call it.
6022 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6023 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6024 * selftest.h (ASSERT_GT): New macro.
6025 (ASSERT_GT_AT): New macro.
6026 (ASSERT_LT): New macro.
6027 (ASSERT_LT_AT): New macro.
6029 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6031 PR rtl-optimization/84780
6032 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6034 2018-03-14 Martin Liska <mliska@suse.cz>
6036 * tree.c (record_node_allocation_statistics): Use
6037 get_stats_node_kind.
6038 (get_stats_node_kind): New function extracted from
6039 record_node_allocation_statistics.
6040 (free_node): Use get_stats_node_kind.
6042 2018-03-14 Richard Biener <rguenther@suse.de>
6044 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6045 that the value-set of ANTIC_IN doesn't grow.
6047 Revert
6048 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6049 member.
6050 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6051 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6053 2018-03-14 Julia Koval <julia.koval@intel.com>
6055 * config.gcc (icelake-client, icelake-server): New.
6056 (icelake): Remove.
6057 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6058 (initial_ix86_arch_features): Ditto.
6059 (PTA_SKYLAKE): Add SGX.
6060 (PTA_ICELAKE): Remove.
6061 (PTA_ICELAKE_CLIENT): New.
6062 (PTA_ICELAKE_SERVER): New.
6063 (ix86_option_override_internal): Split up icelake on icelake client and
6064 icelake server.
6065 (get_builtin_code_for_version): Ditto.
6066 (fold_builtin_cpu): Ditto.
6067 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6068 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6069 * config/i386/i386.h (processor_type): Ditto.
6070 * doc/invoke.texi: Ditto.
6072 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6074 PR sanitizer/83392
6075 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6076 INTEGER_CST offset, add it together with bitpos / 8 and
6077 sign extend based on POINTER_SIZE.
6079 PR target/84844
6080 Revert
6081 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6083 PR target/78090
6084 * config/i386/constraints.md (Yc): New register constraint.
6085 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6086 Use Yc constraint for alternative 2 of operand 0. Remove
6087 preferred_for_speed attribute.
6089 2018-03-14 Richard Biener <rguenther@suse.de>
6091 PR tree-optimization/84830
6092 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6093 with the old one to avoid oscillations.
6095 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6097 PR target/83712
6098 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6099 pseudos.
6100 (assign_by_spills): Return a flag of reload assignment failure.
6101 Do not process the reload assignment failures. Do not spill other
6102 reload pseudos if they has the same reg class. Update n if
6103 necessary.
6104 (lra_assign): Add a return arg. Set up from the result of
6105 assign_by_spills call.
6106 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6107 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6108 usage_insns if it is not NULL.
6109 (spill_hard_reg_in_range): New function.
6110 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6111 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6112 function prototypes.
6113 (lra_assign): Change prototype.
6114 * lra.c (lra): Add code to deal with fails by splitting hard reg
6115 live ranges.
6117 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6119 * config/riscv/riscv.opt (mrelax): New option.
6120 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6121 "norelax" when riscv_mrelax is disabled.
6122 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6124 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6126 PR target/84743
6127 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6128 reassociation for int modes.
6130 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6132 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6133 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6134 for big-endian.
6135 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6136 * config/aarch64/aarch64-sve.md
6137 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6138 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6139 (*extend<mode><Vwide>2): Rename to...
6140 (aarch64_sve_extend<mode><Vwide>2): ...this.
6141 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6142 renaming the old pattern to...
6143 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6144 unsigned packs.
6145 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6146 define_expand, renaming the old pattern to...
6147 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6148 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6149 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6150 account when deciding which SVE instruction the optab should use.
6151 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6153 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6155 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6156 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6157 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6158 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6159 (tlsdesc_small_<mode>): Turn a define_expand and use
6160 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6161 (tlsdesc_small_advsimd_<mode>): ...this.
6162 (tlsdesc_small_sve_<mode>): New pattern.
6164 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6166 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6167 (UNSPEC_UMUL_HIGHPART): New constants.
6168 (MUL_HIGHPART): New int iteraor.
6169 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6170 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6171 define_expand.
6172 (*<su>mul<mode>3_highpart): New define_insn.
6174 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6176 PR lto/84805
6177 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6178 incomplete types.
6180 2018-03-13 Martin Liska <mliska@suse.cz>
6182 PR ipa/84658.
6183 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6184 vector.
6185 (sem_item_optimizer::~sem_item_optimizer): Release it.
6186 (sem_item_optimizer::merge_classes): Register variable aliases.
6187 (sem_item_optimizer::fixup_pt_set): New function.
6188 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6189 * ipa-icf.h: Declare new variables and functions.
6191 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6193 PR middle-end/84834
6194 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6195 integer_pow2p@2 and test integer_pow2p in condition.
6196 (A < 0 ? C : 0): Similarly for @1.
6198 PR middle-end/84831
6199 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6200 characters starting at p contain '\0' character, don't look beyond
6201 that.
6203 PR target/84827
6204 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6205 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6207 PR target/84828
6208 * reg-stack.c (change_stack): Change update_end var from int to
6209 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6210 also call set_block_for_insn on the newly added insns and rescan.
6212 PR target/84786
6213 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6214 on the last operand.
6216 PR c++/84704
6217 * tree.c (stabilize_reference_1): Return save_expr (e) for
6218 STATEMENT_LIST even if it doesn't have side-effects.
6220 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6222 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6224 2018-03-12 Renlin Li <renlin.li@arm.com>
6226 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6227 aarch64_output_scalar_simd_mov_immediate.
6229 2018-03-12 Martin Sebor <msebor@redhat.com>
6231 PR tree-optimization/83456
6232 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6233 for perfectly overlapping calls to memcpy.
6234 (gimple_fold_builtin_memory_chk): Same.
6235 (gimple_fold_builtin_strcpy): Handle no-warning.
6236 (gimple_fold_builtin_stxcpy_chk): Same.
6237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6239 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6241 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6242 parameter. Use it for SFmode.
6243 (rs6000_function_arg_advance_1): Adjust.
6244 (rs6000_function_arg): Adjust.
6245 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6247 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6249 PR rtl-optimization/84169
6250 PR rtl-optimization/84780
6251 * combine.c (can_combine_p): Check for a 2-insn combination whether
6252 the destination register is used between the two insns, too.
6254 2018-03-12 Richard Biener <rguenther@suse.de>
6256 PR tree-optimization/84803
6257 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6258 for refs DR analysis didn't process.
6260 2018-03-12 Richard Biener <rguenther@suse.de>
6262 PR tree-optimization/84777
6263 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6264 force-vectorize loops ignore whether we are optimizing for size.
6266 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6268 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6269 (TARGET_MD_ASM_ADJUST): Define.
6271 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6272 Kito Cheng <kito.cheng@gmail.com>
6273 Chung-Ju Wu <jasonwucj@gmail.com>
6275 * config/nds32/nds32.c (nds32_compute_stack_frame,
6276 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6277 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6278 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6279 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6280 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6281 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6282 * config/nds32/nds32.md (prologue, epilogue): Use macro
6283 NDS32_V3PUSH_AVAILABLE_P to do checking.
6285 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6287 PR debug/58150
6288 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6289 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6290 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6291 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6292 addition of most attributes on !orig_type_die or the attribute not
6293 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6295 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6296 Chung-Ju Wu <jasonwucj@gmail.com>
6298 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6299 __NDS32_VH__ macro.
6300 * config/nds32/nds32.opt (mvh): New option.
6302 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6303 Chung-Ju Wu <jasonwucj@gmail.com>
6305 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6306 function.
6307 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6308 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6309 definition.
6311 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6312 Chung-Ju Wu <jasonwucj@gmail.com>
6314 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6315 function.
6316 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6317 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6319 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6320 Kito Cheng <kito.cheng@gmail.com>
6321 Chung-Ju Wu <jasonwucj@gmail.com>
6323 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6324 UNSPEC_FFMISM and UNSPEC_FLMISM.
6325 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6326 for ffb, ffmism and flmism.
6327 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6328 (unspec_ffmism): Ditto.
6329 (unspec_flmism): Ditto.
6330 (nds32_expand_builtin_impl): Check if string extension is available.
6331 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6332 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6334 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6336 Reverting patch:
6337 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6339 PR target/83712
6340 * lra-assigns.c (assign_by_spills): Return a flag of reload
6341 assignment failure. Do not process the reload assignment
6342 failures. Do not spill other reload pseudos if they has the same
6343 reg class.
6344 (lra_assign): Add a return arg. Set up from the result of
6345 assign_by_spills call.
6346 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6347 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6348 usage_insns if it is not NULL.
6349 (spill_hard_reg_in_range): New function.
6350 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6351 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6352 function prototypes.
6353 (lra_assign): Change prototype.
6354 * lra.c (lra): Add code to deal with fails by splitting hard reg
6355 live ranges.
6357 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6359 PR target/84807
6360 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6362 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6364 PR debug/84620
6365 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6366 (dw_val_node): Add val_symbolic_view.
6367 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6368 (symview_upper_bound): New.
6369 (new_line_info_table): Initialize symviews_since_reset.
6370 (dwarf2out_source_line): Count symviews_since_reset and set
6371 symview_upper_bound.
6372 (dw_val_equal_p): Handle symview.
6373 (add_AT_symview): New.
6374 (print_dw_val): Handle symview.
6375 (attr_checksum, attr_checksum_ordered): Likewise.
6376 (same_dw_val_p, size_of_die): Likewise.
6377 (value_format, output_die): Likewise.
6378 (add_high_low_attributes): Use add_AT_symview for entry_view.
6379 (dwarf2out_finish): Reset symview_upper_bound, clear
6380 zero_view_p.
6382 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6384 PR target/83969
6385 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6386 Add strict argument and use it.
6387 (rs6000_split_multireg_move): Update for new strict argument.
6388 (mem_operand_gpr): Disallow all non-offsettable addresses.
6389 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6391 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6393 PR target/84772
6394 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6395 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6396 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6398 PR c++/84767
6399 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6400 decl, use remap_type if we want to use the type.
6402 2018-03-09 Martin Sebor <msebor@redhat.com>
6404 PR tree-optimization/84526
6405 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6406 Remove dead code.
6407 (builtin_access::generic_overlap): Be prepared to handle non-array
6408 base objects.
6410 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6412 PR rtl-optimization/84682
6413 * lra-constraints.c (process_address_1): Check is_address flag
6414 for address constraints.
6415 (process_alt_operands): Likewise.
6416 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6417 preprocess_constraints.
6418 * recog.h (preprocess_constraints): Add oploc parameter.
6419 Adjust callers.
6420 * recog.c (preprocess_constraints): Test address_operand for
6421 CT_ADDRESS constraints.
6423 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6425 PR target/83712
6426 * lra-assigns.c (assign_by_spills): Return a flag of reload
6427 assignment failure. Do not process the reload assignment
6428 failures. Do not spill other reload pseudos if they has the same
6429 reg class.
6430 (lra_assign): Add a return arg. Set up from the result of
6431 assign_by_spills call.
6432 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6433 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6434 usage_insns if it is not NULL.
6435 (spill_hard_reg_in_range): New function.
6436 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6437 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6438 function prototypes.
6439 (lra_assign): Change prototype.
6440 * lra.c (lra): Add code to deal with fails by splitting hard reg
6441 live ranges.
6443 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6445 PR target/83193
6446 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6447 Accept complain bool parameter. Only emit errors if it is true.
6448 (arm_parse_cpu_option_name): Likewise.
6449 (arm_target_thumb_only): Adjust callers of the above.
6450 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6451 prototype to take a default true bool parameter.
6452 (arm_parse_arch_option_name): Likewise.
6454 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6455 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6457 PR jit/64089
6458 PR jit/84288
6459 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6460 * configure: Regenerate.
6461 * configure.ac ("linker --version-script option"): New.
6462 ("linker soname option"): New.
6464 2018-03-09 Richard Biener <rguenther@suse.de>
6466 PR tree-optimization/84775
6467 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6468 immediate uses of predicate stmts and mark them modified.
6470 Revert
6471 PR tree-optimization/84178
6472 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6473 to caller.
6474 (version_loop_for_if_conversion): Delay update_ssa call.
6475 (tree_if_conversion): Delay update_ssa until after predicate
6476 insertion.
6478 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6480 PR target/84763
6481 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6482 when the function accesses prior frames.
6484 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6486 PR debug/84456
6487 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6488 gen_llsym, otherwise call maybe_gen_llsym.
6490 PR inline-asm/84742
6491 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6492 has ',' character inside of it.
6494 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6496 PR target/84748
6497 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6498 as clobbering CC_REGNUM.
6500 2018-03-08 Richard Biener <rguenther@suse.de>
6502 PR middle-end/84552
6503 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6504 (follow_copies_to_constant): Do not follow SSA names registered
6505 for update.
6507 2018-03-08 Richard Biener <rguenther@suse.de>
6509 PR tree-optimization/84178
6510 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6511 to caller.
6512 (version_loop_for_if_conversion): Delay update_ssa call.
6513 (tree_if_conversion): Delay update_ssa until after predicate
6514 insertion.
6516 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6518 PR tree-optimization/84178
6519 * tree-if-conv.c (release_bb_predicate): Remove the
6520 the assertion that the stmts have NULL use_ops.
6521 Discard the statements, asserting that they haven't
6522 yet been added to a BB.
6524 2018-03-08 Richard Biener <rguenther@suse.de>
6526 PR tree-optimization/84746
6527 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6528 (phi_translate): Pass in destination ANTIC_OUT set.
6529 (phi_translate_1): Likewise. For a simplified result lookup
6530 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6531 (phi_translate_set): Adjust.
6532 (do_pre_regular_insertion): Likewise.
6533 (do_pre_partial_partial_insertion): Likewise.
6535 2018-03-08 Martin Liska <mliska@suse.cz>
6537 PR gcov-profile/84735
6538 * doc/gcov.texi: Document usage of profile files.
6539 * gcov-io.h: Document changes in the format.
6541 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6543 PR debug/84404
6544 PR debug/84408
6545 * dwarf2out.c (struct dw_line_info_table): Update comments for
6546 view == -1.
6547 (FORCE_RESET_NEXT_VIEW): New.
6548 (FORCE_RESETTING_VIEW_P): New.
6549 (RESETTING_VIEW_P): Check for -1 too.
6550 (ZERO_VIEW_P): Likewise.
6551 (new_line_info_table): Force-reset next view.
6552 (dwarf2out_begin_function): Likewise.
6553 (dwarf2out_source_line): Simplify zero_view_p initialization.
6554 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6555 view directly. Omit view when omitting .loc at line 0.
6557 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6559 PR tree-optimization/84740
6560 * tree-switch-conversion.c (process_switch): Call build_constructors
6561 only if info.phi_count is non-zero.
6563 PR tree-optimization/84739
6564 * tree-tailcall.c (find_tail_calls): Check call arguments against
6565 DECL_ARGUMENTS (current_function_decl) rather than
6566 DECL_ARGUMENTS (func) when checking for tail recursion.
6568 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6570 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6571 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6572 Volker Reichelt's entry and add entries for people that perform
6573 GCC fuzzy testing and report numerous bugs.
6575 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6577 PR target/82411
6578 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6579 readonly data in sdata, if that is disabled.
6580 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6581 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6582 -mreadonly-in-sdata option.
6584 2018-03-07 Martin Sebor <msebor@redhat.com>
6586 PR tree-optimization/84468
6587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6588 basic block when looking for nul assignment.
6590 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6592 PR target/84277
6593 * except.h (output_function_exception_table): Adjust prototype.
6594 * except.c (output_function_exception_table): Remove FNNAME parameter
6595 and add SECTION parameter. Ouput one part of the table at a time.
6596 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6597 the first part of the exception table and emit unwind directives.
6598 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6599 (i386_pe_seh_cold_init): Likewise.
6600 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6601 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6602 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6603 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6604 * config/i386/winnt.c: Include except.h.
6605 (struct seh_frame_state): Add reg_offset, after_prologue and
6606 in_cold_section fields.
6607 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6608 (i386_pe_seh_cold_init): New function.
6609 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6610 to seh->in_cold_section.
6611 (seh_emit_push): Record the offset of the push.
6612 (seh_emit_save): Record the offet of the save.
6613 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6614 Test seh->after_prologue to disregard the epilogue.
6615 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6616 (i386_pe_end_cold_function): New function.
6618 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6620 PR fortran/84565
6621 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6622 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6624 PR c++/84704
6625 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6626 on tmp_var.
6627 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6628 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6630 PR middle-end/84723
6631 * multiple_target.c: Include tree-inline.h and intl.h.
6632 (expand_target_clones): Diagnose and fail if node->definition and
6633 !tree_versionable_function_p (node->decl).
6635 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6637 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6638 sprint_ul.
6639 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6640 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6641 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6643 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6645 PR target/84710
6646 * combine.c (try_combine): Use reg_or_subregno instead of handling
6647 just paradoxical SUBREGs and REGs.
6649 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6651 * config/arc/arc.c (arc_finalize_pic): Remove function.
6652 (arc_must_save_register): We use single base PIC register, remove
6653 checks to save/restore the PIC register.
6654 (arc_expand_prologue): Likewise.
6655 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6656 Remove.
6657 (arc_verify_short): Likewise.
6658 (arc_attr_type): Likewise.
6659 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6660 (walk_stores): Likewise.
6661 (arc_address_cost): Make it static.
6662 (arc_verify_short): Likewise.
6663 (branch_dest): Likewise.
6664 (arc_attr_type): Likewise.
6665 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6666 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6667 (arc_final_prescan_insn): Remove inserting the nops due to
6668 hardware hazards. It is done in reorg step.
6669 (insn_length_variant_t): Remove.
6670 (insn_length_parameters_t): Likewise.
6671 (arc_insn_length_parameters): Likewise.
6672 (arc_get_insn_variants): Likewise.
6673 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6675 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6677 PR inline-asm/84683
6678 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6679 assertion failure.
6681 PR tree-optimization/84687
6682 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6683 on new_node->decl.
6684 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6686 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6688 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6689 Rename to ppc_speculation_barrier.
6690 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6691 __builtin_ppc_speculation_barrier.
6693 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6695 PR target/84700
6696 * combine.c (combine_simplify_rtx): Don't try to simplify if
6697 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6698 are equal to x.
6700 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6702 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6703 to 32 bytes when compiling for POWER9.
6705 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6707 PR target/84564
6708 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6709 regparm >= 3 with no arg reg available also for calls with
6710 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6712 PR target/84524
6713 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6714 orig,vex.
6715 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6717 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6719 PR target/84264
6720 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6722 2018-03-05 Richard Biener <rguenther@suse.de>
6724 PR tree-optimization/84486
6725 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6726 When inserting a __builtin_assume_aligned call set the LHS
6727 SSA name alignment info accordingly.
6729 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6731 PR tree-optimization/84114
6732 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6733 Avoid reassociation of FLOAT_MODE addition.
6735 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6737 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6738 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6739 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6740 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6741 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6742 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6743 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6744 and -mwbnoinvd.
6745 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6746 __builtin_ia32_wbinvd): New builtins.
6747 (SPECIAL_ARGS2): New.
6748 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6749 (SPECIAL_ARGS2): New.
6750 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6751 (ix86_valid_target_attribute_inner_p): Ditto.
6752 (ix86_init_mmx_sse_builtins): Add special_args2.
6753 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6754 TARGET_WBNOINVD_P): New.
6755 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6756 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6757 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6758 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6759 * config/i386/pconfigintrin.h: New file.
6760 * config/i386/wbnoinvdintrin.h: Ditto.
6761 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6762 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6764 2018-03-05 Richard Biener <rguenther@suse.de>
6766 PR tree-optimization/84670
6767 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6768 member.
6769 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6770 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6771 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6772 doesn't grow if all successors (recursively) were visited at least
6773 once.
6775 2018-03-05 Richard Biener <rguenther@suse.de>
6777 PR tree-optimization/84650
6778 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6779 if executed in the loop pipeline.
6781 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6783 * doc/configfiles.texi (Configuration Files): Move info about
6784 conditionalizing $target-protos.h to...
6785 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6786 differs from $target-protos.h.
6788 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6789 Chung-Ju Wu <jasonwucj@gmail.com>
6791 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6792 * config/nds32/nds32-multiple.md (setmemsi): Define.
6793 * config/nds32/nds32-memory-manipulation.c
6794 (nds32_gen_dup_4_byte_to_word_value): New.
6795 (emit_setmem_word_loop): New.
6796 (emit_setmem_byte_loop): New.
6797 (nds32_expand_setmem_loop): New.
6798 (nds32_expand_setmem_loop_v3m): New.
6799 (nds32_expand_setmem_unroll): New.
6800 (nds32_expand_setmem): New.
6802 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6803 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config/nds32/nds32-memory-manipulation.c
6806 (nds32_emit_load_store): New.
6807 (nds32_emit_post_inc_load_store): New.
6808 (nds32_emit_mem_move): New.
6809 (nds32_emit_mem_move_block): New.
6810 (nds32_expand_movmemsi_loop_unknown_size): New.
6811 (nds32_expand_movmemsi_loop_known_size): New.
6812 (nds32_expand_movmemsi_loop): New.
6813 (nds32_expand_movmemsi_unroll): New.
6814 (nds32_expand_movmemqi): Rename ...
6815 (nds32_expand_movmemsi): ... to this.
6816 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6817 (movmemsi): ... to this.
6818 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6819 (nds32_expand_movmemsi): ... to this.
6821 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6822 Monk Chiang <sh.chiang04@gmail.com>
6823 Chung-Ju Wu <jasonwucj@gmail.com>
6825 * config/nds32/nds32-protos.h
6826 (nds32_expand_load_multiple): New arguments.
6827 (nds32_expand_store_multiple): Ditto.
6828 (nds32_valid_multiple_load_store): Rename ...
6829 (nds32_valid_multiple_load_store_p): ... to this.
6830 * config/nds32/nds32-memory-manipulation.c
6831 (nds32_expand_load_multiple): Refine implementation.
6832 (nds32_expand_store_multiple): Ditto.
6833 * config/nds32/nds32-multiple.md
6834 (load_multiple): Update nds32_expand_load_multiple interface.
6835 (store_multiple): Update nds32_expand_store_multiple interface.
6836 * config/nds32/nds32-predicates.c
6837 (nds32_valid_multiple_load_store): Rename ...
6838 (nds32_valid_multiple_load_store_p): ... to this and refine
6839 implementation.
6840 * config/nds32/predicates.md
6841 (nds32_load_multiple_and_update_address_operation): New predicate.
6842 (nds32_store_multiple_and_update_address_operation): New predicate.
6844 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6845 Chung-Ju Wu <jasonwucj@gmail.com>
6847 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6848 (combo): New attribute.
6849 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6851 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6853 * config/nds32/nds32.opt: Change -mcmodel= default value.
6855 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6856 Monk Chiang <sh.chiang04@gmail.com>
6857 Chung-Ju Wu <jasonwucj@gmail.com>
6859 * config/nds32/constants.md (unspec_element): New enum.
6860 * config/nds32/constraints.md (Umw): New constraint.
6861 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6862 * config/nds32/nds32-intrinsic.md: Likewise.
6863 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6864 (nds32_valid_smw_lwm_base_p): New.
6865 (nds32_output_smw_single_word): New.
6866 (nds32_output_lmw_single_word): New.
6867 (nds32_expand_unaligned_load): New.
6868 (nds32_expand_unaligned_store): New.
6869 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6870 (nds32_output_smw_single_word): Declare.
6871 (nds32_output_lmw_single_word): Declare.
6872 (nds32_expand_unaligned_load): Declare.
6873 (nds32_expand_unaligned_store): Declare.
6874 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6875 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6876 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6877 NDS32_BUILTIN_UASTORE_DW.
6878 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6879 predicate.
6881 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6882 Kito Cheng <kito.cheng@gmail.com>
6883 Chung-Ju Wu <jasonwucj@gmail.com>
6885 * config/nds32/nds32-intrinsic.c
6886 (nds32_expand_builtin_null_ftype_reg): Delete.
6887 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6888 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6889 (nds32_read_argument): New.
6890 (nds32_legitimize_target): Ditto.
6891 (nds32_legitimize_argument): Ditto.
6892 (nds32_check_constant_argument): Ditto.
6893 (nds32_expand_unop_builtin): Ditto.
6894 (nds32_expand_unopimm_builtin): Ditto.
6895 (nds32_expand_binop_builtin): Ditto.
6896 (nds32_builtin_decl_impl): Ditto.
6897 (builtin_description): Ditto.
6898 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6899 (nds32_init_builtins_impl): Ditto.
6900 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6901 (nds32_builtin_decl): New.
6902 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6903 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6905 2018-03-02 Jeff Law <law@redhat.com>
6907 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6908 (redundant_insn, fill_simple_delay_slots): Likewise.
6909 (fill_slots_from_thread): Likewise.
6910 * resource.c (mark_referenced_resources): Likewise.
6911 (mark_set_resources, find_dead_or_set_registers): Likewise.
6913 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6915 * substring-locations.h (format_warning_va): Formatting fix for
6916 ATTRIBUTE_GCC_DIAG.
6917 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6918 argument.
6919 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6920 * substring-locations.c: Include intl.h.
6921 (format_warning_va): Turned into small wrapper around
6922 format_warning_n_va, renamed to ...
6923 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6924 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6925 use ngettext.
6926 (format_warning_at_substring_n): New function.
6927 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6928 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6929 format_warning_at_substring with just a shorter name instead of
6930 const function pointer.
6931 (fmtwarn_n): New function.
6932 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6933 appropriate, get rid of all the fmtstr temporaries, move conditionals
6934 with G_() wrapped string literals directly into fmtwarn arguments,
6935 cast dir.len to (int), formatting fixes.
6937 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6939 * doc/invoke.texi: Remove "Cilk Plus" references.
6941 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6942 Richard Biener <rguenther@suse.de>
6944 PR ipa/84628
6945 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6946 for error or warning attributes if CALL_FROM_THUNK_P is set.
6947 Formatting fixes.
6949 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6951 PR target/56540
6952 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6953 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6955 PR target/56540
6956 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6957 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6959 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6960 instead of -1U in last predictors element's probability member.
6962 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6964 PR ipa/83983
6965 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6966 arguments if they are comparable.
6968 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6970 PR tree-optimization/84634
6971 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6972 masks and masked_loop_p with a single loop_masks, making sure it's
6973 null for bb vectorization.
6975 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6977 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6978 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6979 loop->force_vectorize to check whether there is no alias.
6981 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6983 PR target/84614
6984 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6985 prototypes.
6986 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6987 comments.
6988 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6989 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6990 instead of a loop around prev_real_insn.
6991 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6992 prev_real_insn.
6994 PR inline-asm/84625
6995 * config/i386/i386.c (ix86_print_operand): Use conditional
6996 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6997 zero vector.
6999 2018-03-02 Richard Biener <rguenther@suse.de>
7001 PR tree-optimization/84427
7002 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7003 (bitmap_set_subtract_values): Rewrite to handle multiple
7004 exprs per value.
7005 (clean): Likewise.
7006 (prune_clobbered_mems): Likewise.
7007 (phi_translate): Take edge instead of pred/phiblock.
7008 (phi_translate_1): Likewise.
7009 (phi_translate_set): Likewise. Insert all translated
7010 exprs for a value into the set, keeping possibly multiple
7011 expressions per value.
7012 (compute_antic_aux): Adjust for phi_translate changes.
7013 When intersecting union the expressions and prune those
7014 not in the final value set, keeping possibly multiple
7015 expressions per value. Do not use value-insertion
7016 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7017 all expressions. Add verification that the value-sets
7018 only shrink during iteration.
7019 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7020 (do_pre_regular_insertion): Likewise.
7021 (do_pre_partial_partial_insertion): Likewise.
7023 2018-03-02 Richard Biener <rguenther@suse.de>
7025 PR target/82005
7026 * config/darwin.c (saved_debug_info_level): New static global.
7027 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7028 (darwin_asm_lto_end): Restore debug info generation settings.
7030 2018-03-01 Martin Liska <mliska@suse.cz>
7032 PR sanitizer/82484
7033 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7034 volatile arguments.
7036 2018-03-01 Richard Biener <rguenther@suse.de>
7038 PR debug/84645
7039 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7040 type annotation with LTO when debug was disabled at compile-time.
7042 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7044 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7045 XINT with INTVAL.
7046 (mips_final_postscan_insn): Likewise.
7048 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7050 PR rtl-optimization/84528
7051 * alias.c (init_alias_target): Add commentary.
7052 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7053 a unique base value if the frame pointer is not eliminated
7054 to the stack pointer.
7056 2018-03-01 Tom de Vries <tom@codesourcery.com>
7058 PR rtl-optimization/83327
7059 * lra-int.h (hard_regs_spilled_into): Declare.
7060 * lra.c (hard_regs_spilled_into): Define.
7061 (init_reg_info): Init hard_regs_spilled_into.
7062 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7063 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7064 (process_bb_lives): Handle hard_regs_spilled_into.
7065 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7066 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7068 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7070 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7071 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7072 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7073 * config/rs6000/aix72.h: New file.
7075 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7077 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7078 instead of warning_at with conditional singular and plural messages
7079 where possible.
7081 PR target/52991
7082 * stor-layout.c (update_alignment_for_field): For
7083 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7084 && !DECL_PACKED (field), do the alignment update, just use
7085 only desired_align instead of MAX (type_align, desired_align)
7086 as the alignment.
7087 (place_field): Don't do known_align < desired_align handling
7088 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7089 is non-NULL, instead do it after rli->prev_field handling and
7090 only if not within a bitfield word. For DECL_PACKED (field)
7091 use type_align of BITS_PER_UNIT.
7093 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7095 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7096 superfluous parentheses and trailing spaces.
7098 2018-02-28 Richard Biener <rguenther@suse.de>
7100 PR tree-optimization/84584
7101 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7102 SCoPs with fake exit edge.
7104 2018-02-28 Martin Liska <mliska@suse.cz>
7106 PR testsuite/84597
7107 * timevar.c (timer::print): Fix format to properly print 100%
7108 values.
7110 2018-02-28 Richard Biener <rguenther@suse.de>
7112 PR middle-end/84607
7113 * genmatch.c (capture_info::walk_match): Do not mark
7114 captured expressions without operands as expr_p given
7115 they act more like predicates and should be subject to
7116 "lost tail" side-effect preserving.
7118 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7120 PR rtl-optimization/81611
7121 * auto-inc-dec.c (attempt_change): Move dead note from
7122 mem_insn if it's the next use of regno
7123 (find_address): Take address use of reg holding
7124 non-incremented value. Add parm to limit search to the named
7125 reg only.
7126 (merge_in_block): Attempt to use a mem insn that is the next
7127 use of the original regno.
7129 2018-02-27 Martin Sebor <msebor@redhat.com>
7131 PR c++/83871
7132 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7133 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7135 2018-02-27 Martin Sebor <msebor@redhat.com>
7137 PR translation/84207
7138 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7139 n argument to unsigned HOST_WIDE_INT.
7140 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7141 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7142 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7143 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7145 2018-02-27 Richard Biener <rguenther@suse.de>
7147 PR tree-optimization/84512
7148 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7149 Do not use the estimate returned from record_stmt_cost for
7150 the scalar iteration cost but sum properly using add_stmt_cost.
7152 2018-02-27 Richard Biener <rguenther@suse.de>
7154 PR tree-optimization/84466
7155 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7156 Adjust last change to less strictly validate use operands.
7158 2018-02-27 Martin Liska <mliska@suse.cz>
7160 PR gcov-profile/84548
7161 * gcov.c (process_file): Allow partial overlap and consider it
7162 also as group functions.
7163 (output_lines): Properly calculate range of lines for a group.
7165 2018-02-27 Martin Liska <mliska@suse.cz>
7167 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7168 'ggc' suffixes. Change first column width.
7169 (timer::print): Fix formatting of the column.
7171 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7173 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7174 preserve inline entry blocks for the sake of debug inline
7175 entry point markers alone.
7176 (remove_unused_locals): Suggest in comments a better place to
7177 force the preservation of inline entry blocks that are
7178 otherwise unused, but do not preserve them.
7180 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7182 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7184 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7186 PR target/84039
7187 * config/i386/constraints.md (Bs): Replace
7188 ix86_indirect_branch_register with
7189 TARGET_INDIRECT_BRANCH_REGISTER.
7190 (Bw): Likewise.
7191 * config/i386/i386.md (indirect_jump): Likewise.
7192 (tablejump): Likewise.
7193 (*sibcall_memory): Likewise.
7194 (*sibcall_value_memory): Likewise.
7195 Peepholes of indirect call and jump via memory: Likewise.
7196 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7197 (*sibcall_value_GOT_32): Likewise.
7198 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7199 (GOT_memory_operand): Likewise.
7200 (call_insn_operand): Likewise.
7201 (sibcall_insn_operand): Likewise.
7202 (GOT32_symbol_operand): Likewise.
7203 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7205 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7207 PR rtl-optimization/83496
7208 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7209 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7210 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7211 redundant insn, if any.
7212 (relax_delay_slots): Likewise.
7213 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7215 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7217 PR tree-optimization/83965
7218 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7219 that grouped statements are part of a reduction chain. Return
7220 true if the statement is not marked as a reduction itself but
7221 is part of a group.
7222 (vect_recog_dot_prod_pattern): Don't check whether the statement
7223 is part of a group here.
7224 (vect_recog_sad_pattern): Likewise.
7225 (vect_recog_widen_sum_pattern): Likewise.
7227 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7229 PR debug/84545
7230 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7232 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7234 PR target/84530
7235 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7236 the bool argument.
7237 (ix86_output_indirect_function_return): New prototype.
7238 (ix86_split_simple_return_pop_internal): Likewise.
7239 * config/i386/i386.c (indirect_return_via_cx): New.
7240 (indirect_return_via_cx_bnd): Likewise.
7241 (indirect_thunk_name): Handle return va CX_REG.
7242 (output_indirect_thunk_function): Create alias for
7243 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7244 (ix86_output_indirect_jmp): Remove the bool argument.
7245 (ix86_output_indirect_function_return): New function.
7246 (ix86_split_simple_return_pop_internal): Likewise.
7247 * config/i386/i386.md (*indirect_jump): Don't pass false
7248 to ix86_output_indirect_jmp.
7249 (*tablejump_1): Likewise.
7250 (simple_return_pop_internal): Change it to define_insn_and_split.
7251 Call ix86_split_simple_return_pop_internal to split it for
7252 -mfunction-return=.
7253 (simple_return_indirect_internal): Call
7254 ix86_output_indirect_function_return instead of
7255 ix86_output_indirect_jmp.
7257 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7259 PR bootstrap/84405
7260 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7261 memset and value initialization afterwards.
7263 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7265 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7267 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7269 PR target/84521
7270 * common/config/aarch64/aarch64-common.c
7271 (aarch_option_optimization_table[]): Switch
7272 off fomit-frame-pointer
7274 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7275 Chung-Ju Wu <jasonwucj@gmail.com>
7277 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7278 volatile memory.
7279 (store_multiple): Ditto.
7281 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7283 * config.gcc: Add --with-cpu support for nds32 target.
7284 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7285 * config/nds32/nds32.opt: Add -mcpu= option.
7287 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7289 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7290 isel=yes): Warn for these deprecated options.
7292 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7294 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7295 ISA_2_5_MASKS_EMBEDDED.
7297 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7299 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7300 p->max as pointers rather than using iterative_hash_expr.
7302 2018-02-23 Carl Love <cel@us.ibm.com>
7304 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7305 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7306 BU_P8V_OVERLOAD_2.
7307 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7308 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7309 P8V_BUILTIN_VEC_VUNSIGNED2.
7311 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7313 PR target/81572
7314 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7315 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7316 LRA_UNKNOWN_ALT.
7317 * lra-constraints.c (curr_insn_transform): Set up
7318 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7319 LRA_UNKNOWN_ALT.
7320 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7321 * lra-eliminations.c (spill_pseudos): Ditto.
7322 (process_insn_for_elimination): Ditto.
7323 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7324 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7325 LRA_NON_CLOBBERED_ALT.
7327 2018-02-22 Martin Sebor <msebor@redhat.com>
7329 PR tree-optimization/84480
7330 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7331 to maybe_diag_stxncpy_trunc. Call it.
7332 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7333 from gimple_fold_builtin_strcpy. Print inlining stack.
7334 (handle_builtin_stxncpy): Print inlining stack.
7335 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7337 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7339 PR target/84176
7340 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7341 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7342 and -fcheck-pointer-bounds are used together.
7343 (indirect_thunk_prefix): New enum.
7344 (indirect_thunk_need_prefix): New function.
7345 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7346 "_nt" instead of "_bnd" for NOTRACK prefix.
7347 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7348 (output_indirect_thunk_function): Likewise.
7349 (): Likewise.
7350 (ix86_code_end): Update output_indirect_thunk_function calls.
7351 (ix86_output_indirect_branch_via_reg): Replace
7352 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7353 (ix86_output_indirect_branch_via_push): Likewise.
7354 (ix86_output_function_return): Likewise.
7355 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7356 incompatible with -fcf-protection=branch and
7357 -fcheck-pointer-bounds.
7359 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7361 PR target/83335
7362 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7363 Change gcc_assert call to output_operand_lossage.
7365 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7367 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7369 2018-02-22 DJ Delorie <dj@redhat.com>
7370 Sebastian Perta <sebastian.perta@renesas.com>
7371 Oleg Endo <olegendo@gcc.gnu.org>
7373 * config/rx/rx.c (rx_rtx_costs): New function.
7374 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7376 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7378 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7380 2018-02-22 Martin Liska <mliska@suse.cz>
7382 PR driver/83193
7383 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7384 Add "native" as a possible value.
7386 2018-02-22 Martin Liska <mliska@suse.cz>
7388 PR driver/83193
7389 * config/i386/i386.c (ix86_option_override_internal):
7390 Add "native" as a possible value for -march and -mtune.
7392 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7394 PR target/84502
7395 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7396 to all type variants.
7398 PR tree-optimization/84503
7399 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7400 width as info->bitpos + info->bitsize - start.
7401 (merged_store_group::merge_overlapping): Simplify width computation.
7402 (check_no_overlap): New function.
7403 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7404 start + width and last_order of the group, fail if check_no_overlap
7405 fails.
7406 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7407 to group if check_no_overlap fails.
7409 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7411 * config/rs6000/altivec.md: Delete contraint arguments to
7412 define_expand, define_split, and define_peephole2, and in
7413 define_insn_and_split if always unused.
7414 * config/rs6000/darwin.md: Ditto.
7415 * config/rs6000/dfp.md: Ditto.
7416 * config/rs6000/rs6000.md: Ditto.
7417 * config/rs6000/sync.md: Ditto.
7418 * config/rs6000/vector.md: Ditto.
7419 * config/rs6000/vsx.md: Ditto.
7421 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7423 * config/rs6000/altivec.md: Write output control strings as braced
7424 blocks instead of double-quoted strings.
7425 * config/rs6000/darwin.md: Ditto.
7426 * config/rs6000/rs6000.md: Ditto.
7427 * config/rs6000/vector.md: Ditto.
7428 * config/rs6000/vsx.md: Ditto.
7430 2018-02-21 Jason Merrill <jason@redhat.com>
7432 PR c++/84314 - ICE with templates and fastcall attribute.
7433 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7435 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7437 * ipa-cp.c (determine_versionability): Fix comment typos.
7439 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7441 PR c/84229
7442 * ipa-cp.c (determine_versionability): Do not version functions caling
7443 va_arg_pack.
7445 2018-02-21 Martin Liska <mliska@suse.cz>
7447 PR driver/83193
7448 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7449 Add "native" as a possible value.
7450 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7451 the macro when native cpu detection is available.
7453 2018-02-21 Martin Liska <mliska@suse.cz>
7455 PR driver/83193
7456 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7457 Add "native" as a possible value.
7458 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7459 when native cpu detection is available.
7461 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7462 Martin Sebor <msebor@redhat.com>
7464 PR tree-optimization/84478
7465 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7466 false.
7467 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7468 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7469 support which is conservatively correct, for 2 only stay conservative
7470 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7471 argument to the 2 argument get_range_strlen, adjust 6 arg
7472 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7473 false.
7474 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7475 (gimple_fold_builtin_strlen): Pass true as last argument to
7476 get_range_strlen.
7478 2018-02-20 Martin Sebor <msebor@redhat.com>
7480 PR middle-end/84095
7481 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7482 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7483 (builtin_memref::builtin_memref): Factor out parts into
7484 set_base_and_offset and call it.
7486 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7488 PR middle-end/84406
7489 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7490 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7491 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7492 search at the associated MODE_INT.
7494 2018-02-20 Jeff Law <law@redhat.com>
7496 PR middle-end/82123
7497 PR tree-optimization/81592
7498 PR middle-end/79257
7499 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7500 for range data rather than using global data.
7501 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7502 range data rather than using global data.
7503 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7504 pass it to children as needed.
7505 (struct directive::fmtresult): Similarly.
7506 (struct directive::set_width): Similarly.
7507 (struct directive::set_precision): Similarly.
7508 (format_integer, format_directive, parse_directive): Similarly.
7509 (format_none): Accept unnamed vr_values parameter.
7510 (format_percent, format_floating, format_character): Similarly.
7511 (format_string, format_plain): Similarly.
7512 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7513 the EVRP range analyzer for range data rather than using global data.
7514 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7515 gimple-ssa-evrp-analyze.h
7516 (class sprintf_dom_walker): Add after_dom_children member function.
7517 Add evrp_range_analyzer member.
7518 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7519 range analyzer as needed.
7520 (sprintf_dom_walker::after_dom_children): New member function.
7521 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7522 if not optimizing.
7523 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7524 (evrp_range_analyzer::pop_to_marker): Likewise.
7526 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7528 PR tree-optimization/84419
7529 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7530 with the required type if its current type is compatible but
7531 different.
7533 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7535 PR middle-end/82004
7536 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7537 after vectorization.
7539 2018-02-20 Martin Liska <mliska@suse.cz>
7541 PR driver/83193
7542 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7543 possible values if we don't have a hint.
7545 2018-02-20 Martin Liska <mliska@suse.cz>
7547 PR c/84310
7548 PR target/79747
7549 * final.c (shorten_branches): Build align_tab array with one
7550 more element.
7551 * opts.c (finish_options): Add alignment option limit check.
7552 (MAX_CODE_ALIGN): Likewise.
7553 (MAX_CODE_ALIGN_VALUE): Likewise.
7554 * doc/invoke.texi: Document maximum allowed option value for
7555 all -falign-* options.
7557 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7559 PR target/84146
7560 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7561 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7562 * var-tracking.c (emit_note_insn_var_location): Remove all references
7563 to NOTE_INSN_CALL_ARG_LOCATION.
7564 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7565 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7566 Use copy_rtx_if_shared.
7567 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7568 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7569 (dwarf2out_var_location): Remove handling of
7570 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7571 on call_insn.
7572 * final.c (final_scan_insn): Remove all references to
7573 NOTE_INSN_CALL_ARG_LOCATION.
7574 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7575 before dumping final insns.
7576 * except.c (emit_note_eh_region_end): Remove all references to
7577 NOTE_INSN_CALL_ARG_LOCATION.
7578 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7579 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7580 * config/arc/arc.c (hwloop_optimize): Likewise.
7581 * config/arm/arm.c (create_fix_barrier): Likewise.
7582 * config/s390/s390.c (s390_chunkify_start): Likewise.
7583 * config/sh/sh.c (find_barrier): Likewise.
7584 * config/i386/i386.c (rest_of_insert_endbranch,
7585 ix86_seh_fixup_eh_fallthru): Likewise.
7586 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7587 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7588 * config/frv/frv.c (frv_function_prologue): Likewise.
7589 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7590 reg note.
7591 (note_outside_basic_block_p): Remove all references to
7592 NOTE_INSN_CALL_ARG_LOCATION.
7593 * gengtype.c (adjust_field_rtx_def): Likewise.
7594 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7595 Likewise.
7596 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7597 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7599 PR c++/84444
7600 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7601 is ADDR_EXPR.
7603 PR tree-optimization/84452
7604 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7605 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7606 is NULL.
7608 2018-02-19 Martin Liska <mliska@suse.cz>
7610 PR sanitizer/82183
7611 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7613 2018-02-19 Martin Liska <mliska@suse.cz>
7614 Richard Sandiford <richard.sandiford@linaro.org>
7616 PR tree-optimization/82491
7617 * gimple-fold.c (get_base_constructor): Make earlier bail out
7618 to prevent ubsan.
7620 2018-02-19 Carl Love <cel@us.ibm.com>
7622 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7623 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7624 BU_P8V_OVERLOAD_1.
7625 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7626 P8V_BUILTIN_VEC_NEG.
7628 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7630 * config/rl78/rl78.md (movdf): New define expand.
7632 2018-02-19 Martin Liska <mliska@suse.cz>
7634 PR other/80589
7635 * doc/invoke.texi: Fix typo.
7636 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7638 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7640 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7641 handle rs6000_single_float and rs6000_double_float specially for
7642 e500 family CPUs.
7644 2018-02-16 Jeff Law <law@redhat.com>
7646 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7647 (pop_regs): Use it.
7649 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7651 PR ipa/84425
7652 * ipa-inline.c (inline_small_functions): Fix a typo.
7654 2018-02-16 Nathan Sidwell <nathan@acm.org>
7656 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7658 2018-02-16 Carl Love <cel@us.ibm.com>
7660 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7661 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7662 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7663 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7664 expansion to P8V_BUILTIN_VEC_FLOAT2.
7666 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7668 PR rtl-optimization/70023
7669 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7670 src_regno into account.
7672 2018-02-16 Carl Love <cel@us.ibm.com>
7674 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7675 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7676 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7677 * config/rs6000/rs6000.c: Remove case statements for
7678 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7679 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7680 and P9V_BUILTIN_VEC_VINSERT4B.
7681 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7682 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7683 * config/rs6000/vsx.md:
7684 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7685 vec_insert4b.
7687 2018-02-16 Carl Love <cel@us.ibm.com>
7689 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7690 vec_insert4b.
7691 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7692 definitions.
7693 * config/rs6000/rs6000-c.c: Add the definitions for
7694 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7695 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7696 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7697 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7698 definition for insert4b and define insn *insert3b_internal.
7699 * doc/extend.texi: Add documentation for vec_extract4b.
7701 2018-02-16 Nathan Sidwell <nathan@acm.org>
7703 * doc/extend.texi (Backwards Compatibility): Mention friend
7704 injection. Note for-scope is deprecated.
7705 * doc/invoke.texi (-ffriend-injection): Deprecate.
7707 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7709 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7710 that moved to I2, also allow destinations that are a paradoxical
7711 subreg (instead of a normal reg).
7713 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7715 PR target/83831
7716 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7717 to QImode.
7719 2018-02-16 Richard Biener <rguenther@suse.de>
7721 PR tree-optimization/84037
7722 PR tree-optimization/84016
7723 PR target/82862
7724 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7725 Adjust vec_construct for the fact we need additional higher latency
7726 128bit inserts for AVX256 and AVX512 vector builds.
7727 (ix86_add_stmt_cost): Scale vector construction cost for
7728 elementwise loads.
7730 2018-02-16 Richard Biener <rguenther@suse.de>
7732 PR tree-optimization/84417
7733 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7734 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7735 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7737 2018-02-16 Martin Liska <mliska@suse.cz>
7739 PR sanitizer/84307
7740 * internal-fn.def (ASAN_CHECK): Set proper flags.
7741 (ASAN_MARK): Likewise.
7743 2018-02-16 Julia Koval <julia.koval@intel.com>
7745 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7746 from PTA_CANNONLAKE.
7748 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7750 PR target/84272
7751 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7752 Use ++iter rather than iter++ for std::list iterators.
7753 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7754 defer deleting them until all nodes in the forest are processed. Do
7755 free even leaf nodes. Change to_process into auto_vec.
7757 PR bootstrap/84405
7758 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7759 * vec.h (vec_default_construct): Use memset instead of placement new
7760 if BROKEN_VALUE_INITIALIZATION is defined.
7761 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7762 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7763 is defined.
7765 PR rtl-optimization/83723
7766 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7767 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7768 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7769 recursive calls.
7770 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7771 callers.
7772 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7774 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7776 PR rtl-optimization/81443
7777 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7778 from inner REGs to paradoxical SUBREGs.
7780 2018-02-16 Richard Biener <rguenther@suse.de>
7782 PR tree-optimization/84399
7783 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7784 For operands we can analyze at their definition make sure we can
7785 analyze them at each use as well.
7787 2018-02-16 Richard Biener <rguenther@suse.de>
7789 PR tree-optimization/84190
7790 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7791 volatile accesses if the decl isn't volatile.
7793 2018-02-15 Jason Merrill <jason@redhat.com>
7795 PR c++/84314 - ICE with templates and fastcall attribute.
7796 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7797 TYPE_CANONICAL on an existing type.
7799 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7801 PR tree-optimization/84383
7802 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7803 dstoff nor call operand_equal_p if dstbase is NULL.
7805 PR tree-optimization/84334
7806 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7807 also a CONSTANT_CLASS_P, punt.
7809 2018-02-14 Jim Wilson <jimw@sifive.com>
7811 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7812 first SMALL_OPERAND check. New local min_second_step. Move assert
7813 to where locals are set. Add TARGET_RVC support.
7814 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7816 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7818 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7820 2018-02-14 Martin Sebor <msebor@redhat.com>
7822 PR tree-optimization/83698
7823 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7824 arrays constrain the offset range to their bounds.
7825 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7826 (builtin_access::overlap): Avoid setting the size of overlap if it's
7827 already been set.
7828 (maybe_diag_overlap): Also consider arrays when deciding what values
7829 of offsets to include in diagnostics.
7831 2018-02-14 Martin Sebor <msebor@redhat.com>
7833 PR c/84108
7834 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7835 that correspond to the kind of a declaration.
7837 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7839 PR target/83984
7840 * config/pa/pa.md: Load address of PIC label using the linkage table
7841 if the label is nonlocal.
7843 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7845 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7846 warning message if user requests -maltivec=be.
7847 * doc/invoke.texi: Document deprecation of -maltivec=be.
7849 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7851 PR target/84220
7852 * config/rs6000/rs6000-c.c: Update definitions for
7853 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7854 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7856 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7858 PR target/84239
7859 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7860 add _get_ssp intrinsics. Remove argument from
7861 __builtin_ia32_rdssp[d|q].
7862 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7863 * config/i386/i386-builtin.def: Remove argument from
7864 __builtin_ia32_rdssp[d|q].
7865 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7866 ix86_expand_special_args_builtin for _rdssp[d|q].
7867 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7868 Clear register before usage.
7869 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7870 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7872 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7874 PR tree-optimization/84357
7875 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7876 operand 1 of an ARRAY_REF too.
7878 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7880 PR target/83831
7881 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7882 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7883 declarations.
7884 (set_of_reg): New struct.
7885 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7886 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7887 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7888 functions.
7889 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7890 Split into bitclr, bitset, bitinvert patterns if appropriate.
7891 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7892 use rx_fuse_in_memory_bitop.
7893 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7894 to named insn, correct maximum insn length.
7896 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7898 PR target/79242
7899 * machmode.def: Define a complex mode for PARTIAL_INT.
7900 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7901 MODE_PARTIAL_INT.
7902 * doc/rtl.texi: Document CSPImode.
7903 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7904 handling.
7905 (msp430_hard_regno_nregs_with_padding): Likewise.
7907 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7909 PR target/84279
7910 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7912 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7914 PR rtl-optimization/84169
7915 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7916 we generated a parallel as new i3 and we split that to new i2 and i3
7917 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7918 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7919 those to i2, not i1. Partially rewrite this scan code.
7921 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7923 PR c/82210
7924 * stor-layout.c (place_field): For variable length fields, adjust
7925 offset_align afterwards not just based on the field's alignment,
7926 but also on the size.
7928 PR middle-end/84309
7929 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7930 of exps and logs in the use_exp2 case.
7932 2018-02-13 Jeff Law <law@redhat.com>
7934 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7935 entry for "vector".
7937 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7938 ARGS as unused.
7940 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7942 PR debug/84342
7943 PR debug/84319
7944 * common.opt (gas-loc-support, gas-locview-support): New.
7945 (ginline-points, ginternal-reset-location-views): New.
7946 * doc/invoke.texi: Document them. Use @itemx where intended.
7947 (gvariable-location-views): Adjust.
7948 * target.def (reset_location_view): New.
7949 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7950 (TARGET_RESET_LOCATION_VIEW): New.
7951 * doc/tm.texi: Rebuilt.
7952 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7953 (dwarf2out_default_as_locview_support): New.
7954 (output_asm_line_debug_info): Use option variables.
7955 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7956 (output_loc_list): Likewise.
7957 (add_high_low_attributes): Check option variables.
7958 Don't output entry view attribute in strict mode.
7959 (gen_inlined_subroutine_die): Check option variables.
7960 (dwarf2out_inline_entry): Likewise.
7961 (init_sections_and_labels): Likewise.
7962 (dwarf2out_early_finish): Likewise.
7963 (maybe_reset_location_view): New, from...
7964 (dwarf2out_var_location): ... here. Call it.
7965 * debug.h (dwarf2out_default_as_loc_support): Declare.
7966 (dwarf2out_default_as_locview_support): Declare.
7967 * hooks.c (hook_int_rtx_insn_0): New.
7968 * hooks.h (hook_int_rtx_insn_0): Declare.
7969 * toplev.c (process_options): Take -gas-loc-support and
7970 -gas-locview-support from dwarf2out. Enable
7971 -gvariable-location-views by default only with locview
7972 assembler support. Enable -ginternal-reset-location-views by
7973 default only if the target defines the corresponding hook.
7974 Enable -ginline-points by default if location views are
7975 enabled; force it disabled if statement frontiers are
7976 disabled.
7977 * tree-inline.c (expand_call_inline): Check option variables.
7978 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7980 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7982 PR tree-optimization/84321
7983 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7984 handling. Also check whether the anti-range contains any values
7985 that satisfy the mask; switch to a VR_RANGE if not.
7987 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7989 PR sanitizer/84340
7990 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7992 2018-02-13 Martin Jambor <mjambor@suse.cz>
7994 PR c++/83990
7995 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7996 of call statements, also set location of a load to a temporary.
7998 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8000 * config/rl78/rl78.c (add_vector_labels): New function.
8001 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8002 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8003 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8004 which checks that no arguments are passed.
8005 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8006 * doc/extend.texi: Documentation for the new attribute.
8008 2018-02-13 Andreas Schwab <schwab@suse.de>
8010 * config/riscv/linux.h (CPP_SPEC): Define.
8012 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8014 PR target/84335
8015 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8016 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8017 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8018 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8019 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8020 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8021 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8022 temporarily for AES and PCLMUL builtins.
8024 PR tree-optimization/84339
8025 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8026 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8027 Formatting fixes.
8029 PR middle-end/84309
8030 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8031 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8032 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8033 inline function.
8034 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8035 inline function.
8036 * omp-simd-clone.h: New file.
8037 * omp-simd-clone.c: Include omp-simd-clone.h.
8038 (expand_simd_clones): No longer static.
8039 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8040 cgraph.h and omp-simd-clone.h.
8041 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8042 (vect_recog_widen_shift_pattern): Formatting fix.
8043 (vect_pattern_recog_1): Don't check optab for calls.
8045 PR target/84336
8046 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8047 operands[2] into a REG before using gen_lowpart on it.
8049 2018-02-12 Jeff Law <law@redhat.com>
8051 PR target/83760
8052 * config/sh/sh.c (find_barrier): Consider a sibling call
8053 a barrier as well.
8055 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8056 successfully back substituting a reg.
8058 2018-02-12 Richard Biener <rguenther@suse.de>
8060 PR tree-optimization/84037
8061 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8062 parameter, move visited init to caller.
8063 (vect_slp_analyze_operations): Separate cost from validity
8064 check, initialize visited once for all instances.
8065 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8066 for all instances.
8067 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8068 out an assert.
8069 (vect_model_promotion_demotion_cost): Likewise.
8070 (vectorizable_bswap): Guard cost modeling with !slp_node
8071 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8072 SLP stmts.
8073 (vectorizable_call): Likewise.
8074 (vectorizable_conversion): Likewise.
8075 (vectorizable_assignment): Likewise.
8076 (vectorizable_shift): Likewise.
8077 (vectorizable_operation): Likewise.
8078 (vectorizable_store): Likewise.
8079 (vectorizable_load): Likewise.
8080 (vectorizable_condition): Likewise.
8081 (vectorizable_comparison): Likewise.
8083 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8085 PR sanitizer/84307
8086 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8087 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8088 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8090 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8092 PR middle-end/83665
8093 * params.def (inline-min-speedup): Increase from 8 to 15.
8094 (max-inline-insns-auto): Decrease from 40 to 30.
8095 * ipa-split.c (consider_split): Add some buffer for function to
8096 be considered inlining candidate.
8097 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8098 default values.
8100 2018-02-12 Richard Biener <rguenther@suse.de>
8102 PR tree-optimization/84037
8103 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8104 matched stmts if we cannot swap the non-matched ones.
8106 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8108 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8109 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8110 _mm_maskz_scalef_round_ss): New intrinsics.
8111 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8112 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8113 __builtin_ia32_scalefss_round): Remove.
8114 (__builtin_ia32_scalefsd_mask_round,
8115 __builtin_ia32_scalefss_mask_round): New intrinsics.
8116 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8117 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8118 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8119 "<round_constraint>")): Changed to ...
8120 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8121 "<round_scalar_constraint>")): ... this.
8122 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8123 %0, %1, %2<round_op3>}"): Changed to ...
8124 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8125 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8126 %2<round_scalar_mask_op3>}"): ... this.
8127 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8129 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8131 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8132 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8133 (_mm_maskz_sqrt_round_ss): New intrinsics.
8134 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8135 (__builtin_ia32_sqrtsd_mask_round)
8136 (__builtin_ia32_sqrtss_mask_round): New builtins.
8137 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8138 (__builtin_ia32_sqrtss_round): Remove.
8139 (__builtin_ia32_sqrtsd_mask_round)
8140 (__builtin_ia32_sqrtss_mask_round): New builtins.
8141 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8142 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8143 ((match_operand:VF_128 1 "vector_operand"
8144 "xBm,<round_constraint>")): Changed to ...
8145 ((match_operand:VF_128 1 "vector_operand"
8146 "xBm,<round_scalar_constraint>")): ... this.
8147 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8148 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8149 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8150 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8151 %<iptr>1<round_scalar_mask_op3>}): ... this.
8152 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8153 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8155 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8157 PR target/84266
8158 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8159 Cast vec_cmpeq result to correct type.
8160 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8161 Cast vec_cmpgt result to correct type.
8163 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8165 * final.c (final_scan_insn_1): Renamed from...
8166 (final_scan_insn): ... this. New wrapper, to recover
8167 seen from the outermost call in recursive ones.
8168 * config/sparc/sparc.c (output_return): Drop seen from call.
8169 (output_sibcall): Likewise.
8170 * config/visium/visium.c (output_branch): Likewise.
8172 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8174 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8175 function label.
8177 2018-02-10 Alan Modra <amodra@gmail.com>
8179 PR target/84300
8180 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8181 Specify LR as an input.
8183 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8185 PR sanitizer/83987
8186 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8187 remove_member_access_dummy_vars): New functions.
8188 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8189 lower_omp_1, execute_lower_omp): Use them.
8191 PR rtl-optimization/84308
8192 * shrink-wrap.c (spread_components): Release todo vector.
8194 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8196 PR rtl-optimization/57193
8197 * ira-color.c (struct allocno_color_data): Add member
8198 conflict_allocno_hard_prefs.
8199 (update_conflict_allocno_hard_prefs): New.
8200 (bucket_allocno_compare_func): Add a preference based on
8201 conflict_allocno_hard_prefs.
8202 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8203 (color_allocnos): Remove a dead code. Initiate
8204 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8206 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8208 PR target/84226
8209 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8210 constraint from =wa to wa. Avoid a subreg on the output operand,
8211 instead use a pseudo and subreg it in a move.
8212 (p9_xxbrd_<mode>): Changed to ...
8213 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8214 (p9_xxbrd_v2df): New expander.
8215 (p9_xxbrw_<mode>): Changed to ...
8216 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8217 (p9_xxbrw_v4sf): New expander.
8219 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8221 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8223 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8225 PR target/83926
8226 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8227 multiply in 32-bit mode.
8228 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8229 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8230 mode.
8232 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8234 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8235 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8236 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8237 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8239 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8241 PR lto/84213
8242 * dwarf2out.c (is_trivial_indirect_ref): New function.
8243 (dwarf2out_late_global_decl): Do not generate a location
8244 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8245 and that are not defined in the current unit.
8247 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8249 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8250 instead of a libcall for UNORDERED.
8252 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8254 PR target/82641
8255 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8256 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8258 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8260 PR target/PR84295
8261 * config/s390/s390.c (s390_set_current_function): Invoke
8262 s390_indirect_branch_settings also if fndecl didn't change.
8264 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8266 * config/rs6000/rs6000.md (blockage): Set length to zero.
8268 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8270 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8272 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8274 PR sanitizer/84285
8275 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8276 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8277 -static-lib*san.
8279 PR debug/84252
8280 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8281 PARALLEL incoming that failed vt_get_decl_and_offset check.
8283 PR middle-end/84237
8284 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8285 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8286 TREE_READONLY bit.
8287 (get_variable_section): For decls in named .bss* sections pass true as
8288 second argument to bss_initializer_p.
8290 2018-02-09 Marek Polacek <polacek@redhat.com>
8291 Jakub Jelinek <jakub@redhat.com>
8293 PR c++/83659
8294 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8295 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8296 Sync some changes from cxx_fold_indirect_ref.
8298 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8300 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8301 markers.
8302 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8303 (BLOCK_INLINE_ENTRY_LABEL): New.
8304 (dwarf2out_var_location): Disregard inline entry markers.
8305 (inline_entry_data): New struct.
8306 (inline_entry_data_hasher): New hashtable type.
8307 (inline_entry_data_hasher::hash): New.
8308 (inline_entry_data_hasher::equal): New.
8309 (inline_entry_data_table): New variable.
8310 (add_high_low_attributes): Add DW_AT_entry_pc and
8311 DW_AT_GNU_entry_view attributes if a pending entry is found
8312 in inline_entry_data_table. Add old entry_pc attribute only
8313 if debug nonbinding markers are disabled.
8314 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8315 markers are enabled.
8316 (block_within_block_p, dwarf2out_inline_entry): New.
8317 (dwarf2out_finish): Check that no entries remained in
8318 inline_entry_data_table.
8319 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8320 (final_scan_insn, notice_source_line): Likewise.
8321 (rest_of_clean_state): Skip inline entry markers.
8322 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8323 markers.
8324 * gimple.c (gimple_build_debug_inline_entry): New.
8325 * gimple.h (enum gimple_debug_subcode): Add
8326 GIMPLE_DEBUG_INLINE_ENTRY.
8327 (gimple_build_debug_inline_entry): Declare.
8328 (gimple_debug_inline_entry_p): New.
8329 (gimple_debug_nonbind_marker_p): Adjust.
8330 * insn-notes.def (INLINE_ENTRY): New.
8331 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8332 inline entry marker notes.
8333 (print_insn): Likewise.
8334 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8335 (INSN_DEBUG_MARKER_KIND): Likewise.
8336 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8337 * tree-inline.c (expand_call_inline): Build and insert
8338 debug_inline_entry stmt.
8339 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8340 inline entry blocks early, if nonbind markers are enabled.
8341 (dump_scope_block): Dump fragment info.
8342 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8343 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8344 (gimple_build_debug_inline_entry): New.
8345 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8346 Enable/disable inline entry points too.
8347 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8348 (DEBUG_INSN): Describe inline entry markers.
8350 * common.opt (gvariable-location-views): New.
8351 (gvariable-location-views=incompat5): New.
8352 * config.in: Rebuilt.
8353 * configure: Rebuilt.
8354 * configure.ac: Test assembler for view support.
8355 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8356 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8357 * dwarf2out.c (var_loc_view): New typedef.
8358 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8359 (dwarf2out_locviews_in_attribute): New.
8360 (dwarf2out_locviews_in_loclist): New.
8361 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8362 (enum dw_line_info_opcode): Add LI_adv_address.
8363 (struct dw_line_info_table): Add view.
8364 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8365 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8366 (zero_view_p): New variable.
8367 (ZERO_VIEW_P): New macro.
8368 (output_asm_line_debug_info): New.
8369 (struct var_loc_node): Add view.
8370 (add_AT_view_list, AT_loc_list): New.
8371 (add_var_loc_to_decl): Add view param. Test it against last.
8372 (new_loc_list): Add view params. Record them.
8373 (AT_loc_list_ptr): Handle loc and view lists.
8374 (view_list_to_loc_list_val_node): New.
8375 (print_dw_val): Handle dw_val_class_view_list.
8376 (size_of_die): Likewise.
8377 (value_format): Likewise.
8378 (loc_list_has_views): New.
8379 (gen_llsym): Set vl_symbol too.
8380 (maybe_gen_llsym, skip_loc_list_entry): New.
8381 (dwarf2out_maybe_output_loclist_view_pair): New.
8382 (output_loc_list): Output view list or entries too.
8383 (output_view_list_offset): New.
8384 (output_die): Handle dw_val_class_view_list.
8385 (output_dwarf_version): New.
8386 (output_compilation_unit_header): Use it.
8387 (output_skeleton_debug_sections): Likewise.
8388 (output_rnglists, output_line_info): Likewise.
8389 (output_pubnames, output_aranges): Update version comments.
8390 (output_one_line_info_table): Output view numbers in asm comments.
8391 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8392 Call maybe_gen_llsym.
8393 (loc_list_from_tree_1): Adjust.
8394 (add_AT_location_description): Create view list attribute if
8395 needed, check it's absent otherwise.
8396 (convert_cfa_to_fb_loc_list): Adjust.
8397 (maybe_emit_file): Call output_asm_line_debug_info for test.
8398 (dwarf2out_var_location): Reset views as needed. Precompute
8399 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8400 attribute. Set view.
8401 (new_line_info_table): Reset next view.
8402 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8403 (dwarf2out_source_line): Likewise. Output view resets and labels to
8404 the assembler, or select appropriate line info opcodes.
8405 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8406 (optimize_string_length): Catch it. Adjust.
8407 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8408 dw_val_class_view_list, and remove it if no longer needed.
8409 (hash_loc_list): Hash view numbers.
8410 (loc_list_hasher::equal): Compare them.
8411 (optimize_location_lists): Check whether a view list symbol is
8412 needed, and whether the locview attribute is present, and
8413 whether they match. Remove the locview attribute if no longer
8414 needed.
8415 (index_location_lists): Call skip_loc_list_entry for test.
8416 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8417 Use output_dwarf_version.
8418 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8419 (struct dw_val_node): Add val_view_list.
8420 * final.c (SEEN_NEXT_VIEW): New.
8421 (set_next_view_needed): New.
8422 (clear_next_view_needed): New.
8423 (maybe_output_next_view): New.
8424 (final_start_function): Rename to...
8425 (final_start_function_1): ... this. Take pointer to FIRST,
8426 add SEEN parameter. Emit param bindings in the initial view.
8427 (final_start_function): Reintroduce SEEN-less interface.
8428 (final): Rename to...
8429 (final_1): ... this. Take SEEN parameter. Output final pending
8430 next view at the end.
8431 (final): Reintroduce seen-less interface.
8432 (final_scan_insn): Output pending next view before switching
8433 sections or ending a block. Mark the next view as needed when
8434 outputting variable locations. Notify debug backend of section
8435 changes, and of location view changes.
8436 (rest_of_handle_final): Adjust.
8437 * toplev.c (process_options): Autodetect value for debug variable
8438 location views option. Warn on incompat5 without -gdwarf-5.
8439 * doc/invoke.texi (gvariable-location-views): New.
8440 (gvariable-location-views=incompat5): New.
8441 (gno-variable-location-views): New.
8443 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8445 PR tree-optimization/84136
8446 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8447 that the result of find_edge is non-NULL.
8449 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8451 PR target/83008
8452 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8453 storing integer register in SImode. Fix cost of 256 and 512
8454 byte aligned SSE register store.
8456 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8458 * config/i386/i386.c (ix86_multiplication_cost): Fix
8459 multiplication cost for TARGET_AVX512DQ.
8461 2018-02-08 Marek Polacek <polacek@redhat.com>
8463 PR tree-optimization/84238
8464 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8465 get_range_strlen.
8467 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8469 PR tree-optimization/84265
8470 * tree-vect-stmts.c (vectorizable_store): Don't treat
8471 VMAT_CONTIGUOUS accesses as grouped.
8472 (vectorizable_load): Likewise.
8474 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8476 PR tree-optimization/81635
8477 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8478 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8479 (test_round_for_mask): New functions.
8480 (wide_int_cc_tests): Call test_round_for_mask.
8481 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8482 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8483 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8484 range returned by get_range_info.
8486 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8488 PR ipa/81360
8489 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8490 * symtab.c: Include builtins.h
8491 (symtab_node::output_to_lto_symbol_table_p): Move here
8492 from lto-streamer-out.c:output_symbol_p.
8493 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8494 (output_symbol_p): Move all logic to symtab.c
8495 (produce_symtab): Update.
8497 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8499 * config/s390/s390-opts.h (enum indirect_branch): Define.
8500 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8501 (s390_indirect_branch_via_thunk)
8502 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8503 (enum s390_indirect_branch_type): Define.
8504 * config/s390/s390.c (struct s390_frame_layout, struct
8505 machine_function): Remove.
8506 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8507 (indirect_branch_table_label_no, indirect_branch_table_name):
8508 Define variables.
8509 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8510 (enum s390_indirect_branch_option): Define.
8511 (s390_return_addr_from_memory): New function.
8512 (s390_handle_string_attribute): New function.
8513 (s390_attribute_table): Add new attribute handler.
8514 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8515 (s390_indirect_branch_via_thunk): New function.
8516 (s390_indirect_branch_via_inline_thunk): New function.
8517 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8518 sibling call optimization for non z10 compiles.
8519 (s390_emit_call): Force indirect branch target to be a single
8520 register. Add r1 clobber for non-z10 compiles.
8521 (s390_emit_epilogue): Emit return jump via return_use expander.
8522 (s390_reorg): Handle JUMP_INSNs as execute targets.
8523 (s390_option_override_internal): Perform validity checks for the
8524 new command line options.
8525 (s390_indirect_branch_attrvalue): New function.
8526 (s390_indirect_branch_settings): New function.
8527 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8528 (s390_output_indirect_thunk_function): New function.
8529 (s390_code_end): Implement target hook.
8530 (s390_case_values_threshold): Implement target hook.
8531 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8532 macros.
8533 * config/s390/s390.h (struct s390_frame_layout)
8534 (struct machine_function): Move here from s390.c.
8535 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8536 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8537 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8538 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8539 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8540 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8541 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8542 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8543 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8544 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8545 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8546 (mnemonic attribute): Add values which aren't recognized
8547 automatically.
8548 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8549 pattern for branch conversion. Fix mnemonic attribute.
8550 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8551 indirect branch via thunk if requested.
8552 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8553 ("*indirect_jump"): Disable for branch conversion using out of
8554 line thunks.
8555 ("indirect_jump_via_thunk<mode>_z10")
8556 ("indirect_jump_via_thunk<mode>")
8557 ("indirect_jump_via_inlinethunk<mode>_z10")
8558 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8559 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8560 ("casesi_jump_via_inlinethunk<mode>_z10")
8561 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8562 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8563 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8564 ("*indirect2_jump"): Disable for branch conversion.
8565 ("casesi_jump"): Turn into expander and expand patterns for branch
8566 conversion.
8567 ("return_use"): New expander.
8568 ("*return"): Emit return via thunk and rename it to ...
8569 ("*return<mode>"): ... this one.
8570 * config/s390/s390.opt: Add new options and and enum for the
8571 option values.
8573 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8575 * lra-constraints.c (match_reload): Unconditionally use
8576 gen_lowpart_SUBREG, rather than selecting between that
8577 and equivalent gen_rtx_SUBREG code.
8579 2018-02-08 Richard Biener <rguenther@suse.de>
8581 PR tree-optimization/84233
8582 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8583 changed flag instead of boguously re-using phi_inserted.
8585 2018-02-08 Martin Jambor <mjambor@suse.cz>
8587 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8588 static local variables.
8590 2018-02-08 Richard Biener <rguenther@suse.de>
8592 PR tree-optimization/84278
8593 * tree-vect-stmts.c (vectorizable_store): When looking for
8594 smaller vector types to perform grouped strided loads/stores
8595 make sure the mode is supported by the target.
8596 (vectorizable_load): Likewise.
8598 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8600 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8601 Increase LDP/STP opportunities by adding adjacent callee-saves.
8603 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8605 PR rtl-optimization/84068
8606 PR rtl-optimization/83459
8607 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8609 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8611 PR tree-optimization/84224
8612 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8613 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8614 non-zero arguments.
8616 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8618 PR target/84113
8619 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8620 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8621 count, remove one USE.
8623 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8625 * doc/install.texi (Configuration): Document the
8626 --with-long-double-format={ibm,ieee} PowerPC configuration
8627 options.
8629 PR target/84154
8630 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8631 Convert from define_expand to be define_insn_and_split. Rework
8632 float/double/_Float128 conversions to QI/HI/SImode to work with
8633 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8634 conversions to QI/HImode types did a store and then a load to
8635 truncate the value. For conversions to VSX registers, don't split
8636 the insn, instead emit the code directly. Use the code iterator
8637 any_fix to combine signed and unsigned conversions.
8638 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8639 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8640 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8641 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8642 (fix_<mode>di2_hw): Likewise.
8643 (fixuns_<mode>di2_hw): Likewise.
8644 (fix_<mode>si2_hw): Likewise.
8645 (fixuns_<mode>si2_hw): Likewise.
8646 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8647 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8648 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8649 fix<uns>_trunc<SFDF:mode>si2_p8.
8650 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8651 used.
8652 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8653 (fix<uns>_<mode>_mem): Likewise.
8654 (fctiw<u>z_<mode>_mem): Likewise.
8655 (fix<uns>_<mode>_mem): Likewise.
8656 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8657 the register allocator from doing a direct move to the GPRs to do
8658 a store, and instead use the ISA 3.0 store byte/half-word from
8659 vector register instruction. For IEEE 128-bit floating point,
8660 also optimize stores of 32-bit ints.
8661 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8663 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8665 * genextract.c (push_pathstr_operand): New function to support
8666 [a-zA-Z].
8667 (walk_rtx): Call push_pathstr_operand.
8668 (print_path): Support [a-zA-Z].
8670 2018-02-07 Richard Biener <rguenther@suse.de>
8672 PR tree-optimization/84037
8673 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8674 (cse_and_gimplify_to_preheader): Declare.
8675 (vect_get_place_in_interleaving_chain): Likewise.
8676 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8677 ivexpr_map.
8678 (_loop_vec_info::~_loop_vec_info): Delete it.
8679 (cse_and_gimplify_to_preheader): New function.
8680 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8681 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8682 (vectorizable_load): Likewise. For grouped stores always base
8683 the IV on the first element.
8684 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8685 condition before gimplifying.
8687 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8689 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8690 *DIV_EXPR and *MOD_EXPR.
8692 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8694 PR target/84248
8695 * config/i386/i386.c (ix86_option_override_internal): Mask out
8696 the CF_SET bit when checking -fcf-protection.
8698 2018-02-07 Tom de Vries <tom@codesourcery.com>
8700 PR libgomp/84217
8701 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8702 enough.
8704 2018-02-07 Richard Biener <rguenther@suse.de>
8706 PR tree-optimization/84204
8707 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8708 this place.
8710 PR tree-optimization/84205
8711 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8712 special-case isl_ast_op_zdiv_r.
8714 PR tree-optimization/84223
8715 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8716 Only add conditions from within the region.
8717 (gather_bbs::after_dom_children): Adjust.
8719 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8721 PR target/84209
8722 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8723 * config/avr/avr.md: Only post-reload split REG-REG moves if
8724 either register is GENERAL_REG_P.
8726 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8728 PR tree-optimization/84235
8729 * tree-ssa-scopedtables.c
8730 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8731 if the subtraction is performed in floating point type where NaNs are
8732 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8733 build 1. Formatting fix.
8735 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8737 PR target/84146
8738 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8739 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8740 and skip it regardless of bb boundaries. Use CALL_P macro,
8741 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8742 unnecessarily, formatting fix.
8744 2018-02-06 Michael Collison <michael.collison@arm.com>
8746 * config/arm/thumb2.md:
8747 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8748 (*thumb_mov_notscc): Ditto.
8750 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8752 PR target/84154
8753 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8754 unsigned_fix, not "s".
8756 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8758 * configure.ac (gcc_fn_eh_frame_ro): New function.
8759 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8760 correct .eh_frame permissions.
8761 * configure: Regenerate.
8763 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8765 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8766 irrelevant options.
8768 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8770 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8771 Display warning message for -mno-speculate-indirect-jumps.
8773 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8775 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8776 Undocumented.
8777 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8779 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8781 PR tree-optimization/84225
8782 * tree-eh.c (find_trapping_overflow): Only call
8783 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8785 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8787 PR target/84145
8788 * config/i386/i386.c: Reimplement the check of possible options
8789 -mibt/-mshstk conbination. Change error messages.
8790 * doc/invoke.texi: Fix a typo: remove extra '='.
8792 2018-02-06 Marek Polacek <polacek@redhat.com>
8794 PR tree-optimization/84228
8795 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8797 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8799 PR target/82641
8800 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8801 emitted arch directives.
8802 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8803 __ARM_FEATURE_COPROC before changing architectures.
8805 2018-02-06 Richard Biener <rguenther@suse.de>
8807 * config/i386/i386.c (print_reg): Fix typo.
8808 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8810 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8812 * configure: Regenerate.
8814 2018-02-05 Martin Sebor <msebor@redhat.com>
8816 PR tree-optimization/83369
8817 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8818 inlining context.
8820 2018-02-05 Martin Liska <mliska@suse.cz>
8822 * doc/invoke.texi: Cherry-pick upstream r323995.
8824 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8826 * ira.c (ira_init_register_move_cost): Adjust comment.
8828 2018-02-05 Martin Liska <mliska@suse.cz>
8830 PR gcov-profile/84137
8831 * doc/gcov.texi: Fix typo in documentation.
8833 2018-02-05 Martin Liska <mliska@suse.cz>
8835 PR gcov-profile/83879
8836 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8837 using dlopen functionality.
8839 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8841 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8842 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8843 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8844 _mm_maskz_range_round_ss): New intrinsics.
8845 (__builtin_ia32_rangesd128_round)
8846 (__builtin_ia32_rangess128_round): Remove.
8847 (__builtin_ia32_rangesd128_mask_round,
8848 __builtin_ia32_rangess128_mask_round): New builtins.
8849 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8850 __builtin_ia32_rangess128_round): Remove.
8851 (__builtin_ia32_rangesd128_mask_round,
8852 __builtin_ia32_rangess128_mask_round): New builtins.
8853 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8854 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8855 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8856 "<round_saeonly_constraint>")): Changed to ...
8857 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8858 "<round_saeonly_scalar_constraint>")): ... this.
8859 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8860 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8861 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8862 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8863 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8865 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8867 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8868 options.
8869 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8870 Remove all values except native, 8540 and 8548.
8872 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8874 * config/i386/i386.c (ix86_output_function_return): Pass
8875 INVALID_REGNUM, instead of -1, as invalid register number to
8876 indirect_thunk_name and output_indirect_thunk.
8878 2018-02-02 Julia Koval <julia.koval@intel.com>
8880 * config.gcc: Add -march=icelake.
8881 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8882 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8883 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8884 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8885 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8886 (processor_target_table): Add icelake.
8887 (ix86_option_override_internal): Handle new PTAs.
8888 (get_builtin_code_for_version): Handle icelake.
8889 (M_INTEL_COREI7_ICELAKE): New.
8890 (fold_builtin_cpu): Handle icelake.
8891 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8892 * doc/invoke.texi: Add -march=icelake.
8894 2018-02-02 Julia Koval <julia.koval@intel.com>
8896 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8897 to wide_int_bitmask.
8898 * wide-int-bitmask.h: New.
8900 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8902 PR target/84066
8903 * config/i386/i386.md: Replace Pmode with word_mode in
8904 builtin_setjmp_setup and builtin_longjmp to support x32.
8906 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8908 PR target/56010
8909 PR target/83743
8910 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8911 #include "opts.h".
8912 (rs6000_supported_cpu_names): New static variable.
8913 (linux_cpu_translation_table): Likewise.
8914 (elf_platform) <cpu>: Define new static variable and use it.
8915 Translate kernel AT_PLATFORM name to canonical name if needed.
8916 Error if platform name is unknown.
8918 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8920 PR target/84089
8921 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8923 2018-02-01 Jeff Law <law@redhat.com>
8925 PR target/84128
8926 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8927 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8928 the scratch if RELEASE_VIA_POP is false.
8929 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8930 If we have to save a temporary register, decrement SIZE appropriately.
8931 Pass new arguments to release_scratch_register_on_entry.
8932 (ix86_adjust_stack_and_probe): Likewise.
8933 (ix86_emit_probe_stack_range): Pass new arguments to
8934 release_scratch_register_on_entry.
8936 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8938 PR rtl-optimization/84157
8939 * combine.c (change_zero_ext): Use REG_P predicate in
8940 front of HARD_REGISTER_P predicate.
8942 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8944 * config/avr/avr.c (avr_option_override): Move disabling of
8945 -fdelete-null-pointer-checks to...
8946 * common/config/avr/avr-common.c (avr_option_optimization_table):
8947 ...here.
8949 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8951 PR tree-optimization/81635
8952 * tree-data-ref.c (split_constant_offset_1): For types that
8953 wrap on overflow, try to use range info to prove that wrapping
8954 cannot occur.
8956 2018-02-01 Renlin Li <renlin.li@arm.com>
8958 PR target/83370
8959 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8960 TAILCALL_ADDR_REGS.
8961 (aarch64_register_move_cost): Likewise.
8962 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8963 TAILCALL_ADDR_REGS.
8964 (REG_CLASS_NAMES): Likewise.
8965 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8966 TAILCALL_ADDR_REGS. Remove IP registers.
8967 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8969 2018-02-01 Richard Biener <rguenther@suse.de>
8971 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8972 for specifying RPO order and allow NULL for that.
8973 * domwalk.c (dom_walker::dom_walker): Likewise.
8974 (dom_walker::walk): Handle NULL RPO order.
8975 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8976 in RPO order.
8977 (rewrite_update_dom_walker): Likewise.
8978 (mark_def_dom_walker): Likewise.
8980 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8982 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8983 (aarch64_maybe_expand_sve_subreg_move): Declare.
8984 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8985 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8986 predicate.
8987 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8988 that are semantically a reverse operation.
8989 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8990 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8991 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8992 functions.
8993 (aarch64_can_change_mode_class): For big-endian, forbid changes
8994 between two SVE modes if they have different element sizes.
8996 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8998 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8999 the TImode handling for big-endian targets.
9001 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9003 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9004 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9005 not just bytes.
9006 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9007 Remove BSWAP handing for big-endian targets and use the form of
9008 LD1RQ appropariate for the mode.
9010 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9012 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9013 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9014 duplicated element.
9016 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9018 PR tearget/83845
9019 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9020 check for operands that need to go through aarch64_sve_reload_be.
9022 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9024 PR tree-optimization/81661
9025 PR tree-optimization/84117
9026 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9027 * tree-eh.c: Include gimplify.h.
9028 (find_trapping_overflow, replace_trapping_overflow,
9029 rewrite_to_non_trapping_overflow): New functions.
9030 * tree-vect-loop.c: Include tree-eh.h.
9031 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9032 * tree-data-ref.c: Include tree-eh.h.
9033 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9035 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9037 PR rtl-optimization/84123
9038 * combine.c (change_zero_ext): Check if hard register satisfies
9039 can_change_dest_mode before calling gen_lowpart_SUBREG.
9041 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9043 PR target/82444
9044 * ira.c (ira_init_register_move_cost): Remove assert.
9046 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9048 PR rtl-optimization/84071
9049 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9050 * doc/tm.texi: Regenerate.
9052 2018-01-31 Richard Biener <rguenther@suse.de>
9054 PR tree-optimization/84132
9055 * tree-data-ref.c (analyze_miv_subscript): Properly
9056 check whether evolution_function_is_affine_multivariate_p
9057 before calling gcd_of_steps_may_divide_p.
9059 2018-01-31 Julia Koval <julia.koval@intel.com>
9061 PR target/83618
9062 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9063 * config/i386/i386.md (rdpid_rex64) New.
9064 (rdpid): Make 32bit only.
9066 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9068 PR lto/84105
9069 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9070 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9072 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9074 Revert
9075 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9077 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9079 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9081 PR rtl-optimization/84071
9082 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9083 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9085 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9087 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9088 (arc_attribute_table): Add 'aux' attribute.
9089 (arc_in_small_data_p): Consider aux like variables.
9090 (arc_is_aux_reg_p): New function.
9091 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9092 (arc_get_aux_arg): New function.
9093 (prepare_move_operands): Handle aux-register access.
9094 (arc_handle_aux_attribute): New function.
9095 * doc/extend.texi (ARC Variable attributes): Add subsection.
9097 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9099 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9100 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9101 (arc_attribute_table): Add 'uncached' attribute.
9102 (arc_print_operand): Print '.di' flag for uncached memory
9103 accesses.
9104 (arc_in_small_data_p): Do not consider for small data the uncached
9105 types.
9106 (arc_is_uncached_mem_p): New function.
9107 * config/arc/predicates.md (compact_store_memory_operand): Check
9108 for uncached memory accesses.
9109 (nonvol_nonimm_operand): Likewise.
9110 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9112 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9114 PR c/84100
9115 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9116 falign-loops=): Add Optimization flag.
9118 2018-01-30 Jeff Law <law@redhat.com>
9120 PR target/84064
9121 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9122 INT_REGISTERS_SAVED. Check it prior to calling
9123 get_scratch_register_on_entry.
9124 (ix86_adjust_stack_and_probe): Similarly.
9125 (ix86_emit_probe_stack_range): Similarly.
9126 (ix86_expand_prologue): Corresponding changes.
9128 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9130 PR target/40411
9131 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9132 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9134 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9136 PR target/84112
9137 * lra-constraints.c (curr_insn_transform): Process AND in the
9138 address.
9140 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9142 PR rtl-optimization/83986
9143 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9144 dependence against last_pending_memory_flush in addition to
9145 pending_jump_insns.
9147 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9149 PR tree-optimization/81611
9150 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9151 copies.
9153 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9155 PR target/83758
9156 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9157 a reg rtx.
9159 2018-01-30 Richard Biener <rguenther@suse.de>
9160 Jakub Jelinek <jakub@redhat.com>
9162 PR tree-optimization/84111
9163 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9164 inner loops added during recursion, as they don't have up-to-date
9165 SSA form.
9167 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9169 PR ipa/81360
9170 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9171 (can_inline_edge_by_limits_p): ... here.
9172 (can_early_inline_edge_p, check_callers,
9173 update_caller_keys, update_callee_keys, recursive_inlining,
9174 add_new_edges_to_heap, speculation_useful_p,
9175 inline_small_functions,
9176 inline_small_functions, flatten_function,
9177 inline_to_all_callers_1): Update.
9179 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9181 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9182 zeros correctly.
9184 2018-01-30 Richard Biener <rguenther@suse.de>
9186 PR tree-optimization/83008
9187 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9188 invariant and constant vector uses in stmts when they need
9189 more than one stmt.
9191 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9193 PR bootstrap/84017
9194 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9195 * configure: Regenerate.
9197 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9199 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9200 pattern.
9201 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9202 Use gen_rtx_REG rather than gen_lowpart.
9204 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9206 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9207 rather than 0 when creating partial subregs.
9209 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9211 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9212 of usage.
9214 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9216 PR target/81550
9217 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9218 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9219 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9220 flags. This restores the settings used before the 2017-07-24.
9221 Turning off pre increment/decrement/modify allows IVOPTS to
9222 optimize DF/SF loops where the index is an int.
9224 2018-01-29 Richard Biener <rguenther@suse.de>
9225 Kelvin Nilsen <kelvin@gcc.gnu.org>
9227 PR bootstrap/80867
9228 * tree-vect-stmts.c (vectorizable_call): Don't call
9229 targetm.vectorize_builtin_md_vectorized_function if callee is
9230 NULL.
9232 2018-01-22 Carl Love <cel@us.ibm.com>
9234 * doc/extend.tex: Fix typo in second arg in
9235 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9237 2018-01-29 Richard Biener <rguenther@suse.de>
9239 PR tree-optimization/84086
9240 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9241 (flush_ssaname_freelist): When SSA names were released reset
9242 the SCEV hash table.
9244 2018-01-29 Richard Biener <rguenther@suse.de>
9246 PR tree-optimization/84057
9247 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9248 removed paths when removing edges.
9250 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9252 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9253 -mfunction-return=@var{choice}.
9255 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9257 PR diagnostic/84034
9258 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9259 Handle CR like TAB.
9260 (layout::print_source_line): Likewise.
9261 (test_get_line_width_without_trailing_whitespace): Add test cases.
9263 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9265 PR middle-end/84040
9266 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9267 debug insns.
9269 2018-01-26 Jim Wilson <jimw@sifive.com>
9271 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9273 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9274 specified.
9276 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9278 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9279 and CMP + SUB-immediate -> SUBS.
9281 2018-01-26 Martin Sebor <msebor@redhat.com>
9283 PR tree-optimization/83896
9284 * tree-ssa-strlen.c (get_string_len): Rename...
9285 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9286 Avoid assuming length is constant.
9287 (handle_char_store): Use HOST_WIDE_INT for string length.
9289 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9291 PR target/81763
9292 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9293 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9295 2018-01-26 Richard Biener <rguenther@suse.de>
9297 PR rtl-optimization/84003
9298 * dse.c (record_store): Only record redundant stores when
9299 the earlier store aliases at least all accesses the later one does.
9301 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9303 PR rtl-optimization/83985
9304 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9305 REG_CFA_RESTORE insns.
9306 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9307 REG_CFA_RESTORE insns here.
9309 PR c/83989
9310 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9311 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9313 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9315 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9316 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9317 (arc_init): Likewise.
9318 (arc_override_options): Likewise.
9319 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9320 value.
9321 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9322 support.
9323 * config/arc/arc.h (TARGET_DBNZ): Define.
9324 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9325 properly set the tune attribute.
9326 (dbnz): Use TARGET_DBNZ guard.
9327 * config/arc/arc.opt (mtune): Add core3 option.
9329 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9331 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9332 recognize new pic like addresses.
9333 (arc_delegitimize_address): Clean up.
9335 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9337 * config/arc/arc-arches.def: Option mrf16 valid for all
9338 architectures.
9339 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9340 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9341 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9342 * config/arc/arc-tables.opt: Regenerate.
9343 * config/arc/arc.c (arc_conditional_register_usage): Handle
9344 reduced register file case.
9345 (arc_file_start): Set must have build attributes.
9346 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9347 mrf16 option value.
9348 * config/arc/arc.opt (mrf16): Add new option.
9349 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9350 * config/arc/genmultilib.awk: Handle new mrf16 option.
9351 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9352 * config/arc/t-multilib: Regenerate.
9353 * doc/invoke.texi (ARC Options): Document mrf16 option.
9355 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9357 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9358 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9359 (arc_attribute_table): Add 'secure_call' attribute.
9360 (arc_print_operand): Print secure call operand.
9361 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9362 secure.
9363 (arc_is_secure_call_p): New function. * config/arc/arc.md
9364 (call_i): Add support for sjli instruction.
9365 (call_value_i): Likewise.
9366 * config/arc/constraints.md (Csc): New constraint.
9368 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9369 John Eric Martin <John.Martin@emmicro-us.com>
9371 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9372 * config/arc/arc.c (_arc_jli_section): New struct.
9373 (arc_jli_section): New type.
9374 (rc_jli_sections): New static variable.
9375 (arc_handle_jli_attribute): New function.
9376 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9377 (arc_file_end): New function.
9378 (TARGET_ASM_FILE_END): Define.
9379 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9380 (arc_add_jli_section): New function.
9381 (jli_call_scan): Likewise.
9382 (arc_reorg): Call jli_call_scan.
9383 (arc_output_addsi): Remove 'S' from printing asm operand.
9384 (arc_is_jli_call_p): New function.
9385 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9386 operand.
9387 (movhi_insn): Likewise.
9388 (movsi_insn): Likewise.
9389 (movsi_set_cc_insn): Likewise.
9390 (loadqi_update): Likewise.
9391 (load_zeroextendqisi_update): Likewise.
9392 (load_signextendqisi_update): Likewise.
9393 (loadhi_update): Likewise.
9394 (load_zeroextendhisi_update): Likewise.
9395 (load_signextendhisi_update): Likewise.
9396 (loadsi_update): Likewise.
9397 (loadsf_update): Likewise.
9398 (movsicc_insn): Likewise.
9399 (bset_insn): Likewise.
9400 (bxor_insn): Likewise.
9401 (bclr_insn): Likewise.
9402 (bmsk_insn): Likewise.
9403 (bicsi3_insn): Likewise.
9404 (cmpsi_cc_c_insn): Likewise.
9405 (movsi_ne): Likewise.
9406 (movsi_cond_exec): Likewise.
9407 (clrsbsi2): Likewise.
9408 (norm_f): Likewise.
9409 (normw): Likewise.
9410 (swap): Likewise.
9411 (divaw): Likewise.
9412 (flag): Likewise.
9413 (sr): Likewise.
9414 (kflag): Likewise.
9415 (ffs): Likewise.
9416 (ffs_f): Likewise.
9417 (fls): Likewise.
9418 (call_i): Remove 'S' asm letter, add jli instruction.
9419 (call_value_i): Likewise.
9420 * config/arc/arc.op (mjli-always): New option.
9421 * config/arc/constraints.md (Cji): New constraint.
9422 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9423 operand.
9424 (subsf3_fpx): Likewise.
9425 (mulsf3_fpx): Likewise.
9426 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9427 asm operand.
9428 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9429 function attrbutes.
9430 * doc/invoke.texi (ARC): Document mjli-always option.
9432 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9434 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9435 avoid addition with 0 and use incw and decw where possible.
9437 2018-01-26 Richard Biener <rguenther@suse.de>
9439 PR tree-optimization/81082
9440 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9441 association if it requires casting to unsigned.
9442 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9443 from fold_plusminus_mult_expr to catch important cases late when
9444 range info is available.
9446 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9448 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9449 * configure.ac (hidden_linkonce): New test.
9450 * configure: Regenerate.
9451 * config.in: Regenerate.
9453 2018-01-26 Julia Koval <julia.koval@intel.com>
9455 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9456 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9457 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9458 _mm_mask_bitshuffle_epi64_mask): Fix type.
9459 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9460 USI_FTYPE_V4DI_V4DI_USI): Remove.
9461 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9462 __builtin_ia32_vpshufbitqmb256_mask,
9463 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9464 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9465 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9467 2018-01-26 Alan Modra <amodra@gmail.com>
9469 PR target/84033
9470 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9471 UNSPEC_VBPERMQ. Sort other unspecs.
9473 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9475 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9477 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9479 PR middle-end/83055
9480 * predict.c (drop_profile): Do not push/pop cfun; update also
9481 node->count.
9482 (handle_missing_profiles): Fix logic looking for zero profiles.
9484 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9486 PR middle-end/83977
9487 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9488 on functions with #pragma omp declare simd or functions with simd
9489 attribute.
9490 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9491 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9492 Remove trailing \n from warning_at calls.
9494 2018-01-25 Tom de Vries <tom@codesourcery.com>
9496 PR target/84028
9497 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9498 for neutered workers.
9500 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9502 PR target/68467
9503 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9504 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9506 2018-01-24 Jeff Law <law@redhat.com>
9508 PR target/83994
9509 * i386.c (get_probe_interval): Move to earlier point.
9510 (ix86_compute_frame_layout): If -fstack-clash-protection and
9511 the frame is larger than the probe interval, then use pushes
9512 to save registers rather than reg->mem moves.
9513 (ix86_expand_prologue): Remove conditional for int_registers_saved
9514 assertion.
9516 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9518 PR target/84014
9519 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9520 min/max for never referenced object.
9522 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9524 PR middle-end/83977
9525 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9526 here.
9527 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9528 attributes from DECL_ATTRIBUTES (decl) without affecting
9529 DECL_ATTRIBUTES (current_function_decl).
9530 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9531 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9533 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9535 PR tree-optimization/83979
9536 * fold-const.c (fold_comparison): Use constant_boolean_node
9537 instead of boolean_{true,false}_node.
9539 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9541 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9542 with zero counts.
9544 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9546 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9547 Simplify the clause that sets the length attribute.
9548 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9549 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9550 clause that sets the length attribute.
9551 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9553 2018-01-24 Tom de Vries <tom@codesourcery.com>
9555 PR target/83589
9556 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9557 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9558 Add strict parameter.
9559 (prevent_branch_around_nothing): Insert dummy insn between branch to
9560 label and label with no ptx insn inbetween.
9561 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9563 2018-01-24 Tom de Vries <tom@codesourcery.com>
9565 PR target/81352
9566 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9567 for neutered threads in warp.
9568 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9570 2018-01-24 Richard Biener <rguenther@suse.de>
9572 PR tree-optimization/83176
9573 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9574 operands.
9576 2018-01-24 Richard Biener <rguenther@suse.de>
9578 PR tree-optimization/82819
9579 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9580 code generating pluses that are no-ops in the target precision.
9582 2018-01-24 Richard Biener <rguenther@suse.de>
9584 PR middle-end/84000
9585 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9587 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9589 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9590 to merge probabilities.
9591 * predict.c (probably_never_executed): Also mark as cold functions
9592 with global 0 profile and guessed local profile.
9593 * profile-count.c (profile_probability::combine_with_count): New
9594 member function.
9595 * profile-count.h (profile_probability::operator*,
9596 profile_probability::operator*=, profile_probability::operator/,
9597 profile_probability::operator/=): Reduce precision to adjusted
9598 and set value to guessed on contradictory divisions.
9599 (profile_probability::combine_with_freq): Remove.
9600 (profile_probability::combine_wiht_count): Declare.
9601 (profile_count::force_nonzero):: Set to adjusted.
9602 (profile_count::probability_in):: Set quality to adjusted.
9603 * tree-ssa-tail-merge.c (replace_block_by): Use
9604 combine_with_count.
9606 2018-01-23 Andrew Waterman <andrew@sifive.com>
9607 Jim Wilson <jimw@sifive.com>
9609 * config/riscv/riscv.c (riscv_stack_boundary): New.
9610 (riscv_option_override): Set riscv_stack_boundary. Handle
9611 riscv_preferred_stack_boundary_arg.
9612 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9613 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9614 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9615 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9616 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9617 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9619 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9621 PR target/83905
9622 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9623 of struct ix86_frame.
9624 (ix86_expand_epilogue): Likewise. Add a local variable for
9625 the reg_save_offset field in struct ix86_frame.
9627 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9629 PR tree-optimization/82604
9630 * tree-loop-distribution.c (enum partition_kind): New enum item
9631 PKIND_PARTIAL_MEMSET.
9632 (partition_builtin_p): Support above new enum item.
9633 (generate_code_for_partition): Ditto.
9634 (compute_access_range): Differentiate cases that equality can be
9635 proven at all loops, the innermost loops or no loops.
9636 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9637 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9638 (finalize_partitions, distribute_loop): Don't fuse partition of
9639 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9640 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9641 parloop is enabled.
9643 2018-01-23 Martin Liska <mliska@suse.cz>
9645 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9646 order to ignore the predictor.
9647 (PRED_POLYMORPHIC_CALL): Likewise.
9648 (PRED_RECURSIVE_CALL): Likewise.
9650 2018-01-23 Martin Liska <mliska@suse.cz>
9652 * tree-profile.c (tree_profiling): Print function header to
9653 aware reader which function we are working on.
9654 * value-prof.c (gimple_find_values_to_profile): Do not print
9655 not interesting value histograms.
9657 2018-01-23 Martin Liska <mliska@suse.cz>
9659 * profile-count.h (enum profile_quality): Add
9660 profile_uninitialized as the first value. Do not number values
9661 as they are zero based.
9662 (profile_count::verify): Update sanity check.
9663 (profile_probability::verify): Likewise.
9665 2018-01-23 Nathan Sidwell <nathan@acm.org>
9667 * doc/invoke.texi (ffor-scope): Deprecate.
9669 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9671 PR tree-optimization/83510
9672 * domwalk.c (set_all_edges_as_executable): New function.
9673 (dom_walker::dom_walker): Convert bool param
9674 "skip_unreachable_blocks" to enum reachability. Move setup of
9675 edge flags to set_all_edges_as_executable and only do it when
9676 reachability is REACHABLE_BLOCKS.
9677 * domwalk.h (enum dom_walker::reachability): New enum.
9678 (dom_walker::dom_walker): Convert bool param
9679 "skip_unreachable_blocks" to enum reachability.
9680 (set_all_edges_as_executable): New decl.
9681 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9682 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9683 "reachability".
9684 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9685 but converting true to REACHABLE_BLOCKS.
9686 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9687 * tree-vrp.c
9688 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9689 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9690 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9691 REACHABLE_BLOCKS.
9692 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9693 if check_all_array_refs will be called.
9695 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9697 * tree.c (selftest::test_location_wrappers): Add more test
9698 coverage.
9700 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9702 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9703 (selftest::test_bit_in_range): Likewise.
9705 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9707 PR testsuite/83888
9708 * doc/sourcebuild.texi (vect_float): Say that the selector
9709 only describes the situation when -funsafe-math-optimizations is on.
9710 (vect_float_strict): Document.
9712 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9714 PR tree-optimization/83965
9715 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9716 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9717 instead of checking only for a reduction.
9718 (vect_recog_widen_sum_pattern): Likewise.
9720 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9722 * predict.c (probably_never_executed): Only use precise profile info.
9723 (compute_function_frequency): Skip after inlining hack since we now
9724 have quality checking.
9726 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9728 * profile-count.h (profile_probability::very_unlikely,
9729 profile_probability::unlikely, profile_probability::even): Set
9730 precision to guessed.
9732 2018-01-23 Richard Biener <rguenther@suse.de>
9734 PR tree-optimization/83963
9735 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9736 Properly terminate dominator walk when crossing the exit edge not
9737 when visiting its source block.
9739 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9741 PR c++/83918
9742 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9743 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9745 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9747 PR tree-optimization/83957
9748 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9749 semicolon after for body surrounded by braces.
9751 PR tree-optimization/83081
9752 * profile-count.h (profile_probability::split): New method.
9753 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9754 Use profile_probability::split.
9755 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9756 when splitting a single conditional jump into 2.
9758 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9760 PR tree-optimization/69452
9761 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9762 decl.
9764 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9766 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9767 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9768 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9770 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9772 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9773 * config/rl78/rl78.md (movdi): New define_expand.
9774 * config/rl78/rl78.c (rl78_split_movdi): New function.
9776 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9778 PR target/83862
9779 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9780 no longer used.
9781 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9782 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9783 128-bit to produce an UNSPEC move to get the double word with the
9784 signbit and then a shift directly to do signbit.
9785 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9786 implementation with a new version that just does either a direct
9787 move or a regular move. Move memory interface to separate insns.
9788 Move insns so they are next to the expander.
9789 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9790 with signbit move. Split big and little endian case.
9791 (signbit<mode>2_dm_mem_le): Likewise.
9792 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9793 (signbit<mode>2_dm2): Likewise.
9795 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9797 * config/rl78/rl78.md (anddi3): New define_expand.
9799 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9801 * config/rl78/rl78.md (umindi3): New define_expand.
9803 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9805 * config/rl78/rl78.md (smindi3): New define_expand.
9807 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9809 * config/rl78/rl78.md (smaxdi3): New define_expand.
9811 2018-01-22 Carl Love <cel@us.ibm.com>
9813 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9814 LVX_V1TI): Add macro expansion.
9815 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9816 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9817 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9818 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9819 Change check to determine if the instruction is a byte reversing
9820 entry. Fix typo in comment.
9821 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9822 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9823 Add def_builtin calls for new builtins.
9824 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9825 Add define_insn expansion.
9827 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9829 * config/rl78/rl78.md (umaxdi3): New define_expand.
9831 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9833 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9834 for non-QImode registers.
9836 2018-01-22 Richard Biener <rguenther@suse.de>
9838 PR tree-optimization/83963
9839 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9840 including the loop exit block.
9841 (scop_detection::merge_sese): Likewise.
9842 (scop_detection::add_scop): Do it here instead.
9844 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9846 * doc/sourcebuild.texi (arm_softfloat): Document.
9848 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9850 PR gcc/77734
9851 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9852 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9853 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9855 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9856 David Edelsohn <dje.gcc@gmail.com>
9858 PR target/83946
9859 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9860 Change "crset eq" to "crset 2".
9861 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9862 (*call_indirect_aix<mode>_nospec): Likewise.
9863 (*call_value_indirect_aix<mode>_nospec): Likewise.
9864 (*call_indirect_elfv2<mode>_nospec): Likewise.
9865 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9866 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9867 change assembly output from . to $.
9868 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9869 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9870 (*tablejump<mode>_internal1_nospec): Likewise.
9872 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9874 PR target/80870
9875 * config/sh/sh_optimize_sett_clrt.cc:
9876 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9878 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9880 PR tree-optimization/83940
9881 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9882 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9883 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9884 use it to pass back the definition type.
9885 (vect_check_store_rhs): Likewise rhs_dt_out.
9886 (vect_build_gather_load_calls): Add a mask_dt argument and use
9887 it instead of a call to vect_is_simple_use.
9888 (vectorizable_store): Update calls to vect_check_load_store_mask
9889 and vect_check_store_rhs. Use the dt returned by the latter instead
9890 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9891 instead of calls to vect_is_simple_use. Pass the scalar rather
9892 than the vector operand to vect_is_simple_use when handling
9893 second and subsequent copies of an rhs value.
9894 (vectorizable_load): Update calls to vect_check_load_store_mask
9895 and vect_build_gather_load_calls. Use the cached mask_dt and
9896 gs_info.offset_dt instead of calls to vect_is_simple_use.
9898 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9900 PR middle-end/83945
9901 * tree-emutls.c: Include gimplify.h.
9902 (lower_emutls_2): New function.
9903 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9904 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9905 it before further processing.
9907 PR target/83930
9908 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9909 UINTVAL (trueop1) instead of INTVAL (op1).
9911 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9913 PR debug/81570
9914 PR debug/83728
9915 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9916 INCOMING_FRAME_SP_OFFSET if not defined.
9917 (scan_trace): Add ENTRY argument. If true and
9918 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9919 emit a note to adjust the CFA offset.
9920 (create_cfi_notes): Adjust scan_trace callers.
9921 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9922 INCOMING_FRAME_SP_OFFSET in the CIE.
9923 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9924 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9925 Likewise.
9926 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9927 * doc/tm.texi: Regenerated.
9929 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9931 PR rtl-optimization/83147
9932 * lra-constraints.c (remove_inheritance_pseudos): Use
9933 lra_substitute_pseudo_within_insn.
9935 2018-01-19 Tom de Vries <tom@codesourcery.com>
9936 Cesar Philippidis <cesar@codesourcery.com>
9938 PR target/83920
9939 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9941 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9943 PR target/83790
9944 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9945 spaces for function labels.
9947 2018-01-19 Martin Liska <mliska@suse.cz>
9949 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9950 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9951 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9952 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9953 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9954 (PRED_CONST_RETURN): Change from 69 to 65.
9955 (PRED_NULL_RETURN): Change from 91 to 71.
9956 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9957 (PRED_LOOP_GUARD): Change from 66 to 73.
9959 2018-01-19 Martin Liska <mliska@suse.cz>
9961 * predict.c (predict_insn_def): Add new assert.
9962 (struct branch_predictor): Change type to signed integer.
9963 (test_prediction_value_range): Amend test to cover
9964 PROB_UNINITIALIZED.
9965 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9966 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9967 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9968 (PRED_LOOP_IV_COMPARE): Likewise.
9969 * predict.h (PROB_UNINITIALIZED): Define new constant.
9971 2018-01-19 Martin Liska <mliska@suse.cz>
9973 * predict.c (dump_prediction): Add new format for
9974 analyze_brprob.py script which is enabled with -details
9975 suboption.
9976 * profile-count.h (precise_p): New function.
9978 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9980 PR tree-optimization/83922
9981 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9982 there are no statements that need masking.
9983 (vect_active_double_reduction_p): New function.
9984 (vect_analyze_loop_operations): Use it when handling phis that
9985 are not in the loop header.
9987 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9989 PR tree-optimization/83914
9990 * tree-vect-loop.c (vectorizable_induction): Don't convert
9991 init_expr or apply the peeling adjustment for inductions
9992 that are nested within the vectorized loop.
9994 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9996 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9997 instead of NEG.
9999 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10001 PR sanitizer/81715
10002 PR testsuite/83882
10003 * function.h (gimplify_parameters): Add gimple_seq * argument.
10004 * function.c: Include gimple.h and options.h.
10005 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10006 for the added local temporaries if needed.
10007 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10008 if there are any parameter cleanups, wrap whole body into a
10009 try/finally with the cleanups.
10011 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10013 PR target/82964
10014 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10015 Use GET_MODE_CLASS for scalar floating point.
10017 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10019 PR ipa/82256
10020 patch by PaX Team
10021 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10022 Fix call of call_cgraph_insertion_hooks.
10024 2018-01-18 Martin Sebor <msebor@redhat.com>
10026 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10028 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10030 PR ipa/83619
10031 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10032 frequencies.
10034 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10036 PR other/70268
10037 * common.opt: (-ffile-prefix-map): New option.
10038 * opts.c (common_handle_option): Defer it.
10039 * opts-global.c (handle_common_deferred_options): Handle it.
10040 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10041 * file-prefix-map.h: New file.
10042 (remap_debug_filename, add_debug_prefix_map): ...here.
10043 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10044 * final.c (debug_prefix_map, add_debug_prefix_map
10045 remap_debug_filename): Move to...
10046 * file-prefix-map.c: New file.
10047 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10048 generalize, get rid of alloca(), use strrchr() instead of strchr().
10049 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10050 Implement in terms of add_prefix_map().
10051 (remap_macro_filename, remap_debug_filename): Implement in term of
10052 remap_filename().
10053 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10054 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10055 * dbxout.c: Include file-prefix-map.h.
10056 * varasm.c: Likewise.
10057 * vmsdbgout.c: Likewise.
10058 * xcoffout.c: Likewise.
10059 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10060 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10061 * doc/invoke.texi (-ffile-prefix-map): Document.
10062 (-fdebug-prefix-map): Update description.
10064 2018-01-18 Martin Liska <mliska@suse.cz>
10066 * config/i386/i386.c (indirect_thunk_name): Document that also
10067 lfence is emitted.
10068 (output_indirect_thunk): Document why both instructions
10069 (pause and lfence) are generated.
10071 2018-01-18 Richard Biener <rguenther@suse.de>
10073 PR tree-optimization/83887
10074 * graphite-scop-detection.c
10075 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10076 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10077 (scop_detection::merge_sese): Re-implement with a flood-fill
10078 algorithm that properly finds a SESE region if it exists.
10080 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10082 PR c/61240
10083 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10084 pointer_diff optimizations use view_convert instead of convert.
10086 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10088 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10089 Generate different code for -mno-speculate-indirect-jumps.
10090 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10091 (*call_indirect_aix<mode>): Disable for
10092 -mno-speculate-indirect-jumps.
10093 (*call_indirect_aix<mode>_nospec): New define_insn.
10094 (*call_value_indirect_aix<mode>): Disable for
10095 -mno-speculate-indirect-jumps.
10096 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10097 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10098 -mno-speculate-indirect-jumps.
10099 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10101 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10103 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10104 long double type, set the flags for noting the default long double
10105 type, even if we don't pass or return a long double type.
10107 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10109 PR ipa/83051
10110 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10111 failure.
10113 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10115 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10116 support for merge[hl].
10117 (fold_mergehl_helper): New helper function.
10118 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10119 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10120 (altivec_vmrglw_direct): Add xxmrglw insn.
10122 2018-01-17 Andrew Waterman <andrew@sifive.com>
10124 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10125 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10127 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10129 PR lto/83121
10130 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10131 call the lto_location_cache before reading the
10132 DECL_SOURCE_LOCATION of the types.
10134 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10135 Richard Sandiford <richard.sandiford@linaro.org>
10137 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10138 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10139 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10140 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10141 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10142 Add declaration.
10143 * config/aarch64/constraints.md (aarch64_movti_operand):
10144 Limit immediates.
10145 * config/aarch64/predicates.md (Uti): Add new constraint.
10147 2018-01-17 Carl Love <cel@us.ibm.com>
10149 * config/rs6000/vsx.md (define_expand xl_len_r,
10150 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10151 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10152 lxvll.
10153 (define_expand, define_insn): Move the shift left from the
10154 define_insn to the define_expand for lxvl and stxvl instructions.
10155 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10156 and XL_LEN_R definitions to PURE.
10158 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10160 * config/i386/i386.c (indirect_thunk_name): Declare regno
10161 as unsigned int. Compare regno with INVALID_REGNUM.
10162 (output_indirect_thunk): Ditto.
10163 (output_indirect_thunk_function): Ditto.
10164 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10165 in the call to output_indirect_thunk_function.
10167 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10169 PR middle-end/83884
10170 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10171 rather than the size of inner_type to determine the stack slot size
10172 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10174 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10176 PR target/83546
10177 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10178 to PTA_SILVERMONT.
10180 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10182 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10183 endian Linux systems to optionally enable multilibs for selecting
10184 the long double type if the user configured an explicit type.
10185 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10186 have no long double multilibs if not defined.
10187 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10188 warn if the user used -mabi={ieee,ibm}longdouble and we built
10189 multilibs for long double.
10190 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10191 appropriate multilib option.
10192 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10193 multilib options.
10194 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10195 for building long double multilibs.
10196 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10198 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10200 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10201 copies.
10203 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10204 64 bits.
10205 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10206 128 bits.
10208 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10209 variables.
10211 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10212 return value.
10214 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10216 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10217 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10219 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10221 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10222 different rtl trees depending on TARGET_64BIT.
10223 (rs6000_gen_lvx): Likewise.
10225 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10227 * config/visium/visium.md (nop): Tweak comment.
10228 (hazard_nop): Likewise.
10230 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10232 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10233 -mspeculate-indirect-jumps.
10234 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10235 for -mno-speculate-indirect-jumps.
10236 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10237 (*call_value_indirect_elfv2<mode>): Disable for
10238 -mno-speculate-indirect-jumps.
10239 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10240 (indirect_jump): Emit different RTL for
10241 -mno-speculate-indirect-jumps.
10242 (*indirect_jump<mode>): Disable for
10243 -mno-speculate-indirect-jumps.
10244 (*indirect_jump<mode>_nospec): New define_insn.
10245 (tablejump): Emit different RTL for
10246 -mno-speculate-indirect-jumps.
10247 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10248 (tablejumpsi_nospec): New define_expand.
10249 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10250 (tablejumpdi_nospec): New define_expand.
10251 (*tablejump<mode>_internal1): Disable for
10252 -mno-speculate-indirect-jumps.
10253 (*tablejump<mode>_internal1_nospec): New define_insn.
10254 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10255 option.
10257 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10259 * caller-save.c (insert_save): Drop unnecessary parameter. All
10260 callers updated.
10262 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10263 Richard Biener <rguenth@suse.de>
10265 PR libgomp/83590
10266 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10267 return early, inline manually is_gimple_sizepos. Make sure if we
10268 call gimplify_expr we don't end up with a gimple constant.
10269 * tree.c (variably_modified_type_p): Don't return true for
10270 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10271 * gimplify.h (is_gimple_sizepos): Remove.
10273 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10275 PR tree-optimization/83857
10276 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10277 vectorizable_live_operation for pure SLP statements.
10278 (vectorizable_live_operation): Handle PHIs.
10280 2018-01-16 Richard Biener <rguenther@suse.de>
10282 PR tree-optimization/83867
10283 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10284 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10286 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10288 PR c/83844
10289 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10290 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10291 If off is not INTEGER_CST, issue a may not be aligned warning
10292 rather than isn't aligned. Use isn%'t rather than isn't.
10293 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10294 into MULT_EXPR.
10295 <case MULT_EXPR>: Improve the case when bottom and one of the
10296 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10297 operand, in that case check if the other operand is multiple of
10298 bottom divided by the INTEGER_CST operand.
10300 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10302 PR target/83858
10303 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10304 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10305 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10306 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10307 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10308 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10309 (pa_function_arg_size): New function.
10311 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10313 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10314 in a separate statement.
10316 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10318 PR tree-optimization/83847
10319 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10320 group gathers and scatters.
10322 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10324 PR rtl-optimization/86620
10325 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10327 PR rtl-optimization/83213
10328 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10329 to last if both are JUMP_INSNs.
10331 PR tree-optimization/83843
10332 * gimple-ssa-store-merging.c
10333 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10334 store_immediate_info for bswap/nop orig_stores.
10336 2018-01-15 Andrew Waterman <andrew@sifive.com>
10338 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10339 !TARGET_MUL.
10340 <UDIV>: Increase cost if !TARGET_DIV.
10342 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10344 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10345 (define_attr "cr_logical_3op"): New.
10346 (cceq_ior_compare): Adjust.
10347 (cceq_ior_compare_complement): Adjust.
10348 (*cceq_rev_compare): Adjust.
10349 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10350 (is_cracked_insn): Adjust.
10351 (insn_must_be_first_in_group): Adjust.
10352 * config/rs6000/40x.md: Adjust.
10353 * config/rs6000/440.md: Adjust.
10354 * config/rs6000/476.md: Adjust.
10355 * config/rs6000/601.md: Adjust.
10356 * config/rs6000/603.md: Adjust.
10357 * config/rs6000/6xx.md: Adjust.
10358 * config/rs6000/7450.md: Adjust.
10359 * config/rs6000/7xx.md: Adjust.
10360 * config/rs6000/8540.md: Adjust.
10361 * config/rs6000/cell.md: Adjust.
10362 * config/rs6000/e300c2c3.md: Adjust.
10363 * config/rs6000/e500mc.md: Adjust.
10364 * config/rs6000/e500mc64.md: Adjust.
10365 * config/rs6000/e5500.md: Adjust.
10366 * config/rs6000/e6500.md: Adjust.
10367 * config/rs6000/mpc.md: Adjust.
10368 * config/rs6000/power4.md: Adjust.
10369 * config/rs6000/power5.md: Adjust.
10370 * config/rs6000/power6.md: Adjust.
10371 * config/rs6000/power7.md: Adjust.
10372 * config/rs6000/power8.md: Adjust.
10373 * config/rs6000/power9.md: Adjust.
10374 * config/rs6000/rs64.md: Adjust.
10375 * config/rs6000/titan.md: Adjust.
10377 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10379 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10380 ix86_indirect_branch_register logic.
10382 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10384 * config/i386/constraints.md (Bs): Update
10385 ix86_indirect_branch_register check. Don't check
10386 ix86_indirect_branch_register with GOT_memory_operand.
10387 (Bw): Likewise.
10388 * config/i386/predicates.md (GOT_memory_operand): Don't check
10389 ix86_indirect_branch_register here.
10390 (GOT32_symbol_operand): Likewise.
10392 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10394 * config/i386/predicates.md (constant_call_address_operand):
10395 Rewrite ix86_indirect_branch_register logic.
10396 (sibcall_insn_operand): Likewise.
10398 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10400 * config/i386/constraints.md (Bs): Replace
10401 ix86_indirect_branch_thunk_register with
10402 ix86_indirect_branch_register.
10403 (Bw): Likewise.
10404 * config/i386/i386.md (indirect_jump): Likewise.
10405 (tablejump): Likewise.
10406 (*sibcall_memory): Likewise.
10407 (*sibcall_value_memory): Likewise.
10408 Peepholes of indirect call and jump via memory: Likewise.
10409 * config/i386/i386.opt: Likewise.
10410 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10411 (GOT_memory_operand): Likewise.
10412 (call_insn_operand): Likewise.
10413 (sibcall_insn_operand): Likewise.
10414 (GOT32_symbol_operand): Likewise.
10416 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10418 PR middle-end/83837
10419 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10420 type rather than type addr's type points to.
10421 (expand_omp_atomic_mutex): Likewise.
10422 (expand_omp_atomic): Likewise.
10424 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10426 PR target/83839
10427 * config/i386/i386.c (output_indirect_thunk_function): Use
10428 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10429 for __x86_return_thunk.
10431 2018-01-15 Richard Biener <rguenther@suse.de>
10433 PR middle-end/83850
10434 * expmed.c (extract_bit_field_1): Fix typo.
10436 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10438 PR target/83687
10439 * config/arm/iterators.md (VF): New mode iterator.
10440 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10441 Remove integer-related logic from pattern.
10442 (neon_vabd<mode>_3): Likewise.
10444 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10446 PR middle-end/82694
10447 * common.opt (fstrict-overflow): No longer an alias.
10448 (fwrapv-pointer): New option.
10449 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10450 also for pointer types based on flag_wrapv_pointer.
10451 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10452 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10453 opts->x_flag_wrapv got set.
10454 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10455 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10456 POINTER_TYPE_OVERFLOW_UNDEFINED.
10457 * match.pd: Likewise in address comparison pattern.
10458 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10460 2018-01-15 Richard Biener <rguenther@suse.de>
10462 PR lto/83804
10463 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10464 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10465 Reset type names to their identifier if their TYPE_DECL doesn't
10466 have linkage (and thus is used for ODR and devirt).
10467 (save_debug_info_for_decl): Remove.
10468 (save_debug_info_for_type): Likewise.
10469 (add_tree_to_fld_list): Adjust.
10470 * tree-pretty-print.c (dump_generic_node): Make dumping of
10471 type names more robust.
10473 2018-01-15 Richard Biener <rguenther@suse.de>
10475 * BASE-VER: Bump to 8.0.1.
10477 2018-01-14 Martin Sebor <msebor@redhat.com>
10479 PR other/83508
10480 * builtins.c (check_access): Avoid warning when the no-warning bit
10481 is set.
10483 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10485 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10486 * ira-color (allocno_hard_regs_compare): Likewise.
10488 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10490 PR target/83013
10491 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10492 Use .pushsection/.popsection.
10494 2018-01-14 Martin Sebor <msebor@redhat.com>
10496 PR c++/81327
10497 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10499 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10501 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10502 entry from extra_headers.
10503 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10504 extra_headers, make the list bitwise identical to the i?86-*-* one.
10506 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10508 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10509 -mcmodel=large with -mindirect-branch=thunk,
10510 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10511 -mfunction-return=thunk-extern.
10512 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10513 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10514 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10516 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10518 * config/i386/i386.c (print_reg): Print the name of the full
10519 integer register without '%'.
10520 (ix86_print_operand): Handle 'V'.
10521 * doc/extend.texi: Document 'V' modifier.
10523 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10525 * config/i386/constraints.md (Bs): Disallow memory operand for
10526 -mindirect-branch-register.
10527 (Bw): Likewise.
10528 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10529 (GOT_memory_operand): Likewise.
10530 (call_insn_operand): Likewise.
10531 (sibcall_insn_operand): Likewise.
10532 (GOT32_symbol_operand): Likewise.
10533 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10534 for -mindirect-branch-register.
10535 (tablejump): Likewise.
10536 (*sibcall_memory): Likewise.
10537 (*sibcall_value_memory): Likewise.
10538 Disallow peepholes of indirect call and jump via memory for
10539 -mindirect-branch-register.
10540 (*call_pop): Replace m with Bw.
10541 (*call_value_pop): Likewise.
10542 (*sibcall_pop_memory): Replace m with Bs.
10543 * config/i386/i386.opt (mindirect-branch-register): New option.
10544 * doc/invoke.texi: Document -mindirect-branch-register option.
10546 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10548 * config/i386/i386-protos.h (ix86_output_function_return): New.
10549 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10550 set function_return_type.
10551 (indirect_thunk_name): Add ret_p to indicate thunk for function
10552 return.
10553 (output_indirect_thunk_function): Pass false to
10554 indirect_thunk_name.
10555 (ix86_output_indirect_branch_via_reg): Likewise.
10556 (ix86_output_indirect_branch_via_push): Likewise.
10557 (output_indirect_thunk_function): Create alias for function
10558 return thunk if regno < 0.
10559 (ix86_output_function_return): New function.
10560 (ix86_handle_fndecl_attribute): Handle function_return.
10561 (ix86_attribute_table): Add function_return.
10562 * config/i386/i386.h (machine_function): Add
10563 function_return_type.
10564 * config/i386/i386.md (simple_return_internal): Use
10565 ix86_output_function_return.
10566 (simple_return_internal_long): Likewise.
10567 * config/i386/i386.opt (mfunction-return=): New option.
10568 (indirect_branch): Mention -mfunction-return=.
10569 * doc/extend.texi: Document function_return function attribute.
10570 * doc/invoke.texi: Document -mfunction-return= option.
10572 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10574 * config/i386/i386-opts.h (indirect_branch): New.
10575 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10576 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10577 with local indirect jump when converting indirect call and jump.
10578 (ix86_set_indirect_branch_type): New.
10579 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10580 (indirectlabelno): New.
10581 (indirect_thunk_needed): Likewise.
10582 (indirect_thunk_bnd_needed): Likewise.
10583 (indirect_thunks_used): Likewise.
10584 (indirect_thunks_bnd_used): Likewise.
10585 (INDIRECT_LABEL): Likewise.
10586 (indirect_thunk_name): Likewise.
10587 (output_indirect_thunk): Likewise.
10588 (output_indirect_thunk_function): Likewise.
10589 (ix86_output_indirect_branch_via_reg): Likewise.
10590 (ix86_output_indirect_branch_via_push): Likewise.
10591 (ix86_output_indirect_branch): Likewise.
10592 (ix86_output_indirect_jmp): Likewise.
10593 (ix86_code_end): Call output_indirect_thunk_function if needed.
10594 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10595 needed.
10596 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10597 (ix86_attribute_table): Add indirect_branch.
10598 * config/i386/i386.h (machine_function): Add indirect_branch_type
10599 and has_local_indirect_jump.
10600 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10601 to true.
10602 (tablejump): Likewise.
10603 (*indirect_jump): Use ix86_output_indirect_jmp.
10604 (*tablejump_1): Likewise.
10605 (simple_return_indirect_internal): Likewise.
10606 * config/i386/i386.opt (mindirect-branch=): New option.
10607 (indirect_branch): New.
10608 (keep): Likewise.
10609 (thunk): Likewise.
10610 (thunk-inline): Likewise.
10611 (thunk-extern): Likewise.
10612 * doc/extend.texi: Document indirect_branch function attribute.
10613 * doc/invoke.texi: Document -mindirect-branch= option.
10615 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10617 PR ipa/83051
10618 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10620 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10622 * ipa-inline.c (want_inline_small_function_p): Return false if
10623 inlining has already failed with CIF_FINAL_ERROR.
10624 (update_caller_keys): Call want_inline_small_function_p before
10625 can_inline_edge_p.
10626 (update_callee_keys): Likewise.
10628 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10630 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10631 New function.
10632 (rs6000_quadword_masked_address_p): Likewise.
10633 (quad_aligned_load_p): Likewise.
10634 (quad_aligned_store_p): Likewise.
10635 (const_load_sequence_p): Add comment to describe the outer-most loop.
10636 (mimic_memory_attributes_and_flags): New function.
10637 (rs6000_gen_stvx): Likewise.
10638 (replace_swapped_aligned_store): Likewise.
10639 (rs6000_gen_lvx): Likewise.
10640 (replace_swapped_aligned_load): Likewise.
10641 (replace_swapped_load_constant): Capitalize argument name in
10642 comment describing this function.
10643 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10644 and stores that access quad-word aligned addresses and replace
10645 with stvx or lvx instructions when appropriate.
10646 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10647 New function prototype.
10648 (rs6000_quadword_masked_address_p): Likewise.
10649 (rs6000_gen_lvx): Likewise.
10650 (rs6000_gen_stvx): Likewise.
10651 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10652 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10653 when memory address is aligned.
10654 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10655 this split to select lvx instruction when memory address is aligned.
10656 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10657 instruction when memory address is aligned.
10658 (*vsx_le_perm_load_v16qi): Likewise.
10659 (four unnamed splitters): Modify to select the stvx instruction
10660 when memory is aligned.
10662 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10664 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10665 which appears in the queue multiple times.
10667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10668 Alan Hayward <alan.hayward@arm.com>
10669 David Sherwood <david.sherwood@arm.com>
10671 * tree-vectorizer.h (vec_lower_bound): New structure.
10672 (_loop_vec_info): Add check_nonzero and lower_bounds.
10673 (LOOP_VINFO_CHECK_NONZERO): New macro.
10674 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10675 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10676 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10677 fields. Make seg_len the distance travelled, not including the
10678 access size.
10679 (dr_direction_indicator): Declare.
10680 (dr_zero_step_indicator): Likewise.
10681 (dr_known_forward_stride_p): Likewise.
10682 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10683 tree-ssanames.h.
10684 (runtime_alias_check_p): Allow runtime alias checks with
10685 variable strides.
10686 (operator ==): Compare access_size and align.
10687 (prune_runtime_alias_test_list): Rework for new distinction between
10688 the access_size and seg_len.
10689 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10690 segment lengths.
10691 (get_segment_min_max): New function.
10692 (create_intersect_range_checks): Use it.
10693 (dr_step_indicator): New function.
10694 (dr_direction_indicator): Likewise.
10695 (dr_zero_step_indicator): Likewise.
10696 (dr_known_forward_stride_p): Likewise.
10697 * tree-loop-distribution.c (data_ref_segment_size): Return
10698 DR_STEP * (niters - 1).
10699 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10700 constructor.
10701 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10702 (vect_preserves_scalar_order_p): New function, split out from...
10703 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10704 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10705 (vect_vfa_access_size): New function.
10706 (vect_vfa_align): Likewise.
10707 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10708 (dump_lower_bound): New function.
10709 (vect_check_lower_bound): Likewise.
10710 (vect_small_gap_p): Likewise.
10711 (vectorizable_with_step_bound_p): Likewise.
10712 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10713 depencies if the vectorization factor is 1. Convert the checks
10714 for nonzero steps into checks on the bounds of DR_STEP. Try using
10715 a bunds check for variable steps if the minimum required step is
10716 relatively small. Update calls to the dr_with_seg_len
10717 constructor and to vect_compile_time_alias.
10718 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10719 function.
10720 (vect_loop_versioning): Call it.
10721 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10722 when retrying.
10723 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10726 Alan Hayward <alan.hayward@arm.com>
10727 David Sherwood <david.sherwood@arm.com>
10729 * doc/sourcebuild.texi (vect_scatter_store): Document.
10730 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10731 optabs.
10732 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10733 Document.
10734 * genopinit.c (main): Add supports_vec_scatter_store and
10735 supports_vec_scatter_store_cached to target_optabs.
10736 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10737 IFN_MASK_SCATTER_STORE.
10738 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10739 functions.
10740 * internal-fn.h (internal_store_fn_p): Declare.
10741 (internal_fn_stored_value_index): Likewise.
10742 * internal-fn.c (scatter_store_direct): New macro.
10743 (expand_scatter_store_optab_fn): New function.
10744 (direct_scatter_store_optab_supported_p): New macro.
10745 (internal_store_fn_p): New function.
10746 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10747 IFN_MASK_SCATTER_STORE.
10748 (internal_fn_mask_index): Likewise.
10749 (internal_fn_stored_value_index): New function.
10750 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10751 for scatter stores.
10752 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10753 * optabs-query.c (supports_vec_scatter_store_p): New function.
10754 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10755 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10756 true for scatter stores.
10757 (vect_gather_scatter_fn_p): Handle scatter stores too.
10758 (vect_check_gather_scatter): Consider using scatter stores if
10759 supports_vec_scatter_store_p.
10760 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10761 scatter stores too.
10762 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10763 internal_fn_stored_value_index.
10764 (check_load_store_masking): Handle scatter stores too.
10765 (vect_get_store_rhs): Make public.
10766 (vectorizable_call): Use internal_store_fn_p.
10767 (vectorizable_store): Handle scatter store internal functions.
10768 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10769 when deciding whether the end of the group has been reached.
10770 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10771 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10772 (mask_scatter_store<mode>): New insns.
10774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10775 Alan Hayward <alan.hayward@arm.com>
10776 David Sherwood <david.sherwood@arm.com>
10778 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10779 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10780 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10781 function.
10782 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10783 Use vect_truncate_gather_scatter_offset if we can't treat the
10784 operation as a normal gather load or scatter store.
10785 (get_group_load_store_type): Take the gather_scatter_info
10786 as argument. Try using a gather load or scatter store for
10787 single-element groups.
10788 (get_load_store_type): Update calls to get_group_load_store_type
10789 and vect_use_strided_gather_scatters_p.
10791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10792 Alan Hayward <alan.hayward@arm.com>
10793 David Sherwood <david.sherwood@arm.com>
10795 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10796 optional tree argument.
10797 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10798 null target hooks.
10799 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10800 but continue to use the current value as a fallback.
10801 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10802 to compare the updates.
10803 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10804 (get_load_store_type): Use it when handling a strided access.
10805 (vect_get_strided_load_store_ops): New function.
10806 (vect_get_data_ptr_increment): Likewise.
10807 (vectorizable_load): Handle strided gather loads. Always pass
10808 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10811 Alan Hayward <alan.hayward@arm.com>
10812 David Sherwood <david.sherwood@arm.com>
10814 * doc/md.texi (gather_load@var{m}): Document.
10815 (mask_gather_load@var{m}): Likewise.
10816 * genopinit.c (main): Add supports_vec_gather_load and
10817 supports_vec_gather_load_cached to target_optabs.
10818 * optabs-tree.c (init_tree_optimization_optabs): Use
10819 ggc_cleared_alloc to allocate target_optabs.
10820 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10821 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10822 functions.
10823 * internal-fn.h (internal_load_fn_p): Declare.
10824 (internal_gather_scatter_fn_p): Likewise.
10825 (internal_fn_mask_index): Likewise.
10826 (internal_gather_scatter_fn_supported_p): Likewise.
10827 * internal-fn.c (gather_load_direct): New macro.
10828 (expand_gather_load_optab_fn): New function.
10829 (direct_gather_load_optab_supported_p): New macro.
10830 (direct_internal_fn_optab): New function.
10831 (internal_load_fn_p): Likewise.
10832 (internal_gather_scatter_fn_p): Likewise.
10833 (internal_fn_mask_index): Likewise.
10834 (internal_gather_scatter_fn_supported_p): Likewise.
10835 * optabs-query.c (supports_at_least_one_mode_p): New function.
10836 (supports_vec_gather_load_p): Likewise.
10837 * optabs-query.h (supports_vec_gather_load_p): Declare.
10838 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10839 and memory_type field.
10840 (NUM_PATTERNS): Bump to 15.
10841 * tree-vect-data-refs.c: Include internal-fn.h.
10842 (vect_gather_scatter_fn_p): New function.
10843 (vect_describe_gather_scatter_call): Likewise.
10844 (vect_check_gather_scatter): Try using internal functions for
10845 gather loads. Recognize existing calls to a gather load function.
10846 (vect_analyze_data_refs): Consider using gather loads if
10847 supports_vec_gather_load_p.
10848 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10849 (vect_get_gather_scatter_offset_type): Likewise.
10850 (vect_convert_mask_for_vectype): Likewise.
10851 (vect_add_conversion_to_patterm): Likewise.
10852 (vect_try_gather_scatter_pattern): Likewise.
10853 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10854 (vect_vect_recog_func_ptrs): Add it.
10855 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10856 internal_fn_mask_index and internal_gather_scatter_fn_p.
10857 (check_load_store_masking): Take the gather_scatter_info as an
10858 argument and handle gather loads.
10859 (vect_get_gather_scatter_ops): New function.
10860 (vectorizable_call): Check internal_load_fn_p.
10861 (vectorizable_load): Likewise. Handle gather load internal
10862 functions.
10863 (vectorizable_store): Update call to check_load_store_masking.
10864 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10865 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10866 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10867 (aarch64_gather_scale_operand_d): New predicates.
10868 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10869 (mask_gather_load<mode>): New insns.
10871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * optabs.def (fold_left_plus_optab): New optab.
10876 * doc/md.texi (fold_left_plus_@var{m}): Document.
10877 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10878 * internal-fn.c (fold_left_direct): Define.
10879 (expand_fold_left_optab_fn): Likewise.
10880 (direct_fold_left_optab_supported_p): Likewise.
10881 * fold-const-call.c (fold_const_fold_left): New function.
10882 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10883 * tree-parloops.c (valid_reduction_p): New function.
10884 (gather_scalar_reductions): Use it.
10885 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10886 (vect_finish_replace_stmt): Declare.
10887 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10888 (needs_fold_left_reduction_p): New function, split out from...
10889 (vect_is_simple_reduction): ...here. Accept reductions that
10890 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10891 (vect_force_simple_reduction): Also store the reduction type in
10892 the assignment's STMT_VINFO_REDUC_TYPE.
10893 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10894 (merge_with_identity): New function.
10895 (vect_expand_fold_left): Likewise.
10896 (vectorize_fold_left_reduction): Likewise.
10897 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10898 scalar phi in place for it. Check for target support and reject
10899 cases that would reassociate the operation. Defer the transform
10900 phase to vectorize_fold_left_reduction.
10901 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10902 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10903 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10907 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10908 call to ifc_temp_var.
10910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10911 Alan Hayward <alan.hayward@arm.com>
10912 David Sherwood <david.sherwood@arm.com>
10914 * target.def (legitimize_address_displacement): Take the original
10915 offset as a poly_int.
10916 * targhooks.h (default_legitimize_address_displacement): Update
10917 accordingly.
10918 * targhooks.c (default_legitimize_address_displacement): Likewise.
10919 * doc/tm.texi: Regenerate.
10920 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10921 as an argument, moving assert of ad->disp == ad->disp_term to...
10922 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10923 Try calling targetm.legitimize_address_displacement before expanding
10924 the address rather than afterwards, and adjust for the new interface.
10925 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10926 Match the new hook interface. Handle SVE addresses.
10927 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10928 new hook interface.
10930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10932 * Makefile.in (OBJS): Add early-remat.o.
10933 * target.def (select_early_remat_modes): New hook.
10934 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10935 * doc/tm.texi: Regenerate.
10936 * targhooks.h (default_select_early_remat_modes): Declare.
10937 * targhooks.c (default_select_early_remat_modes): New function.
10938 * timevar.def (TV_EARLY_REMAT): New timevar.
10939 * passes.def (pass_early_remat): New pass.
10940 * tree-pass.h (make_pass_early_remat): Declare.
10941 * early-remat.c: New file.
10942 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10943 function.
10944 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10947 Alan Hayward <alan.hayward@arm.com>
10948 David Sherwood <david.sherwood@arm.com>
10950 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10951 vfm1 with a bound_epilog parameter.
10952 (vect_do_peeling): Update calls accordingly, and move the prologue
10953 call earlier in the function. Treat the base bound_epilog as 0 for
10954 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10955 this base when peeling for gaps.
10956 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10957 with fully-masked loops.
10958 (vect_estimate_min_profitable_iters): Handle the single peeled
10959 iteration in that case.
10961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10962 Alan Hayward <alan.hayward@arm.com>
10963 David Sherwood <david.sherwood@arm.com>
10965 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10966 single-element interleaving even if the size is not a power of 2.
10967 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10968 accesses for single-element interleaving if the group size is
10969 not a power of 2.
10971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10972 Alan Hayward <alan.hayward@arm.com>
10973 David Sherwood <david.sherwood@arm.com>
10975 * doc/md.texi (fold_extract_last_@var{m}): Document.
10976 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10977 * optabs.def (fold_extract_last_optab): New optab.
10978 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10979 * internal-fn.c (fold_extract_direct): New macro.
10980 (expand_fold_extract_optab_fn): Likewise.
10981 (direct_fold_extract_optab_supported_p): Likewise.
10982 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10983 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10984 EXTRACT_LAST_REDUCTION.
10985 (get_initial_def_for_reduction): Do not create an initial vector
10986 for EXTRACT_LAST_REDUCTION reductions.
10987 (vectorizable_reduction): Leave the scalar phi in place for
10988 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10989 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10990 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10991 transform phase to vectorizable_condition.
10992 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10993 split out from...
10994 (vect_finish_stmt_generation): ...here.
10995 (vect_finish_replace_stmt): New function.
10996 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10997 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10998 pattern.
10999 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11002 Alan Hayward <alan.hayward@arm.com>
11003 David Sherwood <david.sherwood@arm.com>
11005 * doc/md.texi (extract_last_@var{m}): Document.
11006 * optabs.def (extract_last_optab): New optab.
11007 * internal-fn.def (EXTRACT_LAST): New internal function.
11008 * internal-fn.c (cond_unary_direct): New macro.
11009 (expand_cond_unary_optab_fn): Likewise.
11010 (direct_cond_unary_optab_supported_p): Likewise.
11011 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11012 loops using EXTRACT_LAST.
11013 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11014 (extract_last_<mode>): ...this optab.
11015 (vec_extract<mode><Vel>): Update accordingly.
11017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11018 Alan Hayward <alan.hayward@arm.com>
11019 David Sherwood <david.sherwood@arm.com>
11021 * target.def (empty_mask_is_expensive): New hook.
11022 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11023 * doc/tm.texi: Regenerate.
11024 * targhooks.h (default_empty_mask_is_expensive): Declare.
11025 * targhooks.c (default_empty_mask_is_expensive): New function.
11026 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11027 if the target says that empty masks are expensive.
11028 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11029 New function.
11030 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11033 Alan Hayward <alan.hayward@arm.com>
11034 David Sherwood <david.sherwood@arm.com>
11036 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11037 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11038 (vect_use_loop_mask_for_alignment_p): New function.
11039 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11040 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11041 niters_skip argument. Make sure that the first niters_skip elements
11042 of the first iteration are inactive.
11043 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11044 Update call to vect_set_loop_masks_directly.
11045 (get_misalign_in_elems): New function, split out from...
11046 (vect_gen_prolog_loop_niters): ...here.
11047 (vect_update_init_of_dr): Take a code argument that specifies whether
11048 the adjustment should be added or subtracted.
11049 (vect_update_init_of_drs): Likewise.
11050 (vect_prepare_for_masked_peels): New function.
11051 (vect_do_peeling): Skip prologue peeling if we're using a mask
11052 instead. Update call to vect_update_inits_of_drs.
11053 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11054 mask_skip_niters.
11055 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11056 alignment. Do not include the number of peeled iterations in
11057 the minimum threshold in that case.
11058 (vectorizable_induction): Adjust the start value down by
11059 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11060 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11061 Take the number of skipped iterations into account when calculating
11062 the loop bounds.
11063 * tree-vect-stmts.c (vect_gen_while_not): New function.
11065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11066 Alan Hayward <alan.hayward@arm.com>
11067 David Sherwood <david.sherwood@arm.com>
11069 * doc/sourcebuild.texi (vect_fully_masked): Document.
11070 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11071 default value to 0.
11072 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11073 split out from...
11074 (vect_analyze_loop_2): ...here. Don't check the vectorization
11075 factor against the number of loop iterations if the loop is
11076 fully-masked.
11078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11079 Alan Hayward <alan.hayward@arm.com>
11080 David Sherwood <david.sherwood@arm.com>
11082 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11083 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11084 (dump_groups): Update accordingly.
11085 (iv_use::mem_type): New member variable.
11086 (address_p): New function.
11087 (record_use): Add a mem_type argument and initialize the new
11088 mem_type field.
11089 (record_group_use): Add a mem_type argument. Use address_p.
11090 Remove obsolete null checks of base_object. Update call to record_use.
11091 (find_interesting_uses_op): Update call to record_group_use.
11092 (find_interesting_uses_cond): Likewise.
11093 (find_interesting_uses_address): Likewise.
11094 (get_mem_type_for_internal_fn): New function.
11095 (find_address_like_use): Likewise.
11096 (find_interesting_uses_stmt): Try find_address_like_use before
11097 calling find_interesting_uses_op.
11098 (addr_offset_valid_p): Use the iv mem_type field as the type
11099 of the addressed memory.
11100 (add_autoinc_candidates): Likewise.
11101 (get_address_cost): Likewise.
11102 (split_small_address_groups_p): Use address_p.
11103 (split_address_groups): Likewise.
11104 (add_iv_candidate_for_use): Likewise.
11105 (autoinc_possible_for_pair): Likewise.
11106 (rewrite_groups): Likewise.
11107 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11108 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11109 (get_alias_ptr_type_for_ptr_address): New function.
11110 (rewrite_use_address): Rewrite address uses in calls that were
11111 identified by find_address_like_use.
11113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11114 Alan Hayward <alan.hayward@arm.com>
11115 David Sherwood <david.sherwood@arm.com>
11117 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11118 TARGET_MEM_REFs.
11119 * gimple-expr.h (is_gimple_addressable: Likewise.
11120 * gimple-expr.c (is_gimple_address): Likewise.
11121 * internal-fn.c (expand_call_mem_ref): New function.
11122 (expand_mask_load_optab_fn): Use it.
11123 (expand_mask_store_optab_fn): Likewise.
11125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11126 Alan Hayward <alan.hayward@arm.com>
11127 David Sherwood <david.sherwood@arm.com>
11129 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11130 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11131 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11132 (cond_umax@var{mode}): Document.
11133 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11134 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11135 (cond_umin_optab, cond_umax_optab): New optabs.
11136 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11137 (COND_IOR, COND_XOR): New internal functions.
11138 * internal-fn.h (get_conditional_internal_fn): Declare.
11139 * internal-fn.c (cond_binary_direct): New macro.
11140 (expand_cond_binary_optab_fn): Likewise.
11141 (direct_cond_binary_optab_supported_p): Likewise.
11142 (get_conditional_internal_fn): New function.
11143 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11144 Cope with reduction statements that are vectorized as calls rather
11145 than assignments.
11146 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11147 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11148 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11149 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11150 (UNSPEC_COND_EOR): New unspecs.
11151 (optab): Add mappings for them.
11152 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11153 (sve_int_op, sve_fp_op): New int attributes.
11155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11156 Alan Hayward <alan.hayward@arm.com>
11157 David Sherwood <david.sherwood@arm.com>
11159 * optabs.def (while_ult_optab): New optab.
11160 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11161 * internal-fn.def (WHILE_ULT): New internal function.
11162 * internal-fn.h (direct_internal_fn_supported_p): New override
11163 that takes two types as argument.
11164 * internal-fn.c (while_direct): New macro.
11165 (expand_while_optab_fn): New function.
11166 (convert_optab_supported_p): Likewise.
11167 (direct_while_optab_supported_p): New macro.
11168 * wide-int.h (wi::udiv_ceil): New function.
11169 * tree-vectorizer.h (rgroup_masks): New structure.
11170 (vec_loop_masks): New typedef.
11171 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11172 and fully_masked_p.
11173 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11174 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11175 (vect_max_vf): New function.
11176 (slpeel_make_loop_iterate_ntimes): Delete.
11177 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11178 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11179 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11180 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11181 internal-fn.h, stor-layout.h and optabs-query.h.
11182 (vect_set_loop_mask): New function.
11183 (add_preheader_seq): Likewise.
11184 (add_header_seq): Likewise.
11185 (interleave_supported_p): Likewise.
11186 (vect_maybe_permute_loop_masks): Likewise.
11187 (vect_set_loop_masks_directly): Likewise.
11188 (vect_set_loop_condition_masked): Likewise.
11189 (vect_set_loop_condition_unmasked): New function, split out from
11190 slpeel_make_loop_iterate_ntimes.
11191 (slpeel_make_loop_iterate_ntimes): Rename to..
11192 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11193 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11194 (vect_do_peeling): Update call accordingly.
11195 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11196 loops.
11197 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11198 mask_compare_type, can_fully_mask_p and fully_masked_p.
11199 (release_vec_loop_masks): New function.
11200 (_loop_vec_info): Use it to free the loop masks.
11201 (can_produce_all_loop_masks_p): New function.
11202 (vect_get_max_nscalars_per_iter): Likewise.
11203 (vect_verify_full_masking): Likewise.
11204 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11205 retries, and free the mask rgroups before retrying. Check loop-wide
11206 reasons for disallowing fully-masked loops. Make the final decision
11207 about whether use a fully-masked loop or not.
11208 (vect_estimate_min_profitable_iters): Do not assume that peeling
11209 for the number of iterations will be needed for fully-masked loops.
11210 (vectorizable_reduction): Disable fully-masked loops.
11211 (vectorizable_live_operation): Likewise.
11212 (vect_halve_mask_nunits): New function.
11213 (vect_double_mask_nunits): Likewise.
11214 (vect_record_loop_mask): Likewise.
11215 (vect_get_loop_mask): Likewise.
11216 (vect_transform_loop): Handle the case in which the final loop
11217 iteration might handle a partial vector. Call vect_set_loop_condition
11218 instead of slpeel_make_loop_iterate_ntimes.
11219 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11220 (check_load_store_masking): New function.
11221 (prepare_load_store_mask): Likewise.
11222 (vectorizable_store): Handle fully-masked loops.
11223 (vectorizable_load): Likewise.
11224 (supportable_widening_operation): Use vect_halve_mask_nunits for
11225 booleans.
11226 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11227 (vect_gen_while): New function.
11228 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11229 (aarch64_uqdec<mode>): New insn.
11231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11232 Alan Hayward <alan.hayward@arm.com>
11233 David Sherwood <david.sherwood@arm.com>
11235 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11236 (reduc_xor_scal_optab): New optabs.
11237 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11238 (reduc_xor_scal_@var{m}): Document.
11239 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11240 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11241 internal functions.
11242 * fold-const-call.c (fold_const_call): Handle them.
11243 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11244 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11245 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11246 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11247 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11248 (UNSPEC_XORV): New unspecs.
11249 (optab): Add entries for them.
11250 (BITWISEV): New int iterator.
11251 (bit_reduc_op): New int attributes.
11253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11254 Alan Hayward <alan.hayward@arm.com>
11255 David Sherwood <david.sherwood@arm.com>
11257 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11258 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11259 * optabs.def (vec_shl_insert_optab): New optab.
11260 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11261 (duplicate_and_interleave): Likewise.
11262 * tree-vect-loop.c: Include internal-fn.h.
11263 (neutral_op_for_slp_reduction): New function, split out from
11264 get_initial_defs_for_reduction.
11265 (get_initial_def_for_reduction): Handle option 2 for variable-length
11266 vectors by loading the neutral value into a vector and then shifting
11267 the initial value into element 0.
11268 (get_initial_defs_for_reduction): Replace the code argument with
11269 the neutral value calculated by neutral_op_for_slp_reduction.
11270 Use gimple_build_vector for constant-length vectors.
11271 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11272 but the first group_size elements have a neutral value.
11273 Use duplicate_and_interleave otherwise.
11274 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11275 Update call to get_initial_defs_for_reduction. Handle SLP
11276 reductions for variable-length vectors by creating one vector
11277 result for each scalar result, with the elements associated
11278 with other scalar results stubbed out with the neutral value.
11279 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11280 Require IFN_VEC_SHL_INSERT for double reductions on
11281 variable-length vectors, or SLP reductions that have
11282 a neutral value. Require can_duplicate_and_interleave_p
11283 support for variable-length unchained SLP reductions if there
11284 is no neutral value, such as for MIN/MAX reductions. Also require
11285 the number of vector elements to be a multiple of the number of
11286 SLP statements when doing variable-length unchained SLP reductions.
11287 Update call to vect_create_epilog_for_reduction.
11288 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11289 and remove initial values.
11290 (duplicate_and_interleave): Make public.
11291 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11292 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11295 Alan Hayward <alan.hayward@arm.com>
11296 David Sherwood <david.sherwood@arm.com>
11298 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11299 (can_duplicate_and_interleave_p): New function.
11300 (vect_get_and_check_slp_defs): Take the vector of statements
11301 rather than just the current one. Remove excess parentheses.
11302 Restriction rejectinon of vect_constant_def and vect_external_def
11303 for variable-length vectors to boolean types, or types for which
11304 can_duplicate_and_interleave_p is false.
11305 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11306 (duplicate_and_interleave): New function.
11307 (vect_get_constant_vectors): Use gimple_build_vector for
11308 constant-length vectors and suitable variable-length constant
11309 vectors. Use duplicate_and_interleave for other variable-length
11310 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11317 min_profitable_iters doesn't go negative.
11319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11320 Alan Hayward <alan.hayward@arm.com>
11321 David Sherwood <david.sherwood@arm.com>
11323 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11324 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11325 * optabs.def (vec_mask_load_lanes_optab): New optab.
11326 (vec_mask_store_lanes_optab): Likewise.
11327 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11328 (MASK_STORE_LANES): Likewise.
11329 * internal-fn.c (mask_load_lanes_direct): New macro.
11330 (mask_store_lanes_direct): Likewise.
11331 (expand_mask_load_optab_fn): Handle masked operations.
11332 (expand_mask_load_lanes_optab_fn): New macro.
11333 (expand_mask_store_optab_fn): Handle masked operations.
11334 (expand_mask_store_lanes_optab_fn): New macro.
11335 (direct_mask_load_lanes_optab_supported_p): Likewise.
11336 (direct_mask_store_lanes_optab_supported_p): Likewise.
11337 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11338 parameter.
11339 (vect_load_lanes_supported): Likewise.
11340 * tree-vect-data-refs.c (strip_conversion): New function.
11341 (can_group_stmts_p): Likewise.
11342 (vect_analyze_data_ref_accesses): Use it instead of checking
11343 for a pair of assignments.
11344 (vect_store_lanes_supported): Take a masked_p parameter.
11345 (vect_load_lanes_supported): Likewise.
11346 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11347 vect_store_lanes_supported and vect_load_lanes_supported.
11348 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11349 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11350 parameter. Don't allow gaps for masked accesses.
11351 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11352 and vect_load_lanes_supported.
11353 (get_load_store_type): Take a masked_p parameter and update
11354 call to get_group_load_store_type.
11355 (vectorizable_store): Update call to get_load_store_type.
11356 Handle IFN_MASK_STORE_LANES.
11357 (vectorizable_load): Update call to get_load_store_type.
11358 Handle IFN_MASK_LOAD_LANES.
11360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11361 Alan Hayward <alan.hayward@arm.com>
11362 David Sherwood <david.sherwood@arm.com>
11364 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11365 modes for SVE.
11366 * config/aarch64/aarch64-protos.h
11367 (aarch64_sve_struct_memory_operand_p): Declare.
11368 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11369 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11370 (VPRED, vpred): Handle SVE structure modes.
11371 * config/aarch64/constraints.md (Utx): New constraint.
11372 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11373 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11374 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11375 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11376 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11377 structure modes. Split into pieces after RA.
11378 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11379 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11380 New patterns.
11381 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11382 SVE structure modes.
11383 (aarch64_classify_address): Likewise.
11384 (sizetochar): Move earlier in file.
11385 (aarch64_print_operand): Handle SVE register lists.
11386 (aarch64_array_mode): New function.
11387 (aarch64_sve_struct_memory_operand_p): Likewise.
11388 (TARGET_ARRAY_MODE): Redefine.
11390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11391 Alan Hayward <alan.hayward@arm.com>
11392 David Sherwood <david.sherwood@arm.com>
11394 * target.def (array_mode): New target hook.
11395 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11396 * doc/tm.texi: Regenerate.
11397 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11398 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11399 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11400 targetm.array_mode.
11401 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11402 type sizes.
11404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11405 Alan Hayward <alan.hayward@arm.com>
11406 David Sherwood <david.sherwood@arm.com>
11408 * fold-const.c (fold_binary_loc): Check the argument types
11409 rather than the result type when testing for a vector operation.
11411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11413 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11414 * doc/tm.texi: Regenerate.
11416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11417 Alan Hayward <alan.hayward@arm.com>
11418 David Sherwood <david.sherwood@arm.com>
11420 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11421 (sve): Document new AArch64 extension.
11422 * doc/md.texi (w): Extend the description of the AArch64
11423 constraint to include SVE vectors.
11424 (Upl, Upa): Document new AArch64 predicate constraints.
11425 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11426 enum.
11427 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11428 (msve-vector-bits=): New option.
11429 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11430 SVE when these are disabled.
11431 (sve): New extension.
11432 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11433 modes. Adjust their number of units based on aarch64_sve_vg.
11434 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11435 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11436 aarch64_addr_query_type.
11437 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11438 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11439 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11440 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11441 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11442 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11443 (aarch64_simd_imm_zero_p): Delete.
11444 (aarch64_check_zero_based_sve_index_immediate): Declare.
11445 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11446 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11447 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11448 (aarch64_sve_float_mul_immediate_p): Likewise.
11449 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11450 rather than an rtx.
11451 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11452 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11453 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11454 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11455 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11456 (aarch64_regmode_natural_size): Likewise.
11457 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11458 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11459 left one place.
11460 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11461 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11462 for VG and the SVE predicate registers.
11463 (V_ALIASES): Add a "z"-prefixed alias.
11464 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11465 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11466 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11467 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11468 (REG_CLASS_NAMES): Add entries for them.
11469 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11470 and the predicate registers.
11471 (aarch64_sve_vg): Declare.
11472 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11473 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11474 (REGMODE_NATURAL_SIZE): Define.
11475 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11476 SVE macros.
11477 * config/aarch64/aarch64.c: Include cfgrtl.h.
11478 (simd_immediate_info): Add a constructor for series vectors,
11479 and an associated step field.
11480 (aarch64_sve_vg): New variable.
11481 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11482 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11483 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11484 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11485 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11486 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11487 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11488 (aarch64_get_mask_mode): New functions.
11489 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11490 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11491 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11492 predicate modes and predicate registers. Explicitly restrict
11493 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11494 to store a vector mode if it is recognized by
11495 aarch64_classify_vector_mode.
11496 (aarch64_regmode_natural_size): New function.
11497 (aarch64_hard_regno_caller_save_mode): Return the original mode
11498 for predicates.
11499 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11500 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11501 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11502 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11503 functions.
11504 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11505 does not overlap dest if the function is frame-related. Handle
11506 SVE constants.
11507 (aarch64_split_add_offset): New function.
11508 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11509 them aarch64_add_offset.
11510 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11511 and update call to aarch64_sub_sp.
11512 (aarch64_add_cfa_expression): New function.
11513 (aarch64_expand_prologue): Pass extra temporary registers to the
11514 functions above. Handle the case in which we need to emit new
11515 DW_CFA_expressions for registers that were originally saved
11516 relative to the stack pointer, but now have to be expressed
11517 relative to the frame pointer.
11518 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11519 functions above.
11520 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11521 IP0 and IP1 values for SVE frames.
11522 (aarch64_expand_vec_series): New function.
11523 (aarch64_expand_sve_widened_duplicate): Likewise.
11524 (aarch64_expand_sve_const_vector): Likewise.
11525 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11526 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11527 into the register, rather than emitting a SET directly.
11528 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11529 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11530 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11531 (offset_9bit_signed_scaled_p): New functions.
11532 (aarch64_replicate_bitmask_imm): New function.
11533 (aarch64_bitmask_imm): Use it.
11534 (aarch64_cannot_force_const_mem): Reject expressions involving
11535 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11536 (aarch64_classify_index): Handle SVE indices, by requiring
11537 a plain register index with a scale that matches the element size.
11538 (aarch64_classify_address): Handle SVE addresses. Assert that
11539 the mode of the address is VOIDmode or an integer mode.
11540 Update call to aarch64_classify_symbol.
11541 (aarch64_classify_symbolic_expression): Update call to
11542 aarch64_classify_symbol.
11543 (aarch64_const_vec_all_in_range_p): New function.
11544 (aarch64_print_vector_float_operand): Likewise.
11545 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11546 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11547 and the FP immediates 1.0 and 0.5.
11548 (aarch64_print_address_internal): Handle SVE addresses.
11549 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11550 (aarch64_regno_regclass): Handle predicate registers.
11551 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11552 data modes.
11553 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11554 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11555 (aarch64_convert_sve_vector_bits): New function.
11556 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11557 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11558 rather than an rtx.
11559 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11560 Handle SVE vector and predicate modes. Accept VL-based constants
11561 that need only one temporary register, and VL offsets that require
11562 no temporary registers.
11563 (aarch64_conditional_register_usage): Mark the predicate registers
11564 as fixed if SVE isn't available.
11565 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11566 Return true for SVE vector and predicate modes.
11567 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11568 rather than an unsigned int. Handle SVE modes.
11569 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11570 SVE modes.
11571 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11572 if SVE is enabled.
11573 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11574 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11575 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11576 (aarch64_sve_float_mul_immediate_p): New functions.
11577 (aarch64_sve_valid_immediate): New function.
11578 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11579 Explicitly reject structure modes. Check for INDEX constants.
11580 Handle PTRUE and PFALSE constants.
11581 (aarch64_check_zero_based_sve_index_immediate): New function.
11582 (aarch64_simd_imm_zero_p): Delete.
11583 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11584 vector modes. Accept constants in the range of CNT[BHWD].
11585 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11586 ask for an Advanced SIMD mode.
11587 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11588 (aarch64_simd_vector_alignment): Handle SVE predicates.
11589 (aarch64_vectorize_preferred_vector_alignment): New function.
11590 (aarch64_simd_vector_alignment_reachable): Use it instead of
11591 the vector size.
11592 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11593 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11594 functions.
11595 (MAX_VECT_LEN): Delete.
11596 (expand_vec_perm_d): Add a vec_flags field.
11597 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11598 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11599 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11600 for SVE modes.
11601 (aarch64_evpc_rev): Rename to...
11602 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11603 (aarch64_evpc_rev_global): New function.
11604 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11605 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11606 MAX_VECT_LEN.
11607 (aarch64_evpc_sve_tbl): New function.
11608 (aarch64_expand_vec_perm_const_1): Update after rename of
11609 aarch64_evpc_rev. Handle SVE permutes too, trying
11610 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11611 than aarch64_evpc_tbl.
11612 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11613 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11614 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11615 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11616 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11617 (aarch64_expand_sve_vcond): New functions.
11618 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11619 of aarch64_vector_mode_p.
11620 (aarch64_dwarf_poly_indeterminate_value): New function.
11621 (aarch64_compute_pressure_classes): Likewise.
11622 (aarch64_can_change_mode_class): Likewise.
11623 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11624 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11625 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11626 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11627 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11628 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11629 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11630 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11631 constraints.
11632 (Dn, Dl, Dr): Accept const as well as const_vector.
11633 (Dz): Likewise. Compare against CONST0_RTX.
11634 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11635 of "vector" where appropriate.
11636 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11637 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11638 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11639 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11640 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11641 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11642 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11643 (v_int_equiv): Extend to SVE modes.
11644 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11645 mode attributes.
11646 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11647 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11648 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11649 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11650 (SVE_COND_FP_CMP): New int iterators.
11651 (perm_hilo): Handle the new unpack unspecs.
11652 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11653 attributes.
11654 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11655 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11656 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11657 (aarch64_equality_operator, aarch64_constant_vector_operand)
11658 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11659 (aarch64_sve_nonimmediate_operand): Likewise.
11660 (aarch64_sve_general_operand): Likewise.
11661 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11662 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11663 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11664 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11665 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11666 (aarch64_sve_float_arith_immediate): Likewise.
11667 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11668 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11669 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11670 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11671 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11672 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11673 (aarch64_sve_float_arith_operand): Likewise.
11674 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11675 (aarch64_sve_float_mul_operand): Likewise.
11676 (aarch64_sve_vec_perm_operand): Likewise.
11677 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11678 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11679 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11680 as well as const_vector.
11681 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11682 in file. Use CONST0_RTX and CONSTM1_RTX.
11683 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11684 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11685 Use aarch64_simd_imm_zero.
11686 * config/aarch64/aarch64-sve.md: New file.
11687 * config/aarch64/aarch64.md: Include it.
11688 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11689 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11690 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11691 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11692 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11693 (sve): New attribute.
11694 (enabled): Disable instructions with the sve attribute unless
11695 TARGET_SVE.
11696 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11697 aarch64_expand_mov_immediate.
11698 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11699 CNT[BHSD] immediates.
11700 (movti): Split CONST_POLY_INT moves into two halves.
11701 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11702 Split additions that need a temporary here if the destination
11703 is the stack pointer.
11704 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11705 (*add<mode>3_poly_1): New instruction.
11706 (set_clobber_cc): New expander.
11708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11710 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11711 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11712 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11713 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11714 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11715 Change innermode from fixed_mode_size to machine_mode.
11716 (simplify_subreg): Update call accordingly. Handle a constant-sized
11717 subreg of a variable-length CONST_VECTOR.
11719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11720 Alan Hayward <alan.hayward@arm.com>
11721 David Sherwood <david.sherwood@arm.com>
11723 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11724 (add_offset_to_base): New function, split out from...
11725 (create_mem_ref): ...here. When handling a scale other than 1,
11726 check first whether the address is valid without the offset.
11727 Add it into the base if so, leaving the index and scale as-is.
11729 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11731 PR c++/83778
11732 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11733 fold_for_warn before checking if arg2 is INTEGER_CST.
11735 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11737 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11738 (store_multiple_operation): Delete.
11739 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11740 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11741 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11742 guarded by TARGET_STRING.
11743 (rs6000_output_load_multiple): Delete.
11744 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11745 OPTION_MASK_STRING / TARGET_STRING handling.
11746 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11747 (const rs6000_opt_masks) <"string">: Change mask to 0.
11748 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11749 (MASK_STRING): Delete.
11750 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11751 parts. Simplify.
11752 (load_multiple): Delete.
11753 (*ldmsi8): Delete.
11754 (*ldmsi7): Delete.
11755 (*ldmsi6): Delete.
11756 (*ldmsi5): Delete.
11757 (*ldmsi4): Delete.
11758 (*ldmsi3): Delete.
11759 (store_multiple): Delete.
11760 (*stmsi8): Delete.
11761 (*stmsi7): Delete.
11762 (*stmsi6): Delete.
11763 (*stmsi5): Delete.
11764 (*stmsi4): Delete.
11765 (*stmsi3): Delete.
11766 (movmemsi_8reg): Delete.
11767 (corresponding unnamed define_insn): Delete.
11768 (movmemsi_6reg): Delete.
11769 (corresponding unnamed define_insn): Delete.
11770 (movmemsi_4reg): Delete.
11771 (corresponding unnamed define_insn): Delete.
11772 (movmemsi_2reg): Delete.
11773 (corresponding unnamed define_insn): Delete.
11774 (movmemsi_1reg): Delete.
11775 (corresponding unnamed define_insn): Delete.
11776 * config/rs6000/rs6000.opt (mno-string): New.
11777 (mstring): Replace by deprecation warning stub.
11778 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11780 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11782 * regrename.c (regrename_do_replace): If replacing the same
11783 reg multiple times, try to reuse last created gen_raw_REG.
11785 PR debug/81155
11786 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11787 main to workaround a bug in GDB.
11789 2018-01-12 Tom de Vries <tom@codesourcery.com>
11791 PR target/83737
11792 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11794 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11796 PR rtl-optimization/80481
11797 * ira-color.c (get_cap_member): New function.
11798 (allocnos_conflict_by_live_ranges_p): Use it.
11799 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11800 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11802 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11804 PR target/83628
11805 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11806 (*saddl_se_1): Ditto.
11807 (*ssubsi_1): Ditto.
11808 (*ssubl_se_1): Ditto.
11810 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11812 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11813 rather than wi::to_widest for DR_INITs.
11814 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11815 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11816 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11817 INTEGER_CSTs.
11818 (vect_analyze_group_access_1): Note that here.
11820 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11822 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11823 polynomial type sizes.
11825 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11827 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11828 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11829 (gimple_add_tmp_var): Likewise.
11831 2018-01-12 Martin Liska <mliska@suse.cz>
11833 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11834 (gimple_alloc_sizes): Likewise.
11835 (dump_gimple_statistics): Use PRIu64 in printf format.
11836 * gimple.h: Change uint64_t to int.
11838 2018-01-12 Martin Liska <mliska@suse.cz>
11840 * tree-core.h: Use uint64_t instead of int.
11841 * tree.c (tree_node_counts): Likewise.
11842 (tree_node_sizes): Likewise.
11843 (dump_tree_statistics): Use PRIu64 in printf format.
11845 2018-01-12 Martin Liska <mliska@suse.cz>
11847 * Makefile.in: As qsort_chk is implemented in vec.c, add
11848 vec.o to linkage of gencfn-macros.
11849 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11850 passing the info to record_node_allocation_statistics.
11851 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11852 and pass the info.
11853 * ggc-common.c (struct ggc_usage): Add operator== and use
11854 it in operator< and compare function.
11855 * mem-stats.h (struct mem_usage): Likewise.
11856 * vec.c (struct vec_usage): Remove operator< and compare
11857 function. Can be simply inherited.
11859 2018-01-12 Martin Jambor <mjambor@suse.cz>
11861 PR target/81616
11862 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11863 * tree-ssa-math-opts.c: Include domwalk.h.
11864 (convert_mult_to_fma_1): New function.
11865 (fma_transformation_info): New type.
11866 (fma_deferring_state): Likewise.
11867 (cancel_fma_deferring): New function.
11868 (result_of_phi): Likewise.
11869 (last_fma_candidate_feeds_initial_phi): Likewise.
11870 (convert_mult_to_fma): Added deferring logic, split actual
11871 transformation to convert_mult_to_fma_1.
11872 (math_opts_dom_walker): New type.
11873 (math_opts_dom_walker::after_dom_children): New method, body moved
11874 here from pass_optimize_widening_mul::execute, added deferring logic
11875 bits.
11876 (pass_optimize_widening_mul::execute): Moved most of code to
11877 math_opts_dom_walker::after_dom_children.
11878 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11879 * config/i386/i386.c (ix86_option_override_internal): Added
11880 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11882 2018-01-12 Richard Biener <rguenther@suse.de>
11884 PR debug/83157
11885 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11886 inline instance vars.
11888 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11890 PR target/81819
11891 * config/rx/rx.c (rx_is_restricted_memory_address):
11892 Handle SUBREG case.
11894 2018-01-12 Richard Biener <rguenther@suse.de>
11896 PR tree-optimization/80846
11897 * target.def (split_reduction): New target hook.
11898 * targhooks.c (default_split_reduction): New function.
11899 * targhooks.h (default_split_reduction): Declare.
11900 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11901 target requests first reduce vectors by combining low and high
11902 parts.
11903 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11904 (get_vectype_for_scalar_type_and_size): Export.
11905 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11906 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11907 * doc/tm.texi: Regenerate.
11908 * config/i386/i386.c (ix86_split_reduction): Implement
11909 TARGET_VECTORIZE_SPLIT_REDUCTION.
11911 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11913 PR target/83368
11914 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11915 in PIC mode except for TARGET_VXWORKS_RTP.
11916 * config/sparc/sparc.c: Include cfgrtl.h.
11917 (TARGET_INIT_PIC_REG): Define.
11918 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11919 (sparc_pic_register_p): New predicate.
11920 (sparc_legitimate_address_p): Use it.
11921 (sparc_legitimize_pic_address): Likewise.
11922 (sparc_delegitimize_address): Likewise.
11923 (sparc_mode_dependent_address_p): Likewise.
11924 (gen_load_pcrel_sym): Remove 4th parameter.
11925 (load_got_register): Adjust call to above. Remove obsolete stuff.
11926 (sparc_expand_prologue): Do not call load_got_register here.
11927 (sparc_flat_expand_prologue): Likewise.
11928 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11929 (sparc_use_pseudo_pic_reg): New function.
11930 (sparc_init_pic_reg): Likewise.
11931 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11932 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11934 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11936 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11937 Add item for branch_cost.
11939 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11941 PR rtl-optimization/83565
11942 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11943 not extend the result to a larger mode for rotate operations.
11944 (num_sign_bit_copies1): Likewise.
11946 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11948 PR target/40411
11949 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11950 -symbolic.
11951 Use values-Xc.o for -pedantic.
11952 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11954 2018-01-12 Martin Liska <mliska@suse.cz>
11956 PR ipa/83054
11957 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11958 New function.
11959 (possible_polymorphic_call_targets): Use it.
11960 (ipa_devirt): Likewise.
11962 2018-01-12 Martin Liska <mliska@suse.cz>
11964 * profile-count.h (enum profile_quality): Use 0 as invalid
11965 enum value of profile_quality.
11967 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11969 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11970 -mext-string options.
11972 2018-01-12 Richard Biener <rguenther@suse.de>
11974 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11975 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11976 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11977 Likewise.
11978 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11980 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11982 * configure.ac (--with-long-double-format): Add support for the
11983 configuration option to change the default long double format on
11984 PowerPC systems.
11985 * config.gcc (powerpc*-linux*-*): Likewise.
11986 * configure: Regenerate.
11987 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11988 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11989 used without modification.
11991 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11993 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11994 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11995 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11996 MISC_BUILTIN_SPEC_BARRIER.
11997 (rs6000_init_builtins): Likewise.
11998 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11999 enum value.
12000 (speculation_barrier): New define_insn.
12001 * doc/extend.texi: Document __builtin_speculation_barrier.
12003 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12005 PR target/83203
12006 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12007 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12008 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12009 iterators.
12010 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12011 integral modes instead of "ss" and "sd".
12012 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12013 vectors with 32-bit and 64-bit elements.
12014 (vecdupssescalarmodesuffix): New mode attribute.
12015 (vec_dup<mode>): Use it.
12017 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12019 PR target/83330
12020 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12021 frame if argument is passed on stack.
12023 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12025 PR target/82682
12026 * ree.c (combine_reaching_defs): Optimize also
12027 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12028 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12030 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12032 PR middle-end/83189
12033 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12035 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12037 PR middle-end/83718
12038 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12039 after they are computed.
12041 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12043 PR tree-optimization/83695
12044 * gimple-loop-linterchange.cc
12045 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12046 reset cached scev information after interchange.
12047 (pass_linterchange::execute): Remove call to scev_reset_htab.
12049 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12051 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12052 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12053 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12054 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12055 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12056 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12057 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12058 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12059 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12060 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12061 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12062 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12063 (V_lane_reg): Likewise.
12064 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12065 New define_expand.
12066 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12067 (vfmal_lane_low<mode>_intrinsic,
12068 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12069 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12070 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12071 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12072 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12073 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12075 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12077 * config/arm/arm-cpus.in (fp16fml): New feature.
12078 (ALL_SIMD): Add fp16fml.
12079 (armv8.2-a): Add fp16fml as an option.
12080 (armv8.3-a): Likewise.
12081 (armv8.4-a): Add fp16fml as part of fp16.
12082 * config/arm/arm.h (TARGET_FP16FML): Define.
12083 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12084 when appropriate.
12085 * config/arm/arm-modes.def (V2HF): Define.
12086 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12087 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12088 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12089 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12090 vfmsl_low, vfmsl_high): New set of builtins.
12091 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12092 (vfml_op): New code attribute.
12093 (VFMLHALVES): New int iterator.
12094 (VFML, VFMLSEL): New mode attributes.
12095 (V_reg): Define mapping for V2HF.
12096 (V_hi, V_lo): New mode attributes.
12097 (VF_constraint): Likewise.
12098 (vfml_half, vfml_half_selector): New int attributes.
12099 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12100 define_expand.
12101 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12102 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12103 New define_insn.
12104 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12105 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12106 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12107 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12108 documentation.
12109 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12110 Document new effective target and option set.
12112 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12114 * config/arm/arm-cpus.in (armv8_4): New feature.
12115 (ARMv8_4a): New fgroup.
12116 (armv8.4-a): New arch.
12117 * config/arm/arm-tables.opt: Regenerate.
12118 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12119 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12120 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12121 Add matching rules for -march=armv8.4-a and extensions.
12122 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12124 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12126 PR target/81821
12127 * config/rx/rx.md (BW): New mode attribute.
12128 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12130 2018-01-11 Richard Biener <rguenther@suse.de>
12132 PR tree-optimization/83435
12133 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12134 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12135 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12137 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12138 Alan Hayward <alan.hayward@arm.com>
12139 David Sherwood <david.sherwood@arm.com>
12141 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12142 field.
12143 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12144 (aarch64_print_address_internal): Use it to check for a zero offset.
12146 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12147 Alan Hayward <alan.hayward@arm.com>
12148 David Sherwood <david.sherwood@arm.com>
12150 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12151 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12152 Return a poly_int64 rather than a HOST_WIDE_INT.
12153 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12154 rather than a HOST_WIDE_INT.
12155 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12156 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12157 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12158 final_offset from HOST_WIDE_INT to poly_int64.
12159 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12160 to_constant when getting the number of units in an Advanced SIMD
12161 mode.
12162 (aarch64_builtin_vectorized_function): Check for a constant number
12163 of units.
12164 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12165 GET_MODE_SIZE.
12166 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12167 attribute instead of GET_MODE_NUNITS.
12168 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12169 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12170 GET_MODE_SIZE for fixed-size registers.
12171 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12172 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12173 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12174 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12175 (aarch64_print_operand, aarch64_print_address_internal)
12176 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12177 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12178 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12179 Handle polynomial GET_MODE_SIZE.
12180 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12181 wider than SImode without modification.
12182 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12183 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12184 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12185 passing and returning SVE modes.
12186 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12187 rather than GEN_INT.
12188 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12189 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12190 (aarch64_allocate_and_probe_stack_space): Likewise.
12191 (aarch64_layout_frame): Cope with polynomial offsets.
12192 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12193 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12194 polynomial offsets.
12195 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12196 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12197 poly_int64 rather than a HOST_WIDE_INT.
12198 (aarch64_get_separate_components, aarch64_process_components)
12199 (aarch64_expand_prologue, aarch64_expand_epilogue)
12200 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12201 (aarch64_anchor_offset): New function, split out from...
12202 (aarch64_legitimize_address): ...here.
12203 (aarch64_builtin_vectorization_cost): Handle polynomial
12204 TYPE_VECTOR_SUBPARTS.
12205 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12206 GET_MODE_NUNITS.
12207 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12208 number of elements from the PARALLEL rather than the mode.
12209 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12210 rather than GET_MODE_BITSIZE.
12211 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12212 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12213 (aarch64_expand_vec_perm_const_1): Handle polynomial
12214 d->perm.length () and d->perm elements.
12215 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12216 Apply to_constant to d->perm elements.
12217 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12218 polynomial CONST_VECTOR_NUNITS.
12219 (aarch64_move_pointer): Take amount as a poly_int64 rather
12220 than an int.
12221 (aarch64_progress_pointer): Avoid temporary variable.
12222 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12223 the mode attribute instead of GET_MODE.
12225 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12226 Alan Hayward <alan.hayward@arm.com>
12227 David Sherwood <david.sherwood@arm.com>
12229 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12230 x exists before using it.
12231 (aarch64_add_constant_internal): Rename to...
12232 (aarch64_add_offset_1): ...this. Replace regnum with separate
12233 src and dest rtxes. Handle the case in which they're different,
12234 including when the offset is zero. Replace scratchreg with an rtx.
12235 Use 2 additions if there is no spare register into which we can
12236 move a 16-bit constant.
12237 (aarch64_add_constant): Delete.
12238 (aarch64_add_offset): Replace reg with separate src and dest
12239 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12240 Use aarch64_add_offset_1.
12241 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12242 an rtx rather than an int. Take the delta as a poly_int64
12243 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12244 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12245 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12246 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12247 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12248 and aarch64_add_sp.
12249 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12250 aarch64_add_constant.
12252 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12254 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12255 Use scalar_float_mode.
12257 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12259 * config/aarch64/aarch64-simd.md
12260 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12261 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12262 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12263 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12264 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12265 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12266 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12267 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12268 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12269 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12271 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12273 PR target/83514
12274 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12275 targ_options->x_arm_arch_string is non NULL.
12277 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12279 * config/aarch64/aarch64.h
12280 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12282 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12284 PR target/82096
12285 * expmed.c (emit_store_flag_force): Swap if const op0
12286 and change VOIDmode to mode of op0.
12288 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12290 PR rtl-optimization/83761
12291 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12292 than bytes to mode_for_size.
12294 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12296 PR middle-end/83189
12297 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12298 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12299 profile.
12301 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12303 PR middle-end/83575
12304 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12305 when in layout mode.
12306 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12307 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12308 partition fixup.
12310 2018-01-10 Michael Collison <michael.collison@arm.com>
12312 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12313 * config/aarch64/aarch64-option-extension.def: Add
12314 AARCH64_OPT_EXTENSION of 'fp16fml'.
12315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12316 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12317 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12318 * config/aarch64/constraints.md (Ui7): New constraint.
12319 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12320 (VFMLA_SEL_W): Ditto.
12321 (f16quad): Ditto.
12322 (f16mac1): Ditto.
12323 (VFMLA16_LOW): New int iterator.
12324 (VFMLA16_HIGH): Ditto.
12325 (UNSPEC_FMLAL): New unspec.
12326 (UNSPEC_FMLSL): Ditto.
12327 (UNSPEC_FMLAL2): Ditto.
12328 (UNSPEC_FMLSL2): Ditto.
12329 (f16mac): New code attribute.
12330 * config/aarch64/aarch64-simd-builtins.def
12331 (aarch64_fmlal_lowv2sf): Ditto.
12332 (aarch64_fmlsl_lowv2sf): Ditto.
12333 (aarch64_fmlalq_lowv4sf): Ditto.
12334 (aarch64_fmlslq_lowv4sf): Ditto.
12335 (aarch64_fmlal_highv2sf): Ditto.
12336 (aarch64_fmlsl_highv2sf): Ditto.
12337 (aarch64_fmlalq_highv4sf): Ditto.
12338 (aarch64_fmlslq_highv4sf): Ditto.
12339 (aarch64_fmlal_lane_lowv2sf): Ditto.
12340 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12341 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12342 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12343 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12344 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12345 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12346 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12347 (aarch64_fmlal_lane_highv2sf): Ditto.
12348 (aarch64_fmlsl_lane_highv2sf): Ditto.
12349 (aarch64_fmlal_laneq_highv2sf): Ditto.
12350 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12351 (aarch64_fmlalq_lane_highv4sf): Ditto.
12352 (aarch64_fmlsl_lane_highv4sf): Ditto.
12353 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12354 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12355 * config/aarch64/aarch64-simd.md:
12356 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12357 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12358 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12359 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12360 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12361 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12362 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12363 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12364 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12365 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12366 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12367 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12368 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12369 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12370 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12371 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12372 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12373 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12374 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12375 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12376 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12377 (vfmlsl_low_u32): Ditto.
12378 (vfmlalq_low_u32): Ditto.
12379 (vfmlslq_low_u32): Ditto.
12380 (vfmlal_high_u32): Ditto.
12381 (vfmlsl_high_u32): Ditto.
12382 (vfmlalq_high_u32): Ditto.
12383 (vfmlslq_high_u32): Ditto.
12384 (vfmlal_lane_low_u32): Ditto.
12385 (vfmlsl_lane_low_u32): Ditto.
12386 (vfmlal_laneq_low_u32): Ditto.
12387 (vfmlsl_laneq_low_u32): Ditto.
12388 (vfmlalq_lane_low_u32): Ditto.
12389 (vfmlslq_lane_low_u32): Ditto.
12390 (vfmlalq_laneq_low_u32): Ditto.
12391 (vfmlslq_laneq_low_u32): Ditto.
12392 (vfmlal_lane_high_u32): Ditto.
12393 (vfmlsl_lane_high_u32): Ditto.
12394 (vfmlal_laneq_high_u32): Ditto.
12395 (vfmlsl_laneq_high_u32): Ditto.
12396 (vfmlalq_lane_high_u32): Ditto.
12397 (vfmlslq_lane_high_u32): Ditto.
12398 (vfmlalq_laneq_high_u32): Ditto.
12399 (vfmlslq_laneq_high_u32): Ditto.
12400 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12401 (AARCH64_FL_FOR_ARCH8_4): New.
12402 (AARCH64_ISA_F16FML): New ISA flag.
12403 (TARGET_F16FML): New feature flag for fp16fml.
12404 (doc/invoke.texi): Document new fp16fml option.
12406 2018-01-10 Michael Collison <michael.collison@arm.com>
12408 * config/aarch64/aarch64-builtins.c:
12409 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12410 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12411 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12412 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12413 (AARCH64_ISA_SHA3): New ISA flag.
12414 (TARGET_SHA3): New feature flag for sha3.
12415 * config/aarch64/iterators.md (sha512_op): New int attribute.
12416 (CRYPTO_SHA512): New int iterator.
12417 (UNSPEC_SHA512H): New unspec.
12418 (UNSPEC_SHA512H2): Ditto.
12419 (UNSPEC_SHA512SU0): Ditto.
12420 (UNSPEC_SHA512SU1): Ditto.
12421 * config/aarch64/aarch64-simd-builtins.def
12422 (aarch64_crypto_sha512hqv2di): New builtin.
12423 (aarch64_crypto_sha512h2qv2di): Ditto.
12424 (aarch64_crypto_sha512su0qv2di): Ditto.
12425 (aarch64_crypto_sha512su1qv2di): Ditto.
12426 (aarch64_eor3qv8hi): Ditto.
12427 (aarch64_rax1qv2di): Ditto.
12428 (aarch64_xarqv2di): Ditto.
12429 (aarch64_bcaxqv8hi): Ditto.
12430 * config/aarch64/aarch64-simd.md:
12431 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12432 (aarch64_crypto_sha512su0qv2di): Ditto.
12433 (aarch64_crypto_sha512su1qv2di): Ditto.
12434 (aarch64_eor3qv8hi): Ditto.
12435 (aarch64_rax1qv2di): Ditto.
12436 (aarch64_xarqv2di): Ditto.
12437 (aarch64_bcaxqv8hi): Ditto.
12438 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12439 (vsha512h2q_u64): Ditto.
12440 (vsha512su0q_u64): Ditto.
12441 (vsha512su1q_u64): Ditto.
12442 (veor3q_u16): Ditto.
12443 (vrax1q_u64): Ditto.
12444 (vxarq_u64): Ditto.
12445 (vbcaxq_u16): Ditto.
12446 * config/arm/types.md (crypto_sha512): New type attribute.
12447 (crypto_sha3): Ditto.
12448 (doc/invoke.texi): Document new sha3 option.
12450 2018-01-10 Michael Collison <michael.collison@arm.com>
12452 * config/aarch64/aarch64-builtins.c:
12453 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12454 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12455 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12456 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12457 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12458 (AARCH64_ISA_SM4): New ISA flag.
12459 (TARGET_SM4): New feature flag for sm4.
12460 * config/aarch64/aarch64-simd-builtins.def
12461 (aarch64_sm3ss1qv4si): Ditto.
12462 (aarch64_sm3tt1aq4si): Ditto.
12463 (aarch64_sm3tt1bq4si): Ditto.
12464 (aarch64_sm3tt2aq4si): Ditto.
12465 (aarch64_sm3tt2bq4si): Ditto.
12466 (aarch64_sm3partw1qv4si): Ditto.
12467 (aarch64_sm3partw2qv4si): Ditto.
12468 (aarch64_sm4eqv4si): Ditto.
12469 (aarch64_sm4ekeyqv4si): Ditto.
12470 * config/aarch64/aarch64-simd.md:
12471 (aarch64_sm3ss1qv4si): Ditto.
12472 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12473 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12474 (aarch64_sm4eqv4si): Ditto.
12475 (aarch64_sm4ekeyqv4si): Ditto.
12476 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12477 (sm3part_op): Ditto.
12478 (CRYPTO_SM3TT): Ditto.
12479 (CRYPTO_SM3PART): Ditto.
12480 (UNSPEC_SM3SS1): New unspec.
12481 (UNSPEC_SM3TT1A): Ditto.
12482 (UNSPEC_SM3TT1B): Ditto.
12483 (UNSPEC_SM3TT2A): Ditto.
12484 (UNSPEC_SM3TT2B): Ditto.
12485 (UNSPEC_SM3PARTW1): Ditto.
12486 (UNSPEC_SM3PARTW2): Ditto.
12487 (UNSPEC_SM4E): Ditto.
12488 (UNSPEC_SM4EKEY): Ditto.
12489 * config/aarch64/constraints.md (Ui2): New constraint.
12490 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12491 * config/arm/types.md (crypto_sm3): New type attribute.
12492 (crypto_sm4): Ditto.
12493 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12494 (vsm3tt1aq_u32): Ditto.
12495 (vsm3tt1bq_u32): Ditto.
12496 (vsm3tt2aq_u32): Ditto.
12497 (vsm3tt2bq_u32): Ditto.
12498 (vsm3partw1q_u32): Ditto.
12499 (vsm3partw2q_u32): Ditto.
12500 (vsm4eq_u32): Ditto.
12501 (vsm4ekeyq_u32): Ditto.
12502 (doc/invoke.texi): Document new sm4 option.
12504 2018-01-10 Michael Collison <michael.collison@arm.com>
12506 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12507 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12508 (AARCH64_FL_FOR_ARCH8_4): New.
12509 (AARCH64_FL_V8_4): New flag.
12510 (doc/invoke.texi): Document new armv8.4-a option.
12512 2018-01-10 Michael Collison <michael.collison@arm.com>
12514 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12515 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12516 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12517 * config/aarch64/aarch64-option-extension.def: Add
12518 AARCH64_OPT_EXTENSION of 'sha2'.
12519 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12520 (crypto): Disable sha2 and aes if crypto disabled.
12521 (crypto): Enable aes and sha2 if enabled.
12522 (simd): Disable sha2 and aes if simd disabled.
12523 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12524 New flags.
12525 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12526 (TARGET_SHA2): New feature flag for sha2.
12527 (TARGET_AES): New feature flag for aes.
12528 * config/aarch64/aarch64-simd.md:
12529 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12530 conditional on TARGET_AES.
12531 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12532 (aarch64_crypto_sha1hsi): Make pattern conditional
12533 on TARGET_SHA2.
12534 (aarch64_crypto_sha1hv4si): Ditto.
12535 (aarch64_be_crypto_sha1hv4si): Ditto.
12536 (aarch64_crypto_sha1su1v4si): Ditto.
12537 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12538 (aarch64_crypto_sha1su0v4si): Ditto.
12539 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12540 (aarch64_crypto_sha256su0v4si): Ditto.
12541 (aarch64_crypto_sha256su1v4si): Ditto.
12542 (doc/invoke.texi): Document new aes and sha2 options.
12544 2018-01-10 Martin Sebor <msebor@redhat.com>
12546 PR tree-optimization/83781
12547 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12548 as string arrays.
12550 2018-01-11 Martin Sebor <msebor@gmail.com>
12551 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12553 PR tree-optimization/83501
12554 PR tree-optimization/81703
12556 * tree-ssa-strlen.c (get_string_cst): Rename...
12557 (get_string_len): ...to this. Handle global constants.
12558 (handle_char_store): Adjust.
12560 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12561 Jim Wilson <jimw@sifive.com>
12563 * config/riscv/riscv-protos.h (riscv_output_return): New.
12564 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12565 (riscv_attribute_table, riscv_output_return),
12566 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12567 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12568 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12569 function.
12570 (riscv_expand_prologue): Add early return for naked function.
12571 (riscv_expand_epilogue): Likewise.
12572 (riscv_function_ok_for_sibcall): Return false for naked function.
12573 (riscv_set_current_function): New.
12574 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12575 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12576 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12577 * doc/extend.texi (RISC-V Function Attributes): New.
12579 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12581 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12582 check for 128-bit long double before checking TCmode.
12583 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12584 128-bit long doubles before checking TFmode or TCmode.
12585 (FLOAT128_IBM_P): Likewise.
12587 2018-01-10 Martin Sebor <msebor@redhat.com>
12589 PR tree-optimization/83671
12590 * builtins.c (c_strlen): Unconditionally return zero for the empty
12591 string.
12592 Use -Warray-bounds for warnings.
12593 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12594 for non-constant array indices with COMPONENT_REF, arrays of
12595 arrays, and pointers to arrays.
12596 (gimple_fold_builtin_strlen): Determine and set length range for
12597 non-constant character arrays.
12599 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12601 PR middle-end/81897
12602 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12603 empty blocks.
12605 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12607 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12609 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12611 PR target/83399
12612 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12613 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12614 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12615 indexed_or_indirect_operand predicate.
12616 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12617 (*vsx_le_perm_load_v8hi): Likewise.
12618 (*vsx_le_perm_load_v16qi): Likewise.
12619 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12620 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12621 (*vsx_le_perm_store_v8hi): Likewise.
12622 (*vsx_le_perm_store_v16qi): Likewise.
12623 (eight unnamed splitters): Likewise.
12625 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12627 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12628 * config/rs6000/emmintrin.h: Likewise.
12629 * config/rs6000/mmintrin.h: Likewise.
12630 * config/rs6000/xmmintrin.h: Likewise.
12632 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12634 PR c++/43486
12635 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12636 "public_flag".
12637 * tree.c (tree_nop_conversion): Return true for location wrapper
12638 nodes.
12639 (maybe_wrap_with_location): New function.
12640 (selftest::check_strip_nops): New function.
12641 (selftest::test_location_wrappers): New function.
12642 (selftest::tree_c_tests): Call it.
12643 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12644 (maybe_wrap_with_location): New decl.
12645 (EXPR_LOCATION_WRAPPER_P): New macro.
12646 (location_wrapper_p): New inline function.
12647 (tree_strip_any_location_wrapper): New inline function.
12649 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12651 PR target/83735
12652 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12653 stack_realign_offset for the largest alignment of stack slot
12654 actually used.
12655 (ix86_find_max_used_stack_alignment): New function.
12656 (ix86_finalize_stack_frame_flags): Use it. Set
12657 max_used_stack_alignment if we don't realign stack.
12658 * config/i386/i386.h (machine_function): Add
12659 max_used_stack_alignment.
12661 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12663 * config/arm/arm.opt (-mbranch-cost): New option.
12664 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12665 account.
12667 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12669 PR target/83629
12670 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12671 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12673 2018-01-10 Richard Biener <rguenther@suse.de>
12675 PR debug/83765
12676 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12677 early out so it also covers the case where we have a non-NULL
12678 origin.
12680 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12682 PR tree-optimization/83753
12683 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12684 for non-strided grouped accesses if the number of elements is 1.
12686 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12688 PR target/81616
12689 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12690 * i386.h (TARGET_USE_GATHER): Define.
12691 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12693 2018-01-10 Martin Liska <mliska@suse.cz>
12695 PR bootstrap/82831
12696 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12697 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12698 partitioning.
12699 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12700 CLEANUP_NO_PARTITIONING is not set.
12702 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12704 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12705 for vectors, as a partial revert of r254296.
12706 * rtl.h (const_vec_p): Delete.
12707 (const_vec_duplicate_p): Don't test for vector CONSTs.
12708 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12709 * expmed.c (make_tree): Likewise.
12711 Revert:
12712 * common.md (E, F): Use CONSTANT_P instead of checking for
12713 CONST_VECTOR.
12714 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12715 checking for CONST_VECTOR.
12717 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12719 PR middle-end/83575
12720 * predict.c (force_edge_cold): Handle in more sane way edges
12721 with no prediction.
12723 2018-01-09 Carl Love <cel@us.ibm.com>
12725 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12726 V4SI, V4SF types.
12727 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12728 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12729 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12730 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12731 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12732 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12733 * config/rs6000/rs6000-protos.h: Add extern defition for
12734 rs6000_generate_float2_double_code.
12735 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12736 function.
12737 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12738 (float2_v2df): Add define_expand.
12740 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12742 PR target/83628
12743 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12744 op_mode in the force_to_mode call.
12746 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12748 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12749 instead of checking each element individually.
12750 (aarch64_evpc_uzp): Likewise.
12751 (aarch64_evpc_zip): Likewise.
12752 (aarch64_evpc_ext): Likewise.
12753 (aarch64_evpc_rev): Likewise.
12754 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12755 instead of checking each element individually. Return true without
12756 generating rtl if
12757 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12758 whether all selected elements come from the same input, instead of
12759 checking each element individually. Remove calls to gen_rtx_REG,
12760 start_sequence and end_sequence and instead assert that no rtl is
12761 generated.
12763 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12765 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12766 order of HIGH and CONST checks.
12768 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12770 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12771 if the destination isn't an SSA_NAME.
12773 2018-01-09 Richard Biener <rguenther@suse.de>
12775 PR tree-optimization/83668
12776 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12777 move prologue...
12778 (canonicalize_loop_form): ... here, renamed from ...
12779 (canonicalize_loop_closed_ssa_form): ... this and amended to
12780 swap successor edges for loop exit blocks to make us use
12781 the RPO order we need for initial schedule generation.
12783 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12785 PR tree-optimization/64811
12786 * match.pd: When optimizing comparisons with Inf, avoid
12787 introducing or losing exceptions from comparisons with NaN.
12789 2018-01-09 Martin Liska <mliska@suse.cz>
12791 PR sanitizer/82517
12792 * asan.c (shadow_mem_size): Add gcc_assert.
12794 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12796 Don't save registers in main().
12798 PR target/83738
12799 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12800 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12801 * config/avr/avr.c (avr_set_current_function): Don't error if
12802 naked, OS_task or OS_main are specified at the same time.
12803 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12804 OS_main.
12805 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12806 attribute.
12807 * common/config/avr/avr-common.c (avr_option_optimization_table):
12808 Switch on -mmain-is-OS_task for optimizing compilations.
12810 2018-01-09 Richard Biener <rguenther@suse.de>
12812 PR tree-optimization/83572
12813 * graphite.c: Include cfganal.h.
12814 (graphite_transform_loops): Connect infinite loops to exit
12815 and remove fake edges at the end.
12817 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12819 * ipa-inline.c (edge_badness): Revert accidental checkin.
12821 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12823 PR ipa/80763
12824 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12825 symbols; not inline clones.
12827 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12829 PR target/83507
12830 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12831 hard registers. Formatting fixes.
12833 PR preprocessor/83722
12834 * gcc.c (try_generate_repro): Pass
12835 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12836 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12837 do_report_bug.
12839 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12840 Kito Cheng <kito.cheng@gmail.com>
12842 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12843 (riscv_leaf_function_p): Delete.
12844 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12846 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12848 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12849 function.
12850 (do_ifelse): New function.
12851 (do_isel): New function.
12852 (do_sub3): New function.
12853 (do_add3): New function.
12854 (do_load_mask_compare): New function.
12855 (do_overlap_load_compare): New function.
12856 (expand_compare_loop): New function.
12857 (expand_block_compare): Call expand_compare_loop() when appropriate.
12858 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12859 option description.
12860 (-mblock-compare-inline-loop-limit): New option.
12862 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12864 PR target/83677
12865 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12866 Reverse order of second and third operands in first alternative.
12867 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12868 of first and second elements in UNSPEC_VPERMR vector.
12869 (altivec_expand_vec_perm_le): Likewise.
12871 2018-01-08 Jeff Law <law@redhat.com>
12873 PR rtl-optimizatin/81308
12874 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12875 (process_switch): If group_case_labels makes a change, then set
12876 cfg_altered.
12877 (pass_convert_switch::execute): If a switch is converted, then
12878 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12880 PR rtl-optimization/81308
12881 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12882 splitting insns.
12884 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12886 PR target/83663 - Revert r255946
12887 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12888 generation for cases where splatting a value is not useful.
12889 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12890 across a vec_duplicate and a paradoxical subreg forming a vector
12891 mode to a vec_concat.
12893 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12895 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12896 -march=armv8.3-a variants.
12897 * config/arm/t-multilib: Likewise.
12898 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12900 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12902 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12903 to generate rtl.
12904 (cceq_ior_compare_complement): Give it a name so I can use it, and
12905 change boolean_or_operator predicate to boolean_operator so it can
12906 be used to generate a crand.
12907 (eqne): New code iterator.
12908 (bd/bd_neg): New code_attrs.
12909 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12910 a single define_insn.
12911 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12912 decrement (bdnzt/bdnzf/bdzt/bdzf).
12913 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12914 with the new names of the branch decrement patterns, and added the
12915 names of the branch decrement conditional patterns.
12917 2018-01-08 Richard Biener <rguenther@suse.de>
12919 PR tree-optimization/83563
12920 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12921 cache.
12923 2018-01-08 Richard Biener <rguenther@suse.de>
12925 PR middle-end/83713
12926 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12928 2018-01-08 Richard Biener <rguenther@suse.de>
12930 PR tree-optimization/83685
12931 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12932 references to abnormals.
12934 2018-01-08 Richard Biener <rguenther@suse.de>
12936 PR lto/83719
12937 * dwarf2out.c (output_indirect_strings): Handle empty
12938 skeleton_debug_str_hash.
12939 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12941 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12943 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12944 (emit_store_direct): Likewise.
12945 (arc_trampoline_adjust_address): Likewise.
12946 (arc_asm_trampoline_template): New function.
12947 (arc_initialize_trampoline): Use asm_trampoline_template.
12948 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12949 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12950 * config/arc/arc.md (flush_icache): Delete pattern.
12952 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12954 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12955 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12956 munaligned-access.
12958 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12960 PR target/83681
12961 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12962 by not USED_FOR_TARGET.
12963 (make_pass_resolve_sw_modes): Likewise.
12965 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12967 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12968 USED_FOR_TARGET.
12970 2018-01-08 Richard Biener <rguenther@suse.de>
12972 PR middle-end/83580
12973 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12975 2018-01-08 Richard Biener <rguenther@suse.de>
12977 PR middle-end/83517
12978 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12980 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12982 PR middle-end/81897
12983 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12984 basic blocks with a small number of successors.
12985 (convert_control_dep_chain_into_preds): Improve handling of
12986 forwarder blocks.
12987 (dump_predicates): Split apart into...
12988 (dump_pred_chain): ...here...
12989 (dump_pred_info): ...and here.
12990 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12991 (can_chain_union_be_invalidated_p): Improve check for invalidation
12992 of paths.
12993 (uninit_uses_cannot_happen): Avoid unnecessary if
12994 convert_control_dep_chain_into_preds yielded nothing.
12996 2018-01-06 Martin Sebor <msebor@redhat.com>
12998 PR tree-optimization/83640
12999 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13000 subtracting negative offset from size.
13001 (builtin_access::overlap): Adjust offset bounds of the access to fall
13002 within the size of the object if possible.
13004 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13006 PR rtl-optimization/83699
13007 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13008 extract_bit_field_as_subreg to cases in which the extracted
13009 value is also a vector.
13011 * lra-constraints.c (process_alt_operands): Test for the equivalence
13012 substitutions when detecting a possible reload cycle.
13014 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13016 PR debug/83480
13017 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13018 by default if flag_selective_schedling{,2}. Formatting fixes.
13020 PR rtl-optimization/83682
13021 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13022 if it has non-VECTOR_MODE element mode.
13023 (vec_duplicate_p): Likewise.
13025 PR middle-end/83694
13026 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13027 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13029 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13031 PR target/83604
13032 * config/i386/i386-builtin.def
13033 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13034 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13035 Require also OPTION_MASK_ISA_AVX512F in addition to
13036 OPTION_MASK_ISA_GFNI.
13037 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13038 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13039 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13040 to OPTION_MASK_ISA_GFNI.
13041 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13042 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13043 OPTION_MASK_ISA_AVX512BW.
13044 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13045 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13046 addition to OPTION_MASK_ISA_GFNI.
13047 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13048 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13049 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13050 to OPTION_MASK_ISA_GFNI.
13051 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13052 a requirement for all ISAs rather than any of them with a few
13053 exceptions.
13054 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13055 processing.
13056 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13057 bitmasks to be enabled with 3 exceptions, instead of requiring any
13058 enabled ISA with lots of exceptions.
13059 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13060 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13061 Change avx512bw in isa attribute to avx512f.
13062 * config/i386/sgxintrin.h: Add license boilerplate.
13063 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13064 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13065 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13066 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13067 defined.
13068 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13069 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13070 temporarily sse2 rather than sse if not enabled already.
13072 PR target/83604
13073 * config/i386/sse.md (VI248_VLBW): Rename to ...
13074 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13075 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13076 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13077 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13078 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13079 mode iterator instead of VI248_VLBW.
13081 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13083 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13084 (record_modified): Skip clobbers; add debug output.
13085 (param_change_prob): Use sreal frequencies.
13087 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13089 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13090 punt for user-aligned variables.
13092 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13094 * tree-chrec.c (chrec_contains_symbols): Return true for
13095 POLY_INT_CST.
13097 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13099 PR target/82439
13100 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13101 of (x|y) == x for BICS pattern.
13103 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13105 PR tree-optimization/83605
13106 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13107 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13108 can throw.
13110 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13112 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13113 * config/epiphany/rtems.h: New file.
13115 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13116 Uros Bizjak <ubizjak@gmail.com>
13118 PR target/83554
13119 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13120 QIreg_operand instead of register_operand predicate.
13121 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13122 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13123 comments instead of -fmitigate[-_]rop.
13125 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13127 PR bootstrap/81926
13128 * cgraphunit.c (symbol_table::compile): Switch to text_section
13129 before calling assembly_start debug hook.
13130 * run-rtl-passes.c (run_rtl_passes): Likewise.
13131 Include output.h.
13133 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13135 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13136 range_int_cst_p rather than !symbolic_range_p before calling
13137 extract_range_from_multiplicative_op_1.
13139 2018-01-04 Jeff Law <law@redhat.com>
13141 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13142 redundant test in assertion.
13144 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13146 * doc/rtl.texi: Document machine_mode wrapper classes.
13148 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13150 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13151 using tree_to_uhwi.
13153 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13155 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13156 the VEC_PERM_EXPR fold to fail.
13158 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13160 PR debug/83585
13161 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13162 to switched_sections.
13164 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13166 PR target/83680
13167 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13168 test for d.testing.
13170 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13172 PR target/83387
13173 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13174 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13176 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13178 PR debug/83666
13179 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13180 is BLKmode and bitpos not zero or mode change is needed.
13182 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13184 PR target/83675
13185 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13186 TARGET_VIS2.
13188 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13190 PR target/83628
13191 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13192 instead of MULT rtx. Update all corresponding splitters.
13193 (*saddl_se): Ditto.
13194 (*ssub<modesuffix>): Ditto.
13195 (*ssubl_se): Ditto.
13196 (*cmp_sadd_di): Update split patterns.
13197 (*cmp_sadd_si): Ditto.
13198 (*cmp_sadd_sidi): Ditto.
13199 (*cmp_ssub_di): Ditto.
13200 (*cmp_ssub_si): Ditto.
13201 (*cmp_ssub_sidi): Ditto.
13202 * config/alpha/predicates.md (const23_operand): New predicate.
13203 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13204 Look for ASHIFT, not MULT inner operand.
13205 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13207 2018-01-04 Martin Liska <mliska@suse.cz>
13209 PR gcov-profile/83669
13210 * gcov.c (output_intermediate_file): Add version to intermediate
13211 gcov file.
13212 * doc/gcov.texi: Document new field 'version' in intermediate
13213 file format. Fix location of '-k' option of gcov command.
13215 2018-01-04 Martin Liska <mliska@suse.cz>
13217 PR ipa/82352
13218 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13220 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13222 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13224 2018-01-03 Martin Sebor <msebor@redhat.com>
13226 PR tree-optimization/83655
13227 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13228 checking calls with invalid arguments.
13230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13232 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13233 (vectorizable_mask_load_store): Delete.
13234 (vectorizable_call): Return false for masked loads and stores.
13235 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13236 instead of gimple_assign_rhs1.
13237 (vectorizable_load): Handle IFN_MASK_LOAD.
13238 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13242 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13243 split out from..,
13244 (vectorizable_mask_load_store): ...here.
13245 (vectorizable_load): ...and here.
13247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13249 * tree-vect-stmts.c (vect_build_all_ones_mask)
13250 (vect_build_zero_merge_argument): New functions, split out from...
13251 (vectorizable_load): ...here.
13253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13255 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13256 split out from...
13257 (vectorizable_mask_load_store): ...here.
13258 (vectorizable_store): ...and here.
13260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13262 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13263 split out from...
13264 (vectorizable_mask_load_store): ...here.
13266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13268 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13269 (vect_model_store_cost): Take a vec_load_store_type instead of a
13270 vect_def_type.
13271 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13272 (vect_model_store_cost): Take a vec_load_store_type instead of a
13273 vect_def_type.
13274 (vectorizable_mask_load_store): Update accordingly.
13275 (vectorizable_store): Likewise.
13276 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13280 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13281 IFN_MASK_LOAD calls here rather than...
13282 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13285 Alan Hayward <alan.hayward@arm.com>
13286 David Sherwood <david.sherwood@arm.com>
13288 * expmed.c (extract_bit_field_1): For vector extracts,
13289 fall back to extract_bit_field_as_subreg if vec_extract
13290 isn't available.
13292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13293 Alan Hayward <alan.hayward@arm.com>
13294 David Sherwood <david.sherwood@arm.com>
13296 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13297 they are variable or constant sized.
13298 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13299 slots for constant-sized data.
13301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13302 Alan Hayward <alan.hayward@arm.com>
13303 David Sherwood <david.sherwood@arm.com>
13305 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13306 handling COND_EXPRs with boolean comparisons, try to find a better
13307 basis for the mask type than the boolean itself.
13309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13311 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13312 is calculated and how it can be overridden.
13313 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13314 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13315 if defined.
13316 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13317 if nonzero.
13319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13320 Alan Hayward <alan.hayward@arm.com>
13321 David Sherwood <david.sherwood@arm.com>
13323 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13324 Remove the mode argument.
13325 (aarch64_simd_valid_immediate): Remove the mode and inverse
13326 arguments.
13327 * config/aarch64/iterators.md (bitsize): New iterator.
13328 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13329 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13330 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13331 aarch64_simd_valid_immediate.
13332 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13333 (aarch64_reg_or_bic_imm): Likewise.
13334 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13335 with an insn_type enum and msl with a modifier_type enum.
13336 Replace element_width with a scalar_mode. Change the shift
13337 to unsigned int. Add constructors for scalar_float_mode and
13338 scalar_int_mode elements.
13339 (aarch64_vect_float_const_representable_p): Delete.
13340 (aarch64_can_const_movi_rtx_p)
13341 (aarch64_simd_scalar_immediate_valid_for_move)
13342 (aarch64_simd_make_constant): Update call to
13343 aarch64_simd_valid_immediate.
13344 (aarch64_advsimd_valid_immediate_hs): New function.
13345 (aarch64_advsimd_valid_immediate): Likewise.
13346 (aarch64_simd_valid_immediate): Remove mode and inverse
13347 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13348 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13349 and aarch64_float_const_representable_p on the result.
13350 (aarch64_output_simd_mov_immediate): Remove mode argument.
13351 Update call to aarch64_simd_valid_immediate and use of
13352 simd_immediate_info.
13353 (aarch64_output_scalar_simd_mov_immediate): Update call
13354 accordingly.
13356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13357 Alan Hayward <alan.hayward@arm.com>
13358 David Sherwood <david.sherwood@arm.com>
13360 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13361 (mode_nunits): Likewise CONST_MODE_NUNITS.
13362 * machmode.def (ADJUST_NUNITS): Document.
13363 * genmodes.c (mode_data::need_nunits_adj): New field.
13364 (blank_mode): Update accordingly.
13365 (adj_nunits): New variable.
13366 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13367 parameter.
13368 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13369 listed in adj_nunits.
13370 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13371 listed in adj_nunits. Don't emit case statements for such modes.
13372 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13373 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13374 nothing if adj_nunits is nonnull.
13375 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13376 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13377 (emit_mode_fbit): Update use of print_maybe_const_decl.
13378 (emit_move_size): Likewise. Treat the array as non-const
13379 if adj_nunits.
13380 (emit_mode_adjustments): Handle adj_nunits.
13382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13384 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13385 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13386 (VECTOR_MODES): Use it.
13387 (make_vector_modes): Take the prefix as an argument.
13389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13390 Alan Hayward <alan.hayward@arm.com>
13391 David Sherwood <david.sherwood@arm.com>
13393 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13394 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13395 for MODE_VECTOR_BOOL.
13396 * machmode.def (VECTOR_BOOL_MODE): Document.
13397 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13398 (make_vector_bool_mode): New function.
13399 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13400 MODE_VECTOR_BOOL.
13401 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13402 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13403 Likewise.
13404 * stor-layout.c (int_mode_for_mode): Likewise.
13405 * tree.c (build_vector_type_for_mode): Likewise.
13406 * varasm.c (output_constant_pool_2): Likewise.
13407 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13408 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13409 for MODE_VECTOR_BOOL.
13410 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13411 of mode class checks.
13412 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13413 instead of a list of mode class checks.
13414 (expand_vector_scalar_condition): Likewise.
13415 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13418 Alan Hayward <alan.hayward@arm.com>
13419 David Sherwood <david.sherwood@arm.com>
13421 * machmode.h (mode_size): Change from unsigned short to
13422 poly_uint16_pod.
13423 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13424 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13425 or if measurement_type is not polynomial.
13426 (fixed_size_mode::includes_p): Check for constant-sized modes.
13427 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13428 return a poly_uint16 rather than an unsigned short.
13429 (emit_mode_size): Change the type of mode_size from unsigned short
13430 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13431 (emit_mode_adjustments): Cope with polynomial vector sizes.
13432 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13433 for GET_MODE_SIZE.
13434 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13435 for GET_MODE_SIZE.
13436 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13437 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13438 * caller-save.c (setup_save_areas): Likewise.
13439 (replace_reg_with_saved_mem): Likewise.
13440 * calls.c (emit_library_call_value_1): Likewise.
13441 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13442 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13443 (gen_lowpart_for_combine): Likewise.
13444 * convert.c (convert_to_integer_1): Likewise.
13445 * cse.c (equiv_constant, cse_insn): Likewise.
13446 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13447 (cselib_subst_to_values): Likewise.
13448 * dce.c (word_dce_process_block): Likewise.
13449 * df-problems.c (df_word_lr_mark_ref): Likewise.
13450 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13451 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13452 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13453 (rtl_for_decl_location): Likewise.
13454 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13455 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13456 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13457 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13458 (expand_expr_real_1): Likewise.
13459 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13460 (pad_below): Likewise.
13461 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13462 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13463 * ira.c (get_subreg_tracking_sizes): Likewise.
13464 * ira-build.c (ira_create_allocno_objects): Likewise.
13465 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13466 (ira_sort_regnos_for_alter_reg): Likewise.
13467 * ira-costs.c (record_operand_costs): Likewise.
13468 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13469 (resolve_simple_move): Likewise.
13470 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13471 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13472 (lra_constraints): Likewise.
13473 (CONST_POOL_OK_P): Reject variable-sized modes.
13474 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13475 (add_pseudo_to_slot, lra_spill): Likewise.
13476 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13477 * optabs-query.c (get_best_extraction_insn): Likewise.
13478 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13479 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13480 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13481 * recog.c (offsettable_address_addr_space_p): Likewise.
13482 * regcprop.c (maybe_mode_change): Likewise.
13483 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13484 * regrename.c (build_def_use): Likewise.
13485 * regstat.c (dump_reg_info): Likewise.
13486 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13487 (find_reloads, find_reloads_subreg_address): Likewise.
13488 * reload1.c (eliminate_regs_1): Likewise.
13489 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13490 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13491 (simplify_binary_operation_1, simplify_subreg): Likewise.
13492 * targhooks.c (default_function_arg_padding): Likewise.
13493 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13494 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13495 (verify_gimple_assign_ternary): Likewise.
13496 * tree-inline.c (estimate_move_cost): Likewise.
13497 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13498 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13499 (get_address_cost_ainc): Likewise.
13500 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13501 (vect_supportable_dr_alignment): Likewise.
13502 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13503 (vectorizable_reduction): Likewise.
13504 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13505 (vectorizable_operation, vectorizable_load): Likewise.
13506 * tree.c (build_same_sized_truth_vector_type): Likewise.
13507 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13508 * var-tracking.c (emit_note_insn_var_location): Likewise.
13509 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13510 (ADDR_VEC_ALIGN): Likewise.
13512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13513 Alan Hayward <alan.hayward@arm.com>
13514 David Sherwood <david.sherwood@arm.com>
13516 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13517 unsigned short.
13518 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13519 or if measurement_type is polynomial.
13520 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13521 * combine.c (make_extraction): Likewise.
13522 * dse.c (find_shift_sequence): Likewise.
13523 * dwarf2out.c (mem_loc_descriptor): Likewise.
13524 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13525 (extract_bit_field, extract_low_bits): Likewise.
13526 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13527 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13528 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13529 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13530 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13531 * reload.c (find_reloads): Likewise.
13532 * reload1.c (alter_reg): Likewise.
13533 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13534 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13535 * tree-if-conv.c (predicate_mem_writes): Likewise.
13536 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13537 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13538 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13539 * valtrack.c (dead_debug_insert_temp): Likewise.
13540 * varasm.c (mergeable_constant_section): Likewise.
13541 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
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 * expr.c (expand_assignment): Cope with polynomial mode sizes
13548 when assigning to a CONCAT.
13550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13551 Alan Hayward <alan.hayward@arm.com>
13552 David Sherwood <david.sherwood@arm.com>
13554 * machmode.h (mode_precision): Change from unsigned short to
13555 poly_uint16_pod.
13556 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13557 short.
13558 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13559 or if measurement_type is not polynomial.
13560 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13561 in which the mode is already known to be a scalar_int_mode.
13562 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13563 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13564 initializer.
13565 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13566 for GET_MODE_PRECISION.
13567 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13568 for GET_MODE_PRECISION.
13569 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13570 as polynomial.
13571 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13572 (expand_field_assignment, make_extraction): Likewise.
13573 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13574 (get_last_value): Likewise.
13575 * convert.c (convert_to_integer_1): Likewise.
13576 * cse.c (cse_insn): Likewise.
13577 * expr.c (expand_expr_real_1): Likewise.
13578 * lra-constraints.c (simplify_operand_subreg): Likewise.
13579 * optabs-query.c (can_atomic_load_p): Likewise.
13580 * optabs.c (expand_atomic_load): Likewise.
13581 (expand_atomic_store): Likewise.
13582 * ree.c (combine_reaching_defs): Likewise.
13583 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13584 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13585 * tree.h (type_has_mode_precision_p): Likewise.
13586 * ubsan.c (instrument_si_overflow): Likewise.
13588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13589 Alan Hayward <alan.hayward@arm.com>
13590 David Sherwood <david.sherwood@arm.com>
13592 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13593 polynomial numbers of units.
13594 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13595 (valid_vector_subparts_p): New function.
13596 (build_vector_type): Remove temporary shim and take the number
13597 of units as a poly_uint64 rather than an int.
13598 (build_opaque_vector_type): Take the number of units as a
13599 poly_uint64 rather than an int.
13600 * tree.c (build_vector_from_ctor): Handle polynomial
13601 TYPE_VECTOR_SUBPARTS.
13602 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13603 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13604 (build_vector_from_val): If the number of units is variable,
13605 use build_vec_duplicate_cst for constant operands and
13606 VEC_DUPLICATE_EXPR otherwise.
13607 (make_vector_type): Remove temporary is_constant ().
13608 (build_vector_type, build_opaque_vector_type): Take the number of
13609 units as a poly_uint64 rather than an int.
13610 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13611 VECTOR_CST_NELTS.
13612 * cfgexpand.c (expand_debug_expr): Likewise.
13613 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13614 (store_constructor, expand_expr_real_1): Likewise.
13615 (const_scalar_mask_from_tree): Likewise.
13616 * fold-const-call.c (fold_const_reduction): Likewise.
13617 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13618 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13619 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13620 (fold_relational_const): Likewise.
13621 (native_interpret_vector): Likewise. Change the size from an
13622 int to an unsigned int.
13623 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13624 TYPE_VECTOR_SUBPARTS.
13625 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13626 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13627 duplicating a non-constant operand into a variable-length vector.
13628 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13629 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13630 * ipa-icf.c (sem_variable::equals): Likewise.
13631 * match.pd: Likewise.
13632 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13633 * print-tree.c (print_node): Likewise.
13634 * stor-layout.c (layout_type): Likewise.
13635 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13636 * tree-cfg.c (verify_gimple_comparison): Likewise.
13637 (verify_gimple_assign_binary): Likewise.
13638 (verify_gimple_assign_ternary): Likewise.
13639 (verify_gimple_assign_single): Likewise.
13640 * tree-pretty-print.c (dump_generic_node): Likewise.
13641 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13642 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13643 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13644 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13645 (vect_shift_permute_load_chain): Likewise.
13646 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13647 (expand_vector_condition, optimize_vector_constructor): Likewise.
13648 (lower_vec_perm, get_compute_type): Likewise.
13649 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13650 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13651 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13652 (vect_recog_mask_conversion_pattern): Likewise.
13653 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13654 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13655 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13656 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13657 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13658 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13659 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13660 (supportable_widening_operation): Likewise.
13661 (supportable_narrowing_operation): Likewise.
13662 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13663 Likewise.
13664 * varasm.c (output_constant): Likewise.
13666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13667 Alan Hayward <alan.hayward@arm.com>
13668 David Sherwood <david.sherwood@arm.com>
13670 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13671 so that both the length == 3 and length != 3 cases set up their
13672 own permute vectors. Add comments explaining why we know the
13673 number of elements is constant.
13674 (vect_permute_load_chain): Likewise.
13676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 Alan Hayward <alan.hayward@arm.com>
13678 David Sherwood <david.sherwood@arm.com>
13680 * machmode.h (mode_nunits): Change from unsigned char to
13681 poly_uint16_pod.
13682 (ONLY_FIXED_SIZE_MODES): New macro.
13683 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13684 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13685 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13686 New typedefs.
13687 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13688 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13689 or if measurement_type is not polynomial.
13690 * genmodes.c (ZERO_COEFFS): New macro.
13691 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13692 poly_uint16.
13693 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13694 Use ZERO_COEFFS when emitting initializers.
13695 * data-streamer.h (bp_pack_poly_value): New function.
13696 (bp_unpack_poly_value): Likewise.
13697 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13698 for GET_MODE_NUNITS.
13699 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13700 for GET_MODE_NUNITS.
13701 * tree.c (make_vector_type): Remove temporary shim and make
13702 the real function take the number of units as a poly_uint64
13703 rather than an int.
13704 (build_vector_type_for_mode): Handle polynomial nunits.
13705 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13706 * emit-rtl.c (const_vec_series_p_1): Likewise.
13707 (gen_rtx_CONST_VECTOR): Likewise.
13708 * fold-const.c (test_vec_duplicate_folding): Likewise.
13709 * genrecog.c (validate_pattern): Likewise.
13710 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13711 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13712 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13713 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13714 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13715 * rtlanal.c (subreg_get_info): Likewise.
13716 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13717 (vect_grouped_load_supported): Likewise.
13718 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13719 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13720 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13721 (simplify_const_unary_operation, simplify_binary_operation_1)
13722 (simplify_const_binary_operation, simplify_ternary_operation)
13723 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13724 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13725 instead of CONST_VECTOR_NUNITS.
13726 * varasm.c (output_constant_pool_2): Likewise.
13727 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13728 explicit-encoded elements in the XVEC for variable-length vectors.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13732 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13735 Alan Hayward <alan.hayward@arm.com>
13736 David Sherwood <david.sherwood@arm.com>
13738 * coretypes.h (fixed_size_mode): Declare.
13739 (fixed_size_mode_pod): New typedef.
13740 * builtins.h (target_builtins::x_apply_args_mode)
13741 (target_builtins::x_apply_result_mode): Change type to
13742 fixed_size_mode_pod.
13743 * builtins.c (apply_args_size, apply_result_size, result_vector)
13744 (expand_builtin_apply_args_1, expand_builtin_apply)
13745 (expand_builtin_return): Update accordingly.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13749 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13750 * cselib.c (cselib_hash_rtx): Likewise.
13751 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13752 CONST_VECTOR encoding.
13754 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13755 Jeff Law <law@redhat.com>
13757 PR target/83641
13758 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13759 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13760 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13761 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13763 PR target/83641
13764 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13765 explicitly probe *sp in a noreturn function if there were any callee
13766 register saves or frame pointer is needed.
13768 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13770 PR debug/83621
13771 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13772 BLKmode for ternary, binary or unary expressions.
13774 PR debug/83645
13775 * var-tracking.c (delete_vta_debug_insn): New inline function.
13776 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13777 insns from get_insns () to NULL instead of each bb separately.
13778 Use delete_vta_debug_insn. No longer static.
13779 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13780 delete_vta_debug_insns callers.
13781 * rtl.h (delete_vta_debug_insns): Declare.
13782 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13783 instead of variable_tracking_main.
13785 2018-01-03 Martin Sebor <msebor@redhat.com>
13787 PR tree-optimization/83603
13788 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13789 arguments past the endof the argument list in functions declared
13790 without a prototype.
13791 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13792 Avoid checking when arguments are null.
13794 2018-01-03 Martin Sebor <msebor@redhat.com>
13796 PR c/83559
13797 * doc/extend.texi (attribute const): Fix a typo.
13798 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13799 issuing -Wsuggest-attribute for void functions.
13801 2018-01-03 Martin Sebor <msebor@redhat.com>
13803 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13804 offset_int::from instead of wide_int::to_shwi.
13805 (maybe_diag_overlap): Remove assertion.
13806 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13807 * gimple-ssa-sprintf.c (format_directive): Same.
13808 (parse_directive): Same.
13809 (sprintf_dom_walker::compute_format_length): Same.
13810 (try_substitute_return_value): Same.
13812 2018-01-03 Jeff Law <law@redhat.com>
13814 PR middle-end/83654
13815 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13816 non-constant residual for zero at runtime and avoid probing in
13817 that case. Reorganize code for trailing problem to mirror handling
13818 of the residual.
13820 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13822 PR tree-optimization/83501
13823 * tree-ssa-strlen.c (get_string_cst): New.
13824 (handle_char_store): Call get_string_cst.
13826 2018-01-03 Martin Liska <mliska@suse.cz>
13828 PR tree-optimization/83593
13829 * tree-ssa-strlen.c: Include tree-cfg.h.
13830 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13831 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13832 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13833 to false.
13834 (strlen_dom_walker::before_dom_children): Call
13835 gimple_purge_dead_eh_edges. Dump tranformation with details
13836 dump flags.
13837 (strlen_dom_walker::before_dom_children): Update call by adding
13838 new argument cleanup_eh.
13839 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13841 2018-01-03 Martin Liska <mliska@suse.cz>
13843 PR ipa/83549
13844 * cif-code.def (VARIADIC_THUNK): New enum value.
13845 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13846 thunks.
13848 2018-01-03 Jan Beulich <jbeulich@suse.com>
13850 * sse.md (mov<mode>_internal): Tighten condition for when to use
13851 vmovdqu<ssescalarsize> for TI and OI modes.
13853 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13855 Update copyright years.
13857 2018-01-03 Martin Liska <mliska@suse.cz>
13859 PR ipa/83594
13860 * ipa-visibility.c (function_and_variable_visibility): Skip
13861 functions with noipa attribure.
13863 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13865 * gcc.c (process_command): Update copyright notice dates.
13866 * gcov-dump.c (print_version): Ditto.
13867 * gcov.c (print_version): Ditto.
13868 * gcov-tool.c (print_version): Ditto.
13869 * gengtype.c (create_file): Ditto.
13870 * doc/cpp.texi: Bump @copying's copyright year.
13871 * doc/cppinternals.texi: Ditto.
13872 * doc/gcc.texi: Ditto.
13873 * doc/gccint.texi: Ditto.
13874 * doc/gcov.texi: Ditto.
13875 * doc/install.texi: Ditto.
13876 * doc/invoke.texi: Ditto.
13878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13880 * vector-builder.h (vector_builder::m_full_nelts): Change from
13881 unsigned int to poly_uint64.
13882 (vector_builder::full_nelts): Update prototype accordingly.
13883 (vector_builder::new_vector): Likewise.
13884 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13885 (vector_builder::operator ==): Likewise.
13886 (vector_builder::finalize): Likewise.
13887 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13888 Take the number of elements as a poly_uint64 rather than an
13889 unsigned int.
13890 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13891 from unsigned int to poly_uint64.
13892 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13893 (vec_perm_indices::new_vector): Likewise.
13894 (vec_perm_indices::length): Likewise.
13895 (vec_perm_indices::nelts_per_input): Likewise.
13896 (vec_perm_indices::input_nelts): Likewise.
13897 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13898 number of elements per input as a poly_uint64 rather than an
13899 unsigned int. Use the original encoding for variable-length
13900 vectors, rather than clamping each individual element.
13901 For the second and subsequent elements in each pattern,
13902 clamp the step and base before clamping their sum.
13903 (vec_perm_indices::series_p): Handle polynomial element counts.
13904 (vec_perm_indices::all_in_range_p): Likewise.
13905 (vec_perm_indices_to_tree): Likewise.
13906 (vec_perm_indices_to_rtx): Likewise.
13907 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13908 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13909 (tree_vector_builder::new_binary_operation): Handle polynomial
13910 element counts. Return false if we need to know the number
13911 of elements at compile time.
13912 * fold-const.c (fold_vec_perm): Punt if the number of elements
13913 isn't known at compile time.
13915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13917 * vec-perm-indices.h (vec_perm_builder): Change element type
13918 from HOST_WIDE_INT to poly_int64.
13919 (vec_perm_indices::element_type): Update accordingly.
13920 (vec_perm_indices::clamp): Handle polynomial element_types.
13921 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13922 (vec_perm_indices::all_in_range_p): Likewise.
13923 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13924 than shwi trees.
13925 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13926 polynomial vec_perm_indices element types.
13927 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13928 * fold-const.c (fold_vec_perm): Likewise.
13929 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13930 * tree-vect-generic.c (lower_vec_perm): Likewise.
13931 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13932 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13933 element type to HOST_WIDE_INT.
13935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13936 Alan Hayward <alan.hayward@arm.com>
13937 David Sherwood <david.sherwood@arm.com>
13939 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13940 rather than an int. Use plus_constant.
13941 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13942 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 (emit_call_1, expand_call): Change struct_value_size from
13949 a HOST_WIDE_INT to a poly_int64.
13951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13952 Alan Hayward <alan.hayward@arm.com>
13953 David Sherwood <david.sherwood@arm.com>
13955 * calls.c (load_register_parameters): Cope with polynomial
13956 mode sizes. Require a constant size for BLKmode parameters
13957 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13958 forces a parameter to be padded at the lsb end in order to
13959 fill a complete number of words, require the parameter size
13960 to be ordered wrt UNITS_PER_WORD.
13962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13963 Alan Hayward <alan.hayward@arm.com>
13964 David Sherwood <david.sherwood@arm.com>
13966 * reload1.c (spill_stack_slot_width): Change element type
13967 from unsigned int to poly_uint64_pod.
13968 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
13975 (reload_inner_reg_of_subreg, push_reload): Use it.
13977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13978 Alan Hayward <alan.hayward@arm.com>
13979 David Sherwood <david.sherwood@arm.com>
13981 * lra-constraints.c (process_alt_operands): Reject matched
13982 operands whose sizes aren't ordered.
13983 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13990 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13997 Use plus_constant instead of gen_rtx_PLUS.
13999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14000 Alan Hayward <alan.hayward@arm.com>
14001 David Sherwood <david.sherwood@arm.com>
14003 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14004 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14005 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14006 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14007 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14008 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14009 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14010 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14011 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14012 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14013 a poly_int64.
14014 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14015 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14016 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14017 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14018 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14019 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14020 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14021 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14022 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14023 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14024 function.
14025 * expr.c (emit_move_resolve_push): Treat the input and result
14026 of PUSH_ROUNDING as a poly_int64.
14027 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14028 (emit_push_insn): Likewise.
14029 * lra-eliminations.c (mark_not_eliminable): Likewise.
14030 * recog.c (push_operand): Likewise.
14031 * reload1.c (elimination_effects): Likewise.
14032 * rtlanal.c (nonzero_bits1): Likewise.
14033 * calls.c (store_one_arg): Likewise. Require the padding to be
14034 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14041 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14048 rather than an int.
14050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14055 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14056 via stack temporaries. Treat the mode size as polynomial too.
14058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14059 Alan Hayward <alan.hayward@arm.com>
14060 David Sherwood <david.sherwood@arm.com>
14062 * expr.c (expand_expr_real_2): When handling conversions involving
14063 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14064 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14065 as a poly_uint64 too.
14067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14068 Alan Hayward <alan.hayward@arm.com>
14069 David Sherwood <david.sherwood@arm.com>
14071 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14074 Alan Hayward <alan.hayward@arm.com>
14075 David Sherwood <david.sherwood@arm.com>
14077 * combine.c (can_change_dest_mode): Handle polynomial
14078 REGMODE_NATURAL_SIZE.
14079 * expmed.c (store_bit_field_1): Likewise.
14080 * expr.c (store_constructor): Likewise.
14081 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14082 and polynomial REGMODE_NATURAL_SIZE.
14083 (gen_lowpart_common): Likewise.
14084 * reginfo.c (record_subregs_of_mode): Likewise.
14085 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14092 numbers of elements.
14094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14095 Alan Hayward <alan.hayward@arm.com>
14096 David Sherwood <david.sherwood@arm.com>
14098 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14105 in a POINTER_PLUS_EXPR.
14107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14108 Alan Hayward <alan.hayward@arm.com>
14109 David Sherwood <david.sherwood@arm.com>
14111 * omp-simd-clone.c (simd_clone_subparts): New function.
14112 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14113 (ipa_simd_modify_function_body): Likewise.
14115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14116 Alan Hayward <alan.hayward@arm.com>
14117 David Sherwood <david.sherwood@arm.com>
14119 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14120 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14121 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14122 (expand_vector_condition, vector_element): Likewise.
14123 (subparts_gt): New function.
14124 (get_compute_type): Use subparts_gt.
14125 (count_type_subparts): Delete.
14126 (expand_vector_operations_1): Use subparts_gt instead of
14127 count_type_subparts.
14129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14134 (vect_compile_time_alias): ...this new function. Do the calculation
14135 on poly_ints rather than trees.
14136 (vect_prune_runtime_alias_test_list): Update call accordingly.
14138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14139 Alan Hayward <alan.hayward@arm.com>
14140 David Sherwood <david.sherwood@arm.com>
14142 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14143 numbers of units.
14144 (vect_schedule_slp_instance): Likewise.
14146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14147 Alan Hayward <alan.hayward@arm.com>
14148 David Sherwood <david.sherwood@arm.com>
14150 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14151 constant and extern definitions for variable-length vectors.
14152 (vect_get_constant_vectors): Note that the number of units
14153 is known to be constant.
14155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14156 Alan Hayward <alan.hayward@arm.com>
14157 David Sherwood <david.sherwood@arm.com>
14159 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14160 of units as polynomial. Choose between WIDE and NARROW based
14161 on multiple_p.
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 (simd_clone_subparts): New function.
14168 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 Alan Hayward <alan.hayward@arm.com>
14172 David Sherwood <david.sherwood@arm.com>
14174 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14175 vectors as polynomial. Use build_index_vector for
14176 IFN_GOMP_SIMD_LANE.
14178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14179 Alan Hayward <alan.hayward@arm.com>
14180 David Sherwood <david.sherwood@arm.com>
14182 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14183 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14184 for variable-length vectors.
14185 (vectorizable_mask_load_store): Treat the number of units as
14186 polynomial, asserting that it is constant if the condition has
14187 already been enforced.
14188 (vectorizable_store, vectorizable_load): Likewise.
14190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14191 Alan Hayward <alan.hayward@arm.com>
14192 David Sherwood <david.sherwood@arm.com>
14194 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14195 of units as polynomial. Punt if we can't tell at compile time
14196 which vector contains the final result.
14198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14199 Alan Hayward <alan.hayward@arm.com>
14200 David Sherwood <david.sherwood@arm.com>
14202 * tree-vect-loop.c (vectorizable_induction): Treat the number
14203 of units as polynomial. Punt on SLP inductions. Use an integer
14204 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14205 cast of such a series for variable-length floating-point
14206 reductions.
14208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14209 Alan Hayward <alan.hayward@arm.com>
14210 David Sherwood <david.sherwood@arm.com>
14212 * tree.h (build_index_vector): Declare.
14213 * tree.c (build_index_vector): New function.
14214 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14215 of units as polynomial, forcibly converting it to a constant if
14216 vectorizable_reduction has already enforced the condition.
14217 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14218 to create a {1,2,3,...} vector.
14219 (vectorizable_reduction): Treat the number of units as polynomial.
14220 Choose vectype_in based on the largest scalar element size rather
14221 than the smallest number of units. Enforce the restrictions
14222 relied on above.
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 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14229 number of units as polynomial.
14231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14232 Alan Hayward <alan.hayward@arm.com>
14233 David Sherwood <david.sherwood@arm.com>
14235 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14236 * target.def (autovectorize_vector_sizes): Return the vector sizes
14237 by pointer, using vector_sizes rather than a bitmask.
14238 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14239 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14240 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14241 Likewise.
14242 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14243 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14244 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14245 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14246 * omp-general.c (omp_max_vf): Likewise.
14247 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14248 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14249 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14250 * tree-vect-slp.c (vect_slp_bb): Likewise.
14251 * doc/tm.texi: Regenerate.
14252 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14253 to a poly_uint64.
14254 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14255 the vector size as a poly_uint64 rather than an unsigned int.
14256 (current_vector_size): Change from an unsigned int to a poly_uint64.
14257 (get_vectype_for_scalar_type): Update accordingly.
14258 * tree.h (build_truth_vector_type): Take the size and number of
14259 units as a poly_uint64 rather than an unsigned int.
14260 (build_vector_type): Add a temporary overload that takes
14261 the number of units as a poly_uint64 rather than an unsigned int.
14262 * tree.c (make_vector_type): Likewise.
14263 (build_truth_vector_type): Take the number of units as a poly_uint64
14264 rather than an unsigned int.
14266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14267 Alan Hayward <alan.hayward@arm.com>
14268 David Sherwood <david.sherwood@arm.com>
14270 * target.def (get_mask_mode): Take the number of units and length
14271 as poly_uint64s rather than unsigned ints.
14272 * targhooks.h (default_get_mask_mode): Update accordingly.
14273 * targhooks.c (default_get_mask_mode): Likewise.
14274 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14275 * doc/tm.texi: Regenerate.
14277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14278 Alan Hayward <alan.hayward@arm.com>
14279 David Sherwood <david.sherwood@arm.com>
14281 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14282 * omp-general.c (omp_max_vf): Likewise.
14283 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14284 (expand_omp_simd): Handle polynomial safelen.
14285 * omp-low.c (omplow_simd_context): Add a default constructor.
14286 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14287 (lower_rec_simd_input_clauses): Update accordingly.
14288 (lower_rec_input_clauses): Likewise.
14290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14291 Alan Hayward <alan.hayward@arm.com>
14292 David Sherwood <david.sherwood@arm.com>
14294 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14295 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14296 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14297 (vect_analyze_slp_cost): Likewise.
14298 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14299 (vect_model_load_cost): Likewise.
14301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14302 Alan Hayward <alan.hayward@arm.com>
14303 David Sherwood <david.sherwood@arm.com>
14305 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14306 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14307 from an unsigned int * to a poly_uint64_pod *.
14308 (calculate_unrolling_factor): New function.
14309 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14312 Alan Hayward <alan.hayward@arm.com>
14313 David Sherwood <david.sherwood@arm.com>
14315 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14316 from an unsigned int to a poly_uint64.
14317 (_loop_vec_info::slp_unrolling_factor): Likewise.
14318 (_loop_vec_info::vectorization_factor): Change from an int
14319 to a poly_uint64.
14320 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14321 (vect_get_num_vectors): New function.
14322 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14323 (vect_get_num_copies): Use vect_get_num_vectors.
14324 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14325 to an unsigned int *.
14326 (vect_analyze_data_refs): Change min_vf from an int * to a
14327 poly_uint64 *.
14328 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14329 than an unsigned HOST_WIDE_INT.
14330 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14331 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14332 to an unsigned int *.
14333 (vect_analyze_data_ref_dependences): Likewise.
14334 (vect_compute_data_ref_alignment): Handle polynomial vf.
14335 (vect_enhance_data_refs_alignment): Likewise.
14336 (vect_prune_runtime_alias_test_list): Likewise.
14337 (vect_shift_permute_load_chain): Likewise.
14338 (vect_supportable_dr_alignment): Likewise.
14339 (dependence_distance_ge_vf): Take the vectorization factor as a
14340 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14341 (vect_analyze_data_refs): Change min_vf from an int * to a
14342 poly_uint64 *.
14343 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14344 vfm1 as a poly_uint64 rather than an int. Make the same change
14345 for the returned bound_scalar.
14346 (vect_gen_vector_loop_niters): Handle polynomial vf.
14347 (vect_do_peeling): Likewise. Update call to
14348 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14349 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14350 be constant.
14351 * tree-vect-loop.c (vect_determine_vectorization_factor)
14352 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14353 (vect_get_known_peeling_cost): Likewise.
14354 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14355 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14356 (vect_transform_loop): Likewise. Use the lowest possible VF when
14357 updating the upper bounds of the loop.
14358 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14359 rather than an int.
14360 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14361 polynomial unroll factors.
14362 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14363 (vect_make_slp_decision): Likewise.
14364 (vect_supported_load_permutation_p): Likewise, and polynomial
14365 vf too.
14366 (vect_analyze_slp_cost): Handle polynomial vf.
14367 (vect_slp_analyze_node_operations): Likewise.
14368 (vect_slp_analyze_bb_1): Likewise.
14369 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14370 than an unsigned HOST_WIDE_INT.
14371 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14372 (vectorizable_load): Handle polynomial vf.
14373 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14374 a poly_uint64.
14375 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14378 Alan Hayward <alan.hayward@arm.com>
14379 David Sherwood <david.sherwood@arm.com>
14381 * match.pd: Handle bit operations involving three constants
14382 and try to fold one pair.
14384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14386 * tree-vect-loop-manip.c: Include gimple-fold.h.
14387 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14388 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14389 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14390 Add a path that uses a step of VF instead of 1, but disable it
14391 for now.
14392 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14393 and niters_no_overflow parameters. Update calls to
14394 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14395 Create a new SSA name if the latter choses to use a ste other
14396 than zero, and return it via niters_vector_mult_vf_var.
14397 * tree-vect-loop.c (vect_transform_loop): Update calls to
14398 vect_do_peeling, vect_gen_vector_loop_niters and
14399 slpeel_make_loop_iterate_ntimes.
14400 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14401 (vect_gen_vector_loop_niters): Update declarations after above changes.
14403 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14405 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14406 128-bit round to integer instructions.
14407 (ceil<mode>2): Likewise.
14408 (btrunc<mode>2): Likewise.
14409 (round<mode>2): Likewise.
14411 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14413 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14414 unaligned VSX load/store on P8/P9.
14415 (expand_block_clear): Allow the use of unaligned VSX
14416 load/store on P8/P9.
14418 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14420 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14421 New function.
14422 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14423 swap associated with both a load and a store.
14425 2018-01-02 Andrew Waterman <andrew@sifive.com>
14427 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14428 * config/riscv/riscv.md (clear_cache): Use it.
14430 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14432 * web.c: Remove out-of-date comment.
14434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14436 * expr.c (fixup_args_size_notes): Check that any existing
14437 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14438 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14439 (emit_single_push_insn): ...here.
14441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14443 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14444 (const_vector_encoded_nelts): New function.
14445 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14446 (const_vector_int_elt, const_vector_elt): Declare.
14447 * emit-rtl.c (const_vector_int_elt_1): New function.
14448 (const_vector_elt): Likewise.
14449 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14450 of CONST_VECTOR_ELT.
14452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14454 * expr.c: Include rtx-vector-builder.h.
14455 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14456 directly on the tree encoding.
14457 (const_vector_from_tree): Likewise.
14458 * optabs.c: Include rtx-vector-builder.h.
14459 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14460 sequence of "u" values.
14461 * vec-perm-indices.c: Include rtx-vector-builder.h.
14462 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14463 directly on the vec_perm_indices encoding.
14465 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14467 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14468 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14469 * rtx-vector-builder.h: New file.
14470 * rtx-vector-builder.c: Likewise.
14471 * rtl.h (rtx_def::u2): Add a const_vector field.
14472 (CONST_VECTOR_NPATTERNS): New macro.
14473 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14474 (CONST_VECTOR_DUPLICATE_P): Likewise.
14475 (CONST_VECTOR_STEPPED_P): Likewise.
14476 (CONST_VECTOR_ENCODED_ELT): Likewise.
14477 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14478 (unwrap_const_vec_duplicate): Likewise.
14479 (const_vec_series_p): Check for a non-duplicated vector encoding.
14480 Say that the function only returns true for integer vectors.
14481 * emit-rtl.c: Include rtx-vector-builder.h.
14482 (gen_const_vec_duplicate_1): Delete.
14483 (gen_const_vector): Call gen_const_vec_duplicate instead of
14484 gen_const_vec_duplicate_1.
14485 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14486 (gen_const_vec_duplicate): Use rtx_vector_builder.
14487 (gen_const_vec_series): Likewise.
14488 (gen_rtx_CONST_VECTOR): Likewise.
14489 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14490 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14491 Build a new vector rather than modifying a CONST_VECTOR in-place.
14492 (handle_special_swappables): Update call accordingly.
14493 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14494 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14495 Build a new vector rather than modifying a CONST_VECTOR in-place.
14496 (handle_special_swappables): Update call accordingly.
14498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14500 * simplify-rtx.c (simplify_const_binary_operation): Use
14501 CONST_VECTOR_ELT instead of XVECEXP.
14503 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14505 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14506 the selector elements to be different from the data elements
14507 if the selector is a VECTOR_CST.
14508 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14509 ssizetype for the selector.
14511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14513 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14514 before testing each element individually.
14515 * tree-vect-generic.c (lower_vec_perm): Likewise.
14517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14519 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14520 * selftest-run-tests.c (selftest::run_tests): Call it.
14521 * vector-builder.h (vector_builder::operator ==): New function.
14522 (vector_builder::operator !=): Likewise.
14523 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14524 (vec_perm_indices::all_from_input_p): New function.
14525 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14526 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14527 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14528 instead of reading the VECTOR_CST directly. Detect whether both
14529 vector inputs are the same before constructing the vec_perm_indices,
14530 and update the number of inputs argument accordingly. Use the
14531 utility functions added above. Only construct sel2 if we need to.
14533 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14535 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14536 the broadcast of the low byte.
14537 (expand_mult_highpart): Use an explicit encoding for the permutes.
14538 * optabs-query.c (can_mult_highpart_p): Likewise.
14539 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14540 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14541 (vectorizable_bswap): Likewise.
14542 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14543 explicit encoding for the power-of-2 permutes.
14544 (vect_permute_store_chain): Likewise.
14545 (vect_grouped_load_supported): Likewise.
14546 (vect_permute_load_chain): Likewise.
14548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14550 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14551 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14552 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14553 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14554 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14555 (vect_gen_perm_mask_any): Likewise.
14557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14559 * int-vector-builder.h: New file.
14560 * vec-perm-indices.h: Include int-vector-builder.h.
14561 (vec_perm_indices): Redefine as an int_vector_builder.
14562 (auto_vec_perm_indices): Delete.
14563 (vec_perm_builder): Redefine as a stand-alone class.
14564 (vec_perm_indices::vec_perm_indices): New function.
14565 (vec_perm_indices::clamp): Likewise.
14566 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14567 (vec_perm_indices::new_vector): New function.
14568 (vec_perm_indices::new_expanded_vector): Update for new
14569 vec_perm_indices class.
14570 (vec_perm_indices::rotate_inputs): New function.
14571 (vec_perm_indices::all_in_range_p): Operate directly on the
14572 encoded form, without computing elided elements.
14573 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14574 encoding. Update for new vec_perm_indices class.
14575 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14576 the given vec_perm_builder.
14577 (expand_vec_perm_var): Update vec_perm_builder constructor.
14578 (expand_mult_highpart): Use vec_perm_builder instead of
14579 auto_vec_perm_indices.
14580 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14581 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14582 or double series encoding as appropriate.
14583 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14584 vec_perm_indices instead of auto_vec_perm_indices.
14585 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14586 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14587 (vect_permute_store_chain): Likewise.
14588 (vect_grouped_load_supported): Likewise.
14589 (vect_permute_load_chain): Likewise.
14590 (vect_shift_permute_load_chain): Likewise.
14591 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14592 (vect_transform_slp_perm_load): Likewise.
14593 (vect_schedule_slp_instance): Likewise.
14594 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14595 (vectorizable_mask_load_store): Likewise.
14596 (vectorizable_bswap): Likewise.
14597 (vectorizable_store): Likewise.
14598 (vectorizable_load): Likewise.
14599 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14600 vec_perm_indices instead of auto_vec_perm_indices. Use
14601 tree_to_vec_perm_builder to read the vector from a tree.
14602 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14603 vec_perm_builder instead of a vec_perm_indices.
14604 (have_whole_vector_shift): Use vec_perm_builder and
14605 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14606 truncation to calc_vec_perm_mask_for_shift.
14607 (vect_create_epilog_for_reduction): Likewise.
14608 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14609 from auto_vec_perm_indices to vec_perm_indices.
14610 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14611 instead of changing individual elements.
14612 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14613 the vector in d.perm.
14614 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14615 from auto_vec_perm_indices to vec_perm_indices.
14616 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14617 instead of changing individual elements.
14618 (arm_vectorize_vec_perm_const): Use new_vector to install
14619 the vector in d.perm.
14620 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14621 Update vec_perm_builder constructor.
14622 (rs6000_expand_interleave): Likewise.
14623 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14624 (rs6000_expand_interleave): Likewise.
14626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14628 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14629 to qimode could truncate the indices.
14630 * optabs.c (expand_vec_perm_var): Likewise.
14632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14634 * Makefile.in (OBJS): Add vec-perm-indices.o.
14635 * vec-perm-indices.h: New file.
14636 * vec-perm-indices.c: Likewise.
14637 * target.h (vec_perm_indices): Replace with a forward class
14638 declaration.
14639 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14640 * optabs.h: Include vec-perm-indices.h.
14641 (expand_vec_perm): Delete.
14642 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14643 (expand_vec_perm_const): Declare.
14644 * target.def (vec_perm_const_ok): Replace with...
14645 (vec_perm_const): ...this new hook.
14646 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14647 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14648 * doc/tm.texi: Regenerate.
14649 * optabs.def (vec_perm_const): Delete.
14650 * doc/md.texi (vec_perm_const): Likewise.
14651 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14652 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14653 expand_vec_perm for constant permutation vectors. Assert that
14654 the mode of variable permutation vectors is the integer equivalent
14655 of the mode that is being permuted.
14656 * optabs-query.h (selector_fits_mode_p): Declare.
14657 * optabs-query.c: Include vec-perm-indices.h.
14658 (selector_fits_mode_p): New function.
14659 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14660 is defined, instead of checking whether the vec_perm_const_optab
14661 exists. Use targetm.vectorize.vec_perm_const instead of
14662 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14663 fit in the vector mode before using a variable permute.
14664 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14665 vec_perm_indices instead of an rtx.
14666 (expand_vec_perm): Replace with...
14667 (expand_vec_perm_const): ...this new function. Take the selector
14668 as a vec_perm_indices rather than an rtx. Also take the mode of
14669 the selector. Update call to shift_amt_for_vec_perm_mask.
14670 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14671 Use vec_perm_indices::new_expanded_vector to expand the original
14672 selector into bytes. Check whether the indices fit in the vector
14673 mode before using a variable permute.
14674 (expand_vec_perm_var): Make global.
14675 (expand_mult_highpart): Use expand_vec_perm_const.
14676 * fold-const.c: Includes vec-perm-indices.h.
14677 * tree-ssa-forwprop.c: Likewise.
14678 * tree-vect-data-refs.c: Likewise.
14679 * tree-vect-generic.c: Likewise.
14680 * tree-vect-loop.c: Likewise.
14681 * tree-vect-slp.c: Likewise.
14682 * tree-vect-stmts.c: Likewise.
14683 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14684 Delete.
14685 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14686 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14687 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14688 (aarch64_vectorize_vec_perm_const): ...this new function.
14689 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14690 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14691 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14692 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14693 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14694 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14695 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14696 into...
14697 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14698 check for NEON modes.
14699 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14700 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14701 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14702 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14703 into...
14704 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14705 the old VEC_PERM_CONST conditions.
14706 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14707 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14708 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14709 (ia64_vectorize_vec_perm_const_ok): Merge into...
14710 (ia64_vectorize_vec_perm_const): ...this new function.
14711 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14712 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14713 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14714 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14715 * config/mips/mips.c (mips_expand_vec_perm_const)
14716 (mips_vectorize_vec_perm_const_ok): Merge into...
14717 (mips_vectorize_vec_perm_const): ...this new function.
14718 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14719 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14720 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14721 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14722 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14723 (rs6000_expand_vec_perm_const): Delete.
14724 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14725 Delete.
14726 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14727 (altivec_expand_vec_perm_const_le): Take each operand individually.
14728 Operate on constant selectors rather than rtxes.
14729 (altivec_expand_vec_perm_const): Likewise. Update call to
14730 altivec_expand_vec_perm_const_le.
14731 (rs6000_expand_vec_perm_const): Delete.
14732 (rs6000_vectorize_vec_perm_const_ok): Delete.
14733 (rs6000_vectorize_vec_perm_const): New function.
14734 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14735 an element count and rtx array.
14736 (rs6000_expand_extract_even): Update call accordingly.
14737 (rs6000_expand_interleave): Likewise.
14738 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14739 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14740 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14741 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14742 (rs6000_expand_vec_perm_const): Delete.
14743 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14744 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14745 (altivec_expand_vec_perm_const_le): Take each operand individually.
14746 Operate on constant selectors rather than rtxes.
14747 (altivec_expand_vec_perm_const): Likewise. Update call to
14748 altivec_expand_vec_perm_const_le.
14749 (rs6000_expand_vec_perm_const): Delete.
14750 (rs6000_vectorize_vec_perm_const_ok): Delete.
14751 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14752 reference to the SPE evmerge intructions.
14753 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14754 an element count and rtx array.
14755 (rs6000_expand_extract_even): Update call accordingly.
14756 (rs6000_expand_interleave): Likewise.
14757 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14758 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14759 new function.
14760 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14764 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14765 vector mode and that that mode matches the mode of the data
14766 being permuted.
14767 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14768 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14769 directly using expand_vec_perm_1 when forcing selectors into
14770 registers.
14771 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14775 * optabs-query.h (can_vec_perm_p): Delete.
14776 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14777 * optabs-query.c (can_vec_perm_p): Split into...
14778 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14779 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14780 particular selector is valid.
14781 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14782 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14783 (vect_grouped_load_supported): Likewise.
14784 (vect_shift_permute_load_chain): Likewise.
14785 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14786 (vect_transform_slp_perm_load): Likewise.
14787 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14788 (vectorizable_bswap): Likewise.
14789 (vect_gen_perm_mask_checked): Likewise.
14790 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14791 implementations of variable permutation vectors into account
14792 when deciding which selector to use.
14793 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14794 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14795 with a false third argument.
14796 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14797 to test whether the constant selector is valid and can_vec_perm_var_p
14798 to test whether a variable selector is valid.
14800 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14802 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14803 * optabs-query.c (can_vec_perm_p): Likewise.
14804 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14805 instead of vec_perm_indices.
14806 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14807 (vect_gen_perm_mask_checked): Likewise,
14808 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14809 (vect_gen_perm_mask_checked): Likewise,
14811 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14813 * optabs-query.h (qimode_for_vec_perm): Declare.
14814 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14815 (qimode_for_vec_perm): ...this new function.
14816 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14818 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14820 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14821 does not have a conditional at the top.
14823 2018-01-02 Richard Biener <rguenther@suse.de>
14825 * ipa-inline.c (big_speedup_p): Fix expression.
14827 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14829 PR target/81616
14830 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14831 for generic 4->6.
14833 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14835 PR target/81616
14836 Generic tuning.
14837 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14838 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14839 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14840 cond_taken_branch_cost 3->4.
14842 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14844 PR tree-optimization/83581
14845 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14846 TODO_cleanup_cfg if any changes have been made.
14848 PR middle-end/83608
14849 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14850 convert_modes if target mode has the right side, but different mode
14851 class.
14853 PR middle-end/83609
14854 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14855 last argument when extracting from CONCAT. If either from_real or
14856 from_imag is NULL, use expansion through memory. If result is not
14857 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14858 the parts directly to inner mode, if even that fails, use expansion
14859 through memory.
14861 PR middle-end/83623
14862 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14863 check for bswap in mode rather than HImode and use that in expand_unop
14864 too.
14865 \f
14866 Copyright (C) 2018 Free Software Foundation, Inc.
14868 Copying and distribution of this file, with or without modification,
14869 are permitted in any medium without royalty provided the copyright
14870 notice and this notice are preserved.