1 2018-05-30 Martin Sebor <msebor@redhat.com>
4 * builtins.c (expand_builtin_stpcpy_1): New function.
5 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6 only if the former succeeds.
8 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
15 * doc/invoke.texi (-flinker-output): Document
17 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
19 * passes.c (ipa_write_summaries): Only modify statements if body
21 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
22 incrementally linking.
23 (ipa_passes): Likewise.
24 * lto-cgraph.c (lto_output_node): When incrementally linking do not
25 pass down resolution info.
26 * common.opt (flag_incremental_link): Update info.
27 * gcc.c (plugin specs): Turn flinker-output=* to
28 -plugin-opt=-linker-output-known
29 * toplev.c (compile_file): Also cut compilation when doing incremental
31 * flag-types. (enum lto_partition_model): Add
32 LTO_LINKER_OUTPUT_NOLTOREL.
33 (invoke.texi): Add -flinker-output docs.
34 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
35 link same way as WPA; do not stream in dead initializers.
37 * dwarf2out.c (dwarf2out_die_ref_for_decl,
38 darf2out_register_external_decl): Support incremental link.
40 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
42 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
44 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
46 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
47 it down to simple_object_copy_lto_debug_sections.
48 (run_gcc): Determine incremental LTO link time and configure
49 lto1 into non-wpa mode, disable renaming of debug sections.
51 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
53 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
54 descriptions of various incorrectly documented functions.
56 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
59 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
60 address check not strict.
62 2018-05-30 Richard Biener <rguenther@suse.de>
64 PR tree-optimization/85964
65 * tracer.c (better_p): Drop initialized count check, we only
66 call the function with initialized counts now.
67 (find_best_successor): Do find a best edge if one
68 has uninitialized count.
69 (find_best_predecessor): Likewise. Do BB frequency check only
70 if count is initialized.
72 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
74 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
75 (aarch64_ldrstr_offset_compare): New.
76 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
78 (aarch64_gen_adjusted_ldpstp): Likewise.
80 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
82 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
83 Check for subset of GENERAL_REGS and FP_REGS.
84 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
87 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
89 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
90 and wi::to_poly_offset. Add the current offset and then check
91 whether the sum fits, rather than using an unchecked addition of
92 a checked term. Check for a shwi rather than a uhwi.
93 * expr.c (get_bit_range): Use tree_to_poly_uint64.
94 (store_constructor): Use poly_int_tree_p.
95 (expand_expr_real_1): Likewise.
96 * function.c (assign_temp): Likewise.
97 * fold-const.c (const_binop): Use poly_int_tree_p and
99 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
101 * ipa-icf-gimple.c (func_checker::compare_operand): Use
102 to_poly_offset for MEM offsets.
103 * ipa-icf.c (sem_variable::equals): Likewise.
104 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
105 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
106 wi::to_poly_offset for BIT_FIELD_REF offsets.
107 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
109 * var-tracking.c (emit_note_insn_var_location): Use
112 2018-05-29 Jim Wilson <jimw@sifive.com>
114 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
116 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
119 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
120 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
122 (sse4_1_round<mode>2): Use nonimmediate_operand
123 for operand 1 predicate.
125 2018-05-29 Martin Sebor <msebor@redhat.com>
126 Richard Biener <rguenther@suse.de>
129 * calls.c (get_size_range): Call determine_value_range instead
131 * tree-vrp.h (determine_value_range): Declared new function.
132 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
134 2018-05-29 Richard Biener <rguenther@suse.de>
136 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
137 sure to use non-pattern stmts for get_earlier_stmt arguments.
138 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
139 called on pattern stmts.
140 (get_later_stmt): Likewise.
142 2018-05-29 Martin Liska <mliska@suse.cz>
144 PR gcov-profile/85759
145 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
148 2018-05-29 Jakub Jelinek <jakub@redhat.com>
150 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
152 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
156 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
157 VEC_PACK_FLOAT_EXPR): New tree codes.
158 * tree-pretty-print.c (op_code_prio): Handle
159 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
160 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
161 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
162 * tree-inline.c (estimate_operator_cost): Likewise.
163 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
164 * fold-const.c (const_binop): Likewise.
165 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
166 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
167 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
168 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
169 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
170 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
171 * expr.c (expand_expr_real_2): Likewise.
172 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
173 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
174 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
176 * optabs.c (expand_widen_pattern_expr): For
177 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
178 sign from result type rather than operand's type.
179 (expand_binop_directly): For vec_packu_float_optab and
180 vec_packs_float_optab allow result type to be different from operand's
182 * optabs-tree.c (optab_for_tree_code): Handle
183 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
184 VEC_PACK_FLOAT_EXPR. Formatting fixes.
185 * tree-vect-generic.c (expand_vector_operations_1): Handle
186 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
188 * tree-vect-stmts.c (supportable_widening_operation): Handle
190 (supportable_narrowing_operation): Handle FLOAT_EXPR.
191 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
192 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
193 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
194 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
196 (vec_pack<floatprefix>_float_<mode>): New expander.
197 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
199 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
200 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
201 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
202 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
203 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
205 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
206 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
207 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
208 VEC_PACK_FLOAT_EXPR): Document.
210 2018-05-29 Richard Biener <rguenther@suse.de>
212 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
214 (stmt_vec_info_vec): Make pointer.
215 (init_stmt_vec_info_vec): Remove.
216 (free_stmt_vec_info_vec): Likewise.
217 (set_stmt_vec_info_vec): New function.
218 (free_stmt_vec_infos): Likewise.
219 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
220 (set_vinfo_for_stmt): Likewise.
221 (get_earlier_stmt): Likewise.
222 (get_later_stmt): Likewise.
223 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
224 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
225 (vec_info::~vec_info): Free stmt_vec_infos.
226 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
227 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
228 (pass_slp_vectorize::execute): Likewise.
229 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
230 (free_stmt_vec_info_vec): Likewise.
231 (set_stmt_vec_info_vec): New function.
232 (free_stmt_vec_infos): Likewise.
233 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
234 the global stmt_vec_info_vec.
235 * tree-parloops.c (gather_scalar_reductions): Use
236 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
239 2018-05-29 Richard Biener <rguenther@suse.de>
241 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
243 2018-05-29 Martin Liska <mliska@suse.cz>
244 David Malcolm <dmalcolm@redhat.com>
246 * vec.c (test_reverse): New.
247 (vec_c_tests): Add new test.
248 * vec.h (vl_ptr>::reverse): New function.
250 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
252 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
254 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
257 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
259 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
261 2018-05-28 Richard Biener <rguenther@suse.de>
263 PR tree-optimization/85933
264 * tree-vect-data-refs.c (vect_record_base_alignments): Only
265 look at stmts marked as vectorizable.
267 2018-05-28 Richard Biener <rguenther@suse.de>
269 PR tree-optimization/85934
270 * tree-vect-generic.c (expand_vector_operations_1): Hoist
271 vector boolean check before scalar optimization.
273 2018-05-28 Jakub Jelinek <jakub@redhat.com>
275 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
278 2018-05-28 Mark Wielaard <mark@klomp.org>
280 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
281 if it is an expression containing a minus sign.
283 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
285 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
287 2018-05-27 Paul Koning <ni1d@arrl.net>
289 * config/pdp11/pdp11.md (truncsihi2): Remove.
291 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
292 Chung-Ju Wu <jasonwucj@gmail.com>
294 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
296 (unaligned_store_dw): Ditto.
297 * config/nds32/nds32-memory-manipulation.c
298 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
299 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
300 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
301 (emit_setmem_word_loop): Rename to ...
302 (emit_setmem_doubleword_loop): ... this.
303 (nds32_gen_dup_4_byte_to_word_value): New function.
304 (nds32_gen_dup_8_byte_to_double_word_value): New function.
305 (nds32_expand_setmem_loop): Refine implementation.
306 (nds32_expand_setmem_loop_v3m): Ditto.
307 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
310 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
312 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
314 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
316 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
317 (nds32_init_machine_status): Initialize machine->attr_naked_p and
318 machine->attr_no_prologue_p.
319 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
320 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
321 (nds32_expand_epilogue): Consider attr_naked_p.
322 (nds32_expand_epilogue_v3pop): Likewise.
323 (nds32_can_use_return_insn): Likewise.
324 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
325 attr_no_prologue_p fields.
326 * config/nds32/nds32.opt (mret-in-naked-func): New option.
328 2018-05-27 Jakub Jelinek <jakub@redhat.com>
331 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
334 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
336 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
338 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
340 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
342 (*<floatsuffix>floatv2div2sf2): Rename to ...
343 (*float<floatunssuffix>v2div2sf2): ... this.
344 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
345 (float<floatunssuffix>v2div2sf2_mask): ... this.
346 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
347 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
348 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
350 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
352 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
354 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
356 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
358 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
360 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
361 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
362 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
363 gen_ufix_truncv8dfv8si2.
364 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
365 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
366 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
367 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
368 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
369 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
370 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
371 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
373 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
377 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
379 2018-05-25 Jim Wilson <jimw@sifive.com>
381 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
382 * config/riscv/riscv.c (struct machine_function): Add
383 interrupt_handler_p and attribute_checked_p fields.
384 (riscv_attribute_table): Add interrupt.
385 (riscv_interrupt_type_p): New.
386 (riscv_save_reg_p): Save extra regs for interrupt handler.
387 (riscv_use_save_libcall): Return false for interrupt handler.
388 (riscv_first_stack_step): Add forward declaration.
389 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
390 for interrupt handler with large frame. Use it for saved reg list.
391 (riscv_expand_prologue): Move flag_stack_usage_info support to
392 eliminate duplication.
393 (riscv_expand_epilogue): Generate mret for interrupt handler.
394 (riscv_epilogue_uses): New.
395 (riscv_can_use_return_insn): Return false for interrupt handler.
396 (riscv_function_ok_for_sibcall): Likewise.
397 (riscv_set_current_function): Add interrupt handler support.
398 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
399 * config/riscv/riscv.md (UNSPECV_MRET): New.
401 (riscv_frflags, riscv_fsflags): Use tab after opcode.
403 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
405 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
407 PR tree-optimization/85712
408 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
409 this candidate has already been replaced in-situ by a copy.
411 2018-05-25 Jason Merrill <jason@redhat.com>
413 PR c++/80485 - inline function non-zero address.
414 * symtab.c (nonzero_address): Check DECL_COMDAT.
416 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
419 * config/alpha/alpha.md (ashlsi3): New insn pattern.
420 (*ashlsi_se): Rename from *ashldi_se. Define as sign
421 extension of SImode operation. Use const123_operand predicate.
423 (*saddl_se_1): Ditto.
425 (*ssubl_se_1): Ditto.
426 * config/alpha/predicates.md (const123_operand): New predicate.
427 * config/alpha/constraints.md (P): Use IN_RANGE.
429 2018-05-25 Richard Biener <rguenther@suse.de>
431 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
433 (ref_maybe_used_by_stmt_p): Likewise.
434 (stmt_may_clobber_ref_p): Likewise.
435 (stmt_may_clobber_ref_p_1): Likewise.
436 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
438 (ref_maybe_used_by_stmt_p): Likewise.
439 (stmt_may_clobber_ref_p): Likewise.
440 (stmt_may_clobber_ref_p_1): Likewise.
441 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
442 the alias oracle to disambiguate DRs with stmts DR analysis
444 (vect_analyze_data_refs): Do not give up on not analyzable
445 DRs for BB vectorization. Remove code truncating the dataref
448 2018-05-25 Jakub Jelinek <jakub@redhat.com>
451 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
452 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
453 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
455 2018-05-25 Richard Biener <rguenther@suse.de>
457 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
458 function, combining stmt data ref gathering and fatal analysis
460 (vect_analyze_data_refs): Remove now redudnant code and simplify.
461 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
462 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
463 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
464 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
466 2018-05-25 Bin Cheng <bin.cheng@arm.com>
468 PR tree-optimization/85720
469 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
470 SCC if all partitions are builtins.
471 (version_loop_by_alias_check): New parameter. Generate cancelable
472 runtime alias check if all partitions are builtins.
473 (distribute_loop): Update call to above function.
475 2018-05-25 Bin Cheng <bin.cheng@arm.com>
477 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
478 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
479 (parm_default_def_partition_arg): Ditto.
480 (set_parm_default_def_partition): Ditto.
481 (get_parm_default_def_partitions): Ditto and make it static.
482 (get_undefined_value_partitions): Ditto and make it static.
483 (remove_ssa_form): Refactor call to init_var_map here.
484 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
485 computation for loop region.
486 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
487 (register_default_def): Delete.
488 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
489 (parm_default_def_partition_arg): Ditto.
490 (set_parm_default_def_partition): Ditto.
491 (get_parm_default_def_partitions): Ditto and make it static.
492 (get_undefined_value_partitions): Ditto and make it static.
493 (coalesce_with_default, coalesce_with_default): Update comment.
494 (create_coalesce_list_for_region): New func factored out from
495 create_outofssa_var_map.
496 (populate_coalesce_list_for_outofssa): New func factored out from
497 create_outofssa_var_map and coalesce_ssa_name.
498 (create_outofssa_var_map): Delete.
499 (coalesce_ssa_name): Refactor to support live range computation.
500 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
501 (get_parm_default_def_partitions): Delete.
502 (get_undefined_value_partitions): Ditto.
503 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
504 computation for loop region.
505 (new_tree_live_info, loe_visit_block): Ditto.
506 (live_worklist, set_var_live_on_entry): Ditto.
507 (calculate_live_on_exit, verify_live_on_entry): Ditto.
508 * tree-ssa-live.h (struct _var_map): New fields.
509 (init_var_map): Change decl.
510 (region_contains_p): New.
512 2018-05-25 Bin Cheng <bin.cheng@arm.com>
514 * tree-ssa-live.h (live_merge_and_clear): Delete.
516 2018-05-25 Richard Biener <rguenther@suse.de>
519 * tree-dump.c (dequeue_and_dump): Remove access to removed
520 operand 2 of a SWITCH_EXPR.
522 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
524 * doc/sourcebuild.texi (vect_double_cond_arith): Include
525 multiplication and division.
526 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
527 (cond_udiv@var{m}, cond_umod@var{m}): Document.
528 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
529 (cond_udiv_optab, cond_umod_optab): New optabs.
530 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
531 (IFN_COND_RDIV): New internal functions.
532 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
533 TRUNC_MOD_EXPR and RDIV_EXPR.
534 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
535 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
537 (SVE_INT_BINARY): Include mult.
538 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
539 (optab, sve_int_op): Handle mult.
540 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
542 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
543 for SVE_INT_BINARY_SD.
545 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
547 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
548 (optab, sve_int_op): Handle div and udiv.
549 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
550 for SVE_INT_BINARY_SD.
551 (*<optab><mode>3): New insn for the same.
553 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
555 * tree-vect-patterns.c: Include predict.h.
556 (vect_recog_divmod_pattern): Restrict check for division support
557 to when optimizing for size.
559 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
561 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
562 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
563 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
564 (gimple_match_op::set_op): Likewise.
565 (gimple_resimplify4): Declare.
566 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
567 (expr::gen_transform): Likewise.
568 (decision_tree::gen): Generate a simplification routine for 4 operands.
569 * gimple-match-head.c (gimple_simplify): Add an overload for
570 4 operands. In the top-level function, handle up to 4 call
571 arguments and call gimple_resimplify4.
572 (gimple_resimplify4): New function.
573 (build_call_internal): Pass a fourth operand.
574 (maybe_push_to_seq): Likewise.
575 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
576 Fold VEC_COND_EXPRs of an operation and a default value into
577 an IFN_COND_* function if possible.
578 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
580 (SVE_COND_FP_BINARY): Include them.
581 (optab, sve_fp_op): Handle them.
582 (SVE_INT_BINARY_REV): New code iterator.
583 (SVE_COND_FP_BINARY_REV): New int iterator.
584 (commutative): New int attribute.
585 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
587 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
589 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
590 (*cond_<optab><mode>): New patterns for reversed operands.
592 2018-05-25 Richard Biener <rguenther@suse.de>
594 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
595 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
596 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
597 (STMT_VINFO_GROUPED_ACCESS): Adjust.
598 * tree-vect-data-refs.c (everywhere): Adjust users.
599 * tree-vect-loop.c (everywhere): Likewise.
600 * tree-vect-slp.c (everywhere): Likewise.
601 * tree-vect-stmts.c (everywhere): Likewise.
602 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
604 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
606 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
608 (gcc_cv_as_section_exclude): ... this.
609 Try Solaris as #exclude syntax.
610 * configure: Regenerate.
611 * config.in: Regenerate.
612 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
614 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
615 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
617 * varasm.c (default_elf_asm_named_section): Don't check if
618 HAVE_GAS_SECTION_EXCLUDE is defined.
620 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
622 * doc/md.texi: Update the documentation of the cond_* optabs
623 to mention the new final operand. Fix GET_MODE_NUNITS call.
624 Describe the scalar case too.
625 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
626 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
628 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
629 (get_conditional_internal_fn): Update comment.
630 * tree-vect-loop.c (vectorizable_reduction): Pass the original
631 accumulator value as a final argument to conditional functions.
632 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
633 a define_expand and add an "else" operand. Assert for now that
634 the else operand is equal to operand 2. Use SVE_INT_BINARY and
635 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
636 (*cond_<optab><mode>): New patterns.
637 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
638 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
639 (UNSPEC_COND_EOR): Delete.
640 (optab): Remove associated mappings.
641 (SVE_INT_BINARY): New code iterator.
642 (sve_int_op): Remove int attribute and add "minus" to the code
644 (SVE_COND_INT_OP): Delete.
645 (SVE_COND_FP_OP): Rename to...
646 (SVE_COND_FP_BINARY): ...this.
648 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
650 * optabs.c (can_reuse_operands_p): New function.
651 (maybe_legitimize_operands): Try to reuse the results for
654 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
656 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
657 Add {q} suffix to insn mnemonic.
659 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
661 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
662 (msp430_warn_func_return): New.
664 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
666 * fold-const.c (tree_nonzero_bits): New function.
667 * fold-const.h (tree_nonzero_bits): Likewise.
668 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
669 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
671 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
675 * varasm.c (assemble_alias): Check ifunc_resolver only on
678 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
681 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
682 when memory input operand is handled.
684 2018-05-24 Luis Machado <luis.machado@linaro.org>
686 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
688 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
690 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
692 * match.pd: Delay FMA folds until after vectorization.
694 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
697 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
698 address check not strict.
700 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
702 * gimple-match.h (gimple_match_op): New class.
703 (mprts_hook): Replace parameters with a gimple_match_op *.
704 (maybe_build_generic_op): Likewise.
705 (gimple_simplified_result_is_gimple_val): Replace parameters with
706 a const gimple_match_op *.
707 (gimple_simplify): Replace code_helper * and tree * parameters with
708 a gimple_match_op * parameter.
709 (gimple_resimplify1): Replace code_helper *, tree and tree *
710 parameters with a gimple_match_op * parameter.
711 (gimple_resimplify2): Likewise.
712 (gimple_resimplify3): Likewise.
713 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
714 parameters with a gimple_match_op * parameter.
715 * gimple-match-head.c (gimple_simplify): Change prototypes of
716 auto-generated functions to take a gimple_match_op * instead of
717 separate code_helper * and tree * parameters. Make the same
718 change in the top-level overload and update calls to the
719 gimple_resimplify routines. Update calls to the auto-generated
720 functions and to maybe_push_res_to_seq in the publicly-facing
721 operation-specific gimple_simplify overloads.
722 (gimple_match_op::MAX_NUM_OPS): Define.
723 (gimple_resimplify1): Replace rcode and ops with a single res_op
724 parameter. Update call to gimple_simplify.
725 (gimple_resimplify2): Likewise.
726 (gimple_resimplify3): Likewise.
727 (mprts_hook): Replace parameters with a gimple_match_op *.
728 (maybe_build_generic_op): Likewise.
729 (build_call_internal): Replace type, nargs and ops with
731 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
732 with a single gimple_match_op *. Update calls to mprts_hook,
733 build_call_internal and gimple_simplified_result_is_gimple_val.
734 Factor out code that is common to the tree_code and combined_fn cases.
735 * genmatch.c (expr::gen_transform): Replace tem_code and
736 tem_ops with a gimple_match_op called tem_op. Update calls
737 to the gimple_resimplify functions and maybe_push_res_to_seq.
738 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
739 res_ops. Update call to the gimple_resimplify functions.
740 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
741 (decision_tree::gen): Make the functions take a gimple_match_op *
742 called res_op instead of separate res_code and res_ops parameters.
743 Update call accordingly.
744 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
745 and ops with a single res_op parameter. Update calls to
746 maybe_build_generic_op and maybe_push_res_to_seq.
747 (fold_stmt_1): Update calls to gimple_simplify and
748 replace_stmt_with_simplification.
749 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
750 and gimple_simplified_result_is_gimple_val.
751 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
753 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
754 with a gimple_match_op *.
755 (vn_nary_build_or_lookup): Likewise. Update call to
756 vn_nary_build_or_lookup_1.
757 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
758 gimple_match_op *. Update calls to the gimple_resimplify routines
759 and to gimple_simplified_result_is_gimple_val.
760 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
761 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
762 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
763 (visit_nary_op): Likewise.
764 (visit_reference_op_load): Likewise.
766 2018-05-23 Luis Machado <luis.machado@linaro.org>
768 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
769 modifier for printing the step amount.
771 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
774 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
777 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
779 * doc/sourcebuild.texi (Endianness): New subsubsection.
781 2018-05-23 Luis Machado <luis.machado@linaro.org>
783 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
784 <prefetch_dynamic_strides>: New const bool field.
785 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
786 prefetch_dynamic_strides.
787 (exynosm1_prefetch_tune): Likewise.
788 (thunderxt88_prefetch_tune): Likewise.
789 (thunderx_prefetch_tune): Likewise.
790 (thunderx2t99_prefetch_tune): Likewise.
791 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
793 (aarch64_override_options_internal): Update to set
794 PARAM_PREFETCH_DYNAMIC_STRIDES.
795 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
796 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
797 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
798 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
799 prefetch-dynamic-strides setting.
801 2018-05-23 Luis Machado <luis.machado@linaro.org>
803 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
804 <minimum_stride>: New const int field.
805 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
806 minimum_stride field defaulting to -1.
807 (exynosm1_prefetch_tune): Likewise.
808 (thunderxt88_prefetch_tune): Likewise.
809 (thunderx_prefetch_tune): Likewise.
810 (thunderx2t99_prefetch_tune): Likewise.
811 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
812 <default_opt_level>: Set to 3.
813 (aarch64_override_options_internal): Update to set
814 PARAM_PREFETCH_MINIMUM_STRIDE.
815 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
816 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
817 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
818 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
819 stride is constant and is below the minimum stride threshold.
821 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
823 * config/arm/arm-cpus.in (mode26): Delete.
824 (armv4): Delete mode26 reference.
825 * config/arm/arm.c (arm_configure_build_target): Delete use of
828 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
830 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
832 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
833 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
835 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
836 (floatunsdidf2): Ditto.
838 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
840 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
841 (fixuns_trunc<mode>si2_avx512f): Ditto.
842 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
843 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
844 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
846 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
848 PR rtl-optimization/79985
849 * df-scan.c (df_insn_refs_collect): Remove special case for
850 global registers and asm statements.
852 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
854 * extend.texi (Global Register Variables): Rewrite the bullet list.
855 Note that the register is available for allocation. Note that access
856 via inline asm must use constraints. Add note about async-signal
857 handlers. Remove paragraph about automagic register selection.
859 2018-05-23 Richard Biener <rguenther@suse.de>
861 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
862 of fixed offset from memset VN.
864 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
866 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
868 (alloc_cand_and_find_basis): Initialize first_interp field.
869 (slsr_process_mul): Modify first_interp field.
870 (slsr_process_add): Likewise.
871 (slsr_process_cast): Modify first_interp field for each new
873 (slsr_process_copy): Likewise.
874 (dump_candidate): Dump first_interp field.
875 (replace_mult_candidate): Process all interpretations, not just
877 (replace_rhs_if_not_dup): Likewise.
878 (replace_one_candidate): Likewise.
880 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
882 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
884 (aarch64_needs_frame_chain): New function.
885 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
887 2018-05-23 Sudakshina Das <sudi.das@arm.com>
890 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
891 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
892 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
893 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
894 as true for strict-align.
895 (aarch64_can_inline_p): Perform checks even when callee has no
896 attributes to check for strict alignment.
897 * doc/extend.texi (AArch64 Function Attributes): Document
899 * doc/invoke.texi: (AArch64 Options): Likewise.
901 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
903 PR tree-optimization/85853
904 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
905 the handling of the root of the node to...
906 (vect_slp_analyze_node_operations_1): ...this new function,
907 and run the whole thing with the child nodes' def types
908 set according to their SLP node's def type.
910 2018-05-23 Richard Biener <rguenther@suse.de>
913 * tree-data-ref.c (create_runtime_alias_checks): Defer
914 and ignore overflow warnings.
916 2018-05-23 Yury Gribov <tetra2005@gmail.com>
918 PR tree-optimization/85822
919 * tree-vrp.c (is_masked_range_test): Fix handling of negative
922 2018-05-23 Richard Biener <rguenther@suse.de>
924 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
925 memset constants via native_interpret_expr.
927 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
930 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
932 (cgraph_node::create_alias): Likewise.
933 (cgraph_node::get_availability): Check ifunc_resolver instead
934 of looking up ifunc attribute.
935 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
936 * varasm.c (do_assemble_alias): Likewise.
937 (assemble_alias): Likewise.
938 (default_binds_local_p_3): Likewise.
939 * cgraph.h (cgraph_node): Add ifunc_resolver.
940 (cgraph_node::only_called_directly_or_aliased_p): Return false
942 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
944 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
945 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
946 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
947 instead of looking up ifunc attribute.
949 2018-05-22 Luis Machado <luis.machado@linaro.org>
951 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
953 2018-05-22 Martin Sebor <msebor@redhat.com>
956 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
957 only when expasion succeeds.
958 (expand_builtin_strcmp): Same.
959 (expand_builtin_strncmp): Same.
961 2018-05-22 Martin Sebor <msebor@redhat.com>
963 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
965 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
966 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
968 * config/aarch64/aarch64-ldpstp.md: Replace uses of
969 aarch64_mem_pair_operand with memory_operand and delete operand swapping
971 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
972 Add check for legitimate_address.
973 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
974 (aarch64_swap_ldrstr_operands): New.
975 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
978 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
979 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
981 * config/aarch64/aarch64.md: New patterns to generate stp
983 (store_pair_sw, store_pair_dw): New patterns to generate stp for
984 single words and double words.
985 (load_pair_sw, load_pair_dw): Likewise.
986 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
988 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
990 * config/aarch64/aarch64-ldpstp.md: Modify peephole
991 for different mode ldpstp and add peephole for merged zero stores.
993 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
995 (aarch64_gen_store_pair): Rename calls to match new patterns.
996 (aarch64_gen_load_pair): Rename calls to match new patterns.
997 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
998 (load_pair<DREG:mode><DREG2:mode>): ... This.
999 (store_pair<mode>): Rename to...
1000 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1001 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1003 (V_INT_EQUIV): Handle SImode.
1004 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1007 2018-05-22 Martin Sebor <msebor@redhat.com>
1010 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1011 or ajust the presumed bound on an operation to avoid unnecessary
1014 2018-05-22 Martin Sebor <msebor@redhat.com>
1016 PR tree-optimization/85826
1017 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1018 assuming that a DECL necesarily has a constant size.
1020 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1023 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1025 2018-05-22 Richard Biener <rguenther@suse.de>
1027 PR tree-optimization/85834
1028 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1029 non-constant and non-zero memset arguments.
1031 2018-05-22 Martin Liska <mliska@suse.cz>
1034 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1036 2018-05-22 Richard Biener <rguenther@suse.de>
1038 PR tree-optimization/85863
1039 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1040 comparisons when vectype is specified.
1041 (vectorizable_condition): Do not specify vectype for
1042 vect_is_simple_cond when SLP vectorizing.
1044 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1047 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1048 define __ibm128 as long double.
1049 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1050 as a distinct type when IEEE 128-bit support is enabled.
1051 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1052 128-bit types to use the correct functions.
1053 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1054 convert between 128-bit floating point types that have different
1055 modes but the same representation, instead of using gen_lowpart to
1057 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1059 (IFKF_reg): New attributes to give the register constraints for
1061 (extend<mode>tf2_internal): New insns to mark an explicit
1062 conversion between 128-bit floating point types that have a
1063 different mode but share the same representation.
1065 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1067 PR tree-optimization/85814
1068 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1069 a null return from get_strinfo when unsharing the next
1070 strinfo in the chain.
1072 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1075 * varasm.c (weak_finish): Clean up weak_decls.
1077 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1079 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1080 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1081 UNSPEC_UADALP values.
1082 * config/aarch64/iterators.md (ABAL): New int iterator.
1085 (sur): Add mappings for the above.
1086 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1088 (aarch64_<sur>abal<mode>_4): Likewise.
1089 (aarch64_<sur>adalp<mode>_3): Likewise.
1090 (<sur>sadv16qi): New define_expand.
1092 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1094 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1095 (*movdf_internal): Ditto.
1096 (*rcpsf2_sse): Ditto.
1097 (*rsqrtsf2_sse): Ditto.
1098 (*sqrt<mode>2_sse): Ditto.
1100 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1102 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1104 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1105 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1106 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1107 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1109 * config/aarch64/arm_neon.h: Likewise.
1110 * config/aarch64/iterators.md (VQ_I): New.
1112 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1114 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1116 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1119 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1121 * config/nds32/constraints.md (S): New constraint.
1122 * config/nds32/nds32.md (call_internal): Use constraint S.
1123 (call_value_internal): Likewise.
1124 (sibcall_internal): Likewise.
1125 (sibcall_value_internal): Likewise.
1127 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1128 Chung-Ju Wu <jasonwucj@gmail.com>
1130 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1133 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1134 Chung-Ju Wu <jasonwucj@gmail.com>
1136 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1137 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1138 (nds32_rtx_costs_impl): Simplify.
1139 (nds32_address_cost_impl): Simplify.
1140 (nds32_init_rtx_costs): New function.
1141 (nds32_rtx_costs_speed_prefer): Likewise.
1142 (nds32_rtx_costs_size_prefer): Likewise.
1143 (nds32_address_cost_speed_prefer): Likewise.
1144 (nds32_address_cost_speed_fwprop): Likewise.
1145 (nds32_address_cost_size_prefer): Likewise.
1146 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1147 * config/nds32/nds32.c (nds32_option_override): Use
1148 nds32_init_rtx_costs function.
1150 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1152 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1153 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1154 (TARGET_PIPELINE_N8): Likewise.
1155 (TARGET_PIPELINE_N10): Likewise.
1156 (TARGET_PIPELINE_N13): Likewise.
1157 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1159 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1161 * config/nds32/nds32-fpu.md: Update copyright year.
1163 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1165 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1167 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1169 * config/nds32/nds32.c
1170 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1171 * config/nds32/nds32.opt (minline-asm-r15): New option.
1173 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1175 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1177 * config/nds32/nds32.md (abssi2): New pattern.
1179 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1181 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1182 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1183 Merge insn pattern from sse_cvtsi2ss<round_name> and
1184 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1185 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1186 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1187 using SWI48 mode iterator.
1188 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1189 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1190 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1191 pattern from sse_cvttss2si<round_saeonly_name>
1192 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1193 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1194 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1195 using SWI48 mode iterator.
1196 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1197 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1198 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1199 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1200 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1201 using SWI48 mode iterator.
1202 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1203 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1204 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1205 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1206 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1207 SWI48 mode iterator.
1208 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1209 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1210 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1211 pattern from sse_cvttsd2si<round_saeonly_name>
1212 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1214 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1216 * config/nds32/nds32-md-auxiliary.c
1217 (nds32_valid_smw_lwm_base_p): Refine.
1218 (nds32_output_smw_single_word): Refine.
1219 (nds32_output_smw_double_word): New.
1220 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1222 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1224 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1225 (nds32_output_stack_pop): Refine.
1226 (nds32_expand_unaligned_load): Refine.
1227 (nds32_expand_unaligned_store): Refine.
1229 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1230 Chung-Ju Wu <jasonwucj@gmail.com>
1232 * config/nds32/constants.md: Add TP_REGNUM constant.
1233 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1234 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1236 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1237 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1238 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1239 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1240 and PIC code generation.
1241 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1243 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1245 * config/nds32/nds32.md: Support TLS and PIC.
1246 * config/nds32/nds32.c: Support TLS and PIC.
1247 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1248 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1251 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1253 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1254 mode with E_ prefix.
1256 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1257 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1260 * config/nds32/nds32-md-auxiliary.c
1261 (symbolic_reference_mentioned_p): New.
1262 (nds32_legitimize_ict_address): New.
1263 (nds32_expand_ict_move): New.
1264 (nds32_indirect_call_referenced_p): New.
1265 (nds32_symbol_binds_local_p): Delete.
1266 (nds32_long_call_p): Modify.
1267 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1268 * config/nds32/nds32-protos.h
1269 (symbolic_reference_mentioned_p): Declare.
1270 (nds32_legitimize_ict_address): Declare.
1271 (nds32_expand_ict_move): Declare.
1272 (nds32_indirect_call_referenced_p): Declare.
1273 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1274 (nds32_relax_group): Use nds32_ict_const_p as condition.
1275 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1276 (nds32_asm_file_start): Output ict_model directive in asm code.
1277 (nds32_legitimate_address_p): Consider indirect call.
1278 (nds32_print_operand): Consider indirect call.
1279 (nds32_print_operand_address): Consider indirect call.
1280 (nds32_insert_attributes): Handle "indirect_call" attribute.
1281 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1282 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1283 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1284 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1285 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1286 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1287 (TARGET_ICT_MODEL_SMALL): Define.
1288 (TARGET_ICT_MODEL_LARGE): Define.
1289 * config/nds32/nds32.md (movsi): Consider ict model.
1290 (call, call_value): Consider ict model.
1291 (sibcall, sibcall_value): Consider ict model.
1292 * config/nds32/nds32.opt (mict-model): New option.
1293 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1296 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1297 Monk Chiang <sh.chiang04@gmail.com>
1298 Jim Wilson <jimw@sifive.com>
1300 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1301 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1302 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1303 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1304 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1305 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1306 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1307 compute save_libcall_adjustment properly.
1308 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1309 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1310 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1311 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1312 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1313 (ABI_SPEC): Handle mabi=ilp32e.
1314 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1315 (RVE): Add RVE mask.
1316 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1317 <-march>: Add rv32e as an example.
1319 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1322 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1323 (intra_create_variable_infos): Handle C++ constructors.
1325 2018-05-18 Martin Liska <mliska@suse.cz>
1327 * passes.def: Remove a redundant pass.
1329 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1332 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1334 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1336 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1338 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1340 (armv2, armv2a, armv3, armv3m): Delete architectures.
1341 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1342 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1343 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1345 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1346 (*mulsidi3adddi): Likewise.
1347 (mulsidi3): Likewise.
1348 (*mulsidi3_nov6): Likewise.
1349 (umulsidi3): Likewise.
1350 (umulsidi3_nov6): Likewise.
1351 (umaddsidi4): Likewise.
1352 (*umulsidi3adddi): Likewise.
1353 (smulsi3_highpart): Likewise.
1354 (*smulsi3_highpart_nov6): Likewise.
1355 (umulsi3_highpart): Likewise.
1356 (*umulsi3_highpart_nov6): Likewise.
1357 * config/arm/arm.h (arm_arch3m): Delete.
1358 * config/arm/arm.c (arm_arch3m): Delete.
1359 (arm_option_override_internal): Update armv3-related comment.
1360 (arm_configure_build_target): Delete use of isa_bit_mode32.
1361 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1362 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1363 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1365 (mulusa3): Likewise.
1366 * config/arm/arm-protos.h (arm_arch3m): Delete.
1367 * config/arm/arm-tables.opt: Regenerate.
1368 * config/arm/arm-tune.md: Likewise.
1369 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1370 deleted architectures.
1372 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1374 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1375 (armv5t, armv5te): New features.
1376 (ARMv5, ARMv5e): Delete fgroups.
1377 (ARMv5t, ARMv5te): Adjust for above changes.
1379 (armv5, armv5e): Delete arches.
1380 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1382 (*call_reg_arm): Likewise.
1383 (*call_value_reg_armv5): Likewise.
1384 (*call_value_reg_arm): Likewise.
1385 (*call_symbol): Likewise.
1386 (*call_value_symbol): Likewise.
1387 (*sibcall_insn): Likewise.
1388 (*sibcall_value_insn): Likewise.
1390 (prefetch): Likewise.
1391 (define_split and define_peephole2 dependent on arm_arch5):
1393 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1395 (TARGET_ARM_QBIT): Likewise.
1396 (TARGET_DSP_MULTIPLY): Likewise.
1397 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1398 (arm_arch5, arm_arch5e): Delete.
1399 (arm_arch5t, arm_arch5te): Declare.
1400 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1401 (arm_arch5t): Declare.
1402 (arm_option_reconfigure_globals): Update for the above.
1403 (arm_options_perform_arch_sanity_checks): Update comment, replace
1404 use of arm_arch5 with arm_arch5t.
1405 (use_return_insn): Likewise.
1406 (arm_emit_call_insn): Likewise.
1407 (output_return_instruction): Likewise.
1408 (arm_final_prescan_insn): Likewise.
1409 (arm_coproc_builtin_available): Likewise.
1410 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1411 arm_arch5e with arm_arch5t and arm_arch5te.
1412 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1413 (arm_arch5t, arm_arch5te): Declare.
1414 * config/arm/arm-tables.opt: Regenerate.
1415 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1416 * config/arm/t-multilib: Likewise.
1417 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1418 instead of arm_arch5.
1419 (*call_reg_thumb1): Likewise.
1420 (*call_value_reg_thumb1_v5): Likewise.
1421 (*call_value_reg_thumb1): Likewise.
1422 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1424 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1426 2018-05-18 Martin Liska <mliska@suse.cz>
1428 PR gcov-profile/84846
1429 * doc/gcov.texi: Document -t option of gcov tool.
1431 2018-05-18 Martin Liska <mliska@suse.cz>
1433 PR gcov-profile/84846
1434 * gcov.c (print_usage): Add new -t option.
1435 (process_args): Handle the option.
1436 (generate_results): Use stdout as output when requested by
1439 2018-05-18 Martin Liska <mliska@suse.cz>
1441 PR gcov-profile/84846
1442 * coverage.c (coverage_init): Write PWD to .gcno file.
1443 * doc/gcov.texi: Document how working directory is printed.
1444 * gcov-dump.c (dump_gcov_file): Print PWD.
1445 * gcov.c (output_intermediate_file): Likewise.
1446 (read_graph_file): Read PWD string.
1447 (output_lines): Print PWD.
1449 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1452 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1453 for retval and return false if all args to phi are zero.
1455 2018-05-18 Richard Biener <rguenther@suse.de>
1457 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1459 (evrp_dom_walker::before_dom_children): Call it.
1461 2018-05-18 Richard Biener <rguenther@suse.de>
1463 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1464 results when processing array refs with variable index.
1466 2018-05-18 Toon Moene <toon@moene.org>
1468 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1469 directly after that of -floop-interchange. Indicate that both
1470 options are enabled by default when specifying -O3.
1472 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1474 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1475 iterator. Delete separate integer-mode vec_set<mode> expander.
1476 (aarch64_simd_vec_setv2di): Delete.
1477 (vec_setv2di): Delete.
1478 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1479 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1480 the "w, r" alternative.
1482 2018-05-18 Martin Liska <mliska@suse.cz>
1484 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1485 * tree-pass.h (make_pass_lower_switch_O0): New function.
1486 * tree-switch-conversion.c (node_has_low_bound): Remove.
1487 (node_has_high_bound): Likewise.
1488 (node_is_bounded): Likewise.
1489 (class pass_lower_switch): Make it a template type and create
1491 (pass_lower_switch::execute): Add template argument.
1492 (make_pass_lower_switch): New function.
1493 (make_pass_lower_switch_O0): New function.
1494 (do_jump_if_equal): Remove.
1495 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1496 all the hard work to tree optimization passes.
1498 2018-05-18 Martin Liska <mliska@suse.cz>
1500 * dbgcnt.c (limit_low): Renamed from limit.
1501 (limit_high): New variable.
1502 (dbg_cnt_is_enabled): Check for upper limit.
1503 (dbg_cnt): Adjust dumping.
1504 (dbg_cnt_set_limit_by_index): Add new argument for high
1506 (dbg_cnt_set_limit_by_name): Likewise.
1507 (dbg_cnt_process_single_pair): Parse new format.
1508 (dbg_cnt_process_opt): Use strtok.
1509 (dbg_cnt_list_all_counters): Remove 'value' and add
1511 * doc/invoke.texi: Document changes.
1513 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1515 * doc/sourcebuild.texi (scalar_all_fma): Document.
1516 * tree.def (FMA_EXPR): Delete.
1517 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1518 * internal-fn.c (ternary_direct): New macro.
1519 (expand_ternary_optab_fn): Likewise.
1520 (direct_ternary_optab_supported_p): Likewise.
1521 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1522 * builtins.c (fold_builtin_fma): Delete.
1523 (fold_builtin_3): Don't call it.
1524 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1525 * expr.c (expand_expr_real_2): Likewise.
1526 * fold-const.c (operand_equal_p): Likewise.
1527 (fold_ternary_loc): Likewise.
1528 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1529 * gimple.c (DEFTREECODE): Likewise.
1530 * gimplify.c (gimplify_expr): Likewise.
1531 * optabs-tree.c (optab_for_tree_code): Likewise.
1532 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1533 * tree-eh.c (operation_could_trap_p): Likewise.
1534 (stmt_could_throw_1_p): Likewise.
1535 * tree-inline.c (estimate_operator_cost): Likewise.
1536 * tree-pretty-print.c (dump_generic_node): Likewise.
1537 (op_code_prio): Likewise.
1538 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1539 * tree-ssa-operands.c (get_expr_operands): Likewise.
1540 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1541 * fold-const-call.h (fold_fma): Delete.
1542 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1543 CFN_FNMA and CFN_FNMS.
1545 * genmatch.c (combined_fn): New enum.
1546 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1547 (commutative_op): New function.
1548 (commutate): Use it. Handle more than 2 operands.
1549 (dt_operand::gen_gimple_expr): Use commutative_op.
1550 (parser::parse_expr): Allow :c to be used with non-binary
1551 operators if the commutative operand is known.
1552 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1553 CFN_FMS, CFN_FNMA and CFN_FNMS.
1554 (backprop::process_assign_use): Remove FMA_EXPR handling.
1555 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1556 (gen_hsa_fma): New function.
1557 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1558 IFN_FNMA and IFN_FNMS.
1559 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1560 * gimple-fold.h (follow_all_ssa_edges): Declare.
1561 * gimple-fold.c (follow_all_ssa_edges): New function.
1562 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1563 gimple_build interface and use follow_all_ssa_edges to fold the result.
1564 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1565 instead of checking for optabs directly.
1566 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1567 rather than FMA_EXPRs.
1568 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1569 call to IFN_FMA instead of an FMA_EXPR.
1571 2018-05-17 Jim Wilson <jimw@sifive.com>
1573 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1574 sign extended promoted subreg, and the range does not have the sign bit
1575 set, then do a sign extend.
1577 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1578 test, check for sign extended subreg and/or constant operands, and
1579 do a sign extend in that case.
1581 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1583 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1584 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1586 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1587 Change logics_shift_reg to logics_shift_imm.
1588 (thunderx2t99_fp_loadpair_basic): Delete.
1589 (thunderx2t99_fp_storepair_basic): Delete.
1590 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1591 (thunderx2t99_asimd_polynomial): Delete.
1592 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1593 and neon_fp_mul_d_scalar_q.
1594 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1595 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1596 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1597 (thunderx2t99_asimd_lut): Add missing tbl types.
1598 (thunderx2t99_asimd_ext): Delete.
1599 (thunderx2t99_asimd_load1_1_mult): Delete.
1600 (thunderx2t99_asimd_load1_2_mult): Delete.
1601 (thunderx2t99_asimd_load1_ldp): New.
1602 (thunderx2t99_asimd_load1): New.
1603 (thunderx2t99_asimd_load2): Add missing *load2* types.
1604 (thunderx2t99_asimd_load3): New.
1605 (thunderx2t99_asimd_load4): New.
1606 (thunderx2t99_asimd_store1_1_mult): Delete.
1607 (thunderx2t99_asimd_store1_2_mult): Delete.
1608 (thunderx2t99_asimd_store2_mult): Delete.
1609 (thunderx2t99_asimd_store2_onelane): Delete.
1610 (thunderx2t99_asimd_store_stp): New.
1611 (thunderx2t99_asimd_store1): New.
1612 (thunderx2t99_asimd_store2): New.
1613 (thunderx2t99_asimd_store3): New.
1614 (thunderx2t99_asimd_store4): New.
1616 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1618 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1619 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1621 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1622 Segher Boessenkool <segher@kernel.crashing.org>
1625 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1628 2018-05-17 Richard Biener <rguenther@suse.de>
1630 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1631 for pruning loop and prune defs feeding only already visited PHIs.
1633 2018-05-17 Richard Biener <rguenther@suse.de>
1635 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1637 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1638 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/85793
1641 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1642 for VMAT_ELEMENTWISE.
1644 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1646 * internal-fn.h (lookup_internal_fn): Declare
1647 * internal-fn.c (lookup_internal_fn): New function.
1648 * gimple.c (gimple_build_call_from_tree): Handle calls to
1650 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1651 internal function names.
1652 * tree-pretty-print.c (dump_generic_node): Likewise.
1653 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1655 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1657 * gimple-fold.h (gimple_build): Make the function forms take
1658 combined_fn rather than built_in_function.
1659 (gimple_simplify): Likewise.
1660 * gimple-match-head.c (gimple_simplify): Likewise.
1661 * gimple-fold.c (gimple_build): Likewise.
1662 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1663 rather than gimple_build_call_internal.
1664 (get_initial_defs_for_reduction): Likewise.
1665 (vect_create_epilog_for_reduction): Likewise.
1666 (vectorizable_live_operation): Likewise.
1668 2018-05-17 Martin Liska <mliska@suse.cz>
1670 * gimple-ssa-sprintf.c (format_directive): Do not use
1671 space in between 'G_' and '('.
1673 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1676 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1677 even if the mask is not all ones.
1680 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1682 (ix86_gimple_fold_builtin): Likewise.
1685 * config/i386/i386.c: Include tree-vector-builder.h.
1686 (ix86_vector_shift_count): New function.
1687 (ix86_fold_builtin): Fold shift builtins by scalar count.
1688 (ix86_gimple_fold_builtin): Likewise.
1690 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1691 _mm512_setzero): New intrinsics.
1693 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1694 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1696 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1697 code generation for cases where splatting a value is not useful.
1698 * simplify-rtx.c (simplify_ternary_operation): Simplify
1699 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1700 mode to a vec_concat.
1702 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1704 * config.gcc: Support "goldmont-plus".
1705 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1707 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1708 PROCESSOR_GOLDMONT_PLUS.
1709 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1710 (processor_target_table): Add "goldmont-plus".
1711 (PTA_GOLDMONT_PLUS): Define.
1712 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1713 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1714 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1715 (fold_builtin_cpu): Add "goldmont-plus".
1716 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1717 (ix86_option_override_internal): Add "goldmont-plus".
1718 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1719 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1720 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1721 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1723 2018-05-17 Richard Biener <rguenther@suse.de>
1725 PR tree-optimization/85757
1726 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1727 remove defs that only feed that PHI from further processing.
1729 2018-05-16 Jim Wilson <jimw@sifive.com>
1731 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1733 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1735 2018-05-16 Mark Wielaard <mark@klomp.org>
1737 * dwarf2out.c (count_index_strings): New function.
1738 (output_indirect_strings): Call count_index_strings and generate
1739 header for dwarf_version >= 5.
1741 2018-05-16 Mark Wielaard <mark@klomp.org>
1743 * dwarf2out.c (dwarf_FORM): New function.
1744 (set_indirect_string): Use dwarf_FORM.
1745 (reset_indirect_string): Likewise.
1746 (size_of_die): Likewise.
1747 (value_format): Likewise.
1748 (output_die): Likewise.
1749 (add_skeleton_AT_string): Likewise.
1750 (output_macinfo_op): Likewise.
1751 (index_string): Likewise.
1752 (output_index_string_offset): Likewise.
1753 (output_index_string): Likewise.
1754 (count_index_strings): Likewise.
1756 2018-05-16 Carl Love <cel@us.ibm.com>
1758 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1759 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1761 2018-05-16 Martin Jambor <mjambor@suse.cz>
1763 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1764 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1766 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1768 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1769 (fnma<mode>4): Likewise.
1770 (fms<mode>4): Likewise.
1771 (fnms<mode>4): Likewise.
1772 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1773 (aarch64_fnma<mode>4): Likewise.
1774 (aarch64_fms<mode>4): Likewise.
1775 (aarch64_fnms<mode>4): Likewise.
1776 (aarch64_fnmadd<mode>4): Likewise.
1778 2018-05-16 Jason Merrill <jason@redhat.com>
1780 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1782 2018-05-16 Richard Biener <rguenther@suse.de>
1784 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1785 (dump_stmt_cost): Declare.
1786 (add_stmt_cost): Dump cost we add.
1787 (add_stmt_costs): New function.
1788 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1790 (vect_analyze_stmt): Adjust prototype.
1791 (vectorizable_condition): Likewise.
1792 (vectorizable_live_operation): Likewise.
1793 (vectorizable_reduction): Likewise.
1794 (vectorizable_induction): Likewise.
1795 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1796 cost vector to pass to vectorizable_ and record afterwards.
1797 (vect_model_reduction_cost): Take cost vector argument and adjust.
1798 (vect_model_induction_cost): Likewise.
1799 (vectorizable_reduction): Likewise.
1800 (vectorizable_induction): Likewise.
1801 (vectorizable_live_operation): Likewise.
1802 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1803 SLP_TREE_NUMBER_OF_VEC_STMTS.
1804 (vect_analyze_slp_cost_1): Remove.
1805 (vect_analyze_slp_cost): Likewise.
1806 (vect_slp_analyze_node_operations): Take visited args and
1807 a target cost vector. Avoid processing already visited stmt sets.
1808 (vect_slp_analyze_operations): Use a local cost vector to gather
1809 costs and register those of non-discarded instances.
1810 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1811 (vect_schedule_slp_instance): Remove copying of
1812 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1814 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1815 adding cost. Record cost entry location.
1816 (vect_prologue_cost_for_slp_op): Function to compute cost of
1817 a constant or invariant generated for SLP vect in the prologue,
1818 split out from vect_analyze_slp_cost_1.
1819 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1820 (vect_model_promotion_demotion_cost): Likewise.
1821 (vect_model_store_cost): Likewise, make static.
1822 (vect_model_load_cost): Likewise.
1823 (vectorizable_bswap): Add cost vector arg and adjust.
1824 (vectorizable_call): Likewise.
1825 (vectorizable_simd_clone_call): Likewise.
1826 (vectorizable_conversion): Likewise.
1827 (vectorizable_assignment): Likewise.
1828 (vectorizable_shift): Likewise.
1829 (vectorizable_operation): Likewise.
1830 (vectorizable_store): Likewise.
1831 (vectorizable_load): Likewise.
1832 (vectorizable_condition): Likewise.
1833 (vectorizable_comparison): Likewise.
1834 (can_vectorize_live_stmts): Likewise.
1835 (vect_analyze_stmt): Likewise.
1836 (vect_transform_stmt): Adjust calls to vectorizable_*.
1837 * tree-vectorizer.c: Include gimple-pretty-print.h.
1838 (dump_stmt_cost): New function.
1840 2018-05-16 Richard Biener <rguenther@suse.de>
1842 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1843 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1844 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1845 (check_name): New callback.
1846 (dse_classify_store): Track cycles via a visited bitmap of PHI
1847 defs and simplify handling of in-loop and across loop dead stores
1848 and properly fail for loop-variant refs. Handle byte-tracking with
1849 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1852 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1854 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1855 (vect_get_mask_type_for_stmt): Likewise.
1856 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1858 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1859 to determine the statement's vector type and the vector type that
1860 should be used for calculating nunits. Deal with cases in which
1861 the type has to be deferred.
1862 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1863 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1864 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1865 (vect_determine_vf_for_stmt): New functions, split out from...
1866 (vect_determine_vectorization_factor): ...here.
1867 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1868 (vect_get_mask_type_for_stmt): New functions, split out from
1869 vect_determine_vectorization_factor.
1871 2018-05-16 Richard Biener <rguenther@suse.de>
1873 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1874 verify the [VEC_]COND_EXPR embedded comparison.
1876 2018-05-15 Martin Sebor <msebor@redhat.com>
1878 PR tree-optimization/85753
1879 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1880 RECORD_TYPE in addition to ARRAY_TYPE.
1882 2018-05-15 Martin Sebor <msebor@redhat.com>
1885 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1887 2018-05-15 Richard Biener <rguenther@suse.de>
1889 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1890 add by_clobber_p one. Change algorithm to collect all defs
1891 representing uses we need to walk and try reducing them to
1892 a single one before failing.
1893 (dse_dom_walker::dse_optimize_stmt): Adjust.
1895 2018-05-13 Mark Wielaard <mark@klomp.org>
1897 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1898 (size_of_loc_descr): Likewise.
1899 (output_loc_operands): Likewise.
1900 (output_loc_operands_raw): Likewise.
1901 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1902 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1903 (hash_loc_operands): Likewise.
1904 (compare_loc_operands): Likewise.
1906 2018-05-14 Mark Wielaard <mark@klomp.org>
1908 * dwarf2out.c (count_index_addrs): New function.
1909 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1911 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1913 PR tree-optimization/83648
1914 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1915 return value as malloc candidate.
1917 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1920 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1921 as true in call to suggest_attribute.
1923 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1925 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1926 -mreadonly-in-sdata.
1928 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1930 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1932 (aarch64_crypto_aesd_fused): Likewise.
1934 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1936 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1937 (movsi_aarch64): Likewise.
1938 (load_pairsi): Likewise.
1939 (load_pairdi): Likewise.
1940 (store_pairsi): Likewise.
1941 (store_pairdi): Likewise.
1942 (load_pairsf): Likewise.
1943 (load_pairdf): Likewise.
1944 (store_pairsf): Likewise.
1945 (store_pairdf): Likewise.
1946 (zero_extend): Likewise.
1947 (trunc): Swap alternatives.
1948 (fcvt_target): Add '?' to prefer w over r.
1950 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1953 * config/i386/i386.md: Disallow non-commutative arithmetics in
1954 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1955 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1956 in the peephole2 before it.
1958 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1960 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1961 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1962 (ix86_handle_option): Handle -mcldemote.
1963 * config.gcc: New header.
1964 * config/i386/cldemoteintrin.h: New file.
1965 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1966 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1968 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1969 OPTION_MASK_ISA_CLDEMOTE.
1970 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1971 (ix86_valid_target_attribute_inner_p): Ditto.
1972 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1973 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1974 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1975 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1976 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1978 * config/i386/i386.opt: Add -mcldemote.
1979 * config/i386/x86intrin.h: New header.
1980 * doc/invoke.texi: Add -mcldemote.
1982 2018-05-14 Richard Biener <rguenther@suse.de>
1984 * doc/match-and-simplify.texi: Adjust :s documentation.
1986 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1988 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1989 intended memcpy size.
1990 (REORDER_45): Likewise.
1992 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1994 * sort.cc: New file.
1995 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1996 * vec.c (qsort_chk): Use gcc_qsort.
1997 * Makefile.in (OBJS-libcommon): Add sort.o.
1998 (build/sort.o): New target. Use it...
1999 (BUILD_RTL): ... here, and...
2000 (build/gencfn-macros): ... here, and...
2001 (build/genmatch): ... here.
2003 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2004 Chung-Ju Wu <jasonwucj@gmail.com>
2006 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2007 * config/nds32/nds32-graywolf.md: New file.
2008 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2009 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2011 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2012 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2013 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2014 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2015 * config/nds32/pipelines.md: Include n15 settings.
2017 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2018 Chung-Ju Wu <jasonwucj@gmail.com>
2020 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2021 * config/nds32/nds32-n13.md: New file.
2022 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2023 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2025 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2026 * config/nds32/nds32.md (pipeline_model): Add n13.
2027 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2028 * config/nds32/pipelines.md: Include n13 settings.
2030 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2031 Chung-Ju Wu <jasonwucj@gmail.com>
2033 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2034 * config/nds32/nds32-n10.md: New file.
2035 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2036 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2038 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2039 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2040 * config/nds32/nds32.md (pipeline_model): Add n10.
2041 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2042 * config/nds32/pipelines.md: Include n10 settings.
2044 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2045 Kito Cheng <kito.cheng@gmail.com>
2046 Chung-Ju Wu <jasonwucj@gmail.com>
2048 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2049 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2050 Add enum values for DSP extension instructions.
2051 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2053 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2054 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2056 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2057 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2058 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2059 * config/nds32/nds32-intrinsic.md: Likewise.
2060 * config/nds32/nds32_intrinsic.h: Likewise.
2061 * config/nds32/nds32-md-auxiliary.c: Likewise.
2062 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2063 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2064 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2065 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2066 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2067 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2068 TYPE_DMAC in switch statement.
2069 * config/nds32/nds32.c: New checking and implementation for DSP
2070 extension instructions.
2071 * config/nds32/nds32.h: Likewise.
2072 * config/nds32/nds32.md: Likewise.
2073 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2074 * config/nds32/predicates.md: Implement new predicates for DSP
2077 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2079 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2080 Reformat alternatives and attributes so it is easier to identify
2081 which constraints/attributes go with which instruction.
2082 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2083 (mov<mode>_softfloat32, FMOVE64): Likewise.
2084 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2085 (mov<mode>_softfloat64, FMOVE64): Likewise.
2087 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2089 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2091 (Basic PowerPC Built-in Functions): The new name of the
2092 subsection previously known as "PowerPC Built-in Functions".
2093 (Basic PowerPC Built-in Functions Available on all Configurations):
2095 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2096 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2097 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2098 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2100 2018-05-11 Martin Jambor <mjambor@suse.cz>
2103 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2106 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2109 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2111 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2113 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2114 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2115 (ix86_handle_option): Handle -mwaitpkg.
2116 * config.gcc: New header.
2117 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2118 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2119 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2121 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2122 OPTION_MASK_ISA_WAITPKG.
2123 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2124 (ix86_option_override_internal): Add PTA_WAITPKG.
2125 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2126 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2127 IX86_BUILTIN_TPAUSE.
2128 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2129 __builtin_ia32_umwait and __builtin_ia32_tpause.
2130 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2131 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2132 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2133 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2134 UNSPECV_TPAUSE): New.
2135 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2136 * config/i386/i386.opt: Add -mwaitpkg.
2137 * config/i386/waitpkgintrin.h: New file.
2138 * config/i386/x86intrin.h: New header.
2139 * doc/invoke.texi: Add -mwaitpkg.
2141 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2144 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2146 (cortex-m0): Use armv6s-m isa.
2147 (cortex-m0plus): Likewise.
2148 (cortex-m1): Likewise.
2149 (cortex-m0.small-multiply): Likewise.
2150 (cortex-m0plus.small-multiply): Likewise.
2151 (cortex-m1.small-multiply): Likewise.
2153 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2154 Jakub Jelinek <jakub@redhat.com>
2156 PR tree-optimization/85692
2157 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2158 source permute as well.
2160 2018-05-11 Martin Liska <mliska@suse.cz>
2163 * doc/extend.texi: Document LLVM style format for no_sanitize
2166 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2168 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2169 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2170 (mode_supports_vsx_dform_quad): Likewise.
2171 (mode_supports_vmx_dform): Move these functions to be next to the
2172 other mode_supports functions.
2173 (mode_supports_dq_form): Likewise.
2174 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2175 mode_supports_dq_form.
2176 (reg_offset_addressing_ok_p): Likewise.
2177 (offsettable_ok_by_alignment): Likewise.
2178 (rs6000_legitimate_offset_address_p): Likewise.
2179 (legitimate_lo_sum_address_p): Likewise.
2180 (rs6000_legitimize_address): Likewise.
2181 (rs6000_legitimize_reload_address): Likewise.
2182 (rs6000_secondary_reload_inner): Likewise.
2183 (rs6000_preferred_reload_class): Likewise.
2184 (rs6000_output_move_128bit): Likewise.
2186 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2188 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2189 Generate SImode target register for null target.
2190 <case IX86_BUILTIN_XGETBV>: Ditto.
2191 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2192 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2194 2018-05-10 Carl Love <cel@us.ibm.com>
2196 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2197 dcbtt and dcbtstt if operands[2] is 0.
2199 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2202 * config/i386/sse.md (usadv64qi): New expander.
2204 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2206 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2207 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2208 -maltivec=be support.
2209 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2210 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2211 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2212 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2213 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2214 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2215 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2216 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2217 altivec_vsumsws): Adjust.
2218 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2219 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2220 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2222 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2223 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2224 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2225 (altivec_lve<VI_char>x): Delete expand.
2226 (*altivec_lve<VI_char>x_internal): Rename to...
2227 (altivec_lve<VI_char>x): ... this.
2228 (altivec_lvxl_<mode>): Delete expand.
2229 (*altivec_lvxl_<mode>_internal): Rename to ...
2230 (altivec_lvxl_<mode>): ... this.
2231 (altivec_stvxl_<mode>): Delete expand.
2232 (*altivec_stvxl_<mode>_internal): Rename to ...
2233 (altivec_stvxl_<mode>): ... this.
2234 (altivec_stve<VI_char>x): Delete expand.
2235 (*altivec_stve<VI_char>x_internal): Rename to ...
2236 (altivec_stve<VI_char>x): ... this.
2237 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2238 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2239 reduc_plus_scal_<mode>): Adjust.
2240 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2242 (rs6000_cpu_cpp_builtins): Adjust.
2243 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2244 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2245 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2246 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2247 -maltivec=be support.
2248 (rs6000_split_vec_extract_var): Adjust.
2249 (rs6000_split_v4si_init): Adjust.
2250 (swap_selector_for_mode): Delete.
2251 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2252 altivec_expand_stvex_be): Delete.
2253 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2254 -maltivec=be support.
2255 (rs6000_gimple_fold_builtin): Ditto.
2256 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2258 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2259 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2260 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2261 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2262 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2263 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2264 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2265 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2266 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2267 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2268 anonymous split): Adjust.
2269 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2270 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2272 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2274 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2275 when --with-gxx-include-dir is also specified.
2276 * configure: Regenerate.
2278 2018-05-09 Jim Wilson <jimw@sifive.com>
2281 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2282 * config/riscv/t-withmultilib: New.
2283 * config/riscv/withmultilib.h: New.
2284 * doc/install.texi: Document RISC-V --with-multilib-list support.
2286 2018-05-09 Richard Biener <rguenther@suse.de>
2288 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2290 (vect_bb_vectorization_profitable_p): Adjust. Compute
2291 actual scalar cost using the cost vector and the add_stmt_cost
2294 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2296 PR rtl-optimization/85645
2297 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2298 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2300 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2302 PR rtl-optimization/85645
2303 * shrink-wrap.c (spread_components): Return a boolean saying if
2304 anything was changed.
2305 (try_shrink_wrapping_separate): Iterate spread_components until
2306 nothing changes anymore.
2308 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2310 PR rtl-optimization/85645
2311 * regrename.c (build_def_use): Also kill the chains that include the
2312 destination of a REG_CFA_REGISTER note.
2314 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2316 PR rtl-optimization/85645
2317 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2318 insn that has a REG_CFA_REGISTER note.
2320 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2322 * cfgexpand.c (expand_clobber): New function.
2323 (expand_gimple_stmt_1): Use it.
2324 * tree-vect-stmts.c (vect_clobber_variable): New function,
2326 (vectorizable_simd_clone_call): ...here.
2327 (vectorizable_store): Emit a clobber either side of an
2328 IFN_STORE_LANES sequence.
2329 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2331 2018-05-09 Tom de Vries <tom@codesourcery.com>
2334 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2335 (define_insn "trap_if_false"): Add exit after trap.
2337 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2339 PR rtl-optimization/85638
2340 * bb-reorder.c: Include common/common-target.h.
2341 (create_forwarder_block): New function extracted from...
2342 (fix_up_crossing_landing_pad): ...here. Rename into...
2343 (dw2_fix_up_crossing_landing_pad): ...this.
2344 (sjlj_fix_up_crossing_landing_pad): New function.
2345 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2346 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2347 from both partitions and exit the loop after one iteration.
2349 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2352 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2354 (Basic PowerPC Built-in Functions): The new name of the
2355 subsection previously known as "PowerPC Built-in Functions".
2356 (Basic PowerPC Built-in Functions Available on all Configurations):
2358 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2360 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2362 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2364 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2367 2018-05-08 Jim Wilson <jimw@sifive.com>
2369 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2370 (LD_EMUL_SUFFIX): New.
2371 (LINK_SPEC): Use it.
2373 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2375 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2377 (Basic PowerPC Built-in Functions): The new name of the
2378 subsection previously known as "PowerPC Built-in Functions".
2379 (Basic PowerPC Built-in Functions Available on all Configurations):
2381 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2383 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2385 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2387 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2390 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2393 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2394 after cmpelim optimization.
2396 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2398 * config.gcc: Support "goldmont".
2399 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2400 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2402 * config/i386/i386.c (m_GOLDMONT): Define.
2403 (processor_target_table): Add "goldmont".
2404 (PTA_GOLDMONT): Define.
2405 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2406 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2407 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2408 (fold_builtin_cpu): Add "goldmont".
2409 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2410 (ix86_option_override_internal): Add "goldmont".
2411 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2412 (processor_type): Add PROCESSOR_GOLDMONT.
2413 * config/i386/i386.md: Add CPU "glm".
2414 * config/i386/glm.md: New file.
2415 * config/i386/x86-tune.def: Add m_GOLDMONT.
2416 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2418 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2421 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2423 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2424 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2425 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2428 * config/i386/i386.c (ix86_fold_builtin): Handle
2429 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2432 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2433 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2435 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2438 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2439 (check_arch): Likewise.
2440 (check_fpu): Return the result rather than printing it.
2441 (end arch): Fix operator precedence.
2442 (end cpu): Likewise.
2443 (END): Print the result from check_fpu.
2445 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2446 Alan Hayward <alan.hayward@arm.com>
2447 David Sherwood <david.sherwood@arm.com>
2449 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2450 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2451 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2452 (*fcmuo<mode>_and): New patterns.
2454 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2456 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2457 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2458 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2459 (cmp_op, sve_imm_con): New code attributes.
2460 (SVE_COND_INT_CMP, imm_con): Delete.
2461 (cmp_op): Remove above unspecs from int attribute.
2462 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2464 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2465 comparison-specific unspecs.
2466 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2467 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2468 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2469 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2470 (*vec_fcm<cmp_op><mode>): Rename to...
2471 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2472 (*vec_fcmuo<mode>): Rename to...
2473 (*fcmuo<mode>): ...this and adjust likewise.
2474 (*pred_fcm<cmp_op><mode>): New pattern.
2475 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2476 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2478 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2480 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2481 (aarch64_emit_sve_predicated_cond): New function.
2482 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2483 (aarch64_emit_unspec_cond_or): Replace with...
2484 (aarch64_emit_sve_or_conds): ...this new function. Use
2485 aarch64_emit_sve_ptrue_op for the individual comparisons and
2486 aarch64_emit_binop to OR them together.
2487 (aarch64_emit_inverted_unspec_cond): Replace with...
2488 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2489 aarch64_emit_sve_ptrue_op for the comparison and
2490 aarch64_emit_unop to invert the result.
2491 (aarch64_expand_sve_vec_cmp_float): Update after the above
2492 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2494 2018-05-07 Nathan Sidwell <nathan@acm.org>
2496 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2497 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2498 (Backwards Compatibility): Likewise.
2500 2018-05-07 Luis Machado <luis.machado@linaro.org>
2504 2018-05-07 Luis Machado <luis.machado@linaro.org>
2506 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2507 <prefetch_dynamic_strides>: New const bool field.
2508 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2509 prefetch_dynamic_strides.
2510 (exynosm1_prefetch_tune): Likewise.
2511 (thunderxt88_prefetch_tune): Likewise.
2512 (thunderx_prefetch_tune): Likewise.
2513 (thunderx2t99_prefetch_tune): Likewise.
2514 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2515 (aarch64_override_options_internal): Update to set
2516 PARAM_PREFETCH_DYNAMIC_STRIDES.
2517 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2518 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2519 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2520 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2521 prefetch-dynamic-strides setting.
2523 2018-05-07 Luis Machado <luis.machado@linaro.org>
2525 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2526 <minimum_stride>: New const int field.
2527 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2528 minimum_stride field.
2529 (exynosm1_prefetch_tune): Likewise.
2530 (thunderxt88_prefetch_tune): Likewise.
2531 (thunderx_prefetch_tune): Likewise.
2532 (thunderx2t99_prefetch_tune): Likewise.
2533 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2534 (aarch64_override_options_internal): Update to set
2535 PARAM_PREFETCH_MINIMUM_STRIDE.
2536 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2537 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2538 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2539 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2540 stride is constant and is below the minimum stride threshold.
2542 2018-05-07 Luis Machado <luis.machado@linaro.org>
2544 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2547 2018-05-07 Luis Machado <luis.machado@linaro.org>
2549 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2550 <prefetch_dynamic_strides>: New const bool field.
2551 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2552 prefetch_dynamic_strides.
2553 (exynosm1_prefetch_tune): Likewise.
2554 (thunderxt88_prefetch_tune): Likewise.
2555 (thunderx_prefetch_tune): Likewise.
2556 (thunderx2t99_prefetch_tune): Likewise.
2557 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2558 (aarch64_override_options_internal): Update to set
2559 PARAM_PREFETCH_DYNAMIC_STRIDES.
2560 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2561 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2562 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2563 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2564 prefetch-dynamic-strides setting.
2566 2018-05-07 Luis Machado <luis.machado@linaro.org>
2568 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2569 <minimum_stride>: New const int field.
2570 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2571 minimum_stride field.
2572 (exynosm1_prefetch_tune): Likewise.
2573 (thunderxt88_prefetch_tune): Likewise.
2574 (thunderx_prefetch_tune): Likewise.
2575 (thunderx2t99_prefetch_tune): Likewise.
2576 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2577 (aarch64_override_options_internal): Update to set
2578 PARAM_PREFETCH_MINIMUM_STRIDE.
2579 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2580 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2581 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2582 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2583 stride is constant and is below the minimum stride threshold.
2585 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2588 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2589 the type is addressable. Don't force op into register if it has
2592 2018-05-05 Roland McGrath <mcgrathr@google.com>
2595 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2596 any section for which we don't know a specific type it should have,
2597 regardless of name. Previously this was done only for the exact
2598 names ".init_array", ".fini_array", and ".preinit_array".
2599 (default_elf_asm_named_section): Add comment about
2600 relationship with default_section_type_flags and SECTION_NOTYPE.
2601 (get_section): Don't consider it a type conflict if one side has
2602 SECTION_NOTYPE and the other doesn't, as long as neither has the
2603 SECTION_BSS et al used in the default_section_type_flags logic.
2605 2018-05-05 Tom de Vries <tom@codesourcery.com>
2608 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2609 (workaround_barsyncs): New function.
2610 (nvptx_reorg): Use workaround_barsyncs.
2611 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2612 (define_expand "nvptx_membar_cta"): New define_expand.
2613 (define_insn "*nvptx_membar_cta"): New insn.
2615 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2617 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2618 To improve optimization opportunities.
2619 * builtin-types.def: The new needed builtin types for the above.
2621 2018-05-04 Richard Biener <rguenther@suse.de>
2623 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2624 * gimple-ssa-store-merging.c
2625 (imm_store_chain_info::output_merged_store): Remove redundant create,
2626 release split_store vector contents on failure.
2627 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2628 scalar stmt vector on cache hit.
2630 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2632 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2634 * config.gcc (powerpc-xilinx-eabi*): Remove.
2635 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2637 (fusion_addis_mem_combo_load): Ditto.
2638 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2640 (rs6000_cpu_cpp_builtins): Ditto.
2641 * config/rs6000/rs6000-linux.c
2642 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2643 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2644 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2646 (rs6000_setup_reg_addr_masks): Ditto.
2647 (rs6000_init_hard_regno_mode_ok): Ditto.
2648 (rs6000_option_override_internal): Ditto.
2649 (legitimate_lo_sum_address_p): Ditto.
2650 (rs6000_legitimize_address): Ditto.
2651 (rs6000_legitimize_reload_address): Ditto.
2652 (rs6000_legitimate_address_p): Ditto.
2653 (abi_v4_pass_in_fpr): Ditto.
2654 (setup_incoming_varargs): Ditto.
2655 (rs6000_gimplify_va_arg): Ditto.
2656 (rs6000_split_multireg_move): Ditto.
2657 (rs6000_savres_strategy): Ditto.
2658 (rs6000_emit_prologue_components): Ditto.
2659 (rs6000_emit_epilogue_components): Ditto.
2660 (rs6000_emit_prologue): Ditto.
2661 (rs6000_emit_epilogue): Ditto.
2662 (rs6000_elf_file_end): Ditto.
2663 (rs6000_function_value): Ditto.
2664 (rs6000_libcall_value): Ditto.
2665 * config/rs6000/rs6000.h: Ditto.
2666 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2667 (TARGET_MINMAX): ... this. New.
2668 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2669 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2670 (*movsi_internal1_single): Delete.
2671 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2672 mfpu=, mxilinx-fpu): Delete.
2673 * config/rs6000/singlefp.h: Delete.
2674 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2675 * config/rs6000/t-rs6000: Ditto.
2676 * config/rs6000/t-xilinx: Delete.
2677 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2678 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2679 (VStype_simple): Delete.
2680 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2681 * config/rs6000/xfpu.h: Delete.
2682 * config/rs6000/xfpu.md: Delete.
2683 * config/rs6000/xilinx.h: Delete.
2684 * config/rs6000/xilinx.opt: Delete.
2685 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2686 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2688 2018-05-04 Tom de Vries <tom@codesourcery.com>
2691 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2694 2018-05-04 Richard Biener <rguenther@suse.de>
2697 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2698 (expand_complex_div_wide): Likewise.
2699 (expand_complex_operations_1): Likewise.
2700 (expand_complex_libcall): Preserve EH info of the original stmt.
2701 (tree_lower_complex): Handle removed blocks.
2702 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2703 on complex multiplication and division libcall builtins.
2705 2018-05-04 Richard Biener <rguenther@suse.de>
2708 * fold-const.c (negate_expr_p): Restrict negation of operand
2709 zero of a division to when we know that can happen without
2711 (fold_negate_expr_1): Likewise.
2713 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2716 * real.h (real_nextafter): Declare.
2717 * real.c (real_nextafter): New function.
2718 * fold-const-call.c (fold_const_nextafter): New function.
2719 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2720 CASE_CFN_NEXTTOWARD.
2721 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2722 even when arg1_mode is different from arg0_mode.
2724 2018-05-03 Nathan Sidwell <nathan@acm.org>
2726 * doc/extend.texi (Deprecated Features): Remove
2728 (Backwards Compatibility): Likewise.
2729 * doc/invoke.texi (C++ Language Options): Likewise.
2730 (C++ Dialect Options): Likewise.
2732 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2735 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2736 _mm512_mask_mullox_epi64): New intrinsics.
2738 2018-05-03 Tom de Vries <tom@codesourcery.com>
2741 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2742 dump files): Add offload-tree.
2744 2018-05-03 Richard Biener <rguenther@suse.de>
2746 PR tree-optimization/85615
2747 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2748 to loops not nested in BBs loop father to avoid creating multi-entry
2751 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2753 PR tree-optimization/70291
2754 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2755 arguments. Change return type to tree. Emit libcall as a new
2756 statement rather than replacing existing one when inplace_p is true.
2757 (expand_complex_multiplication_components): New function.
2758 (expand_complex_multiplication): Expand floating-point complex
2759 multiplication using the above.
2760 (expand_complex_division): Rename inner_type parameter to type.
2761 Update expand_complex_libcall call-site.
2762 (expand_complex_operations_1): Update expand_complex_multiplication
2763 and expand_complex_division call-sites.
2765 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2768 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2769 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2770 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2771 the highest significant bit of the shift count mask is clear. In
2772 check whether and[sq]i3 is needed verify that all significant bits
2773 of the shift count other than the highest are set.
2775 2018-05-02 Tom de Vries <tom@codesourcery.com>
2778 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2779 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2780 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2781 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2782 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2783 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2784 __builtin_goacc_parlevel_size.
2786 2018-05-02 Richard Biener <rguenther@suse.de>
2788 PR tree-optimization/85597
2789 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2790 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2793 2018-05-02 Tom de Vries <tom@codesourcery.com>
2796 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2797 dump files): Add ltrans-tree.
2799 2018-05-02 Tom de Vries <tom@codesourcery.com>
2802 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2803 dump files): Add wpa-ipa.
2805 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2807 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2808 powerpc*-*-linux*paired* target.
2809 * config/rs6000/750cl.h: Delete.
2810 * config/rs6000/paired.h: Delete.
2811 * config/rs6000/paired.md: Delete.
2812 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2814 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2815 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2816 comment. Remove paired float support.
2817 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2818 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2820 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2821 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2823 * config/rs6000/rs6000.c: Remove paired float support.
2824 (paired_expand_vector_init, paired_expand_vector_move,
2825 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2826 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2827 paired_expand_builtin, paired_expand_predicate_builtin,
2828 paired_init_builtins): Delete.
2829 * config/rs6000/rs6000.h: Remove paired float support.
2830 * config/rs6000/rs6000.md: Remove paired float support.
2831 (move_from_CR_ov_bit): Delete.
2832 * config/rs6000/rs6000.opt (mpaired): Delete.
2833 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2834 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2836 2018-05-02 Richard Biener <rguenther@suse.de>
2839 * gimplify.c (gimplify_save_expr): When in SSA form allow
2840 SAVE_EXPRs to compute to SSA vars.
2842 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2845 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2846 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2847 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2848 clobber operands[2], instead use a new pseudo. Formatting fixes.
2850 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2852 PR tree-optimization/85586
2853 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2854 exit early for statements in the same group if the accesses are
2857 2018-05-02 Tom de Vries <tom@codesourcery.com>
2860 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2863 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2865 PR tree-optimization/85143
2866 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2868 2018-05-01 Tom de Vries <tom@codesourcery.com>
2871 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2872 not found" error message.
2874 2018-05-01 Tom de Vries <tom@codesourcery.com>
2877 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2878 * vec.c (test_ordered_remove_if): New function.
2879 (vec_c_tests): Call test_ordered_remove_if.
2880 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2881 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2882 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2883 VEC_ORDERED_REMOVE_IF.
2885 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2887 PR tree-optimization/82665
2888 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2889 pointer subtraction where arguments come from a memchr call.
2891 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2893 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2894 --push-state --as-needed and --pop-state instead of --as-needed and
2895 --no-as-needed if ld supports it.
2896 * configure: Regenerated.
2899 * doc/install.texi2html: Replace _002d with - and _002a with * in
2900 generated html files using sed.
2902 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2905 * gcc-rich-location.c (blank_line_before_p): New function.
2906 (use_new_line): New function.
2907 (gcc_rich_location::add_fixit_insert_formatted): New function.
2908 * gcc-rich-location.h
2909 (gcc_rich_location::add_fixit_insert_formatted): New function.
2911 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2913 * selftest.c (assert_streq): Rename "expected" and "actual" to
2914 "val1" and "val2". Extend NULL-handling to cover both inputs
2915 symmetrically, while still requiring both to be non-NULL for a pass.
2916 * selftest.h (assert_streq): Rename "expected" and "actual" to
2918 (ASSERT_EQ): Likewise.
2919 (ASSERT_EQ_AT): Likewise.
2920 (ASSERT_KNOWN_EQ): Likewise.
2921 (ASSERT_KNOWN_EQ_AT): Likewise.
2922 (ASSERT_NE): Likewise.
2923 (ASSERT_MAYBE_NE): Likewise.
2924 (ASSERT_MAYBE_NE_AT): Likewise.
2925 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2926 the assertion to pass.
2927 (ASSERT_STREQ_AT): Likewise.
2929 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2931 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2932 interaction with -pie.
2934 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2936 * selftest.h: Fix alphabetization of per-source-file selftest
2939 2018-04-30 Jason Merrill <jason@redhat.com>
2941 PR c++/61982 - dead stores to destroyed objects.
2942 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2945 2018-04-30 Jason Merrill <jason@redhat.com>
2947 * tree.c (build_clobber): New.
2948 * tree.h: Declare it.
2949 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2951 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2953 * diagnostic-show-locus.c (layout::layout): Update for
2954 location_get_source_line returning a char_span.
2955 (struct char_span): Move to input.h.
2956 (struct correction): Update for fields in char_span becoming
2958 (struct source_line): Update for location_get_source_line
2959 returning a char_span.
2960 (layout::print_line): Likewise.
2961 * edit-context.c (edited_file::print_content): Likewise.
2962 (edited_file::print_diff_hunk): Likewise.
2963 (edited_file::print_run_of_changed_lines): Likewise.
2964 (edited_file::get_num_lines): Likewise.
2965 (edited_line::edited_line): Likewise.
2966 * final.c (asm_show_source): Likewise.
2967 * input.c (location_get_source_line): Convert return type
2968 from const char * to char_span, losing the final "line_len"
2970 (dump_location_info): Update for the above.
2971 (get_substring_ranges_for_loc): Likewise. Use a char_span
2972 when handling the literal within the line.
2973 (test_reading_source_line): Update for location_get_source_line
2974 returning a char_span.
2975 * input.h (class char_span): Move here from
2976 diagnostic-show-locus.c, converting from a struct to a class.
2977 Make data members private.
2978 (char_span::operator bool): New.
2979 (char_span::length): New.
2980 (char_span::get_buffer): New.
2981 (char_span::operator[]): New.
2982 (char_span::subspan): Make const.
2983 (char_span::xstrdup): New.
2984 (location_get_source_line): Convert return type from const char *
2985 to char_span, losing the final "line_size" param.
2987 2018-04-30 Jan Hubicka <jh@suse.cz>
2989 * lto-wrapper.c (ltrans_priorities): New static var.
2990 (cmp_priority): New.
2991 (run_gcc): Read priorities and if doing parallel build order
2992 the Makefile by them.
2994 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2996 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2998 2018-04-30 Richard Biener <rguenther@suse.de>
3000 * tree-cfg.c (verify_address): Remove base argument, add
3001 flag whether to check TREE_ADDRESSABLE and do that.
3002 (verify_expr): Remove.
3003 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3004 (verify_gimple_assign_single): Likewise.
3005 (verify_gimple_switch): Likewise.
3006 (verify_expr_location_1): Dereference tp once. Add (disabled)
3007 piece from verify_expr.
3008 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3010 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3012 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3014 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3016 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3017 (small_data_pattern): Likewise.
3018 (arc_rewrite_small_data): Likewise.
3019 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3020 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3021 (get_symbol_alignment): New function.
3022 (legitimate_small_data_address_p): Likewise.
3023 (legitimate_scaled_address): Update, call
3024 legitimate_small_data_address_p.
3025 (output_sdata): New static variable.
3026 (arc_print_operand): Update how we handle small data operands.
3027 (arc_print_operand_address): Likewise.
3028 (arc_legitimate_address_p): Update, use
3029 legitimate_small_data_address_p.
3030 (arc_rewrite_small_data_p): Remove.
3031 (arc_rewrite_small_data_1): Likewise.
3032 (arc_rewrite_small_data): Likewise.
3033 (small_data_pattern): Likewise.
3034 (compact_sda_memory_operand): Update to use
3035 legitimate_small_data_address_p and get_symbol_alignment.
3036 (prepare_move_operands): Don't rewite sdata pattern.
3037 (prepare_extend_operands): Remove.
3038 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3040 (zero_extendqisi2): Likewise.
3041 (zero_extendhisi2): Likewise.
3042 (extendqihi2): Likewise.
3043 (extendqisi2): Likewise.
3044 (extendhisi2): Likewise.
3048 * config/arc/constraints.md (Usd): Change it to memory constraint.
3050 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3052 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3053 as source of std instructions.
3054 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3055 allow 6-bit constants as source for store instructions.
3056 (movdi_insn): Update instruction pattern to allow 6-bit constants
3057 as source for store instructions.
3059 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3061 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3063 2018-04-30 Nathan Sidwell <nathan@acm.org>
3064 Sandra Loosemore <sandra@codesourcery.com>
3066 * dumpfile.c (dump_open): Allow '-' for stdout.
3067 * doc/invoke.texi (Developer Options): Document dump filename
3068 determination early. Document stdin/stdout selection.
3070 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3072 Microblaze Target: PIC data text relative
3074 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3075 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3077 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3078 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3079 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3080 New addressing mode for data-text relative position indepenedent code.
3081 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3082 'ADDRESS_SYMBOLIC_TXT_REL'.
3083 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3084 (microblaze_legitimate_pic_operand): Exclude function calls from
3085 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3086 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3088 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3089 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3090 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3091 for 'address + offset'.
3092 (microblaze_expand_prologue): Add new function prologue call for
3094 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3095 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3096 table in case of TARGET_PIC_DATA_TEXT_REL.
3097 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3098 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3099 Add new macros 'UNSPEC_TEXT',
3100 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3101 + exclude function calls from 'UNSPEC_PLT' in case of data text
3103 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3104 new target hook for generating address diff vector tables in case of
3106 * doc/tm.texi : Regenerate.
3107 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3108 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3109 of addr diff vector generation.
3110 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3111 target hook definition.
3112 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3113 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3114 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3115 option for data text relative.
3117 2018-04-30 Richard Biener <rguenther@suse.de>
3119 * tree-chrec.h (evolution_function_is_constant_p): Remove
3121 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3123 2018-04-30 Richard Biener <rguenther@suse.de>
3126 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3128 2018-04-30 Richard Biener <rguenther@suse.de>
3130 PR tree-optimization/28364
3131 PR tree-optimization/85275
3132 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3133 copying first exit test.
3135 2018-04-28 Mark Wielaard <mark@klomp.org>
3137 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3139 (dwarf_AT): Handle DW_AT_addr_base.
3140 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3142 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3145 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3146 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3147 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3148 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3150 2018-04-28 Richard Biener <rguenther@suse.de>
3152 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3153 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3154 to reflect use. Only add interesting stmts.
3156 2018-04-27 Martin Jambor <mjambor@suse.cz>
3159 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3160 the jump function allows for passing through aggregate values.
3162 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3164 * input.h (in_system_header_at): Convert from macro to inline
3166 (from_macro_expansion_at): Likewise.
3167 (from_macro_definition_at): Likewise.
3169 2018-04-27 Jeff Law <law@redhat.com>
3171 * config.gcc: Mark tile* targets as deprecated/obsolete.
3173 2018-04-27 Richard Biener <rguenther@suse.de>
3175 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3178 2018-04-27 Richard Biener <rguenther@suse.de>
3180 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3182 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3184 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3185 with Yd constraint. Set "preferred_for_speed" attribute from
3186 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3188 (*movdi_internal): Ditto.
3189 (movti_interunit splitters): Remove
3190 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3191 (movdi_interunit splitters): Ditto.
3192 * config/i386/constraints.md (Ye): Remove.
3193 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3195 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3198 * config/aarch64/constraints.md (Usg): Limit to 31.
3201 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3203 PR tree-optimization/85529
3204 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3205 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3206 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3207 zero extension or masking of the MSB bit.
3208 (optimize_range_tests): Add FIRST_BB argument, pass it through
3209 to optimize_range_tests_var_bound.
3210 (maybe_optimize_range_tests, reassociate_bb): Adjust
3211 optimize_range_tests callers.
3213 2018-04-26 Richard Biener <rguenther@suse.de>
3214 Jakub Jelinek <jakub@redhat.com>
3216 * cgraph.h (symbol_table): Just declare debug method here.
3217 * symtab.c (symbol_table::debug): Define.
3219 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3221 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3223 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3225 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3226 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3227 (*movdi_internal): Substitute Yi and Yj constraint with x
3228 and Ym and Yn constraint with y constraint. Update "isa"
3229 attribute and set "preferred_for_speed" attribute from
3230 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3231 (*movsi_internal): Ditto.
3232 (*movdf_internal): Ditto.
3233 (*movsf_internal): Ditto.
3234 (*zero_extendsidi2): Ditto.
3235 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3236 (sse2_loadld): Ditto.
3237 (*vec_extract<ssevecmodelower>_0): Ditto.
3238 (*vec_extractv4si_0_zext_sse4): Ditto.
3239 (vec_concatv2di): Ditto.
3240 (*vec_dup<mode>): Ditto.
3241 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3242 * config/i386/constraints.md (Yi): Remove.
3247 2018-04-26 Nathan Sidwell <nathan@acm.org>
3249 * dumpfile.c (dump_open): New.
3250 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3251 (dump_finish): Detect stdio/stderr by value not name.
3253 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3255 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3257 2018-04-26 Tom de Vries <tom@codesourcery.com>
3260 * config/nvptx/nvptx.c (verify_neutering_jumps)
3261 (verify_neutering_labels): New function
3262 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3264 2018-04-26 Tom de Vries <tom@codesourcery.com>
3267 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3268 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3271 2018-04-26 Richard Biener <rguenther@suse.de>
3272 Tom de Vries <tom@codesourcery.com>
3275 * lto-streamer-out.c (output_function): Fixup loops if required to match
3276 discovery done in the reader.
3278 2018-04-26 Richard Biener <rguenther@suse.de>
3280 PR tree-optimization/85116
3281 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3282 have a loop exit from the single latch predecessor. Remove
3283 case of header with just condition.
3284 (ch_base::copy_headers): Exclude infinite loops from any
3286 (pass_ch::execute): Record exits.
3288 2018-04-26 Richard Biener <rguenther@suse.de>
3290 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3291 prologue cost vector and pass it to vect_get_load_cost.
3292 (vect_get_peeling_costs_all_drs): Likewise.
3293 (vect_peeling_hash_get_lowest_cost): Likewise.
3294 (vect_enhance_data_refs_alignment): Likewise.
3296 2018-04-26 Richard Biener <rguenther@suse.de>
3299 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3300 checking of integer<->pointer conversions.
3301 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3302 sign-/zero-extending pointer types.
3303 (expand_omp_for_static_chunk): Likewise.
3305 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3306 Jean Lee <xiaoyur347@gmail.com>
3308 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3309 (TARGET_ASAN_SHADOW_OFFSET): Define.
3310 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3311 true for -fsanitize=address.
3313 2018-04-25 Mark Wielaard <mark@klomp.org>
3315 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3318 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3320 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3321 than "alu", remove explicit "memory" and "imm_disp" attributes.
3322 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3325 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3326 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3327 gen_lowpart_no_emit.
3329 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3332 * config/i386/i386.c (ix86_expand_builtin): Change memory
3333 operand to XI, extend p0 to Pmode.
3334 * config/i386/i386.md: Change unspec volatile and operand
3335 1 mode to XI, change operand 0 mode to P.
3337 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3339 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3340 GET_MODE_MASK before any checking.
3341 (nds32_can_use_bset_p): Likewise.
3342 (nds32_can_use_btgl_p): Likewise.
3344 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3346 * config/nds32/nds32-doubleword.md: New define_split pattern for
3347 illegal register number.
3349 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3351 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3353 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3355 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3357 2018-04-25 Richard Biener <rguenther@suse.de>
3359 * lto-streamer.h (LTO_major_version): Bump to 8.
3361 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3363 * BASE-VER: Set to 9.0.0.
3365 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3367 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3368 in __abskf2 and __powikf2.
3370 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3373 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3374 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3375 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3376 Use the above on operand 2. Reindent.
3377 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3379 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3382 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3384 * config/i386/i386.opt (mcet): Removed.
3385 * doc/install.texi: Remove -mcet documentation.
3386 * doc/invoke.texi: Likewise.
3388 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3391 * doc/install.texi: Remove -mcet from bootstrap-cet.
3393 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3396 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3397 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3401 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3402 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3403 containing a CONST_VECTOR.
3405 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3407 * doc/install.texi: Update newlib dependency for nvptx.
3409 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3412 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3413 instead of INTVAL when shifting x left.
3415 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3417 PR tree-optimization/85478
3418 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3419 vect_grouped_store_supported for single element vectors.
3421 2018-04-24 Richard Biener <rguenther@suse.de>
3424 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3425 load cost increase to the case of non-constant step.
3427 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3430 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3431 destination if any_malformed_asm.
3433 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3436 * expr.c (store_field): In the bitfield case, if the value comes from
3437 a function call and is returned in registers by means of a PARALLEL,
3438 do not change the mode of the temporary unless BLKmode and VOIDmode.
3440 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3442 PR rtl-optimization/85423
3443 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3444 dependencies to debug insns when the previous insn is non-debug.
3446 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3448 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3449 enums into a single definition.
3450 (fls): Fix predicates and printing.
3453 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3455 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3456 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3457 and short u6 immediate.
3458 (check_if_valid_sleep_operand): Remove.
3459 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3461 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3463 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3464 flag_always_save_lp condition.
3465 * config/nds32/nds32.opt (malways-save-lp): New option.
3467 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3469 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3470 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3471 * config/nds32/nds32.h
3472 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3473 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3475 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3477 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3478 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3480 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3481 Chung-Ju Wu <jasonwucj@gmail.com>
3483 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3485 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3486 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3488 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3490 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3492 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3494 * config/nds32/nds32-protos.h (nds32_data_alignment,
3495 nds32_local_alignment): Declare.
3496 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3497 nds32_local_alignment): New functions.
3498 (TARGET_CONSTANT_ALIGNMENT): Define.
3499 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3501 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3503 * config/nds32/nds32.c
3504 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3505 (TARGET_MODES_TIEABLE_P): Likewise.
3507 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3509 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3512 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3513 Chung-Ju Wu <jasonwucj@gmail.com>
3515 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3516 for unaligned access.
3517 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3519 * config/nds32/nds32-intrinsic.md: Likewise.
3520 * config/nds32/nds32_intrinsic.h: Likewise.
3521 * config/nds32/nds32.h (nds32_builtins): Likewise.
3522 * config/nds32/nds32.opt (munaligned-access): New option.
3523 * config/nds32/nds32.c (nds32_asm_file_start): Display
3524 flag_unaligned_access status.
3526 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3528 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3529 -mno-relax is present.
3530 * config/riscv/linux.h (LINK_SPEC): Ditto.
3532 2018-04-20 Martin Sebor <msebor@redhat.com>
3535 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3537 (gimple_fold_builtin_stxcpy_chk): Same.
3538 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3540 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3543 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3544 __powikf2 when long double is IEEE 128-bit.
3546 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3548 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3549 step to make sure stack always aligned.
3551 2018-04-20 Carl Love <cel@us.ibm.com>
3554 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3555 size check for arg0.
3557 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3558 Tom de Vries <tom@codesourcery.com>
3561 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3562 Emit insns for calls too.
3563 (nvptx_find_par): Always look for worker-level predecessor insn.
3564 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3566 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3567 (nvptx_process_pars): Propagate frames for calls.
3569 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3572 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3574 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3575 (ix86_handle_option): Don't handle OPT_mibt.
3576 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3578 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3580 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3581 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3582 (ix86_target_macros): Define __CET__ with flag_cf_protection
3583 for -fcf-protection.
3584 * config/i386/i386.c (isa2_opts): Remove -mibt.
3585 * config/i386/i386.h (TARGET_IBT): Removed.
3586 (TARGET_IBT_P): Likewise.
3587 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3588 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3589 * config/i386/i386.opt (mcet): Update help message.
3592 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3593 -mcet as an alias for -mshstk.
3595 2018-04-20 Richard Biener <rguenther@suse.de>
3598 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3599 complexity by forcing a single use of the multiply operand.
3601 2018-04-20 Martin Jambor <mjambor@suse.cz>
3604 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3605 recursion dependency to only apply to non-clones.
3607 2018-04-20 Martin Jambor <mjambor@suse.cz>
3610 * ipa-cp.c (create_specialized_node): Check that clones of
3611 self-recursive edges exist during IPA-CP.
3613 2018-04-19 Toon Moene <toon@moene.org>
3615 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3618 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3620 PR tree-optimization/85467
3621 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3622 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3623 VECTOR_CST element to type.
3625 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3628 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3629 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3630 (builtin_longjmp): Likewise.
3631 (save_stack_nonlocal): New pattern.
3632 (restore_stack_nonlocal): Likewise.
3634 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3637 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3638 Replace ASM_OUTPUT_LABEL with fprintf.
3640 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3643 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3644 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3645 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3646 define __IBT__ and __SHSTK__ for -fcf-protection.
3647 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3649 (ix86_trampoline_init): Likewise.
3650 (x86_output_mi_thunk): Likewise.
3651 (ix86_notrack_prefixed_insn_p): Likewise.
3652 (ix86_option_override_internal): Don't disallow -fcf-protection.
3653 * config/i386/i386.md (rdssp<mode>): Also enable for
3655 (incssp<mode>): Likewise.
3656 (nop_endbr): Likewise.
3657 * config/i386/i386.opt (mcet): Change help message to built-in
3661 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3662 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3663 enable CET built-in functions.
3665 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3667 * common/config/i386/i386-common.c
3668 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3669 OPTION_MASK_ISA_MOVDIRI_UNSET,
3670 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3671 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3672 * config.gcc (movdirintrin.h): New header.
3673 * config/i386/cpuid.h (bit_MOVDIRI,
3674 bit_MOVDIR64B): New bits.
3675 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3677 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3678 (VOID, PVOID, PCVOID)): New function types.
3679 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3680 __builtin_ia32_directstoreu_u64,
3681 __builtin_ia32_movdir64b): New builtins.
3682 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3683 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3685 (ix86_valid_target_attribute_inner_p): Ditto.
3686 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3687 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3688 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3689 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3690 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3691 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3692 (movdiri<mode>, movdir64b_<mode>): New.
3693 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3694 * config/i386/immintrin.h: Include movdirintrin.h.
3695 * config/i386/movdirintrin.h: New file.
3696 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3698 2018-04-19 Richard Biener <rguenther@suse.de>
3701 * cfg.c (clear_bb_flags): When loop state says we have
3702 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3704 2018-04-19 Richard Biener <rguenther@suse.de>
3706 PR tree-optimization/84737
3707 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3708 copying restrict info.
3709 (vect_setup_realignment): Use it.
3710 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3711 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3712 the first DR to all generated stores.
3713 (vectorizable_load): Likewise for loads.
3715 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3717 PR tree-optimization/85446
3718 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3719 the integral and pointer types to have the same precision.
3721 * doc/install.texi: Document --disable-cet being the default and
3724 2018-04-18 Martin Liska <mliska@suse.cz>
3726 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3729 2018-04-18 Martin Liska <mliska@suse.cz>
3732 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3735 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3736 arguments if they are comparable.
3738 2018-04-18 Martin Liska <mliska@suse.cz>
3741 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3744 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3747 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3750 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3751 ENDBR after calling __morestack.
3753 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3756 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3757 by using gcc_base_ver to generate a gcc_driver_version, and use
3758 it when generating GCC_DRIVER_NAME.
3759 * configure: Regenerate.
3761 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3764 * config.gcc: Obsolete powerpc*-*-*spe*.
3766 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3769 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3770 (stabstr_D): Change type of unum from unsigned int to
3771 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3774 2018-04-17 Jim Wilson <jimw@sifive.com>
3777 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3778 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3779 Set arg_pointer_offset after using pretend_args_size.
3781 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3783 PR rtl-optimization/85431
3784 * dse.c (record_store): Ignore zero width stores.
3787 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3788 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3789 __builtin_stack_restore rather than after it.
3790 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3791 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3792 argument instead of virtual_dynamic_stack_rtx.
3794 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3796 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3798 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3799 Add note to error message to explain internal mapping of overloaded
3800 built-in function name to non-overloaded built-in function name.
3801 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3804 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3807 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3808 where the inputs overlap with the output.
3810 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3812 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3813 (=v, v) alternative and explicit "memory" attribute.
3814 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3815 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3817 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3818 "sselog1" type instead of "sselog".
3819 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3820 "sselog". Remove explicit "memory" attribute.
3821 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3822 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3824 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3825 "sselog1" type instead of "sselog". Remove explicit "memory"
3827 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3828 use "sselog1" type instead of "sselog". Remove explicit "memory"
3830 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3831 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3833 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3834 "sselog1" type instead of "sselog". Remove explicit "memory"
3836 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3837 use "sselog1" type instead of "sselog". Remove explicit "memory"
3841 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3844 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3847 2018-04-17 Martin Jambor <mjambor@suse.cz>
3850 * ipa-cp.c (create_specialized_node): Call
3851 expand_all_artificial_thunks if necessary.
3853 2018-04-17 Martin Liska <mliska@suse.cz>
3856 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3857 in message, remote space in between '_G' and '('.
3859 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3862 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3863 avx512f_vmcmp<mode>3<round_saeonly_name>,
3864 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3865 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3866 avx512f_rndscale<mode><round_saeonly_name>,
3867 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3868 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3869 Use %<iptr>2 instead of %2 for -masm=intel.
3870 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3871 avx512f_vcvttss2usi<round_saeonly_name>,
3872 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3874 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3875 avx512f_vcvttsd2usi<round_saeonly_name>,
3876 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3877 Use %q1 instead of %1 for -masm=intel.
3878 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3879 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3880 of %3 for -masm=intel.
3881 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3883 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3885 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3887 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3888 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3890 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3892 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3893 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3894 %g1 and one with %0 and %1.
3895 (avx512er_vmrcp28<mode><round_saeonly_name>,
3896 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3898 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3899 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3900 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3901 of %0 and %{%4%} for -masm=intel.
3902 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3903 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3904 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3905 order of %0 and %{%5%}%{z%} for -masm=intel.
3907 2018-04-17 Jan Hubicka <jh@suse.cz>
3910 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3912 2018-04-17 Martin Liska <mliska@suse.cz>
3915 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3916 for target_clone error message. Make default implementation
3917 clone to be a local declaration.
3918 (separate_attrs): Add new argument and check for an empty
3920 (expand_target_clones): Handle it.
3921 (ipa_target_clone): Make redirection just for target_clones
3924 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3925 Tom de Vries <tom@codesourcery.com>
3928 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3929 tiled basic blocks without omp continue statements.
3931 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3934 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3935 vec_extract expression as having side effects to make sure it gets
3938 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3941 * config/i386/i386.c (get_builtin_code_for_version): Check
3944 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3947 * gcc/config.gcc: Support "skylake".
3948 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3950 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3951 (processor_target_table): Add "skylake".
3952 (ix86_option_override_internal): Add "skylake".
3953 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3954 PROCESSOR_CANNONLAKE.
3955 (get_builtin_code_for_version): Fix priority for
3956 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3957 PROCESSOR_SKYLAKE-AVX512.
3958 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3959 (processor_type): Add PROCESSOR_SKYLAKE.
3961 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3962 Jason Merrill <jason@redhat.com>
3965 * convert.c (convert_to_integer_1): Use direct recursion for
3966 enumeral types and types with a precision less than the number
3967 of bits in their mode.
3969 2018-04-16 Julia Koval <julia.koval@intel.com>
3972 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3973 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3975 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3978 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3979 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3980 and -mno-direct-move.
3982 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3985 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3986 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3987 Ensure negative shifts result in {0}.
3989 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3991 PR rtl-optimization/79916
3992 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3993 regs (if any) to define how to gnerate SD moves when LRA is in
3996 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3998 PR rtl-optimization/85393
3999 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4000 * except.c (expand_dw2_landing_pad_for_region): Make static.
4001 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4002 a label and unconditional jump to old_bb, rather than
4003 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4006 PR rtl-optimization/85376
4007 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4008 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4009 instead of a specific value.
4011 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4012 Bin Cheng <bin.cheng@arm.com>
4014 PR tree-optimization/82965
4015 PR tree-optimization/83991
4016 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4017 by_profile_only parameter.
4018 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4019 information if the loop was predicted to iterate too many times.
4020 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4022 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4025 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4028 2018-04-13 Martin Liska <mliska@suse.cz>
4029 Jakub Jelinek <jakub@redhat.com>
4032 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4033 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4034 * builtins.c (expand_builtin_memory_copy_args): Use
4035 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4036 handle dest_addr == pc_rtx.
4038 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4041 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4042 asked to not generate direct moves.
4043 (fix_trunc<mode>si2_stfiwx): Similar.
4044 (fix_trunc<mode>si2_internal): Similar.
4046 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4049 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4050 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4051 lookup if dest in some wider mode is known to be const0_rtx and
4052 if so, record permanent equivalence for it to be ZERO_EXTEND of
4053 the narrower mode destination.
4055 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4057 * lto-streamer-out.c (output_function): Revert 259346.
4058 * omp-expand.c (expand_oacc_for): Likewise.
4060 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4062 PR rtl-optimization/85354
4063 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4064 * sel-sched.c (sel_global_init): ... here.
4066 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4069 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4070 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4071 mode for PE-COFF targets.
4072 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4073 (i386_pe_asm_lto_end): Likewise.
4074 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4075 (TARGET_ASM_LTO_END): Likewise.
4076 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4077 (i386_pe_asm_lto_start): New function.
4078 (i386_pe_asm_lto_end): Likewise.
4080 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4081 Richard Biener <rguenther@suse.de>
4084 * lto-streamer-out.c (output_function): Fix CFG loop state before
4086 * omp-expand.c (expand_oacc_for): Handle calls to internal
4087 functions like regular functions.
4089 2018-04-12 Richard Biener <rguenther@suse.de>
4092 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4093 for the early LTO debug to properly generate references to it
4094 during DIE emission. Do not re-use that for the skeleton for
4096 (dwarf2out_early_finish): Likewise.
4098 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4101 * config/i386/sse.md
4102 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4103 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4104 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4105 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4106 and output is a reg, avoid creating invalid lowpart subreg, but
4107 instead split into a 512-bit move. Don't split if not AVX512VL,
4108 input is xmm16+ reg and output is a mem.
4109 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4110 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4111 xmm16+ reg and output is a mem.
4113 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4115 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4116 also for flag_dwarf2_cfi_asm.
4118 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4120 PR rtl-optimization/85342
4121 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4122 a bool scalar var inside of the loop instead. Don't try to update
4123 recog_data.operand after failed apply_change_group.
4125 2018-04-12 Tom de Vries <tom@codesourcery.com>
4128 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4129 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4130 array with flexible array member as array without given dimension.
4131 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4132 argument for undefined param to true.
4134 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4137 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4138 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4139 from PowerPC section.
4140 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4141 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4142 help text that is too long.
4143 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4144 help text that is too long.
4145 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4146 help text that is too long.
4148 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4150 * config/alpha/alpha.md (stack_probe_internal): Rename
4151 from "probe_stack". Update all callers.
4153 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4155 PR rtl-optimization/84566
4156 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4157 sched_macro_fuse_insns.
4159 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4162 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4163 (compute_block_dependences): ... from here.
4165 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4167 PR tree-optimization/85331
4168 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4169 from int to HOST_WIDE_INT.
4171 2018-04-11 Martin Jambor <mjambor@suse.cz>
4174 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4175 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4176 not the same as the source val.
4177 (cgraph_edge_brings_value_p): New parameter.
4178 (gather_edges_for_value): Pass destination value to
4179 cgraph_edge_brings_value_p.
4180 (perhaps_add_new_callers): Likewise.
4181 (get_info_about_necessary_edges): Likewise and exclude values brought
4182 only by self-recursive edges.
4183 (create_specialized_node): Redirect only clones of self-calling edges.
4184 (+self_recursive_pass_through_p): New function.
4185 (find_more_scalar_values_for_callers_subset): Use it.
4186 (find_aggregate_values_for_callers_subset): Likewise.
4187 (known_aggs_to_agg_replacement_list): Removed.
4188 (decide_whether_version_node): Re-calculate known constants for all
4189 remaining context clones.
4191 2018-04-11 Richard Biener <rguenther@suse.de>
4194 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4195 from early DWARF output.
4196 (dwarf2out_early_finish): Output line info unconditionally into
4197 early DWARF and add reference to it.
4199 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4202 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4203 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4204 other than V2DFmode using iptr mode attribute.
4205 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4207 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4209 PR rtl-optimization/84659
4210 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4212 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4215 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4217 (output_loc_list): Pass address of a dummy size variable even in the
4218 locview handling loop.
4219 (index_location_lists): Add comment on why skip_loc_list_entry can't
4222 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4225 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4228 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4231 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4232 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4233 and -mstring-compare-inline-limit.
4235 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4238 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4239 for stack clash protection in a register whenever we need it to be in
4242 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4244 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4245 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4247 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4250 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4252 (mlong-double-): Ditto.
4253 * config/rs6000/sysv4.opt (msdata=): Ditto.
4254 (mtls-size=): Ditto.
4256 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4258 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4259 erroneous entries for
4260 "vector int vec_ldl (int, long int *)", and
4261 "vector unsigned int vec_ldl (int, unsigned long int *)".
4262 Add comments and entries for
4263 "vector bool char vec_ldl (int, bool char *)",
4264 "vector bool short vec_ldl (int, bool short *)",
4265 "vector bool int vec_ldl (int, bool int *)",
4266 "vector bool long long vec_ldl (int, bool long long *)",
4267 "vector pixel vec_ldl (int, pixel *)",
4268 "vector long long vec_ldl (int, long long *)",
4269 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4270 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4271 type tree bool_long_long_type_node and correct definition of
4272 bool_V2DI_type_node to make reference to this new type tree.
4273 (rs6000_mangle_type): Replace erroneous reference to
4274 bool_long_type_node with bool_long_long_type_node.
4275 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4276 comments to emphasize sign distinctions for char and int types and
4277 replace RS6000_BTI_bool_long constant with
4278 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4279 use of RS6000_BTI_pixel.
4280 (bool_long_type_node): Remove this macro definition.
4281 (bool_long_long_type_node): New macro definition
4283 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4285 PR rtl-optimization/85300
4286 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4287 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4288 simplify_unary_operation fails.
4290 2018-04-10 Martin Liska <mliska@suse.cz>
4292 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4293 cgraph_edge and ipa_ref.
4295 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4299 * config/i386/sse.md
4300 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4301 computation of the VEC_MERGE selector from mask.
4302 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4303 Fix decoding of the VEC_MERGE selector into mask.
4305 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4307 PR tree-optimization/85286
4308 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4310 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4312 * final.c (final_1): Set insn_last_address as well as
4313 insn_current_address.
4315 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4318 * explow.c (emit_stack_probe): Call validize_mem on memory location
4319 before passing it to gen_probe_stack. Create address operand and
4320 legitimize it for the probe_stack_address case.
4322 2018-04-09 Jan Hubicka <jh@suse.cz>
4325 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4326 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4327 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4329 (free_lang_data): Rebuild type inheritance graph.
4331 2018-04-09 Martin Sebor <msebor@redhat.com>
4333 * invoke.texi (-finline-small-functions): Mention other optimization
4335 (-findirect-inlining, -fpartial-inlining): Same.
4336 (-finline-functions-called-once): Same.
4337 (-freorder-blocks-and-partition): Same.
4339 2018-04-09 Jan Hubicka <jh@suse.cz>
4342 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4343 jumps; choose last target that matches the criteria (i.e.
4344 no partition changes for non-crossing jumps).
4345 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4346 support for redirecting crossing jumps to non-crossing.
4348 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4350 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4351 also for naked functions.
4353 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4355 * config/arc/arc.md (add_shift): New pattern.
4356 (add_shift2): Likewise.
4357 (sub_shift): Likewise.
4358 (sub_shift_cmp0_noout): Likewise.
4359 (compare_si_ashiftsi): Likewise.
4360 (xbfu_cmp0_noout): New combine pattern.
4361 (xbfu_cmp0"): Likewise.
4362 (movsi_set_cc_insn): Place the predicable variant first.
4363 (commutative_binary_cmp0_noout): Remove clobber.
4364 (commutative_binary_cmp0): New pattern.
4365 (noncommutative_binary_cmp0): Likewise.
4366 (noncommutative_binary_cmp0_noout): Likewise.
4367 (noncommutative_binary_comparison_result_used): Removed.
4368 (rsub_cmp0): New pattern.
4369 (rsub_cmp0_noout): Likewise.
4370 (extzvsi): Changed, keep only meaningful variants.
4371 (SQH, SEZ): New iterators.
4372 (SQH_postfix): New mode attribute.
4373 (SEZ_prefix): New code attribute.
4374 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4375 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4376 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4378 (noncommutative_operator): Check the availability of barrel
4381 2018-04-09 Richard Biener <rguenther@suse.de>
4383 PR tree-optimization/85284
4384 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4385 Only use the niter constraining form of simple_iv when the exit
4388 2018-04-09 Tom de Vries <tom@codesourcery.com>
4391 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4392 (define_expand "*memory_barrier"): New define_expand.
4393 (define_insn "memory_barrier"): New insn.
4395 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4397 PR rtl-optimization/80463
4398 PR rtl-optimization/83972
4399 PR rtl-optimization/83480
4401 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4402 correct producer for the insn.
4403 (tidy_control_flow): Fixup seqnos in case of debug insns.
4405 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4407 PR rtl-optimization/83913
4409 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4410 different sched-times when merging exprs.
4412 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4414 PR rtl-optimization/83962
4416 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4417 tidy_fallthru_edge and tidy_control_flow.
4419 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4421 PR rtl-optimization/83530
4423 * sel-sched.c (force_next_insn): New global variable.
4424 (remove_insn_for_debug): When force_next_insn is true, also leave only
4425 next insn in the ready list.
4426 (sel_sched_region): When the region wasn't scheduled, make another pass
4427 over it with force_next_insn set to 1.
4429 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4431 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4433 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4434 for interrupt control.
4435 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4436 functions for interrupt control.
4437 * config/nds32/nds32-intrinsic.md: Likewise.
4438 * config/nds32/nds32_intrinsic.h: Likewise.
4439 * config/nds32/nds32.h (nds32_builtins): Likewise.
4441 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4443 * config/nds32/nds32.c (nds32_init_machine_status,
4444 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4445 strict_aligned_p field.
4446 (nds32_expand_to_rtl_hook): New function.
4447 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4448 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4450 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4451 Chung-Ju Wu <jasonwucj@gmail.com>
4453 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4454 * config/nds32/nds32-n7.md: New file.
4455 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4456 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4458 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4459 * config/nds32/nds32.md (pipeline_model): Add n7.
4460 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4461 * config/nds32/pipelines.md: Include n7 settings.
4463 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4464 Chung-Ju Wu <jasonwucj@gmail.com>
4466 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4467 * config/nds32/nds32-e8.md: New file.
4468 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4469 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4471 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4472 * config/nds32/nds32.md (pipeline_model): Add e8.
4473 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4474 * config/nds32/pipelines.md: Include e8 settings.
4476 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4477 Chung-Ju Wu <jasonwucj@gmail.com>
4479 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4480 * config/nds32/nds32-n8.md: New file.
4481 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4482 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4484 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4485 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4486 * config/nds32/nds32.md (pipeline_model): Add n8.
4487 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4488 * config/nds32/pipelines.md: Include n8 settings.
4490 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4491 Chung-Ju Wu <jasonwucj@gmail.com>
4493 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4494 * config/nds32/nds32-n9-2r1w.md: New file.
4495 * config/nds32/nds32-n9-3r2w.md: New file.
4496 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4497 nds32_register_ports): New or modify for cpu n9.
4498 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4500 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4501 * config/nds32/nds32-utils.c: New file.
4502 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4503 TARGET_MUL_SLOW): Define.
4504 * config/nds32/nds32.md (pipeline_model): New attribute.
4505 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4506 New options that support cpu n9.
4507 * config/nds32/pipelines.md: Include n9 settings.
4508 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4510 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4513 information if necessary.
4514 (output_cond_branch_compare_zero): Likewise.
4515 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4516 (nds32_target_alignment): Refine for alignment.
4517 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4518 (FUNCTION_BOUNDARY): Modify.
4519 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4521 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4523 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4525 * config/nds32/constants.md (unspec_volatile_element): Add values for
4526 TLB operation and data prefetch.
4527 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4528 functions for TLB operation and data prefetch.
4529 * config/nds32/nds32-intrinsic.md: Likewise.
4530 * config/nds32/nds32_intrinsic.h: Likewise.
4531 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4532 (nds32_print_operand): Likewise.
4533 * config/nds32/nds32.h (nds32_builtins): Likewise.
4535 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4536 Andrew Pinski <pinsika@gcc.gnu.org>
4539 * match.pd: Use constant_boolean_node of correct type instead of
4540 boolean_true_node or boolean_false_node for simplifying
4541 pointer comparisons to zero.
4543 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4545 PR tree-optimization/80021
4546 * tree.c (verify_type_variant): Make error call in verify_variant_match
4547 translatable and remove final full stop.
4549 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4551 * config/nds32/constants.md (unspec_volatile_element): Add
4552 UNSPEC_VOLATILE_EH_RETURN.
4553 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4554 nds32_output_stack_pop): Support dwarf exception handling process.
4555 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4556 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4557 exception handling process.
4558 (nds32_compute_stack_frame): Likewise.
4559 (nds32_return_addr_rtx): Likewise.
4560 (nds32_initial_elimination_offset): Likewise.
4561 (nds32_expand_prologue): Likewise.
4562 (nds32_expand_epilogue): Likewise.
4563 (nds32_dynamic_chain_address): New function.
4564 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4566 (DYNAMIC_CHAIN_ADDRESS): Define.
4567 (EH_RETURN_DATA_REGNO): Define.
4568 (EH_RETURN_STACKADJ_RTX): Define.
4569 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4570 patterns for dwarf exception handling.
4572 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4574 * config/nds32/nds32.h: Clean up obsolete macros.
4576 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4578 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4579 Add enum values for particular instructions.
4580 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4581 particular intrinsic functions.
4582 * config/nds32/nds32-intrinsic.md: Likewise.
4583 * config/nds32/nds32_intrinsic.h: Likewise.
4584 * config/nds32/nds32.h (nds32_builtins): Likewise.
4585 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4586 (btst, ave): New patterns for particular instructions.
4588 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4590 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4591 Add enum values for atomic load/store and memory sync.
4592 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4594 * config/nds32/nds32-intrinsic.md: Likewise.
4595 * config/nds32/nds32_intrinsic.h: Likewise.
4596 * config/nds32/nds32.h (nds32_builtins): Likewise.
4598 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4600 PR tree-optimization/85257
4601 * fold-const.c (native_encode_vector): If not all elts could fit
4602 and off is -1, return 0 rather than offset.
4603 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4604 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4605 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4606 adjust buffer in native_interpret_expr call.
4608 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4610 * config/nds32/constants.md (unspec_volatile_element): Add cache
4611 control enum values.
4612 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4613 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4614 * config/nds32/nds32.c (nds32_cctl_names): New.
4615 (nds32_print_operand): Handle cache control register names.
4616 * config/nds32/nds32.h (nds32_builtins): New enum values.
4617 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4619 * config/nds32/nds32.md (type): Add mmu.
4620 * config/nds32/pipelines.md (simple_insn): Add mmu.
4622 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4624 * config/nds32/nds32.md (type): Remove call.
4625 * config/nds32/pipelines.md (simple_insn): Likewise.
4627 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4629 * config/nds32/constants.md (unspec_volatile_element): Add
4630 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4631 UNSPEC_VOLATILE_FMFCFG.
4632 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4633 description for fmfcfg and fmfcsr.
4634 (bdesc_1arg): Add fmtcsr.
4635 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4636 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4637 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4638 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4639 unspec_fmfcfg): New patterns.
4640 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4641 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4642 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4643 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4644 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4645 __nds32__fmfcfg): Define.
4647 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4649 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4650 intrinsic register names.
4651 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4652 intrinsic register enum values and macros.
4654 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4656 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4657 for load/store addressing form.
4658 (nds32_print_operand_address): Likewise.
4660 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4663 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4664 based on LABEL_REF. Remove useless assertion.
4665 (pic_address_needs_scratch): Fix formatting.
4666 (sparc_legitimize_pic_address): Minor tweaks.
4667 (sparc_delegitimize_address): Adjust assertion accordingly.
4668 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4669 into symbolic_operand.
4670 (movsi_high_pic_label_ref): Likewise.
4671 (movsi_lo_sum_pic_label_ref): Likewise.
4672 (movdi_pic_label_ref): Likewise.
4673 (movdi_high_pic_label_ref): Likewise.
4674 (movdi_lo_sum_pic_label_ref): Likewise.
4676 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4678 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4679 custom LIB_SPEC setup.
4681 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4682 Kito Cheng <kito.cheng@gmail.com>
4684 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4685 * config/riscv/freebsd.h: New.
4687 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4689 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4690 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4693 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4694 Kito Cheng <kito.cheng@gmail.com>
4696 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4697 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4698 * config/nds32/nds32-protos.h (nds32_output_call,
4699 nds32_output_return): Declare.
4700 * config/nds32/nds32.md: Refine all the call and return patterns.
4702 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4705 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4706 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4708 PR rtl-optimization/84872
4709 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4710 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4713 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4715 * expr.c (copy_blkmode_to_reg): Revert 254862.
4716 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4718 2018-04-06 Richard Biener <rguenther@suse.de>
4721 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4722 after seeing a component reference with an adjacent field. Treat
4723 refs to arrays at struct end of external decls similar to
4724 refs to unconstrained commons.
4726 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4729 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4730 look through SAVE_EXPRs with non-side-effects argument. Adjust
4732 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4735 2018-04-06 Richard Biener <rguenther@suse.de>
4738 * alias.c (find_base_term): New wrapper around find_base_term
4739 unwinding CSELIB_VAL_PTR changes.
4740 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4743 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4745 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4747 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4748 constant definitions.
4749 ("nop"): lr 0,0 -> nopr r0
4750 ("nop_lr0", "nop_lr1"): New insn definitions.
4752 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4754 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4755 NDS32_V3PUSH_AVAILABLE_P macro.
4757 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4758 Chung-Ju Wu <jasonwucj@gmail.com>
4760 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4761 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4762 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4763 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4764 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4765 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4766 * config/nds32/constraints.md: New constraints and checking for hard
4767 float configuration.
4768 * config/nds32/iterators.md: New mode iterator and attribute for hard
4769 float configuration.
4770 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4772 * config/nds32/nds32-fpu.md: New file.
4773 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4774 deal with hard float code generation.
4775 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4777 (abi_type, float_reg_number): New enum type.
4778 * config/nds32/nds32-predicates.c: New predicates for hard float.
4779 * config/nds32/nds32-protos.h: Declare functions for hard float.
4780 * config/nds32/nds32.c: Implementation for hard float configuration.
4781 * config/nds32/nds32.h: Definitions for hard float configuration.
4782 * config/nds32/nds32.md: Include hard float machine description and
4783 modify patterns for hard float configuration.
4784 * config/nds32/nds32.opt: New options for hard float configuration.
4785 * config/nds32/predicates.md: New predicates for hard float
4788 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4790 * common/config/nds32/nds32-common.c
4791 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4793 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4796 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4797 CONSTRUCTOR_ELT (ctor, ...)->value.
4799 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4802 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4804 2018-04-05 Tom de Vries <tom@codesourcery.com>
4807 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4810 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4811 Kito Cheng <kito.cheng@gmail.com>
4813 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4814 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4815 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4816 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4818 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4819 Kito Cheng <kito.cheng@gmail.com>
4821 * config/nds32/constraints.md (Ufe): New memory constraint.
4822 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4823 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4824 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4826 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4827 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4829 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4831 * config/nds32/nds32.md: Use optimize_size in the condition for
4832 alu-shift instructions.
4834 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4836 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4838 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4840 * config/nds32/nds32.md (negsi2): Refine pattern.
4842 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4843 Chung-Ju Wu <jasonwucj@gmail.com>
4845 * config/nds32/iterators.md (shift_rotate): New code iterator.
4846 (shift): New code attribute.
4847 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4848 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4849 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4850 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4851 bit-wise operations.
4852 (andsi3, *andsi3): Ditto.
4853 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4854 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4855 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4856 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4857 nds32_ior_operand, nds32_xor_operand): New predicates.
4859 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4861 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4862 (addsi3, subsi3): ... this.
4864 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4866 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4868 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4870 * config/nds32/nds32.md: Adjust indention.
4872 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4874 * config/nds32/nds32.md (feature): New attribute.
4876 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4878 * config/nds32/nds32.md (subtype): New attribute.
4880 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4883 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4884 expansion to perform a bitwise AND of the argument followed by a
4885 boolean negation of the result.
4887 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4889 PR rtl-optimization/84878
4890 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4891 the basic block. Assert the use reference is not artificial and that
4892 it has an associated insn.
4894 2018-04-04 Michael Matz <matz@suse.de>
4896 * builtins.c (compute_objsize): Pass correct operand
4897 to array_at_struct_end_p.
4899 2018-04-04 Richard Biener <rguenther@suse.de>
4902 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4903 from contexts for DINFO_LEVEL_TERSE and below.
4905 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4907 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4908 resiter_operand condition.
4909 * config/nds32/nds32.md (*move<mode>): Ditto.
4911 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4912 Monk Chiang <sh.chiang04@gmail.com>
4914 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4916 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4918 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4920 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4921 Kito Cheng <kito.cheng@gmail.com>
4923 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4924 nds32_cond_code_str, output_cond_branch,
4925 output_cond_branch_compare_zero, nds32_expand_cbranch,
4926 nds32_expand_cstore, nds32_expand_movcc,
4927 nds32_output_cbranchsi4_equality_zero,
4928 nds32_output_cbranchsi4_equality_reg,
4929 nds32_output_cbranchsi4_equality_reg_or_const_int,
4930 nds32_output_cbranchsi4_greater_less_zero: New functions.
4931 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4932 nds32_expand_cstore, nds32_expand_movcc,
4933 nds32_output_cbranchsi4_equality_zero,
4934 nds32_output_cbranchsi4_equality_reg,
4935 nds32_output_cbranchsi4_equality_reg_or_const_int,
4936 nds32_output_cbranchsi4_greater_less_zero): Declare.
4937 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4938 nds32_rimm11s_operand): New predicates.
4939 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4940 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4943 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4945 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4946 * config/nds32/nds32.md: Ditto.
4947 * config/nds32/pipelines.md: Ditto.
4949 2018-04-04 Richard Biener <rguenther@suse.de>
4951 PR tree-optimization/85168
4952 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4953 propagating abnormals.
4955 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4957 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4959 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4960 Kito Cheng <kito.cheng@gmail.com>
4962 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4963 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4964 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4965 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4966 * config/nds32/nds32.md (sibcall_internal): New.
4967 (sibcall_register): Remove.
4968 (sibcall_immediate): Remove.
4969 (sibcall_value_internal): New.
4970 (sibcall_value_register): Remove.
4971 (sibcall_value_immediate): Remove.
4972 * config/nds32/predicates.md (nds32_general_register_operand): New.
4973 (nds32_call_address_operand): New.
4975 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4977 PR rtl-optimization/85167
4978 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4979 bb_defs if *split_p, instead preinitialize it to NULL.
4981 PR tree-optimization/85156
4982 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4983 evaluating the argument multiple times.
4985 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4987 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4989 (_mm_cvtpd_ps): Likewise.
4990 (_mm_cvttpd_epi32): Likewise.
4991 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4992 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4993 vector, pixel, and bool following altivec.h include.
4995 2018-04-03 Martin Sebor <msebor@redhat.com>
4997 * doc/extend.texi (Common Function Attributes): Clarify.
4998 (const attribute): Likewise.
4999 (pure attribute): Likewise.
5001 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5004 * config/i386/i386.c (ix86_expand_vector_set): Use
5005 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5007 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5009 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5010 instructions when changing rounding bits to preserve precision bits
5011 in the x87 control word.
5013 2018-04-03 Martin Liska <mliska@suse.cz>
5015 PR tree-optimization/82491
5016 * rtl.h (strip_offset_and_add): Replace += suboffset with
5017 poly_uint64 () + suboffset.
5019 2018-03-29 Martin Liska <mliska@suse.cz>
5020 Martin Jambor <mjambor@suse.cz>
5023 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5024 param_type is not an integral or pointer type.
5026 2018-04-03 Richard Biener <rguenther@suse.de>
5028 * sese.h (recompute_all_dominators): Remove.
5030 2018-04-02 Martin Sebor <msebor@redhat.com>
5032 * doc/invoke.texi (-Wrestrict): Fix typos.
5034 2018-04-02 Jim Wilson <jimw@sifive.com>
5036 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5037 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5038 (<optab>di3, <optab>si3_extend): Likewise.
5039 (<optab>si3_mask, <optab>si3_mask_1): New.
5040 (<optab>di3_mask, <optab>di3_mask_1): New.
5041 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5042 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5043 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5045 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5047 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5050 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5052 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5053 (nds32_canonicalize_comparison): New function.
5055 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5056 Kito Cheng <kito.cheng@gmail.com>
5057 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5059 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5060 * config/nds32/constants.md (unspec_volatile_element): Add
5061 UNSPEC_VOLATILE_RELAX_GROUP.
5062 * config/nds32/nds32-relax-opt.c: New file.
5063 * config/nds32/nds32-predicates.c
5064 (nds32_symbol_load_store_p): New function.
5065 * config/nds32/nds32-protos.h
5066 (nds32_symbol_load_store_p): Declare function.
5067 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5068 * config/nds32/nds32.c
5069 (nds32_register_pass): New function to register pass.
5070 (nds32_register_passes): New function to register passes.
5071 * config/nds32/nds32.md (relax_group): New pattern.
5072 * config/nds32/nds32.opt (mrelax-hint): New option.
5073 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5075 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5077 * config/nds32/t-nds32: Modify files dependency.
5079 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5081 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5082 (PROFILE_HOOK): Define its implementation.
5084 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5086 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5087 type and 32-bit size.
5089 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5092 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5093 (V_128_256): New mode iterator.
5094 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5095 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5096 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5098 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5099 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5101 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5104 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5105 NaN inputs correctly.
5107 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5110 * config/rs6000/vsx.md (??r): New mode attribute.
5111 (*vsx_mov<mode>_64bit): Use it.
5112 (*vsx_mov<mode>_32bit): Likewise.
5114 2018-03-30 Martin Sebor <msebor@redhat.com>
5116 PR tree-optimization/84818
5117 * builtins.c (check_access): Use warning_n.
5119 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5122 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5124 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5127 2018-03-30 Julia Koval <julia.koval@intel.com>
5130 * x86-tune.def (movx, partial_reg_dependency): Enable for
5133 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5136 * lra-constraints.c (process_alt_operands): Move setting
5137 this_alternative_matches below.
5139 2018-03-29 Martin Liska <mliska@suse.cz>
5142 * doc/invoke.texi: Document how LTO works with debug info.
5143 Describe auto-load support of binutils. Mention 'x86-64'
5144 as valid option value of -march option.
5146 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5148 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5151 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5152 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5153 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5156 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5159 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5160 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5161 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5162 for RS6000_BTM_POWERPC64.
5163 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5164 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5165 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5170 2018-03-28 Carl Love <cel@us.ibm.com>
5173 2017-09-27 Carl Love <cel@us.ibm.com>
5175 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5176 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5177 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5180 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5182 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5183 instead of __vector bool.
5184 (_mm_max_pu8): Likewise.
5185 (_mm_min_pi16): Likewise.
5187 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5190 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5191 (DIVWEUO): Likewise.
5193 (DIVDEUO): Likewise.
5194 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5195 DIVWEUO and DIVDEUO.
5196 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5197 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5198 (div_extend): Likewise.
5199 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5201 (__builtin_divweuo): Likewise.
5202 (__builtin_divdeo): Likewise.
5203 (__builtin_divdeuo): Likewise.
5205 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5208 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5209 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5211 PR tree-optimization/82004
5212 * gimple-match-head.c (optimize_pow_to_exp): New function.
5213 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5214 Don't fold to exp if optimize_pow_to_exp is false.
5216 2018-03-28 Martin Liska <mliska@suse.cz>
5219 * calls.c (initialize_argument_information): Fix trailing space.
5220 * common.opt: Fix typo and provide better explanation for
5221 -fsanitize-coverage option.
5222 * config/i386/i386.opt: Fix typo.
5224 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5225 Martin Liska <mliska@suse.cz>
5228 * gimplify.c (asan_poison_variable): Don't do the check for
5229 gimplify_omp_ctxp here.
5230 (gimplify_decl_expr): Do it here.
5231 (gimplify_target_expr): Likewise.
5233 2018-03-28 Martin Liska <mliska@suse.cz>
5236 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5237 chkp_type_bounds_count if MPX is not enabled.
5239 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5241 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5243 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5246 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5247 function to create the function decl for complex long double
5248 multiply and divide for -mabi=ieeelongdouble.
5249 (init_float128_ieee): Call it.
5251 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5254 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5255 -fcf-protection=branch -mibt.
5256 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5258 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5261 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5263 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5266 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5267 extern array declarations.
5269 2018-03-27 Richard Biener <rguenther@suse.de>
5272 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5273 explicit single_use checks.
5275 2018-03-27 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/85082
5278 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5281 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5283 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5284 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5285 Turn on fasynchronous-unwind-tables and funwind-tables.
5287 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5290 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5291 (*bmi_blsr_<mode>_ccz): Ditto.
5293 2018-03-26 Tom de Vries <tom@codesourcery.com>
5295 PR tree-optimization/85063
5296 * omp-general.c (offloading_function_p): New function. Factor out
5298 * omp-offload.c (pass_omp_target_link::gate): ... here.
5299 * omp-general.h (offloading_function_p): Declare.
5300 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5301 with attribute omp declare target for offloading functions.
5303 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5305 PR tree-optimization/84005
5306 * tree-data-ref.h (get_base_for_alignment): Declare.
5307 * tree-data-ref.c (get_base_for_alignment_1): New function.
5308 (get_base_for_alignment): Likewise.
5309 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5310 get_base_for_alignment to find a suitable base object, instead
5311 of always using drb->base_address.
5313 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5316 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5317 known size by default.
5319 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5322 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5323 and non BLKmode operands.
5325 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5328 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5329 Clean up attributes.
5331 2018-03-23 Richard Biener <rguenther@suse.de>
5334 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5335 we are going to emit early debug for LTO.
5337 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5340 * function.c (match_asm_constraints_1): Don't optimize if input
5341 doesn't satisfy general_operand predicate for output's mode.
5344 * alias.c (write_dependence_p): Don't require for x_canonicalized
5345 non-VOIDmode if x has VOIDmode.
5348 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5349 just don't try to optimize it rather than assert it never happens.
5351 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5353 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5354 macro expansions for definition of ST_INTERNAL_<mode> and
5355 LD_INTERNAL_<mode> builtins.
5356 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5358 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5360 (altivec_expand_st_builtin): Likewise.
5361 (altivec_expand_builtin): Remove calls to deleted functions.
5362 (rs6000_address_for_altivec): Delete this function.
5363 * config/rs6000/vector.md: Remove expands for
5364 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5366 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5369 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5370 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5371 re-computing once computed.
5372 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5373 (arm_init_machine_status): Initialize
5374 machine->static_chain_stack_bytes.
5376 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5379 * doc/extend.texi: Add four new prototypes for vec_ld.
5380 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5381 definitions for more logical presentation.
5382 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5383 entries for V1TI variants of __builtin_altivec_ld builtin.
5384 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5385 handling of V1TI variant of LVX icode pattern.
5386 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5387 (rs6000_gimple_fold_builtin): Likewise.
5388 (altivec_init_builtins): Add code to define
5389 __builtin_altivec_lvx_v1ti function.
5391 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5394 * function.c (match_asm_constraints_1): Don't do the optimization
5395 if input isn't a REG, SUBREG, MEM or constant.
5397 2018-03-22 Tom de Vries <tom@codesourcery.com>
5399 PR tree-optimization/84956
5400 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5401 bb_has_abnormal_pred.
5403 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5406 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5407 DECL_INITIAL (decl) to decl at the end.
5408 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5411 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5413 * doc/extend.texi (__builtin_tgmath): Document when complex
5414 integer types are treated as _Complex _Float64.
5416 2018-03-21 Tom de Vries <tom@codesourcery.com>
5418 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5420 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5422 PR tree-optimization/84960
5423 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5424 if it is ENTRY block, move them into single succ of ENTRY in that case.
5426 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5428 PR tree-optimization/84811
5429 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5430 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5431 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5432 (known_subrange_p): Update accordingly. Cast each value involved
5433 in the size comparison, rather than casting the result of the
5436 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5438 PR tree-optimization/84982
5439 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5440 by flipping the least significant bit rather than all bits from
5441 bitpos to bitpos + bitsize - 1.
5443 2018-03-21 Nathan Sidwell <nathan@acm.org>
5445 * doc/extend.texi (Deprecated Features): Remove mention of
5446 long-deleted deprecations.
5448 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5451 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5452 * configure: Regenerate.
5454 2018-03-21 Tom de Vries <tom@codesourcery.com>
5456 PR tree-optimization/83126
5457 * tree-parloops.c (num_phis): New function.
5458 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5460 2018-03-21 Nathan Sidwell <nathan@acm.org>
5462 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5463 mention anon-struct/union members and trailing attributes.
5465 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5467 PR tree-optimization/84969
5468 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5469 builtin memset partitions if they set different rhs values.
5471 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5473 PR rtl-optimization/84989
5474 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5475 VEC_DUPLICATE with scalar result mode.
5477 2018-03-21 Martin Liska <mliska@suse.cz>
5480 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5481 not intended return statement.
5483 2018-03-21 Martin Liska <mliska@suse.cz>
5486 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5487 (chkp_find_bound_slots_1): Limit number of iterations.
5489 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5492 * Minor grammar fixes for x86 options.
5494 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5497 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5498 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5500 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5503 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5504 (altivec_lvx_<mode>_1op): Likewise.
5505 (altivec_stvx_<mode>_2op): Likewise.
5506 (altivec_stvx_<mode>_1op): Likewise.
5507 (altivec_lvx_<VM2:mode>): New define_expand.
5508 (altivec_stvx_<VM2:mode>): Likewise.
5509 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5510 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5511 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5512 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5513 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5514 (rs6000_gen_lvx): Likewise.
5515 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5516 (altivec_expand_stv_builtin): Likewise.
5517 (altivec_expand_builtin): Likewise.
5518 * config/rs6000/vector.md: Likewise.
5520 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5523 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5526 2018-03-20 Richard Biener <rguenther@suse.de>
5529 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5530 sign-conversions as zero, fall back to standard scalar_stmt
5533 2018-03-20 Martin Liska <mliska@suse.cz>
5536 * predict.c (rebuild_frequencies): Handle case when we have
5537 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5539 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5542 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5543 flag_section_anchors.
5544 * varasm.c (use_blocks_for_decl_p): Remove hack for
5545 dw2_force_const_mem.
5548 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5550 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5551 be created, use lowpart_subreg of operands[0] rather than operands[0]
5553 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5554 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5555 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5556 and n constraint instead of aarch64_shift_imm_di and Usd.
5557 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5558 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5560 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5563 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5564 to favor GPR over NEON registers.
5565 (<shift>di3_neon): Likewise.
5567 2018-03-20 Tom de Vries <tom@codesourcery.com>
5570 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5571 (nvptx_process_pars): Emit bar.sync asap and alap.
5573 2018-03-20 Tom de Vries <tom@codesourcery.com>
5576 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5577 seen_label if seen_label is already set.
5579 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5582 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5583 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5584 Use 1U instead of 1. Formatting fixes.
5587 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5588 instead of TREE_TYPE (s1) for the return value.
5590 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5592 PR tree-optimization/84946
5593 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5594 bitsize + bitsize in poly_uint64 rather than poly_int64.
5597 * dwarf2asm.c: Include fold-const.c.
5598 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5599 of decl rather than decl itself.
5601 PR rtl-optimization/84643
5602 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5604 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5607 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5608 calling assemble_variable.
5610 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5613 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5614 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5616 2018-03-19 Jim Wilson <jimw@sifive.com>
5619 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5620 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5621 (riscv_first_stack_step): Likewise.
5622 (riscv_option_override): Use STACK_BOUNDARY instead of
5624 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5626 (BIGGEST_ALIGNMENT): Set to 128.
5627 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5628 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5631 2018-03-19 Richard Biener <rguenther@suse.de>
5633 PR tree-optimization/84933
5634 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5635 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5637 2018-03-19 Richard Biener <rguenther@suse.de>
5639 PR tree-optimization/84859
5640 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5641 (cond_if_else_store_replacement): Perform sinking operation on
5642 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5643 Generalize what a BB with a single eligible store is.
5645 2018-03-19 Richard Biener <rguenther@suse.de>
5647 PR tree-optimization/84929
5648 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5649 chrec_is_positive against non-chrec arg.
5651 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5654 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5656 2018-03-18 Martin Liska <mliska@suse.cz>
5658 PR rtl-optimization/84635
5659 * regrename.c (build_def_use): Use matches_mode only when
5662 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5664 PR tree-optimization/84913
5665 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5666 vectorize chains of COND_EXPRs.
5668 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5670 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5672 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5674 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5676 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5678 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5680 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5681 Kito Cheng <kito.cheng@gmail.com>
5683 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5684 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5685 (nds32_adjust_reg_alloc_order): New function.
5686 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5688 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5690 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5691 nds32_print_operand, nds32_print_operand_address): Use
5692 HOST_WIDE_INT_PRINT_DEC instead.
5694 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5696 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5698 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5701 * config/i386/i386.c (initial_ix86_tune_features,
5702 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5704 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5705 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5706 rather than 1u << ix86_tune. Formatting fix.
5707 (ix86_option_override_internal): Change ix86_arch_mask from
5708 unsigned int to unsigned HOST_WIDE_INT, initialize to
5709 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5710 (ix86_function_specific_restore): Likewise.
5712 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5715 * postreload.c (reload_combine_recognize_pattern): Perform
5716 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5717 truncate_int_for_mode the result for the destination's mode.
5720 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5721 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5723 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5726 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5729 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5730 regno_allocno_class_array and sorted_pseudos.
5731 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5732 insns where regno is used.
5734 2018-03-16 Martin Liska <mliska@suse.cz>
5737 * multiple_target.c (create_dispatcher_calls): Redirect
5738 reference in the symbol table.
5740 2018-03-16 Martin Liska <mliska@suse.cz>
5743 * multiple_target.c (create_dispatcher_calls): Redirect also
5746 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5750 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5751 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5752 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5754 2018-03-16 Julia Koval <julia.koval@intel.com>
5756 * doc/invoke.texi (Skylake Server): Add CLWB.
5757 Cannonlake): Remove CLWB.
5759 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5761 PR tree-optimization/84841
5762 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5764 (FLOAT_ONE_CONST_TYPE): Define.
5765 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5766 (sort_by_operand_rank): Put entries with higher constant_type last
5767 rather than first to match comments.
5769 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5771 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5774 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5777 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5778 check and use address of target always.
5780 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5783 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5784 (indirect_thunk_bnd_needed): Likewise.
5785 (indirect_thunks_used): Likewise.
5786 (indirect_thunks_bnd_used): Likewise.
5787 (indirect_return_needed): New.
5788 (indirect_return_bnd_needed): Likewise.
5789 (output_indirect_thunk_function): Add a bool argument for
5791 (output_indirect_thunk_function): Don't generate alias for
5792 function return thunk.
5793 (ix86_code_end): Call output_indirect_thunk_function to generate
5794 function return thunks.
5795 (ix86_output_function_return): Set indirect_return_bnd_needed
5796 and indirect_return_needed instead of indirect_thunk_bnd_needed
5797 and indirect_thunk_needed.
5799 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5801 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5802 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5803 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5805 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5806 Paul Hua <paul.hua.gm@gmail.com>
5809 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5811 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5813 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5814 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5817 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5820 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5821 instead of GET_MODE_SIZE when comparing Units.
5823 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5826 * varasm.c (hash_section): Return an unchangeble hash value
5827 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5828 Return !aarch64_can_use_per_function_literal_pools_p ().
5830 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5833 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5834 rather than address of cmode as last argument to prepare_cmp_insn.
5836 2018-03-15 Julia Koval <julia.koval@intel.com>
5838 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5839 F_AVX512VNNI, F_AVX512BITALG): New.
5841 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5844 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5845 insn for floating-point loads and stores.
5847 2018-03-14 Carl Love <cel@us.ibm.com>
5849 * config/rs6000/rs6000-c.c: Add macro definitions for
5850 ALTIVEC_BUILTIN_VEC_PERMXOR.
5851 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5852 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5853 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5854 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5856 * config/doc/extend.texi: Add prototypes for vec_permxor.
5858 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5861 * diagnostic-show-locus.c (class layout_point): Convert m_line
5862 from int to linenum_type.
5863 (line_span::comparator): Use linenum "compare" function when
5864 comparing line numbers.
5865 (test_line_span): New function.
5866 (layout_range::contains_point): Convert param "row" from int to
5868 (layout_range::intersects_line_p): Likewise.
5869 (layout::will_show_line_p): Likewise.
5870 (layout::print_source_line): Likewise.
5871 (layout::should_print_annotation_line_p): Likewise.
5872 (layout::print_annotation_line): Likewise.
5873 (layout::print_leading_fixits): Likewise.
5874 (layout::annotation_line_showed_range_p): Likewise.
5875 (struct line_corrections): Likewise for field m_row.
5876 (line_corrections::line_corrections): Likewise for param "row".
5877 (layout::print_trailing_fixits): Likewise.
5878 (layout::get_state_at_point): Likewise.
5879 (layout::get_x_bound_for_row): Likewise.
5880 (layout::print_line): Likewise.
5881 (diagnostic_show_locus): Likewise for locals "last_line" and
5883 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5884 * input.c (selftest::test_linenum_comparisons): New function.
5885 (selftest::input_c_tests): Call it.
5886 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5887 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5888 * selftest.h (ASSERT_GT): New macro.
5889 (ASSERT_GT_AT): New macro.
5890 (ASSERT_LT): New macro.
5891 (ASSERT_LT_AT): New macro.
5893 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5895 PR rtl-optimization/84780
5896 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5898 2018-03-14 Martin Liska <mliska@suse.cz>
5900 * tree.c (record_node_allocation_statistics): Use
5901 get_stats_node_kind.
5902 (get_stats_node_kind): New function extracted from
5903 record_node_allocation_statistics.
5904 (free_node): Use get_stats_node_kind.
5906 2018-03-14 Richard Biener <rguenther@suse.de>
5908 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5909 that the value-set of ANTIC_IN doesn't grow.
5912 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5914 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5915 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5917 2018-03-14 Julia Koval <julia.koval@intel.com>
5919 * config.gcc (icelake-client, icelake-server): New.
5921 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5922 (initial_ix86_arch_features): Ditto.
5923 (PTA_SKYLAKE): Add SGX.
5924 (PTA_ICELAKE): Remove.
5925 (PTA_ICELAKE_CLIENT): New.
5926 (PTA_ICELAKE_SERVER): New.
5927 (ix86_option_override_internal): Split up icelake on icelake client and
5929 (get_builtin_code_for_version): Ditto.
5930 (fold_builtin_cpu): Ditto.
5931 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5932 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5933 * config/i386/i386.h (processor_type): Ditto.
5934 * doc/invoke.texi: Ditto.
5936 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5939 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5940 INTEGER_CST offset, add it together with bitpos / 8 and
5941 sign extend based on POINTER_SIZE.
5945 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5948 * config/i386/constraints.md (Yc): New register constraint.
5949 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5950 Use Yc constraint for alternative 2 of operand 0. Remove
5951 preferred_for_speed attribute.
5953 2018-03-14 Richard Biener <rguenther@suse.de>
5955 PR tree-optimization/84830
5956 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5957 with the old one to avoid oscillations.
5959 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5962 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5964 (assign_by_spills): Return a flag of reload assignment failure.
5965 Do not process the reload assignment failures. Do not spill other
5966 reload pseudos if they has the same reg class. Update n if
5968 (lra_assign): Add a return arg. Set up from the result of
5969 assign_by_spills call.
5970 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5971 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5972 usage_insns if it is not NULL.
5973 (spill_hard_reg_in_range): New function.
5974 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5975 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5976 function prototypes.
5977 (lra_assign): Change prototype.
5978 * lra.c (lra): Add code to deal with fails by splitting hard reg
5981 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5983 * config/riscv/riscv.opt (mrelax): New option.
5984 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5985 "norelax" when riscv_mrelax is disabled.
5986 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5988 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5991 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5992 reassociation for int modes.
5994 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5996 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5997 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5999 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6000 * config/aarch64/aarch64-sve.md
6001 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6002 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6003 (*extend<mode><Vwide>2): Rename to...
6004 (aarch64_sve_extend<mode><Vwide>2): ...this.
6005 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6006 renaming the old pattern to...
6007 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6009 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6010 define_expand, renaming the old pattern to...
6011 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6012 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6013 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6014 account when deciding which SVE instruction the optab should use.
6015 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6017 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6019 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6020 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6021 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6022 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6023 (tlsdesc_small_<mode>): Turn a define_expand and use
6024 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6025 (tlsdesc_small_advsimd_<mode>): ...this.
6026 (tlsdesc_small_sve_<mode>): New pattern.
6028 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6030 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6031 (UNSPEC_UMUL_HIGHPART): New constants.
6032 (MUL_HIGHPART): New int iteraor.
6033 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6034 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6036 (*<su>mul<mode>3_highpart): New define_insn.
6038 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6041 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6044 2018-03-13 Martin Liska <mliska@suse.cz>
6047 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6049 (sem_item_optimizer::~sem_item_optimizer): Release it.
6050 (sem_item_optimizer::merge_classes): Register variable aliases.
6051 (sem_item_optimizer::fixup_pt_set): New function.
6052 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6053 * ipa-icf.h: Declare new variables and functions.
6055 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6058 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6059 integer_pow2p@2 and test integer_pow2p in condition.
6060 (A < 0 ? C : 0): Similarly for @1.
6063 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6064 characters starting at p contain '\0' character, don't look beyond
6068 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6069 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6072 * reg-stack.c (change_stack): Change update_end var from int to
6073 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6074 also call set_block_for_insn on the newly added insns and rescan.
6077 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6078 on the last operand.
6081 * tree.c (stabilize_reference_1): Return save_expr (e) for
6082 STATEMENT_LIST even if it doesn't have side-effects.
6084 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6086 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6088 2018-03-12 Renlin Li <renlin.li@arm.com>
6090 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6091 aarch64_output_scalar_simd_mov_immediate.
6093 2018-03-12 Martin Sebor <msebor@redhat.com>
6095 PR tree-optimization/83456
6096 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6097 for perfectly overlapping calls to memcpy.
6098 (gimple_fold_builtin_memory_chk): Same.
6099 (gimple_fold_builtin_strcpy): Handle no-warning.
6100 (gimple_fold_builtin_stxcpy_chk): Same.
6101 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6103 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6105 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6106 parameter. Use it for SFmode.
6107 (rs6000_function_arg_advance_1): Adjust.
6108 (rs6000_function_arg): Adjust.
6109 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6111 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6113 PR rtl-optimization/84169
6114 PR rtl-optimization/84780
6115 * combine.c (can_combine_p): Check for a 2-insn combination whether
6116 the destination register is used between the two insns, too.
6118 2018-03-12 Richard Biener <rguenther@suse.de>
6120 PR tree-optimization/84803
6121 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6122 for refs DR analysis didn't process.
6124 2018-03-12 Richard Biener <rguenther@suse.de>
6126 PR tree-optimization/84777
6127 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6128 force-vectorize loops ignore whether we are optimizing for size.
6130 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6132 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6133 (TARGET_MD_ASM_ADJUST): Define.
6135 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6136 Kito Cheng <kito.cheng@gmail.com>
6137 Chung-Ju Wu <jasonwucj@gmail.com>
6139 * config/nds32/nds32.c (nds32_compute_stack_frame,
6140 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6141 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6142 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6143 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6144 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6145 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6146 * config/nds32/nds32.md (prologue, epilogue): Use macro
6147 NDS32_V3PUSH_AVAILABLE_P to do checking.
6149 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6152 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6153 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6154 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6155 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6156 addition of most attributes on !orig_type_die or the attribute not
6157 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6159 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6160 Chung-Ju Wu <jasonwucj@gmail.com>
6162 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6164 * config/nds32/nds32.opt (mvh): New option.
6166 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6167 Chung-Ju Wu <jasonwucj@gmail.com>
6169 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6171 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6172 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6175 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6176 Chung-Ju Wu <jasonwucj@gmail.com>
6178 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6180 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6181 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6183 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6184 Kito Cheng <kito.cheng@gmail.com>
6185 Chung-Ju Wu <jasonwucj@gmail.com>
6187 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6188 UNSPEC_FFMISM and UNSPEC_FLMISM.
6189 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6190 for ffb, ffmism and flmism.
6191 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6192 (unspec_ffmism): Ditto.
6193 (unspec_flmism): Ditto.
6194 (nds32_expand_builtin_impl): Check if string extension is available.
6195 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6196 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6198 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6201 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6204 * lra-assigns.c (assign_by_spills): Return a flag of reload
6205 assignment failure. Do not process the reload assignment
6206 failures. Do not spill other reload pseudos if they has the same
6208 (lra_assign): Add a return arg. Set up from the result of
6209 assign_by_spills call.
6210 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6211 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6212 usage_insns if it is not NULL.
6213 (spill_hard_reg_in_range): New function.
6214 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6215 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6216 function prototypes.
6217 (lra_assign): Change prototype.
6218 * lra.c (lra): Add code to deal with fails by splitting hard reg
6221 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6224 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6226 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6229 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6230 (dw_val_node): Add val_symbolic_view.
6231 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6232 (symview_upper_bound): New.
6233 (new_line_info_table): Initialize symviews_since_reset.
6234 (dwarf2out_source_line): Count symviews_since_reset and set
6235 symview_upper_bound.
6236 (dw_val_equal_p): Handle symview.
6237 (add_AT_symview): New.
6238 (print_dw_val): Handle symview.
6239 (attr_checksum, attr_checksum_ordered): Likewise.
6240 (same_dw_val_p, size_of_die): Likewise.
6241 (value_format, output_die): Likewise.
6242 (add_high_low_attributes): Use add_AT_symview for entry_view.
6243 (dwarf2out_finish): Reset symview_upper_bound, clear
6246 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6249 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6250 Add strict argument and use it.
6251 (rs6000_split_multireg_move): Update for new strict argument.
6252 (mem_operand_gpr): Disallow all non-offsettable addresses.
6253 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6255 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6258 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6259 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6260 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6263 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6264 decl, use remap_type if we want to use the type.
6266 2018-03-09 Martin Sebor <msebor@redhat.com>
6268 PR tree-optimization/84526
6269 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6271 (builtin_access::generic_overlap): Be prepared to handle non-array
6274 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6276 PR rtl-optimization/84682
6277 * lra-constraints.c (process_address_1): Check is_address flag
6278 for address constraints.
6279 (process_alt_operands): Likewise.
6280 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6281 preprocess_constraints.
6282 * recog.h (preprocess_constraints): Add oploc parameter.
6284 * recog.c (preprocess_constraints): Test address_operand for
6285 CT_ADDRESS constraints.
6287 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6290 * lra-assigns.c (assign_by_spills): Return a flag of reload
6291 assignment failure. Do not process the reload assignment
6292 failures. Do not spill other reload pseudos if they has the same
6294 (lra_assign): Add a return arg. Set up from the result of
6295 assign_by_spills call.
6296 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6297 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6298 usage_insns if it is not NULL.
6299 (spill_hard_reg_in_range): New function.
6300 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6301 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6302 function prototypes.
6303 (lra_assign): Change prototype.
6304 * lra.c (lra): Add code to deal with fails by splitting hard reg
6307 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6310 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6311 Accept complain bool parameter. Only emit errors if it is true.
6312 (arm_parse_cpu_option_name): Likewise.
6313 (arm_target_thumb_only): Adjust callers of the above.
6314 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6315 prototype to take a default true bool parameter.
6316 (arm_parse_arch_option_name): Likewise.
6318 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6319 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6323 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6324 * configure: Regenerate.
6325 * configure.ac ("linker --version-script option"): New.
6326 ("linker soname option"): New.
6328 2018-03-09 Richard Biener <rguenther@suse.de>
6330 PR tree-optimization/84775
6331 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6332 immediate uses of predicate stmts and mark them modified.
6335 PR tree-optimization/84178
6336 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6338 (version_loop_for_if_conversion): Delay update_ssa call.
6339 (tree_if_conversion): Delay update_ssa until after predicate
6342 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6345 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6346 when the function accesses prior frames.
6348 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6351 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6352 gen_llsym, otherwise call maybe_gen_llsym.
6355 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6356 has ',' character inside of it.
6358 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6361 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6362 as clobbering CC_REGNUM.
6364 2018-03-08 Richard Biener <rguenther@suse.de>
6367 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6368 (follow_copies_to_constant): Do not follow SSA names registered
6371 2018-03-08 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/84178
6374 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6376 (version_loop_for_if_conversion): Delay update_ssa call.
6377 (tree_if_conversion): Delay update_ssa until after predicate
6380 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6382 PR tree-optimization/84178
6383 * tree-if-conv.c (release_bb_predicate): Remove the
6384 the assertion that the stmts have NULL use_ops.
6385 Discard the statements, asserting that they haven't
6386 yet been added to a BB.
6388 2018-03-08 Richard Biener <rguenther@suse.de>
6390 PR tree-optimization/84746
6391 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6392 (phi_translate): Pass in destination ANTIC_OUT set.
6393 (phi_translate_1): Likewise. For a simplified result lookup
6394 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6395 (phi_translate_set): Adjust.
6396 (do_pre_regular_insertion): Likewise.
6397 (do_pre_partial_partial_insertion): Likewise.
6399 2018-03-08 Martin Liska <mliska@suse.cz>
6401 PR gcov-profile/84735
6402 * doc/gcov.texi: Document usage of profile files.
6403 * gcov-io.h: Document changes in the format.
6405 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6409 * dwarf2out.c (struct dw_line_info_table): Update comments for
6411 (FORCE_RESET_NEXT_VIEW): New.
6412 (FORCE_RESETTING_VIEW_P): New.
6413 (RESETTING_VIEW_P): Check for -1 too.
6414 (ZERO_VIEW_P): Likewise.
6415 (new_line_info_table): Force-reset next view.
6416 (dwarf2out_begin_function): Likewise.
6417 (dwarf2out_source_line): Simplify zero_view_p initialization.
6418 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6419 view directly. Omit view when omitting .loc at line 0.
6421 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6423 PR tree-optimization/84740
6424 * tree-switch-conversion.c (process_switch): Call build_constructors
6425 only if info.phi_count is non-zero.
6427 PR tree-optimization/84739
6428 * tree-tailcall.c (find_tail_calls): Check call arguments against
6429 DECL_ARGUMENTS (current_function_decl) rather than
6430 DECL_ARGUMENTS (func) when checking for tail recursion.
6432 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6434 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6435 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6436 Volker Reichelt's entry and add entries for people that perform
6437 GCC fuzzy testing and report numerous bugs.
6439 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6442 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6443 readonly data in sdata, if that is disabled.
6444 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6445 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6446 -mreadonly-in-sdata option.
6448 2018-03-07 Martin Sebor <msebor@redhat.com>
6450 PR tree-optimization/84468
6451 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6452 basic block when looking for nul assignment.
6454 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6457 * except.h (output_function_exception_table): Adjust prototype.
6458 * except.c (output_function_exception_table): Remove FNNAME parameter
6459 and add SECTION parameter. Ouput one part of the table at a time.
6460 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6461 the first part of the exception table and emit unwind directives.
6462 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6463 (i386_pe_seh_cold_init): Likewise.
6464 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6465 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6466 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6467 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6468 * config/i386/winnt.c: Include except.h.
6469 (struct seh_frame_state): Add reg_offset, after_prologue and
6470 in_cold_section fields.
6471 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6472 (i386_pe_seh_cold_init): New function.
6473 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6474 to seh->in_cold_section.
6475 (seh_emit_push): Record the offset of the push.
6476 (seh_emit_save): Record the offet of the save.
6477 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6478 Test seh->after_prologue to disregard the epilogue.
6479 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6480 (i386_pe_end_cold_function): New function.
6482 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6485 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6486 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6489 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6491 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6492 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6495 * multiple_target.c: Include tree-inline.h and intl.h.
6496 (expand_target_clones): Diagnose and fail if node->definition and
6497 !tree_versionable_function_p (node->decl).
6499 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6501 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6503 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6504 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6505 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6507 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6510 * combine.c (try_combine): Use reg_or_subregno instead of handling
6511 just paradoxical SUBREGs and REGs.
6513 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6515 * config/arc/arc.c (arc_finalize_pic): Remove function.
6516 (arc_must_save_register): We use single base PIC register, remove
6517 checks to save/restore the PIC register.
6518 (arc_expand_prologue): Likewise.
6519 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6521 (arc_verify_short): Likewise.
6522 (arc_attr_type): Likewise.
6523 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6524 (walk_stores): Likewise.
6525 (arc_address_cost): Make it static.
6526 (arc_verify_short): Likewise.
6527 (branch_dest): Likewise.
6528 (arc_attr_type): Likewise.
6529 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6530 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6531 (arc_final_prescan_insn): Remove inserting the nops due to
6532 hardware hazards. It is done in reorg step.
6533 (insn_length_variant_t): Remove.
6534 (insn_length_parameters_t): Likewise.
6535 (arc_insn_length_parameters): Likewise.
6536 (arc_get_insn_variants): Likewise.
6537 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6539 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6542 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6545 PR tree-optimization/84687
6546 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6548 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6550 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6552 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6553 Rename to ppc_speculation_barrier.
6554 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6555 __builtin_ppc_speculation_barrier.
6557 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6560 * combine.c (combine_simplify_rtx): Don't try to simplify if
6561 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6564 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6566 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6567 to 32 bytes when compiling for POWER9.
6569 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6572 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6573 regparm >= 3 with no arg reg available also for calls with
6574 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6577 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6579 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6581 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6584 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6586 2018-03-05 Richard Biener <rguenther@suse.de>
6588 PR tree-optimization/84486
6589 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6590 When inserting a __builtin_assume_aligned call set the LHS
6591 SSA name alignment info accordingly.
6593 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6595 PR tree-optimization/84114
6596 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6597 Avoid reassociation of FLOAT_MODE addition.
6599 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6601 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6602 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6603 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6604 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6605 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6606 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6607 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6609 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6610 __builtin_ia32_wbinvd): New builtins.
6611 (SPECIAL_ARGS2): New.
6612 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6613 (SPECIAL_ARGS2): New.
6614 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6615 (ix86_valid_target_attribute_inner_p): Ditto.
6616 (ix86_init_mmx_sse_builtins): Add special_args2.
6617 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6618 TARGET_WBNOINVD_P): New.
6619 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6620 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6621 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6622 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6623 * config/i386/pconfigintrin.h: New file.
6624 * config/i386/wbnoinvdintrin.h: Ditto.
6625 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6626 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6628 2018-03-05 Richard Biener <rguenther@suse.de>
6630 PR tree-optimization/84670
6631 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6633 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6634 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6635 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6636 doesn't grow if all successors (recursively) were visited at least
6639 2018-03-05 Richard Biener <rguenther@suse.de>
6641 PR tree-optimization/84650
6642 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6643 if executed in the loop pipeline.
6645 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6647 * doc/configfiles.texi (Configuration Files): Move info about
6648 conditionalizing $target-protos.h to...
6649 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6650 differs from $target-protos.h.
6652 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6653 Chung-Ju Wu <jasonwucj@gmail.com>
6655 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6656 * config/nds32/nds32-multiple.md (setmemsi): Define.
6657 * config/nds32/nds32-memory-manipulation.c
6658 (nds32_gen_dup_4_byte_to_word_value): New.
6659 (emit_setmem_word_loop): New.
6660 (emit_setmem_byte_loop): New.
6661 (nds32_expand_setmem_loop): New.
6662 (nds32_expand_setmem_loop_v3m): New.
6663 (nds32_expand_setmem_unroll): New.
6664 (nds32_expand_setmem): New.
6666 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6667 Chung-Ju Wu <jasonwucj@gmail.com>
6669 * config/nds32/nds32-memory-manipulation.c
6670 (nds32_emit_load_store): New.
6671 (nds32_emit_post_inc_load_store): New.
6672 (nds32_emit_mem_move): New.
6673 (nds32_emit_mem_move_block): New.
6674 (nds32_expand_movmemsi_loop_unknown_size): New.
6675 (nds32_expand_movmemsi_loop_known_size): New.
6676 (nds32_expand_movmemsi_loop): New.
6677 (nds32_expand_movmemsi_unroll): New.
6678 (nds32_expand_movmemqi): Rename ...
6679 (nds32_expand_movmemsi): ... to this.
6680 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6681 (movmemsi): ... to this.
6682 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6683 (nds32_expand_movmemsi): ... to this.
6685 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6686 Monk Chiang <sh.chiang04@gmail.com>
6687 Chung-Ju Wu <jasonwucj@gmail.com>
6689 * config/nds32/nds32-protos.h
6690 (nds32_expand_load_multiple): New arguments.
6691 (nds32_expand_store_multiple): Ditto.
6692 (nds32_valid_multiple_load_store): Rename ...
6693 (nds32_valid_multiple_load_store_p): ... to this.
6694 * config/nds32/nds32-memory-manipulation.c
6695 (nds32_expand_load_multiple): Refine implementation.
6696 (nds32_expand_store_multiple): Ditto.
6697 * config/nds32/nds32-multiple.md
6698 (load_multiple): Update nds32_expand_load_multiple interface.
6699 (store_multiple): Update nds32_expand_store_multiple interface.
6700 * config/nds32/nds32-predicates.c
6701 (nds32_valid_multiple_load_store): Rename ...
6702 (nds32_valid_multiple_load_store_p): ... to this and refine
6704 * config/nds32/predicates.md
6705 (nds32_load_multiple_and_update_address_operation): New predicate.
6706 (nds32_store_multiple_and_update_address_operation): New predicate.
6708 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6709 Chung-Ju Wu <jasonwucj@gmail.com>
6711 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6712 (combo): New attribute.
6713 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6715 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6717 * config/nds32/nds32.opt: Change -mcmodel= default value.
6719 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6720 Monk Chiang <sh.chiang04@gmail.com>
6721 Chung-Ju Wu <jasonwucj@gmail.com>
6723 * config/nds32/constants.md (unspec_element): New enum.
6724 * config/nds32/constraints.md (Umw): New constraint.
6725 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6726 * config/nds32/nds32-intrinsic.md: Likewise.
6727 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6728 (nds32_valid_smw_lwm_base_p): New.
6729 (nds32_output_smw_single_word): New.
6730 (nds32_output_lmw_single_word): New.
6731 (nds32_expand_unaligned_load): New.
6732 (nds32_expand_unaligned_store): New.
6733 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6734 (nds32_output_smw_single_word): Declare.
6735 (nds32_output_lmw_single_word): Declare.
6736 (nds32_expand_unaligned_load): Declare.
6737 (nds32_expand_unaligned_store): Declare.
6738 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6739 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6740 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6741 NDS32_BUILTIN_UASTORE_DW.
6742 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6745 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6746 Kito Cheng <kito.cheng@gmail.com>
6747 Chung-Ju Wu <jasonwucj@gmail.com>
6749 * config/nds32/nds32-intrinsic.c
6750 (nds32_expand_builtin_null_ftype_reg): Delete.
6751 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6752 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6753 (nds32_read_argument): New.
6754 (nds32_legitimize_target): Ditto.
6755 (nds32_legitimize_argument): Ditto.
6756 (nds32_check_constant_argument): Ditto.
6757 (nds32_expand_unop_builtin): Ditto.
6758 (nds32_expand_unopimm_builtin): Ditto.
6759 (nds32_expand_binop_builtin): Ditto.
6760 (nds32_builtin_decl_impl): Ditto.
6761 (builtin_description): Ditto.
6762 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6763 (nds32_init_builtins_impl): Ditto.
6764 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6765 (nds32_builtin_decl): New.
6766 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6767 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6769 2018-03-02 Jeff Law <law@redhat.com>
6771 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6772 (redundant_insn, fill_simple_delay_slots): Likewise.
6773 (fill_slots_from_thread): Likewise.
6774 * resource.c (mark_referenced_resources): Likewise.
6775 (mark_set_resources, find_dead_or_set_registers): Likewise.
6777 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6779 * substring-locations.h (format_warning_va): Formatting fix for
6781 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6783 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6784 * substring-locations.c: Include intl.h.
6785 (format_warning_va): Turned into small wrapper around
6786 format_warning_n_va, renamed to ...
6787 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6788 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6790 (format_warning_at_substring_n): New function.
6791 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6792 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6793 format_warning_at_substring with just a shorter name instead of
6794 const function pointer.
6795 (fmtwarn_n): New function.
6796 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6797 appropriate, get rid of all the fmtstr temporaries, move conditionals
6798 with G_() wrapped string literals directly into fmtwarn arguments,
6799 cast dir.len to (int), formatting fixes.
6801 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6803 * doc/invoke.texi: Remove "Cilk Plus" references.
6805 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6806 Richard Biener <rguenther@suse.de>
6809 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6810 for error or warning attributes if CALL_FROM_THUNK_P is set.
6813 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6816 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6817 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6820 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6821 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6823 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6824 instead of -1U in last predictors element's probability member.
6826 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6829 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6830 arguments if they are comparable.
6832 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6834 PR tree-optimization/84634
6835 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6836 masks and masked_loop_p with a single loop_masks, making sure it's
6837 null for bb vectorization.
6839 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6841 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6842 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6843 loop->force_vectorize to check whether there is no alias.
6845 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6848 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6850 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6852 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6853 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6854 instead of a loop around prev_real_insn.
6855 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6859 * config/i386/i386.c (ix86_print_operand): Use conditional
6860 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6863 2018-03-02 Richard Biener <rguenther@suse.de>
6865 PR tree-optimization/84427
6866 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6867 (bitmap_set_subtract_values): Rewrite to handle multiple
6870 (prune_clobbered_mems): Likewise.
6871 (phi_translate): Take edge instead of pred/phiblock.
6872 (phi_translate_1): Likewise.
6873 (phi_translate_set): Likewise. Insert all translated
6874 exprs for a value into the set, keeping possibly multiple
6875 expressions per value.
6876 (compute_antic_aux): Adjust for phi_translate changes.
6877 When intersecting union the expressions and prune those
6878 not in the final value set, keeping possibly multiple
6879 expressions per value. Do not use value-insertion
6880 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6881 all expressions. Add verification that the value-sets
6882 only shrink during iteration.
6883 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6884 (do_pre_regular_insertion): Likewise.
6885 (do_pre_partial_partial_insertion): Likewise.
6887 2018-03-02 Richard Biener <rguenther@suse.de>
6890 * config/darwin.c (saved_debug_info_level): New static global.
6891 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6892 (darwin_asm_lto_end): Restore debug info generation settings.
6894 2018-03-01 Martin Liska <mliska@suse.cz>
6897 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6900 2018-03-01 Richard Biener <rguenther@suse.de>
6903 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6904 type annotation with LTO when debug was disabled at compile-time.
6906 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6908 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6910 (mips_final_postscan_insn): Likewise.
6912 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6914 PR rtl-optimization/84528
6915 * alias.c (init_alias_target): Add commentary.
6916 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6917 a unique base value if the frame pointer is not eliminated
6918 to the stack pointer.
6920 2018-03-01 Tom de Vries <tom@codesourcery.com>
6922 PR rtl-optimization/83327
6923 * lra-int.h (hard_regs_spilled_into): Declare.
6924 * lra.c (hard_regs_spilled_into): Define.
6925 (init_reg_info): Init hard_regs_spilled_into.
6926 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6927 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6928 (process_bb_lives): Handle hard_regs_spilled_into.
6929 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6930 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6932 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6934 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6935 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6936 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6937 * config/rs6000/aix72.h: New file.
6939 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6941 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6942 instead of warning_at with conditional singular and plural messages
6946 * stor-layout.c (update_alignment_for_field): For
6947 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6948 && !DECL_PACKED (field), do the alignment update, just use
6949 only desired_align instead of MAX (type_align, desired_align)
6951 (place_field): Don't do known_align < desired_align handling
6952 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6953 is non-NULL, instead do it after rli->prev_field handling and
6954 only if not within a bitfield word. For DECL_PACKED (field)
6955 use type_align of BITS_PER_UNIT.
6957 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6959 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6960 superfluous parentheses and trailing spaces.
6962 2018-02-28 Richard Biener <rguenther@suse.de>
6964 PR tree-optimization/84584
6965 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6966 SCoPs with fake exit edge.
6968 2018-02-28 Martin Liska <mliska@suse.cz>
6971 * timevar.c (timer::print): Fix format to properly print 100%
6974 2018-02-28 Richard Biener <rguenther@suse.de>
6977 * genmatch.c (capture_info::walk_match): Do not mark
6978 captured expressions without operands as expr_p given
6979 they act more like predicates and should be subject to
6980 "lost tail" side-effect preserving.
6982 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6984 PR rtl-optimization/81611
6985 * auto-inc-dec.c (attempt_change): Move dead note from
6986 mem_insn if it's the next use of regno
6987 (find_address): Take address use of reg holding
6988 non-incremented value. Add parm to limit search to the named
6990 (merge_in_block): Attempt to use a mem insn that is the next
6991 use of the original regno.
6993 2018-02-27 Martin Sebor <msebor@redhat.com>
6996 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6997 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6999 2018-02-27 Martin Sebor <msebor@redhat.com>
7001 PR translation/84207
7002 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7003 n argument to unsigned HOST_WIDE_INT.
7004 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7005 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7006 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7007 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7009 2018-02-27 Richard Biener <rguenther@suse.de>
7011 PR tree-optimization/84512
7012 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7013 Do not use the estimate returned from record_stmt_cost for
7014 the scalar iteration cost but sum properly using add_stmt_cost.
7016 2018-02-27 Richard Biener <rguenther@suse.de>
7018 PR tree-optimization/84466
7019 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7020 Adjust last change to less strictly validate use operands.
7022 2018-02-27 Martin Liska <mliska@suse.cz>
7024 PR gcov-profile/84548
7025 * gcov.c (process_file): Allow partial overlap and consider it
7026 also as group functions.
7027 (output_lines): Properly calculate range of lines for a group.
7029 2018-02-27 Martin Liska <mliska@suse.cz>
7031 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7032 'ggc' suffixes. Change first column width.
7033 (timer::print): Fix formatting of the column.
7035 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7037 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7038 preserve inline entry blocks for the sake of debug inline
7039 entry point markers alone.
7040 (remove_unused_locals): Suggest in comments a better place to
7041 force the preservation of inline entry blocks that are
7042 otherwise unused, but do not preserve them.
7044 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7046 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7048 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7051 * config/i386/constraints.md (Bs): Replace
7052 ix86_indirect_branch_register with
7053 TARGET_INDIRECT_BRANCH_REGISTER.
7055 * config/i386/i386.md (indirect_jump): Likewise.
7056 (tablejump): Likewise.
7057 (*sibcall_memory): Likewise.
7058 (*sibcall_value_memory): Likewise.
7059 Peepholes of indirect call and jump via memory: Likewise.
7060 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7061 (*sibcall_value_GOT_32): Likewise.
7062 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7063 (GOT_memory_operand): Likewise.
7064 (call_insn_operand): Likewise.
7065 (sibcall_insn_operand): Likewise.
7066 (GOT32_symbol_operand): Likewise.
7067 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7069 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7071 PR rtl-optimization/83496
7072 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7073 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7074 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7075 redundant insn, if any.
7076 (relax_delay_slots): Likewise.
7077 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7079 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7081 PR tree-optimization/83965
7082 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7083 that grouped statements are part of a reduction chain. Return
7084 true if the statement is not marked as a reduction itself but
7086 (vect_recog_dot_prod_pattern): Don't check whether the statement
7087 is part of a group here.
7088 (vect_recog_sad_pattern): Likewise.
7089 (vect_recog_widen_sum_pattern): Likewise.
7091 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7094 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7096 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7099 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7101 (ix86_output_indirect_function_return): New prototype.
7102 (ix86_split_simple_return_pop_internal): Likewise.
7103 * config/i386/i386.c (indirect_return_via_cx): New.
7104 (indirect_return_via_cx_bnd): Likewise.
7105 (indirect_thunk_name): Handle return va CX_REG.
7106 (output_indirect_thunk_function): Create alias for
7107 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7108 (ix86_output_indirect_jmp): Remove the bool argument.
7109 (ix86_output_indirect_function_return): New function.
7110 (ix86_split_simple_return_pop_internal): Likewise.
7111 * config/i386/i386.md (*indirect_jump): Don't pass false
7112 to ix86_output_indirect_jmp.
7113 (*tablejump_1): Likewise.
7114 (simple_return_pop_internal): Change it to define_insn_and_split.
7115 Call ix86_split_simple_return_pop_internal to split it for
7117 (simple_return_indirect_internal): Call
7118 ix86_output_indirect_function_return instead of
7119 ix86_output_indirect_jmp.
7121 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7124 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7125 memset and value initialization afterwards.
7127 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7129 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7131 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7134 * common/config/aarch64/aarch64-common.c
7135 (aarch_option_optimization_table[]): Switch
7136 off fomit-frame-pointer
7138 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7139 Chung-Ju Wu <jasonwucj@gmail.com>
7141 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7143 (store_multiple): Ditto.
7145 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7147 * config.gcc: Add --with-cpu support for nds32 target.
7148 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7149 * config/nds32/nds32.opt: Add -mcpu= option.
7151 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7153 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7154 isel=yes): Warn for these deprecated options.
7156 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7158 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7159 ISA_2_5_MASKS_EMBEDDED.
7161 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7163 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7164 p->max as pointers rather than using iterative_hash_expr.
7166 2018-02-23 Carl Love <cel@us.ibm.com>
7168 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7169 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7171 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7172 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7173 P8V_BUILTIN_VEC_VUNSIGNED2.
7175 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7178 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7179 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7181 * lra-constraints.c (curr_insn_transform): Set up
7182 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7184 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7185 * lra-eliminations.c (spill_pseudos): Ditto.
7186 (process_insn_for_elimination): Ditto.
7187 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7188 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7189 LRA_NON_CLOBBERED_ALT.
7191 2018-02-22 Martin Sebor <msebor@redhat.com>
7193 PR tree-optimization/84480
7194 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7195 to maybe_diag_stxncpy_trunc. Call it.
7196 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7197 from gimple_fold_builtin_strcpy. Print inlining stack.
7198 (handle_builtin_stxncpy): Print inlining stack.
7199 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7201 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7204 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7205 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7206 and -fcheck-pointer-bounds are used together.
7207 (indirect_thunk_prefix): New enum.
7208 (indirect_thunk_need_prefix): New function.
7209 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7210 "_nt" instead of "_bnd" for NOTRACK prefix.
7211 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7212 (output_indirect_thunk_function): Likewise.
7214 (ix86_code_end): Update output_indirect_thunk_function calls.
7215 (ix86_output_indirect_branch_via_reg): Replace
7216 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7217 (ix86_output_indirect_branch_via_push): Likewise.
7218 (ix86_output_function_return): Likewise.
7219 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7220 incompatible with -fcf-protection=branch and
7221 -fcheck-pointer-bounds.
7223 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7226 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7227 Change gcc_assert call to output_operand_lossage.
7229 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7231 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7233 2018-02-22 DJ Delorie <dj@redhat.com>
7234 Sebastian Perta <sebastian.perta@renesas.com>
7235 Oleg Endo <olegendo@gcc.gnu.org>
7237 * config/rx/rx.c (rx_rtx_costs): New function.
7238 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7240 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7242 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7244 2018-02-22 Martin Liska <mliska@suse.cz>
7247 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7248 Add "native" as a possible value.
7250 2018-02-22 Martin Liska <mliska@suse.cz>
7253 * config/i386/i386.c (ix86_option_override_internal):
7254 Add "native" as a possible value for -march and -mtune.
7256 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7259 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7260 to all type variants.
7262 PR tree-optimization/84503
7263 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7264 width as info->bitpos + info->bitsize - start.
7265 (merged_store_group::merge_overlapping): Simplify width computation.
7266 (check_no_overlap): New function.
7267 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7268 start + width and last_order of the group, fail if check_no_overlap
7270 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7271 to group if check_no_overlap fails.
7273 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7275 * config/rs6000/altivec.md: Delete contraint arguments to
7276 define_expand, define_split, and define_peephole2, and in
7277 define_insn_and_split if always unused.
7278 * config/rs6000/darwin.md: Ditto.
7279 * config/rs6000/dfp.md: Ditto.
7280 * config/rs6000/rs6000.md: Ditto.
7281 * config/rs6000/sync.md: Ditto.
7282 * config/rs6000/vector.md: Ditto.
7283 * config/rs6000/vsx.md: Ditto.
7285 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7287 * config/rs6000/altivec.md: Write output control strings as braced
7288 blocks instead of double-quoted strings.
7289 * config/rs6000/darwin.md: Ditto.
7290 * config/rs6000/rs6000.md: Ditto.
7291 * config/rs6000/vector.md: Ditto.
7292 * config/rs6000/vsx.md: Ditto.
7294 2018-02-21 Jason Merrill <jason@redhat.com>
7296 PR c++/84314 - ICE with templates and fastcall attribute.
7297 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7299 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7301 * ipa-cp.c (determine_versionability): Fix comment typos.
7303 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7306 * ipa-cp.c (determine_versionability): Do not version functions caling
7309 2018-02-21 Martin Liska <mliska@suse.cz>
7312 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7313 Add "native" as a possible value.
7314 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7315 the macro when native cpu detection is available.
7317 2018-02-21 Martin Liska <mliska@suse.cz>
7320 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7321 Add "native" as a possible value.
7322 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7323 when native cpu detection is available.
7325 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7326 Martin Sebor <msebor@redhat.com>
7328 PR tree-optimization/84478
7329 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7331 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7332 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7333 support which is conservatively correct, for 2 only stay conservative
7334 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7335 argument to the 2 argument get_range_strlen, adjust 6 arg
7336 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7338 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7339 (gimple_fold_builtin_strlen): Pass true as last argument to
7342 2018-02-20 Martin Sebor <msebor@redhat.com>
7345 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7346 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7347 (builtin_memref::builtin_memref): Factor out parts into
7348 set_base_and_offset and call it.
7350 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7353 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7354 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7355 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7356 search at the associated MODE_INT.
7358 2018-02-20 Jeff Law <law@redhat.com>
7361 PR tree-optimization/81592
7363 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7364 for range data rather than using global data.
7365 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7366 range data rather than using global data.
7367 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7368 pass it to children as needed.
7369 (struct directive::fmtresult): Similarly.
7370 (struct directive::set_width): Similarly.
7371 (struct directive::set_precision): Similarly.
7372 (format_integer, format_directive, parse_directive): Similarly.
7373 (format_none): Accept unnamed vr_values parameter.
7374 (format_percent, format_floating, format_character): Similarly.
7375 (format_string, format_plain): Similarly.
7376 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7377 the EVRP range analyzer for range data rather than using global data.
7378 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7379 gimple-ssa-evrp-analyze.h
7380 (class sprintf_dom_walker): Add after_dom_children member function.
7381 Add evrp_range_analyzer member.
7382 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7383 range analyzer as needed.
7384 (sprintf_dom_walker::after_dom_children): New member function.
7385 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7387 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7388 (evrp_range_analyzer::pop_to_marker): Likewise.
7390 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7392 PR tree-optimization/84419
7393 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7394 with the required type if its current type is compatible but
7397 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7400 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7401 after vectorization.
7403 2018-02-20 Martin Liska <mliska@suse.cz>
7406 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7407 possible values if we don't have a hint.
7409 2018-02-20 Martin Liska <mliska@suse.cz>
7413 * final.c (shorten_branches): Build align_tab array with one
7415 * opts.c (finish_options): Add alignment option limit check.
7416 (MAX_CODE_ALIGN): Likewise.
7417 (MAX_CODE_ALIGN_VALUE): Likewise.
7418 * doc/invoke.texi: Document maximum allowed option value for
7419 all -falign-* options.
7421 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7424 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7425 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7426 * var-tracking.c (emit_note_insn_var_location): Remove all references
7427 to NOTE_INSN_CALL_ARG_LOCATION.
7428 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7429 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7430 Use copy_rtx_if_shared.
7431 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7432 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7433 (dwarf2out_var_location): Remove handling of
7434 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7436 * final.c (final_scan_insn): Remove all references to
7437 NOTE_INSN_CALL_ARG_LOCATION.
7438 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7439 before dumping final insns.
7440 * except.c (emit_note_eh_region_end): Remove all references to
7441 NOTE_INSN_CALL_ARG_LOCATION.
7442 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7443 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7444 * config/arc/arc.c (hwloop_optimize): Likewise.
7445 * config/arm/arm.c (create_fix_barrier): Likewise.
7446 * config/s390/s390.c (s390_chunkify_start): Likewise.
7447 * config/sh/sh.c (find_barrier): Likewise.
7448 * config/i386/i386.c (rest_of_insert_endbranch,
7449 ix86_seh_fixup_eh_fallthru): Likewise.
7450 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7451 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7452 * config/frv/frv.c (frv_function_prologue): Likewise.
7453 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7455 (note_outside_basic_block_p): Remove all references to
7456 NOTE_INSN_CALL_ARG_LOCATION.
7457 * gengtype.c (adjust_field_rtx_def): Likewise.
7458 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7460 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7461 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7464 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7467 PR tree-optimization/84452
7468 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7469 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7472 2018-02-19 Martin Liska <mliska@suse.cz>
7475 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7477 2018-02-19 Martin Liska <mliska@suse.cz>
7478 Richard Sandiford <richard.sandiford@linaro.org>
7480 PR tree-optimization/82491
7481 * gimple-fold.c (get_base_constructor): Make earlier bail out
7484 2018-02-19 Carl Love <cel@us.ibm.com>
7486 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7487 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7489 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7490 P8V_BUILTIN_VEC_NEG.
7492 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7494 * config/rl78/rl78.md (movdf): New define expand.
7496 2018-02-19 Martin Liska <mliska@suse.cz>
7499 * doc/invoke.texi: Fix typo.
7500 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7502 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7504 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7505 handle rs6000_single_float and rs6000_double_float specially for
7508 2018-02-16 Jeff Law <law@redhat.com>
7510 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7513 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7516 * ipa-inline.c (inline_small_functions): Fix a typo.
7518 2018-02-16 Nathan Sidwell <nathan@acm.org>
7520 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7522 2018-02-16 Carl Love <cel@us.ibm.com>
7524 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7525 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7526 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7527 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7528 expansion to P8V_BUILTIN_VEC_FLOAT2.
7530 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7532 PR rtl-optimization/70023
7533 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7534 src_regno into account.
7536 2018-02-16 Carl Love <cel@us.ibm.com>
7538 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7539 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7540 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7541 * config/rs6000/rs6000.c: Remove case statements for
7542 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7543 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7544 and P9V_BUILTIN_VEC_VINSERT4B.
7545 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7546 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7547 * config/rs6000/vsx.md:
7548 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7551 2018-02-16 Carl Love <cel@us.ibm.com>
7553 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7555 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7557 * config/rs6000/rs6000-c.c: Add the definitions for
7558 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7559 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7560 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7561 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7562 definition for insert4b and define insn *insert3b_internal.
7563 * doc/extend.texi: Add documentation for vec_extract4b.
7565 2018-02-16 Nathan Sidwell <nathan@acm.org>
7567 * doc/extend.texi (Backwards Compatibility): Mention friend
7568 injection. Note for-scope is deprecated.
7569 * doc/invoke.texi (-ffriend-injection): Deprecate.
7571 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7573 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7574 that moved to I2, also allow destinations that are a paradoxical
7575 subreg (instead of a normal reg).
7577 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7580 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7583 2018-02-16 Richard Biener <rguenther@suse.de>
7585 PR tree-optimization/84037
7586 PR tree-optimization/84016
7588 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7589 Adjust vec_construct for the fact we need additional higher latency
7590 128bit inserts for AVX256 and AVX512 vector builds.
7591 (ix86_add_stmt_cost): Scale vector construction cost for
7594 2018-02-16 Richard Biener <rguenther@suse.de>
7596 PR tree-optimization/84417
7597 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7598 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7599 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7601 2018-02-16 Martin Liska <mliska@suse.cz>
7604 * internal-fn.def (ASAN_CHECK): Set proper flags.
7605 (ASAN_MARK): Likewise.
7607 2018-02-16 Julia Koval <julia.koval@intel.com>
7609 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7610 from PTA_CANNONLAKE.
7612 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7615 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7616 Use ++iter rather than iter++ for std::list iterators.
7617 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7618 defer deleting them until all nodes in the forest are processed. Do
7619 free even leaf nodes. Change to_process into auto_vec.
7622 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7623 * vec.h (vec_default_construct): Use memset instead of placement new
7624 if BROKEN_VALUE_INITIALIZATION is defined.
7625 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7626 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7629 PR rtl-optimization/83723
7630 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7631 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7632 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7634 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7636 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7638 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7640 PR rtl-optimization/81443
7641 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7642 from inner REGs to paradoxical SUBREGs.
7644 2018-02-16 Richard Biener <rguenther@suse.de>
7646 PR tree-optimization/84399
7647 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7648 For operands we can analyze at their definition make sure we can
7649 analyze them at each use as well.
7651 2018-02-16 Richard Biener <rguenther@suse.de>
7653 PR tree-optimization/84190
7654 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7655 volatile accesses if the decl isn't volatile.
7657 2018-02-15 Jason Merrill <jason@redhat.com>
7659 PR c++/84314 - ICE with templates and fastcall attribute.
7660 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7661 TYPE_CANONICAL on an existing type.
7663 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7665 PR tree-optimization/84383
7666 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7667 dstoff nor call operand_equal_p if dstbase is NULL.
7669 PR tree-optimization/84334
7670 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7671 also a CONSTANT_CLASS_P, punt.
7673 2018-02-14 Jim Wilson <jimw@sifive.com>
7675 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7676 first SMALL_OPERAND check. New local min_second_step. Move assert
7677 to where locals are set. Add TARGET_RVC support.
7678 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7680 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7682 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7684 2018-02-14 Martin Sebor <msebor@redhat.com>
7686 PR tree-optimization/83698
7687 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7688 arrays constrain the offset range to their bounds.
7689 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7690 (builtin_access::overlap): Avoid setting the size of overlap if it's
7692 (maybe_diag_overlap): Also consider arrays when deciding what values
7693 of offsets to include in diagnostics.
7695 2018-02-14 Martin Sebor <msebor@redhat.com>
7698 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7699 that correspond to the kind of a declaration.
7701 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7704 * config/pa/pa.md: Load address of PIC label using the linkage table
7705 if the label is nonlocal.
7707 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7709 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7710 warning message if user requests -maltivec=be.
7711 * doc/invoke.texi: Document deprecation of -maltivec=be.
7713 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7716 * config/rs6000/rs6000-c.c: Update definitions for
7717 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7718 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7720 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7723 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7724 add _get_ssp intrinsics. Remove argument from
7725 __builtin_ia32_rdssp[d|q].
7726 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7727 * config/i386/i386-builtin.def: Remove argument from
7728 __builtin_ia32_rdssp[d|q].
7729 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7730 ix86_expand_special_args_builtin for _rdssp[d|q].
7731 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7732 Clear register before usage.
7733 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7734 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7736 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7738 PR tree-optimization/84357
7739 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7740 operand 1 of an ARRAY_REF too.
7742 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7745 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7746 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7748 (set_of_reg): New struct.
7749 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7750 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7751 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7753 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7754 Split into bitclr, bitset, bitinvert patterns if appropriate.
7755 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7756 use rx_fuse_in_memory_bitop.
7757 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7758 to named insn, correct maximum insn length.
7760 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7763 * machmode.def: Define a complex mode for PARTIAL_INT.
7764 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7766 * doc/rtl.texi: Document CSPImode.
7767 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7769 (msp430_hard_regno_nregs_with_padding): Likewise.
7771 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7774 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7776 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7778 PR rtl-optimization/84169
7779 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7780 we generated a parallel as new i3 and we split that to new i2 and i3
7781 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7782 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7783 those to i2, not i1. Partially rewrite this scan code.
7785 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7788 * stor-layout.c (place_field): For variable length fields, adjust
7789 offset_align afterwards not just based on the field's alignment,
7790 but also on the size.
7793 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7794 of exps and logs in the use_exp2 case.
7796 2018-02-13 Jeff Law <law@redhat.com>
7798 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7801 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7804 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7808 * common.opt (gas-loc-support, gas-locview-support): New.
7809 (ginline-points, ginternal-reset-location-views): New.
7810 * doc/invoke.texi: Document them. Use @itemx where intended.
7811 (gvariable-location-views): Adjust.
7812 * target.def (reset_location_view): New.
7813 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7814 (TARGET_RESET_LOCATION_VIEW): New.
7815 * doc/tm.texi: Rebuilt.
7816 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7817 (dwarf2out_default_as_locview_support): New.
7818 (output_asm_line_debug_info): Use option variables.
7819 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7820 (output_loc_list): Likewise.
7821 (add_high_low_attributes): Check option variables.
7822 Don't output entry view attribute in strict mode.
7823 (gen_inlined_subroutine_die): Check option variables.
7824 (dwarf2out_inline_entry): Likewise.
7825 (init_sections_and_labels): Likewise.
7826 (dwarf2out_early_finish): Likewise.
7827 (maybe_reset_location_view): New, from...
7828 (dwarf2out_var_location): ... here. Call it.
7829 * debug.h (dwarf2out_default_as_loc_support): Declare.
7830 (dwarf2out_default_as_locview_support): Declare.
7831 * hooks.c (hook_int_rtx_insn_0): New.
7832 * hooks.h (hook_int_rtx_insn_0): Declare.
7833 * toplev.c (process_options): Take -gas-loc-support and
7834 -gas-locview-support from dwarf2out. Enable
7835 -gvariable-location-views by default only with locview
7836 assembler support. Enable -ginternal-reset-location-views by
7837 default only if the target defines the corresponding hook.
7838 Enable -ginline-points by default if location views are
7839 enabled; force it disabled if statement frontiers are
7841 * tree-inline.c (expand_call_inline): Check option variables.
7842 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7844 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7846 PR tree-optimization/84321
7847 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7848 handling. Also check whether the anti-range contains any values
7849 that satisfy the mask; switch to a VR_RANGE if not.
7851 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7854 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7856 2018-02-13 Martin Jambor <mjambor@suse.cz>
7859 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7860 of call statements, also set location of a load to a temporary.
7862 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7864 * config/rl78/rl78.c (add_vector_labels): New function.
7865 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7866 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7867 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7868 which checks that no arguments are passed.
7869 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7870 * doc/extend.texi: Documentation for the new attribute.
7872 2018-02-13 Andreas Schwab <schwab@suse.de>
7874 * config/riscv/linux.h (CPP_SPEC): Define.
7876 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7879 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7880 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7881 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7882 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7883 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7884 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7885 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7886 temporarily for AES and PCLMUL builtins.
7888 PR tree-optimization/84339
7889 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7890 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7894 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7895 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7896 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7898 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7900 * omp-simd-clone.h: New file.
7901 * omp-simd-clone.c: Include omp-simd-clone.h.
7902 (expand_simd_clones): No longer static.
7903 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7904 cgraph.h and omp-simd-clone.h.
7905 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7906 (vect_recog_widen_shift_pattern): Formatting fix.
7907 (vect_pattern_recog_1): Don't check optab for calls.
7910 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7911 operands[2] into a REG before using gen_lowpart on it.
7913 2018-02-12 Jeff Law <law@redhat.com>
7916 * config/sh/sh.c (find_barrier): Consider a sibling call
7919 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7920 successfully back substituting a reg.
7922 2018-02-12 Richard Biener <rguenther@suse.de>
7924 PR tree-optimization/84037
7925 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7926 parameter, move visited init to caller.
7927 (vect_slp_analyze_operations): Separate cost from validity
7928 check, initialize visited once for all instances.
7929 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7931 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7933 (vect_model_promotion_demotion_cost): Likewise.
7934 (vectorizable_bswap): Guard cost modeling with !slp_node
7935 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7937 (vectorizable_call): Likewise.
7938 (vectorizable_conversion): Likewise.
7939 (vectorizable_assignment): Likewise.
7940 (vectorizable_shift): Likewise.
7941 (vectorizable_operation): Likewise.
7942 (vectorizable_store): Likewise.
7943 (vectorizable_load): Likewise.
7944 (vectorizable_condition): Likewise.
7945 (vectorizable_comparison): Likewise.
7947 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7950 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7951 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7952 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7954 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7957 * params.def (inline-min-speedup): Increase from 8 to 15.
7958 (max-inline-insns-auto): Decrease from 40 to 30.
7959 * ipa-split.c (consider_split): Add some buffer for function to
7960 be considered inlining candidate.
7961 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7964 2018-02-12 Richard Biener <rguenther@suse.de>
7966 PR tree-optimization/84037
7967 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7968 matched stmts if we cannot swap the non-matched ones.
7970 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7972 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7973 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7974 _mm_maskz_scalef_round_ss): New intrinsics.
7975 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7976 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7977 __builtin_ia32_scalefss_round): Remove.
7978 (__builtin_ia32_scalefsd_mask_round,
7979 __builtin_ia32_scalefss_mask_round): New intrinsics.
7980 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7981 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7982 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7983 "<round_constraint>")): Changed to ...
7984 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7985 "<round_scalar_constraint>")): ... this.
7986 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7987 %0, %1, %2<round_op3>}"): Changed to ...
7988 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7989 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7990 %2<round_scalar_mask_op3>}"): ... this.
7991 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7993 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7995 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7996 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7997 (_mm_maskz_sqrt_round_ss): New intrinsics.
7998 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7999 (__builtin_ia32_sqrtsd_mask_round)
8000 (__builtin_ia32_sqrtss_mask_round): New builtins.
8001 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8002 (__builtin_ia32_sqrtss_round): Remove.
8003 (__builtin_ia32_sqrtsd_mask_round)
8004 (__builtin_ia32_sqrtss_mask_round): New builtins.
8005 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8006 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8007 ((match_operand:VF_128 1 "vector_operand"
8008 "xBm,<round_constraint>")): Changed to ...
8009 ((match_operand:VF_128 1 "vector_operand"
8010 "xBm,<round_scalar_constraint>")): ... this.
8011 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8012 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8013 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8014 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8015 %<iptr>1<round_scalar_mask_op3>}): ... this.
8016 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8017 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8019 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8022 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8023 Cast vec_cmpeq result to correct type.
8024 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8025 Cast vec_cmpgt result to correct type.
8027 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8029 * final.c (final_scan_insn_1): Renamed from...
8030 (final_scan_insn): ... this. New wrapper, to recover
8031 seen from the outermost call in recursive ones.
8032 * config/sparc/sparc.c (output_return): Drop seen from call.
8033 (output_sibcall): Likewise.
8034 * config/visium/visium.c (output_branch): Likewise.
8036 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8038 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8041 2018-02-10 Alan Modra <amodra@gmail.com>
8044 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8045 Specify LR as an input.
8047 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8050 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8051 remove_member_access_dummy_vars): New functions.
8052 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8053 lower_omp_1, execute_lower_omp): Use them.
8055 PR rtl-optimization/84308
8056 * shrink-wrap.c (spread_components): Release todo vector.
8058 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8060 PR rtl-optimization/57193
8061 * ira-color.c (struct allocno_color_data): Add member
8062 conflict_allocno_hard_prefs.
8063 (update_conflict_allocno_hard_prefs): New.
8064 (bucket_allocno_compare_func): Add a preference based on
8065 conflict_allocno_hard_prefs.
8066 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8067 (color_allocnos): Remove a dead code. Initiate
8068 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8070 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8073 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8074 constraint from =wa to wa. Avoid a subreg on the output operand,
8075 instead use a pseudo and subreg it in a move.
8076 (p9_xxbrd_<mode>): Changed to ...
8077 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8078 (p9_xxbrd_v2df): New expander.
8079 (p9_xxbrw_<mode>): Changed to ...
8080 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8081 (p9_xxbrw_v4sf): New expander.
8083 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8085 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8087 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8090 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8091 multiply in 32-bit mode.
8092 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8093 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8096 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8098 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8099 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8100 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8101 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8103 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8106 * dwarf2out.c (is_trivial_indirect_ref): New function.
8107 (dwarf2out_late_global_decl): Do not generate a location
8108 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8109 and that are not defined in the current unit.
8111 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8113 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8114 instead of a libcall for UNORDERED.
8116 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8119 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8120 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8122 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8125 * config/s390/s390.c (s390_set_current_function): Invoke
8126 s390_indirect_branch_settings also if fndecl didn't change.
8128 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8130 * config/rs6000/rs6000.md (blockage): Set length to zero.
8132 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8134 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8136 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8139 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8140 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8144 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8145 PARALLEL incoming that failed vt_get_decl_and_offset check.
8148 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8149 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8151 (get_variable_section): For decls in named .bss* sections pass true as
8152 second argument to bss_initializer_p.
8154 2018-02-09 Marek Polacek <polacek@redhat.com>
8155 Jakub Jelinek <jakub@redhat.com>
8158 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8159 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8160 Sync some changes from cxx_fold_indirect_ref.
8162 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8164 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8166 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8167 (BLOCK_INLINE_ENTRY_LABEL): New.
8168 (dwarf2out_var_location): Disregard inline entry markers.
8169 (inline_entry_data): New struct.
8170 (inline_entry_data_hasher): New hashtable type.
8171 (inline_entry_data_hasher::hash): New.
8172 (inline_entry_data_hasher::equal): New.
8173 (inline_entry_data_table): New variable.
8174 (add_high_low_attributes): Add DW_AT_entry_pc and
8175 DW_AT_GNU_entry_view attributes if a pending entry is found
8176 in inline_entry_data_table. Add old entry_pc attribute only
8177 if debug nonbinding markers are disabled.
8178 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8179 markers are enabled.
8180 (block_within_block_p, dwarf2out_inline_entry): New.
8181 (dwarf2out_finish): Check that no entries remained in
8182 inline_entry_data_table.
8183 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8184 (final_scan_insn, notice_source_line): Likewise.
8185 (rest_of_clean_state): Skip inline entry markers.
8186 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8188 * gimple.c (gimple_build_debug_inline_entry): New.
8189 * gimple.h (enum gimple_debug_subcode): Add
8190 GIMPLE_DEBUG_INLINE_ENTRY.
8191 (gimple_build_debug_inline_entry): Declare.
8192 (gimple_debug_inline_entry_p): New.
8193 (gimple_debug_nonbind_marker_p): Adjust.
8194 * insn-notes.def (INLINE_ENTRY): New.
8195 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8196 inline entry marker notes.
8197 (print_insn): Likewise.
8198 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8199 (INSN_DEBUG_MARKER_KIND): Likewise.
8200 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8201 * tree-inline.c (expand_call_inline): Build and insert
8202 debug_inline_entry stmt.
8203 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8204 inline entry blocks early, if nonbind markers are enabled.
8205 (dump_scope_block): Dump fragment info.
8206 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8207 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8208 (gimple_build_debug_inline_entry): New.
8209 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8210 Enable/disable inline entry points too.
8211 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8212 (DEBUG_INSN): Describe inline entry markers.
8214 * common.opt (gvariable-location-views): New.
8215 (gvariable-location-views=incompat5): New.
8216 * config.in: Rebuilt.
8217 * configure: Rebuilt.
8218 * configure.ac: Test assembler for view support.
8219 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8220 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8221 * dwarf2out.c (var_loc_view): New typedef.
8222 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8223 (dwarf2out_locviews_in_attribute): New.
8224 (dwarf2out_locviews_in_loclist): New.
8225 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8226 (enum dw_line_info_opcode): Add LI_adv_address.
8227 (struct dw_line_info_table): Add view.
8228 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8229 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8230 (zero_view_p): New variable.
8231 (ZERO_VIEW_P): New macro.
8232 (output_asm_line_debug_info): New.
8233 (struct var_loc_node): Add view.
8234 (add_AT_view_list, AT_loc_list): New.
8235 (add_var_loc_to_decl): Add view param. Test it against last.
8236 (new_loc_list): Add view params. Record them.
8237 (AT_loc_list_ptr): Handle loc and view lists.
8238 (view_list_to_loc_list_val_node): New.
8239 (print_dw_val): Handle dw_val_class_view_list.
8240 (size_of_die): Likewise.
8241 (value_format): Likewise.
8242 (loc_list_has_views): New.
8243 (gen_llsym): Set vl_symbol too.
8244 (maybe_gen_llsym, skip_loc_list_entry): New.
8245 (dwarf2out_maybe_output_loclist_view_pair): New.
8246 (output_loc_list): Output view list or entries too.
8247 (output_view_list_offset): New.
8248 (output_die): Handle dw_val_class_view_list.
8249 (output_dwarf_version): New.
8250 (output_compilation_unit_header): Use it.
8251 (output_skeleton_debug_sections): Likewise.
8252 (output_rnglists, output_line_info): Likewise.
8253 (output_pubnames, output_aranges): Update version comments.
8254 (output_one_line_info_table): Output view numbers in asm comments.
8255 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8256 Call maybe_gen_llsym.
8257 (loc_list_from_tree_1): Adjust.
8258 (add_AT_location_description): Create view list attribute if
8259 needed, check it's absent otherwise.
8260 (convert_cfa_to_fb_loc_list): Adjust.
8261 (maybe_emit_file): Call output_asm_line_debug_info for test.
8262 (dwarf2out_var_location): Reset views as needed. Precompute
8263 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8264 attribute. Set view.
8265 (new_line_info_table): Reset next view.
8266 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8267 (dwarf2out_source_line): Likewise. Output view resets and labels to
8268 the assembler, or select appropriate line info opcodes.
8269 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8270 (optimize_string_length): Catch it. Adjust.
8271 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8272 dw_val_class_view_list, and remove it if no longer needed.
8273 (hash_loc_list): Hash view numbers.
8274 (loc_list_hasher::equal): Compare them.
8275 (optimize_location_lists): Check whether a view list symbol is
8276 needed, and whether the locview attribute is present, and
8277 whether they match. Remove the locview attribute if no longer
8279 (index_location_lists): Call skip_loc_list_entry for test.
8280 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8281 Use output_dwarf_version.
8282 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8283 (struct dw_val_node): Add val_view_list.
8284 * final.c (SEEN_NEXT_VIEW): New.
8285 (set_next_view_needed): New.
8286 (clear_next_view_needed): New.
8287 (maybe_output_next_view): New.
8288 (final_start_function): Rename to...
8289 (final_start_function_1): ... this. Take pointer to FIRST,
8290 add SEEN parameter. Emit param bindings in the initial view.
8291 (final_start_function): Reintroduce SEEN-less interface.
8292 (final): Rename to...
8293 (final_1): ... this. Take SEEN parameter. Output final pending
8294 next view at the end.
8295 (final): Reintroduce seen-less interface.
8296 (final_scan_insn): Output pending next view before switching
8297 sections or ending a block. Mark the next view as needed when
8298 outputting variable locations. Notify debug backend of section
8299 changes, and of location view changes.
8300 (rest_of_handle_final): Adjust.
8301 * toplev.c (process_options): Autodetect value for debug variable
8302 location views option. Warn on incompat5 without -gdwarf-5.
8303 * doc/invoke.texi (gvariable-location-views): New.
8304 (gvariable-location-views=incompat5): New.
8305 (gno-variable-location-views): New.
8307 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8309 PR tree-optimization/84136
8310 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8311 that the result of find_edge is non-NULL.
8313 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8316 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8317 storing integer register in SImode. Fix cost of 256 and 512
8318 byte aligned SSE register store.
8320 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8322 * config/i386/i386.c (ix86_multiplication_cost): Fix
8323 multiplication cost for TARGET_AVX512DQ.
8325 2018-02-08 Marek Polacek <polacek@redhat.com>
8327 PR tree-optimization/84238
8328 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8331 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8333 PR tree-optimization/84265
8334 * tree-vect-stmts.c (vectorizable_store): Don't treat
8335 VMAT_CONTIGUOUS accesses as grouped.
8336 (vectorizable_load): Likewise.
8338 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8340 PR tree-optimization/81635
8341 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8342 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8343 (test_round_for_mask): New functions.
8344 (wide_int_cc_tests): Call test_round_for_mask.
8345 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8346 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8347 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8348 range returned by get_range_info.
8350 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8353 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8354 * symtab.c: Include builtins.h
8355 (symtab_node::output_to_lto_symbol_table_p): Move here
8356 from lto-streamer-out.c:output_symbol_p.
8357 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8358 (output_symbol_p): Move all logic to symtab.c
8359 (produce_symtab): Update.
8361 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8363 * config/s390/s390-opts.h (enum indirect_branch): Define.
8364 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8365 (s390_indirect_branch_via_thunk)
8366 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8367 (enum s390_indirect_branch_type): Define.
8368 * config/s390/s390.c (struct s390_frame_layout, struct
8369 machine_function): Remove.
8370 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8371 (indirect_branch_table_label_no, indirect_branch_table_name):
8373 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8374 (enum s390_indirect_branch_option): Define.
8375 (s390_return_addr_from_memory): New function.
8376 (s390_handle_string_attribute): New function.
8377 (s390_attribute_table): Add new attribute handler.
8378 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8379 (s390_indirect_branch_via_thunk): New function.
8380 (s390_indirect_branch_via_inline_thunk): New function.
8381 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8382 sibling call optimization for non z10 compiles.
8383 (s390_emit_call): Force indirect branch target to be a single
8384 register. Add r1 clobber for non-z10 compiles.
8385 (s390_emit_epilogue): Emit return jump via return_use expander.
8386 (s390_reorg): Handle JUMP_INSNs as execute targets.
8387 (s390_option_override_internal): Perform validity checks for the
8388 new command line options.
8389 (s390_indirect_branch_attrvalue): New function.
8390 (s390_indirect_branch_settings): New function.
8391 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8392 (s390_output_indirect_thunk_function): New function.
8393 (s390_code_end): Implement target hook.
8394 (s390_case_values_threshold): Implement target hook.
8395 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8397 * config/s390/s390.h (struct s390_frame_layout)
8398 (struct machine_function): Move here from s390.c.
8399 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8400 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8401 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8402 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8403 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8404 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8405 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8406 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8407 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8408 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8409 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8410 (mnemonic attribute): Add values which aren't recognized
8412 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8413 pattern for branch conversion. Fix mnemonic attribute.
8414 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8415 indirect branch via thunk if requested.
8416 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8417 ("*indirect_jump"): Disable for branch conversion using out of
8419 ("indirect_jump_via_thunk<mode>_z10")
8420 ("indirect_jump_via_thunk<mode>")
8421 ("indirect_jump_via_inlinethunk<mode>_z10")
8422 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8423 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8424 ("casesi_jump_via_inlinethunk<mode>_z10")
8425 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8426 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8427 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8428 ("*indirect2_jump"): Disable for branch conversion.
8429 ("casesi_jump"): Turn into expander and expand patterns for branch
8431 ("return_use"): New expander.
8432 ("*return"): Emit return via thunk and rename it to ...
8433 ("*return<mode>"): ... this one.
8434 * config/s390/s390.opt: Add new options and and enum for the
8437 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8439 * lra-constraints.c (match_reload): Unconditionally use
8440 gen_lowpart_SUBREG, rather than selecting between that
8441 and equivalent gen_rtx_SUBREG code.
8443 2018-02-08 Richard Biener <rguenther@suse.de>
8445 PR tree-optimization/84233
8446 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8447 changed flag instead of boguously re-using phi_inserted.
8449 2018-02-08 Martin Jambor <mjambor@suse.cz>
8451 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8452 static local variables.
8454 2018-02-08 Richard Biener <rguenther@suse.de>
8456 PR tree-optimization/84278
8457 * tree-vect-stmts.c (vectorizable_store): When looking for
8458 smaller vector types to perform grouped strided loads/stores
8459 make sure the mode is supported by the target.
8460 (vectorizable_load): Likewise.
8462 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8464 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8465 Increase LDP/STP opportunities by adding adjacent callee-saves.
8467 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8469 PR rtl-optimization/84068
8470 PR rtl-optimization/83459
8471 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8473 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8475 PR tree-optimization/84224
8476 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8477 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8480 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8483 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8484 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8485 count, remove one USE.
8487 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8489 * doc/install.texi (Configuration): Document the
8490 --with-long-double-format={ibm,ieee} PowerPC configuration
8494 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8495 Convert from define_expand to be define_insn_and_split. Rework
8496 float/double/_Float128 conversions to QI/HI/SImode to work with
8497 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8498 conversions to QI/HImode types did a store and then a load to
8499 truncate the value. For conversions to VSX registers, don't split
8500 the insn, instead emit the code directly. Use the code iterator
8501 any_fix to combine signed and unsigned conversions.
8502 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8503 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8504 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8505 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8506 (fix_<mode>di2_hw): Likewise.
8507 (fixuns_<mode>di2_hw): Likewise.
8508 (fix_<mode>si2_hw): Likewise.
8509 (fixuns_<mode>si2_hw): Likewise.
8510 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8511 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8512 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8513 fix<uns>_trunc<SFDF:mode>si2_p8.
8514 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8516 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8517 (fix<uns>_<mode>_mem): Likewise.
8518 (fctiw<u>z_<mode>_mem): Likewise.
8519 (fix<uns>_<mode>_mem): Likewise.
8520 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8521 the register allocator from doing a direct move to the GPRs to do
8522 a store, and instead use the ISA 3.0 store byte/half-word from
8523 vector register instruction. For IEEE 128-bit floating point,
8524 also optimize stores of 32-bit ints.
8525 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8527 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8529 * genextract.c (push_pathstr_operand): New function to support
8531 (walk_rtx): Call push_pathstr_operand.
8532 (print_path): Support [a-zA-Z].
8534 2018-02-07 Richard Biener <rguenther@suse.de>
8536 PR tree-optimization/84037
8537 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8538 (cse_and_gimplify_to_preheader): Declare.
8539 (vect_get_place_in_interleaving_chain): Likewise.
8540 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8542 (_loop_vec_info::~_loop_vec_info): Delete it.
8543 (cse_and_gimplify_to_preheader): New function.
8544 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8545 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8546 (vectorizable_load): Likewise. For grouped stores always base
8547 the IV on the first element.
8548 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8549 condition before gimplifying.
8551 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8553 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8554 *DIV_EXPR and *MOD_EXPR.
8556 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8559 * config/i386/i386.c (ix86_option_override_internal): Mask out
8560 the CF_SET bit when checking -fcf-protection.
8562 2018-02-07 Tom de Vries <tom@codesourcery.com>
8565 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8568 2018-02-07 Richard Biener <rguenther@suse.de>
8570 PR tree-optimization/84204
8571 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8574 PR tree-optimization/84205
8575 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8576 special-case isl_ast_op_zdiv_r.
8578 PR tree-optimization/84223
8579 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8580 Only add conditions from within the region.
8581 (gather_bbs::after_dom_children): Adjust.
8583 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8586 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8587 * config/avr/avr.md: Only post-reload split REG-REG moves if
8588 either register is GENERAL_REG_P.
8590 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8592 PR tree-optimization/84235
8593 * tree-ssa-scopedtables.c
8594 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8595 if the subtraction is performed in floating point type where NaNs are
8596 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8597 build 1. Formatting fix.
8599 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8602 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8603 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8604 and skip it regardless of bb boundaries. Use CALL_P macro,
8605 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8606 unnecessarily, formatting fix.
8608 2018-02-06 Michael Collison <michael.collison@arm.com>
8610 * config/arm/thumb2.md:
8611 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8612 (*thumb_mov_notscc): Ditto.
8614 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8617 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8618 unsigned_fix, not "s".
8620 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8622 * configure.ac (gcc_fn_eh_frame_ro): New function.
8623 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8624 correct .eh_frame permissions.
8625 * configure: Regenerate.
8627 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8629 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8632 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8634 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8635 Display warning message for -mno-speculate-indirect-jumps.
8637 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8639 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8641 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8643 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8645 PR tree-optimization/84225
8646 * tree-eh.c (find_trapping_overflow): Only call
8647 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8649 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8652 * config/i386/i386.c: Reimplement the check of possible options
8653 -mibt/-mshstk conbination. Change error messages.
8654 * doc/invoke.texi: Fix a typo: remove extra '='.
8656 2018-02-06 Marek Polacek <polacek@redhat.com>
8658 PR tree-optimization/84228
8659 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8661 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8664 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8665 emitted arch directives.
8666 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8667 __ARM_FEATURE_COPROC before changing architectures.
8669 2018-02-06 Richard Biener <rguenther@suse.de>
8671 * config/i386/i386.c (print_reg): Fix typo.
8672 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8674 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8676 * configure: Regenerate.
8678 2018-02-05 Martin Sebor <msebor@redhat.com>
8680 PR tree-optimization/83369
8681 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8684 2018-02-05 Martin Liska <mliska@suse.cz>
8686 * doc/invoke.texi: Cherry-pick upstream r323995.
8688 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8690 * ira.c (ira_init_register_move_cost): Adjust comment.
8692 2018-02-05 Martin Liska <mliska@suse.cz>
8694 PR gcov-profile/84137
8695 * doc/gcov.texi: Fix typo in documentation.
8697 2018-02-05 Martin Liska <mliska@suse.cz>
8699 PR gcov-profile/83879
8700 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8701 using dlopen functionality.
8703 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8705 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8706 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8707 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8708 _mm_maskz_range_round_ss): New intrinsics.
8709 (__builtin_ia32_rangesd128_round)
8710 (__builtin_ia32_rangess128_round): Remove.
8711 (__builtin_ia32_rangesd128_mask_round,
8712 __builtin_ia32_rangess128_mask_round): New builtins.
8713 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8714 __builtin_ia32_rangess128_round): Remove.
8715 (__builtin_ia32_rangesd128_mask_round,
8716 __builtin_ia32_rangess128_mask_round): New builtins.
8717 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8718 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8719 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8720 "<round_saeonly_constraint>")): Changed to ...
8721 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8722 "<round_saeonly_scalar_constraint>")): ... this.
8723 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8724 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8725 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8726 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8727 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8729 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8731 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8733 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8734 Remove all values except native, 8540 and 8548.
8736 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8738 * config/i386/i386.c (ix86_output_function_return): Pass
8739 INVALID_REGNUM, instead of -1, as invalid register number to
8740 indirect_thunk_name and output_indirect_thunk.
8742 2018-02-02 Julia Koval <julia.koval@intel.com>
8744 * config.gcc: Add -march=icelake.
8745 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8746 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8747 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8748 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8749 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8750 (processor_target_table): Add icelake.
8751 (ix86_option_override_internal): Handle new PTAs.
8752 (get_builtin_code_for_version): Handle icelake.
8753 (M_INTEL_COREI7_ICELAKE): New.
8754 (fold_builtin_cpu): Handle icelake.
8755 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8756 * doc/invoke.texi: Add -march=icelake.
8758 2018-02-02 Julia Koval <julia.koval@intel.com>
8760 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8761 to wide_int_bitmask.
8762 * wide-int-bitmask.h: New.
8764 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8767 * config/i386/i386.md: Replace Pmode with word_mode in
8768 builtin_setjmp_setup and builtin_longjmp to support x32.
8770 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8774 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8776 (rs6000_supported_cpu_names): New static variable.
8777 (linux_cpu_translation_table): Likewise.
8778 (elf_platform) <cpu>: Define new static variable and use it.
8779 Translate kernel AT_PLATFORM name to canonical name if needed.
8780 Error if platform name is unknown.
8782 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8785 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8787 2018-02-01 Jeff Law <law@redhat.com>
8790 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8791 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8792 the scratch if RELEASE_VIA_POP is false.
8793 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8794 If we have to save a temporary register, decrement SIZE appropriately.
8795 Pass new arguments to release_scratch_register_on_entry.
8796 (ix86_adjust_stack_and_probe): Likewise.
8797 (ix86_emit_probe_stack_range): Pass new arguments to
8798 release_scratch_register_on_entry.
8800 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8802 PR rtl-optimization/84157
8803 * combine.c (change_zero_ext): Use REG_P predicate in
8804 front of HARD_REGISTER_P predicate.
8806 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8808 * config/avr/avr.c (avr_option_override): Move disabling of
8809 -fdelete-null-pointer-checks to...
8810 * common/config/avr/avr-common.c (avr_option_optimization_table):
8813 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8815 PR tree-optimization/81635
8816 * tree-data-ref.c (split_constant_offset_1): For types that
8817 wrap on overflow, try to use range info to prove that wrapping
8820 2018-02-01 Renlin Li <renlin.li@arm.com>
8823 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8825 (aarch64_register_move_cost): Likewise.
8826 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8828 (REG_CLASS_NAMES): Likewise.
8829 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8830 TAILCALL_ADDR_REGS. Remove IP registers.
8831 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8833 2018-02-01 Richard Biener <rguenther@suse.de>
8835 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8836 for specifying RPO order and allow NULL for that.
8837 * domwalk.c (dom_walker::dom_walker): Likewise.
8838 (dom_walker::walk): Handle NULL RPO order.
8839 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8841 (rewrite_update_dom_walker): Likewise.
8842 (mark_def_dom_walker): Likewise.
8844 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8846 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8847 (aarch64_maybe_expand_sve_subreg_move): Declare.
8848 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8849 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8851 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8852 that are semantically a reverse operation.
8853 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8854 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8855 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8857 (aarch64_can_change_mode_class): For big-endian, forbid changes
8858 between two SVE modes if they have different element sizes.
8860 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8862 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8863 the TImode handling for big-endian targets.
8865 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8867 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8868 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8870 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8871 Remove BSWAP handing for big-endian targets and use the form of
8872 LD1RQ appropariate for the mode.
8874 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8876 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8877 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8880 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8883 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8884 check for operands that need to go through aarch64_sve_reload_be.
8886 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8888 PR tree-optimization/81661
8889 PR tree-optimization/84117
8890 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8891 * tree-eh.c: Include gimplify.h.
8892 (find_trapping_overflow, replace_trapping_overflow,
8893 rewrite_to_non_trapping_overflow): New functions.
8894 * tree-vect-loop.c: Include tree-eh.h.
8895 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8896 * tree-data-ref.c: Include tree-eh.h.
8897 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8899 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8901 PR rtl-optimization/84123
8902 * combine.c (change_zero_ext): Check if hard register satisfies
8903 can_change_dest_mode before calling gen_lowpart_SUBREG.
8905 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8908 * ira.c (ira_init_register_move_cost): Remove assert.
8910 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8912 PR rtl-optimization/84071
8913 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8914 * doc/tm.texi: Regenerate.
8916 2018-01-31 Richard Biener <rguenther@suse.de>
8918 PR tree-optimization/84132
8919 * tree-data-ref.c (analyze_miv_subscript): Properly
8920 check whether evolution_function_is_affine_multivariate_p
8921 before calling gcd_of_steps_may_divide_p.
8923 2018-01-31 Julia Koval <julia.koval@intel.com>
8926 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8927 * config/i386/i386.md (rdpid_rex64) New.
8928 (rdpid): Make 32bit only.
8930 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8933 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8934 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8936 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8939 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8941 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8943 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8945 PR rtl-optimization/84071
8946 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8947 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8949 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8951 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8952 (arc_attribute_table): Add 'aux' attribute.
8953 (arc_in_small_data_p): Consider aux like variables.
8954 (arc_is_aux_reg_p): New function.
8955 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8956 (arc_get_aux_arg): New function.
8957 (prepare_move_operands): Handle aux-register access.
8958 (arc_handle_aux_attribute): New function.
8959 * doc/extend.texi (ARC Variable attributes): Add subsection.
8961 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8963 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8964 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8965 (arc_attribute_table): Add 'uncached' attribute.
8966 (arc_print_operand): Print '.di' flag for uncached memory
8968 (arc_in_small_data_p): Do not consider for small data the uncached
8970 (arc_is_uncached_mem_p): New function.
8971 * config/arc/predicates.md (compact_store_memory_operand): Check
8972 for uncached memory accesses.
8973 (nonvol_nonimm_operand): Likewise.
8974 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8976 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8979 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8980 falign-loops=): Add Optimization flag.
8982 2018-01-30 Jeff Law <law@redhat.com>
8985 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8986 INT_REGISTERS_SAVED. Check it prior to calling
8987 get_scratch_register_on_entry.
8988 (ix86_adjust_stack_and_probe): Similarly.
8989 (ix86_emit_probe_stack_range): Similarly.
8990 (ix86_expand_prologue): Corresponding changes.
8992 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8995 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8996 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8998 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9001 * lra-constraints.c (curr_insn_transform): Process AND in the
9004 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9006 PR rtl-optimization/83986
9007 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9008 dependence against last_pending_memory_flush in addition to
9011 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9013 PR tree-optimization/81611
9014 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9017 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9020 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9023 2018-01-30 Richard Biener <rguenther@suse.de>
9024 Jakub Jelinek <jakub@redhat.com>
9026 PR tree-optimization/84111
9027 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9028 inner loops added during recursion, as they don't have up-to-date
9031 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9034 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9035 (can_inline_edge_by_limits_p): ... here.
9036 (can_early_inline_edge_p, check_callers,
9037 update_caller_keys, update_callee_keys, recursive_inlining,
9038 add_new_edges_to_heap, speculation_useful_p,
9039 inline_small_functions,
9040 inline_small_functions, flatten_function,
9041 inline_to_all_callers_1): Update.
9043 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9045 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9048 2018-01-30 Richard Biener <rguenther@suse.de>
9050 PR tree-optimization/83008
9051 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9052 invariant and constant vector uses in stmts when they need
9055 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9058 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9059 * configure: Regenerate.
9061 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9063 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9065 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9066 Use gen_rtx_REG rather than gen_lowpart.
9068 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9070 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9071 rather than 0 when creating partial subregs.
9073 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9075 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9078 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9081 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9082 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9083 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9084 flags. This restores the settings used before the 2017-07-24.
9085 Turning off pre increment/decrement/modify allows IVOPTS to
9086 optimize DF/SF loops where the index is an int.
9088 2018-01-29 Richard Biener <rguenther@suse.de>
9089 Kelvin Nilsen <kelvin@gcc.gnu.org>
9092 * tree-vect-stmts.c (vectorizable_call): Don't call
9093 targetm.vectorize_builtin_md_vectorized_function if callee is
9096 2018-01-22 Carl Love <cel@us.ibm.com>
9098 * doc/extend.tex: Fix typo in second arg in
9099 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9101 2018-01-29 Richard Biener <rguenther@suse.de>
9103 PR tree-optimization/84086
9104 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9105 (flush_ssaname_freelist): When SSA names were released reset
9106 the SCEV hash table.
9108 2018-01-29 Richard Biener <rguenther@suse.de>
9110 PR tree-optimization/84057
9111 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9112 removed paths when removing edges.
9114 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9116 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9117 -mfunction-return=@var{choice}.
9119 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9122 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9124 (layout::print_source_line): Likewise.
9125 (test_get_line_width_without_trailing_whitespace): Add test cases.
9127 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9130 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9133 2018-01-26 Jim Wilson <jimw@sifive.com>
9135 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9137 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9140 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9142 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9143 and CMP + SUB-immediate -> SUBS.
9145 2018-01-26 Martin Sebor <msebor@redhat.com>
9147 PR tree-optimization/83896
9148 * tree-ssa-strlen.c (get_string_len): Rename...
9149 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9150 Avoid assuming length is constant.
9151 (handle_char_store): Use HOST_WIDE_INT for string length.
9153 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9156 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9157 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9159 2018-01-26 Richard Biener <rguenther@suse.de>
9161 PR rtl-optimization/84003
9162 * dse.c (record_store): Only record redundant stores when
9163 the earlier store aliases at least all accesses the later one does.
9165 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9167 PR rtl-optimization/83985
9168 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9169 REG_CFA_RESTORE insns.
9170 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9171 REG_CFA_RESTORE insns here.
9174 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9175 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9177 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9179 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9180 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9181 (arc_init): Likewise.
9182 (arc_override_options): Likewise.
9183 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9185 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9187 * config/arc/arc.h (TARGET_DBNZ): Define.
9188 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9189 properly set the tune attribute.
9190 (dbnz): Use TARGET_DBNZ guard.
9191 * config/arc/arc.opt (mtune): Add core3 option.
9193 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9195 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9196 recognize new pic like addresses.
9197 (arc_delegitimize_address): Clean up.
9199 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9201 * config/arc/arc-arches.def: Option mrf16 valid for all
9203 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9204 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9205 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9206 * config/arc/arc-tables.opt: Regenerate.
9207 * config/arc/arc.c (arc_conditional_register_usage): Handle
9208 reduced register file case.
9209 (arc_file_start): Set must have build attributes.
9210 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9212 * config/arc/arc.opt (mrf16): Add new option.
9213 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9214 * config/arc/genmultilib.awk: Handle new mrf16 option.
9215 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9216 * config/arc/t-multilib: Regenerate.
9217 * doc/invoke.texi (ARC Options): Document mrf16 option.
9219 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9221 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9222 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9223 (arc_attribute_table): Add 'secure_call' attribute.
9224 (arc_print_operand): Print secure call operand.
9225 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9227 (arc_is_secure_call_p): New function. * config/arc/arc.md
9228 (call_i): Add support for sjli instruction.
9229 (call_value_i): Likewise.
9230 * config/arc/constraints.md (Csc): New constraint.
9232 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9233 John Eric Martin <John.Martin@emmicro-us.com>
9235 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9236 * config/arc/arc.c (_arc_jli_section): New struct.
9237 (arc_jli_section): New type.
9238 (rc_jli_sections): New static variable.
9239 (arc_handle_jli_attribute): New function.
9240 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9241 (arc_file_end): New function.
9242 (TARGET_ASM_FILE_END): Define.
9243 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9244 (arc_add_jli_section): New function.
9245 (jli_call_scan): Likewise.
9246 (arc_reorg): Call jli_call_scan.
9247 (arc_output_addsi): Remove 'S' from printing asm operand.
9248 (arc_is_jli_call_p): New function.
9249 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9251 (movhi_insn): Likewise.
9252 (movsi_insn): Likewise.
9253 (movsi_set_cc_insn): Likewise.
9254 (loadqi_update): Likewise.
9255 (load_zeroextendqisi_update): Likewise.
9256 (load_signextendqisi_update): Likewise.
9257 (loadhi_update): Likewise.
9258 (load_zeroextendhisi_update): Likewise.
9259 (load_signextendhisi_update): Likewise.
9260 (loadsi_update): Likewise.
9261 (loadsf_update): Likewise.
9262 (movsicc_insn): Likewise.
9263 (bset_insn): Likewise.
9264 (bxor_insn): Likewise.
9265 (bclr_insn): Likewise.
9266 (bmsk_insn): Likewise.
9267 (bicsi3_insn): Likewise.
9268 (cmpsi_cc_c_insn): Likewise.
9269 (movsi_ne): Likewise.
9270 (movsi_cond_exec): Likewise.
9271 (clrsbsi2): Likewise.
9282 (call_i): Remove 'S' asm letter, add jli instruction.
9283 (call_value_i): Likewise.
9284 * config/arc/arc.op (mjli-always): New option.
9285 * config/arc/constraints.md (Cji): New constraint.
9286 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9288 (subsf3_fpx): Likewise.
9289 (mulsf3_fpx): Likewise.
9290 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9292 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9294 * doc/invoke.texi (ARC): Document mjli-always option.
9296 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9298 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9299 avoid addition with 0 and use incw and decw where possible.
9301 2018-01-26 Richard Biener <rguenther@suse.de>
9303 PR tree-optimization/81082
9304 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9305 association if it requires casting to unsigned.
9306 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9307 from fold_plusminus_mult_expr to catch important cases late when
9308 range info is available.
9310 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9312 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9313 * configure.ac (hidden_linkonce): New test.
9314 * configure: Regenerate.
9315 * config.in: Regenerate.
9317 2018-01-26 Julia Koval <julia.koval@intel.com>
9319 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9320 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9321 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9322 _mm_mask_bitshuffle_epi64_mask): Fix type.
9323 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9324 USI_FTYPE_V4DI_V4DI_USI): Remove.
9325 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9326 __builtin_ia32_vpshufbitqmb256_mask,
9327 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9328 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9329 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9331 2018-01-26 Alan Modra <amodra@gmail.com>
9334 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9335 UNSPEC_VBPERMQ. Sort other unspecs.
9337 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9339 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9341 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9344 * predict.c (drop_profile): Do not push/pop cfun; update also
9346 (handle_missing_profiles): Fix logic looking for zero profiles.
9348 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9351 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9352 on functions with #pragma omp declare simd or functions with simd
9354 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9355 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9356 Remove trailing \n from warning_at calls.
9358 2018-01-25 Tom de Vries <tom@codesourcery.com>
9361 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9362 for neutered workers.
9364 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9367 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9368 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9370 2018-01-24 Jeff Law <law@redhat.com>
9373 * i386.c (get_probe_interval): Move to earlier point.
9374 (ix86_compute_frame_layout): If -fstack-clash-protection and
9375 the frame is larger than the probe interval, then use pushes
9376 to save registers rather than reg->mem moves.
9377 (ix86_expand_prologue): Remove conditional for int_registers_saved
9380 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9383 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9384 min/max for never referenced object.
9386 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9389 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9391 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9392 attributes from DECL_ATTRIBUTES (decl) without affecting
9393 DECL_ATTRIBUTES (current_function_decl).
9394 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9395 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9397 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9399 PR tree-optimization/83979
9400 * fold-const.c (fold_comparison): Use constant_boolean_node
9401 instead of boolean_{true,false}_node.
9403 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9405 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9408 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9410 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9411 Simplify the clause that sets the length attribute.
9412 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9413 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9414 clause that sets the length attribute.
9415 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9417 2018-01-24 Tom de Vries <tom@codesourcery.com>
9420 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9421 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9422 Add strict parameter.
9423 (prevent_branch_around_nothing): Insert dummy insn between branch to
9424 label and label with no ptx insn inbetween.
9425 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9427 2018-01-24 Tom de Vries <tom@codesourcery.com>
9430 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9431 for neutered threads in warp.
9432 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9434 2018-01-24 Richard Biener <rguenther@suse.de>
9436 PR tree-optimization/83176
9437 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9440 2018-01-24 Richard Biener <rguenther@suse.de>
9442 PR tree-optimization/82819
9443 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9444 code generating pluses that are no-ops in the target precision.
9446 2018-01-24 Richard Biener <rguenther@suse.de>
9449 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9451 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9453 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9454 to merge probabilities.
9455 * predict.c (probably_never_executed): Also mark as cold functions
9456 with global 0 profile and guessed local profile.
9457 * profile-count.c (profile_probability::combine_with_count): New
9459 * profile-count.h (profile_probability::operator*,
9460 profile_probability::operator*=, profile_probability::operator/,
9461 profile_probability::operator/=): Reduce precision to adjusted
9462 and set value to guessed on contradictory divisions.
9463 (profile_probability::combine_with_freq): Remove.
9464 (profile_probability::combine_wiht_count): Declare.
9465 (profile_count::force_nonzero):: Set to adjusted.
9466 (profile_count::probability_in):: Set quality to adjusted.
9467 * tree-ssa-tail-merge.c (replace_block_by): Use
9470 2018-01-23 Andrew Waterman <andrew@sifive.com>
9471 Jim Wilson <jimw@sifive.com>
9473 * config/riscv/riscv.c (riscv_stack_boundary): New.
9474 (riscv_option_override): Set riscv_stack_boundary. Handle
9475 riscv_preferred_stack_boundary_arg.
9476 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9477 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9478 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9479 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9480 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9481 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9483 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9486 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9487 of struct ix86_frame.
9488 (ix86_expand_epilogue): Likewise. Add a local variable for
9489 the reg_save_offset field in struct ix86_frame.
9491 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9493 PR tree-optimization/82604
9494 * tree-loop-distribution.c (enum partition_kind): New enum item
9495 PKIND_PARTIAL_MEMSET.
9496 (partition_builtin_p): Support above new enum item.
9497 (generate_code_for_partition): Ditto.
9498 (compute_access_range): Differentiate cases that equality can be
9499 proven at all loops, the innermost loops or no loops.
9500 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9501 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9502 (finalize_partitions, distribute_loop): Don't fuse partition of
9503 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9504 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9507 2018-01-23 Martin Liska <mliska@suse.cz>
9509 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9510 order to ignore the predictor.
9511 (PRED_POLYMORPHIC_CALL): Likewise.
9512 (PRED_RECURSIVE_CALL): Likewise.
9514 2018-01-23 Martin Liska <mliska@suse.cz>
9516 * tree-profile.c (tree_profiling): Print function header to
9517 aware reader which function we are working on.
9518 * value-prof.c (gimple_find_values_to_profile): Do not print
9519 not interesting value histograms.
9521 2018-01-23 Martin Liska <mliska@suse.cz>
9523 * profile-count.h (enum profile_quality): Add
9524 profile_uninitialized as the first value. Do not number values
9525 as they are zero based.
9526 (profile_count::verify): Update sanity check.
9527 (profile_probability::verify): Likewise.
9529 2018-01-23 Nathan Sidwell <nathan@acm.org>
9531 * doc/invoke.texi (ffor-scope): Deprecate.
9533 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9535 PR tree-optimization/83510
9536 * domwalk.c (set_all_edges_as_executable): New function.
9537 (dom_walker::dom_walker): Convert bool param
9538 "skip_unreachable_blocks" to enum reachability. Move setup of
9539 edge flags to set_all_edges_as_executable and only do it when
9540 reachability is REACHABLE_BLOCKS.
9541 * domwalk.h (enum dom_walker::reachability): New enum.
9542 (dom_walker::dom_walker): Convert bool param
9543 "skip_unreachable_blocks" to enum reachability.
9544 (set_all_edges_as_executable): New decl.
9545 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9546 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9548 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9549 but converting true to REACHABLE_BLOCKS.
9550 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9552 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9553 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9554 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9556 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9557 if check_all_array_refs will be called.
9559 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9561 * tree.c (selftest::test_location_wrappers): Add more test
9564 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9566 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9567 (selftest::test_bit_in_range): Likewise.
9569 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9572 * doc/sourcebuild.texi (vect_float): Say that the selector
9573 only describes the situation when -funsafe-math-optimizations is on.
9574 (vect_float_strict): Document.
9576 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9578 PR tree-optimization/83965
9579 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9580 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9581 instead of checking only for a reduction.
9582 (vect_recog_widen_sum_pattern): Likewise.
9584 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9586 * predict.c (probably_never_executed): Only use precise profile info.
9587 (compute_function_frequency): Skip after inlining hack since we now
9588 have quality checking.
9590 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9592 * profile-count.h (profile_probability::very_unlikely,
9593 profile_probability::unlikely, profile_probability::even): Set
9594 precision to guessed.
9596 2018-01-23 Richard Biener <rguenther@suse.de>
9598 PR tree-optimization/83963
9599 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9600 Properly terminate dominator walk when crossing the exit edge not
9601 when visiting its source block.
9603 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9606 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9607 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9609 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9611 PR tree-optimization/83957
9612 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9613 semicolon after for body surrounded by braces.
9615 PR tree-optimization/83081
9616 * profile-count.h (profile_probability::split): New method.
9617 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9618 Use profile_probability::split.
9619 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9620 when splitting a single conditional jump into 2.
9622 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9624 PR tree-optimization/69452
9625 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9630 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9631 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9632 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9634 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9636 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9637 * config/rl78/rl78.md (movdi): New define_expand.
9638 * config/rl78/rl78.c (rl78_split_movdi): New function.
9640 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9643 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9645 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9646 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9647 128-bit to produce an UNSPEC move to get the double word with the
9648 signbit and then a shift directly to do signbit.
9649 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9650 implementation with a new version that just does either a direct
9651 move or a regular move. Move memory interface to separate insns.
9652 Move insns so they are next to the expander.
9653 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9654 with signbit move. Split big and little endian case.
9655 (signbit<mode>2_dm_mem_le): Likewise.
9656 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9657 (signbit<mode>2_dm2): Likewise.
9659 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9661 * config/rl78/rl78.md (anddi3): New define_expand.
9663 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9665 * config/rl78/rl78.md (umindi3): New define_expand.
9667 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9669 * config/rl78/rl78.md (smindi3): New define_expand.
9671 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9673 * config/rl78/rl78.md (smaxdi3): New define_expand.
9675 2018-01-22 Carl Love <cel@us.ibm.com>
9677 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9678 LVX_V1TI): Add macro expansion.
9679 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9680 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9681 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9682 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9683 Change check to determine if the instruction is a byte reversing
9684 entry. Fix typo in comment.
9685 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9686 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9687 Add def_builtin calls for new builtins.
9688 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9689 Add define_insn expansion.
9691 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9693 * config/rl78/rl78.md (umaxdi3): New define_expand.
9695 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9697 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9698 for non-QImode registers.
9700 2018-01-22 Richard Biener <rguenther@suse.de>
9702 PR tree-optimization/83963
9703 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9704 including the loop exit block.
9705 (scop_detection::merge_sese): Likewise.
9706 (scop_detection::add_scop): Do it here instead.
9708 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9710 * doc/sourcebuild.texi (arm_softfloat): Document.
9712 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9715 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9716 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9717 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9719 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9720 David Edelsohn <dje.gcc@gmail.com>
9723 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9724 Change "crset eq" to "crset 2".
9725 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9726 (*call_indirect_aix<mode>_nospec): Likewise.
9727 (*call_value_indirect_aix<mode>_nospec): Likewise.
9728 (*call_indirect_elfv2<mode>_nospec): Likewise.
9729 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9730 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9731 change assembly output from . to $.
9732 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9733 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9734 (*tablejump<mode>_internal1_nospec): Likewise.
9736 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9739 * config/sh/sh_optimize_sett_clrt.cc:
9740 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9742 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9744 PR tree-optimization/83940
9745 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9746 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9747 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9748 use it to pass back the definition type.
9749 (vect_check_store_rhs): Likewise rhs_dt_out.
9750 (vect_build_gather_load_calls): Add a mask_dt argument and use
9751 it instead of a call to vect_is_simple_use.
9752 (vectorizable_store): Update calls to vect_check_load_store_mask
9753 and vect_check_store_rhs. Use the dt returned by the latter instead
9754 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9755 instead of calls to vect_is_simple_use. Pass the scalar rather
9756 than the vector operand to vect_is_simple_use when handling
9757 second and subsequent copies of an rhs value.
9758 (vectorizable_load): Update calls to vect_check_load_store_mask
9759 and vect_build_gather_load_calls. Use the cached mask_dt and
9760 gs_info.offset_dt instead of calls to vect_is_simple_use.
9762 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9765 * tree-emutls.c: Include gimplify.h.
9766 (lower_emutls_2): New function.
9767 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9768 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9769 it before further processing.
9772 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9773 UINTVAL (trueop1) instead of INTVAL (op1).
9775 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9779 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9780 INCOMING_FRAME_SP_OFFSET if not defined.
9781 (scan_trace): Add ENTRY argument. If true and
9782 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9783 emit a note to adjust the CFA offset.
9784 (create_cfi_notes): Adjust scan_trace callers.
9785 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9786 INCOMING_FRAME_SP_OFFSET in the CIE.
9787 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9788 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9790 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9791 * doc/tm.texi: Regenerated.
9793 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9795 PR rtl-optimization/83147
9796 * lra-constraints.c (remove_inheritance_pseudos): Use
9797 lra_substitute_pseudo_within_insn.
9799 2018-01-19 Tom de Vries <tom@codesourcery.com>
9800 Cesar Philippidis <cesar@codesourcery.com>
9803 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9805 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9808 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9809 spaces for function labels.
9811 2018-01-19 Martin Liska <mliska@suse.cz>
9813 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9814 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9815 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9816 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9817 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9818 (PRED_CONST_RETURN): Change from 69 to 65.
9819 (PRED_NULL_RETURN): Change from 91 to 71.
9820 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9821 (PRED_LOOP_GUARD): Change from 66 to 73.
9823 2018-01-19 Martin Liska <mliska@suse.cz>
9825 * predict.c (predict_insn_def): Add new assert.
9826 (struct branch_predictor): Change type to signed integer.
9827 (test_prediction_value_range): Amend test to cover
9829 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9830 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9831 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9832 (PRED_LOOP_IV_COMPARE): Likewise.
9833 * predict.h (PROB_UNINITIALIZED): Define new constant.
9835 2018-01-19 Martin Liska <mliska@suse.cz>
9837 * predict.c (dump_prediction): Add new format for
9838 analyze_brprob.py script which is enabled with -details
9840 * profile-count.h (precise_p): New function.
9842 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9844 PR tree-optimization/83922
9845 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9846 there are no statements that need masking.
9847 (vect_active_double_reduction_p): New function.
9848 (vect_analyze_loop_operations): Use it when handling phis that
9849 are not in the loop header.
9851 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9853 PR tree-optimization/83914
9854 * tree-vect-loop.c (vectorizable_induction): Don't convert
9855 init_expr or apply the peeling adjustment for inductions
9856 that are nested within the vectorized loop.
9858 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9860 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9863 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9867 * function.h (gimplify_parameters): Add gimple_seq * argument.
9868 * function.c: Include gimple.h and options.h.
9869 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9870 for the added local temporaries if needed.
9871 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9872 if there are any parameter cleanups, wrap whole body into a
9873 try/finally with the cleanups.
9875 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9878 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9879 Use GET_MODE_CLASS for scalar floating point.
9881 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9885 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9886 Fix call of call_cgraph_insertion_hooks.
9888 2018-01-18 Martin Sebor <msebor@redhat.com>
9890 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9892 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9895 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9898 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9901 * common.opt: (-ffile-prefix-map): New option.
9902 * opts.c (common_handle_option): Defer it.
9903 * opts-global.c (handle_common_deferred_options): Handle it.
9904 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9905 * file-prefix-map.h: New file.
9906 (remap_debug_filename, add_debug_prefix_map): ...here.
9907 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9908 * final.c (debug_prefix_map, add_debug_prefix_map
9909 remap_debug_filename): Move to...
9910 * file-prefix-map.c: New file.
9911 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9912 generalize, get rid of alloca(), use strrchr() instead of strchr().
9913 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9914 Implement in terms of add_prefix_map().
9915 (remap_macro_filename, remap_debug_filename): Implement in term of
9917 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9918 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9919 * dbxout.c: Include file-prefix-map.h.
9920 * varasm.c: Likewise.
9921 * vmsdbgout.c: Likewise.
9922 * xcoffout.c: Likewise.
9923 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9924 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9925 * doc/invoke.texi (-ffile-prefix-map): Document.
9926 (-fdebug-prefix-map): Update description.
9928 2018-01-18 Martin Liska <mliska@suse.cz>
9930 * config/i386/i386.c (indirect_thunk_name): Document that also
9932 (output_indirect_thunk): Document why both instructions
9933 (pause and lfence) are generated.
9935 2018-01-18 Richard Biener <rguenther@suse.de>
9937 PR tree-optimization/83887
9938 * graphite-scop-detection.c
9939 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9940 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9941 (scop_detection::merge_sese): Re-implement with a flood-fill
9942 algorithm that properly finds a SESE region if it exists.
9944 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9947 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9948 pointer_diff optimizations use view_convert instead of convert.
9950 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9952 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9953 Generate different code for -mno-speculate-indirect-jumps.
9954 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9955 (*call_indirect_aix<mode>): Disable for
9956 -mno-speculate-indirect-jumps.
9957 (*call_indirect_aix<mode>_nospec): New define_insn.
9958 (*call_value_indirect_aix<mode>): Disable for
9959 -mno-speculate-indirect-jumps.
9960 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9961 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9962 -mno-speculate-indirect-jumps.
9963 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9965 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9967 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9968 long double type, set the flags for noting the default long double
9969 type, even if we don't pass or return a long double type.
9971 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9974 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9977 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9979 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9980 support for merge[hl].
9981 (fold_mergehl_helper): New helper function.
9982 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9983 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9984 (altivec_vmrglw_direct): Add xxmrglw insn.
9986 2018-01-17 Andrew Waterman <andrew@sifive.com>
9988 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9989 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9991 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9994 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9995 call the lto_location_cache before reading the
9996 DECL_SOURCE_LOCATION of the types.
9998 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9999 Richard Sandiford <richard.sandiford@linaro.org>
10001 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10002 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10003 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10004 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10005 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10007 * config/aarch64/constraints.md (aarch64_movti_operand):
10009 * config/aarch64/predicates.md (Uti): Add new constraint.
10011 2018-01-17 Carl Love <cel@us.ibm.com>
10013 * config/rs6000/vsx.md (define_expand xl_len_r,
10014 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10015 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10017 (define_expand, define_insn): Move the shift left from the
10018 define_insn to the define_expand for lxvl and stxvl instructions.
10019 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10020 and XL_LEN_R definitions to PURE.
10022 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10024 * config/i386/i386.c (indirect_thunk_name): Declare regno
10025 as unsigned int. Compare regno with INVALID_REGNUM.
10026 (output_indirect_thunk): Ditto.
10027 (output_indirect_thunk_function): Ditto.
10028 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10029 in the call to output_indirect_thunk_function.
10031 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10033 PR middle-end/83884
10034 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10035 rather than the size of inner_type to determine the stack slot size
10036 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10038 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10041 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10044 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10046 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10047 endian Linux systems to optionally enable multilibs for selecting
10048 the long double type if the user configured an explicit type.
10049 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10050 have no long double multilibs if not defined.
10051 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10052 warn if the user used -mabi={ieee,ibm}longdouble and we built
10053 multilibs for long double.
10054 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10055 appropriate multilib option.
10056 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10058 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10059 for building long double multilibs.
10060 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10062 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10064 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10067 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10069 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10072 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10075 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10078 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10080 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10081 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10083 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10085 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10086 different rtl trees depending on TARGET_64BIT.
10087 (rs6000_gen_lvx): Likewise.
10089 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10091 * config/visium/visium.md (nop): Tweak comment.
10092 (hazard_nop): Likewise.
10094 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10096 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10097 -mspeculate-indirect-jumps.
10098 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10099 for -mno-speculate-indirect-jumps.
10100 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10101 (*call_value_indirect_elfv2<mode>): Disable for
10102 -mno-speculate-indirect-jumps.
10103 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10104 (indirect_jump): Emit different RTL for
10105 -mno-speculate-indirect-jumps.
10106 (*indirect_jump<mode>): Disable for
10107 -mno-speculate-indirect-jumps.
10108 (*indirect_jump<mode>_nospec): New define_insn.
10109 (tablejump): Emit different RTL for
10110 -mno-speculate-indirect-jumps.
10111 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10112 (tablejumpsi_nospec): New define_expand.
10113 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10114 (tablejumpdi_nospec): New define_expand.
10115 (*tablejump<mode>_internal1): Disable for
10116 -mno-speculate-indirect-jumps.
10117 (*tablejump<mode>_internal1_nospec): New define_insn.
10118 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10121 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10123 * caller-save.c (insert_save): Drop unnecessary parameter. All
10126 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10127 Richard Biener <rguenth@suse.de>
10130 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10131 return early, inline manually is_gimple_sizepos. Make sure if we
10132 call gimplify_expr we don't end up with a gimple constant.
10133 * tree.c (variably_modified_type_p): Don't return true for
10134 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10135 * gimplify.h (is_gimple_sizepos): Remove.
10137 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10139 PR tree-optimization/83857
10140 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10141 vectorizable_live_operation for pure SLP statements.
10142 (vectorizable_live_operation): Handle PHIs.
10144 2018-01-16 Richard Biener <rguenther@suse.de>
10146 PR tree-optimization/83867
10147 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10148 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10150 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10153 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10154 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10155 If off is not INTEGER_CST, issue a may not be aligned warning
10156 rather than isn't aligned. Use isn%'t rather than isn't.
10157 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10159 <case MULT_EXPR>: Improve the case when bottom and one of the
10160 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10161 operand, in that case check if the other operand is multiple of
10162 bottom divided by the INTEGER_CST operand.
10164 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10167 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10168 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10169 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10170 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10171 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10172 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10173 (pa_function_arg_size): New function.
10175 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10177 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10178 in a separate statement.
10180 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10182 PR tree-optimization/83847
10183 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10184 group gathers and scatters.
10186 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10188 PR rtl-optimization/86620
10189 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10191 PR rtl-optimization/83213
10192 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10193 to last if both are JUMP_INSNs.
10195 PR tree-optimization/83843
10196 * gimple-ssa-store-merging.c
10197 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10198 store_immediate_info for bswap/nop orig_stores.
10200 2018-01-15 Andrew Waterman <andrew@sifive.com>
10202 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10204 <UDIV>: Increase cost if !TARGET_DIV.
10206 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10208 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10209 (define_attr "cr_logical_3op"): New.
10210 (cceq_ior_compare): Adjust.
10211 (cceq_ior_compare_complement): Adjust.
10212 (*cceq_rev_compare): Adjust.
10213 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10214 (is_cracked_insn): Adjust.
10215 (insn_must_be_first_in_group): Adjust.
10216 * config/rs6000/40x.md: Adjust.
10217 * config/rs6000/440.md: Adjust.
10218 * config/rs6000/476.md: Adjust.
10219 * config/rs6000/601.md: Adjust.
10220 * config/rs6000/603.md: Adjust.
10221 * config/rs6000/6xx.md: Adjust.
10222 * config/rs6000/7450.md: Adjust.
10223 * config/rs6000/7xx.md: Adjust.
10224 * config/rs6000/8540.md: Adjust.
10225 * config/rs6000/cell.md: Adjust.
10226 * config/rs6000/e300c2c3.md: Adjust.
10227 * config/rs6000/e500mc.md: Adjust.
10228 * config/rs6000/e500mc64.md: Adjust.
10229 * config/rs6000/e5500.md: Adjust.
10230 * config/rs6000/e6500.md: Adjust.
10231 * config/rs6000/mpc.md: Adjust.
10232 * config/rs6000/power4.md: Adjust.
10233 * config/rs6000/power5.md: Adjust.
10234 * config/rs6000/power6.md: Adjust.
10235 * config/rs6000/power7.md: Adjust.
10236 * config/rs6000/power8.md: Adjust.
10237 * config/rs6000/power9.md: Adjust.
10238 * config/rs6000/rs64.md: Adjust.
10239 * config/rs6000/titan.md: Adjust.
10241 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10243 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10244 ix86_indirect_branch_register logic.
10246 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10248 * config/i386/constraints.md (Bs): Update
10249 ix86_indirect_branch_register check. Don't check
10250 ix86_indirect_branch_register with GOT_memory_operand.
10252 * config/i386/predicates.md (GOT_memory_operand): Don't check
10253 ix86_indirect_branch_register here.
10254 (GOT32_symbol_operand): Likewise.
10256 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10258 * config/i386/predicates.md (constant_call_address_operand):
10259 Rewrite ix86_indirect_branch_register logic.
10260 (sibcall_insn_operand): Likewise.
10262 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10264 * config/i386/constraints.md (Bs): Replace
10265 ix86_indirect_branch_thunk_register with
10266 ix86_indirect_branch_register.
10268 * config/i386/i386.md (indirect_jump): Likewise.
10269 (tablejump): Likewise.
10270 (*sibcall_memory): Likewise.
10271 (*sibcall_value_memory): Likewise.
10272 Peepholes of indirect call and jump via memory: Likewise.
10273 * config/i386/i386.opt: Likewise.
10274 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10275 (GOT_memory_operand): Likewise.
10276 (call_insn_operand): Likewise.
10277 (sibcall_insn_operand): Likewise.
10278 (GOT32_symbol_operand): Likewise.
10280 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10282 PR middle-end/83837
10283 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10284 type rather than type addr's type points to.
10285 (expand_omp_atomic_mutex): Likewise.
10286 (expand_omp_atomic): Likewise.
10288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10291 * config/i386/i386.c (output_indirect_thunk_function): Use
10292 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10293 for __x86_return_thunk.
10295 2018-01-15 Richard Biener <rguenther@suse.de>
10297 PR middle-end/83850
10298 * expmed.c (extract_bit_field_1): Fix typo.
10300 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10303 * config/arm/iterators.md (VF): New mode iterator.
10304 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10305 Remove integer-related logic from pattern.
10306 (neon_vabd<mode>_3): Likewise.
10308 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10310 PR middle-end/82694
10311 * common.opt (fstrict-overflow): No longer an alias.
10312 (fwrapv-pointer): New option.
10313 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10314 also for pointer types based on flag_wrapv_pointer.
10315 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10316 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10317 opts->x_flag_wrapv got set.
10318 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10319 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10320 POINTER_TYPE_OVERFLOW_UNDEFINED.
10321 * match.pd: Likewise in address comparison pattern.
10322 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10324 2018-01-15 Richard Biener <rguenther@suse.de>
10327 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10328 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10329 Reset type names to their identifier if their TYPE_DECL doesn't
10330 have linkage (and thus is used for ODR and devirt).
10331 (save_debug_info_for_decl): Remove.
10332 (save_debug_info_for_type): Likewise.
10333 (add_tree_to_fld_list): Adjust.
10334 * tree-pretty-print.c (dump_generic_node): Make dumping of
10335 type names more robust.
10337 2018-01-15 Richard Biener <rguenther@suse.de>
10339 * BASE-VER: Bump to 8.0.1.
10341 2018-01-14 Martin Sebor <msebor@redhat.com>
10344 * builtins.c (check_access): Avoid warning when the no-warning bit
10347 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10349 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10350 * ira-color (allocno_hard_regs_compare): Likewise.
10352 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10355 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10356 Use .pushsection/.popsection.
10358 2018-01-14 Martin Sebor <msebor@redhat.com>
10361 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10363 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10365 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10366 entry from extra_headers.
10367 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10368 extra_headers, make the list bitwise identical to the i?86-*-* one.
10370 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10372 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10373 -mcmodel=large with -mindirect-branch=thunk,
10374 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10375 -mfunction-return=thunk-extern.
10376 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10377 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10378 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10380 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10382 * config/i386/i386.c (print_reg): Print the name of the full
10383 integer register without '%'.
10384 (ix86_print_operand): Handle 'V'.
10385 * doc/extend.texi: Document 'V' modifier.
10387 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10389 * config/i386/constraints.md (Bs): Disallow memory operand for
10390 -mindirect-branch-register.
10392 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10393 (GOT_memory_operand): Likewise.
10394 (call_insn_operand): Likewise.
10395 (sibcall_insn_operand): Likewise.
10396 (GOT32_symbol_operand): Likewise.
10397 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10398 for -mindirect-branch-register.
10399 (tablejump): Likewise.
10400 (*sibcall_memory): Likewise.
10401 (*sibcall_value_memory): Likewise.
10402 Disallow peepholes of indirect call and jump via memory for
10403 -mindirect-branch-register.
10404 (*call_pop): Replace m with Bw.
10405 (*call_value_pop): Likewise.
10406 (*sibcall_pop_memory): Replace m with Bs.
10407 * config/i386/i386.opt (mindirect-branch-register): New option.
10408 * doc/invoke.texi: Document -mindirect-branch-register option.
10410 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10412 * config/i386/i386-protos.h (ix86_output_function_return): New.
10413 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10414 set function_return_type.
10415 (indirect_thunk_name): Add ret_p to indicate thunk for function
10417 (output_indirect_thunk_function): Pass false to
10418 indirect_thunk_name.
10419 (ix86_output_indirect_branch_via_reg): Likewise.
10420 (ix86_output_indirect_branch_via_push): Likewise.
10421 (output_indirect_thunk_function): Create alias for function
10422 return thunk if regno < 0.
10423 (ix86_output_function_return): New function.
10424 (ix86_handle_fndecl_attribute): Handle function_return.
10425 (ix86_attribute_table): Add function_return.
10426 * config/i386/i386.h (machine_function): Add
10427 function_return_type.
10428 * config/i386/i386.md (simple_return_internal): Use
10429 ix86_output_function_return.
10430 (simple_return_internal_long): Likewise.
10431 * config/i386/i386.opt (mfunction-return=): New option.
10432 (indirect_branch): Mention -mfunction-return=.
10433 * doc/extend.texi: Document function_return function attribute.
10434 * doc/invoke.texi: Document -mfunction-return= option.
10436 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10438 * config/i386/i386-opts.h (indirect_branch): New.
10439 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10440 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10441 with local indirect jump when converting indirect call and jump.
10442 (ix86_set_indirect_branch_type): New.
10443 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10444 (indirectlabelno): New.
10445 (indirect_thunk_needed): Likewise.
10446 (indirect_thunk_bnd_needed): Likewise.
10447 (indirect_thunks_used): Likewise.
10448 (indirect_thunks_bnd_used): Likewise.
10449 (INDIRECT_LABEL): Likewise.
10450 (indirect_thunk_name): Likewise.
10451 (output_indirect_thunk): Likewise.
10452 (output_indirect_thunk_function): Likewise.
10453 (ix86_output_indirect_branch_via_reg): Likewise.
10454 (ix86_output_indirect_branch_via_push): Likewise.
10455 (ix86_output_indirect_branch): Likewise.
10456 (ix86_output_indirect_jmp): Likewise.
10457 (ix86_code_end): Call output_indirect_thunk_function if needed.
10458 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10460 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10461 (ix86_attribute_table): Add indirect_branch.
10462 * config/i386/i386.h (machine_function): Add indirect_branch_type
10463 and has_local_indirect_jump.
10464 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10466 (tablejump): Likewise.
10467 (*indirect_jump): Use ix86_output_indirect_jmp.
10468 (*tablejump_1): Likewise.
10469 (simple_return_indirect_internal): Likewise.
10470 * config/i386/i386.opt (mindirect-branch=): New option.
10471 (indirect_branch): New.
10474 (thunk-inline): Likewise.
10475 (thunk-extern): Likewise.
10476 * doc/extend.texi: Document indirect_branch function attribute.
10477 * doc/invoke.texi: Document -mindirect-branch= option.
10479 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10482 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10484 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10486 * ipa-inline.c (want_inline_small_function_p): Return false if
10487 inlining has already failed with CIF_FINAL_ERROR.
10488 (update_caller_keys): Call want_inline_small_function_p before
10490 (update_callee_keys): Likewise.
10492 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10494 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10496 (rs6000_quadword_masked_address_p): Likewise.
10497 (quad_aligned_load_p): Likewise.
10498 (quad_aligned_store_p): Likewise.
10499 (const_load_sequence_p): Add comment to describe the outer-most loop.
10500 (mimic_memory_attributes_and_flags): New function.
10501 (rs6000_gen_stvx): Likewise.
10502 (replace_swapped_aligned_store): Likewise.
10503 (rs6000_gen_lvx): Likewise.
10504 (replace_swapped_aligned_load): Likewise.
10505 (replace_swapped_load_constant): Capitalize argument name in
10506 comment describing this function.
10507 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10508 and stores that access quad-word aligned addresses and replace
10509 with stvx or lvx instructions when appropriate.
10510 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10511 New function prototype.
10512 (rs6000_quadword_masked_address_p): Likewise.
10513 (rs6000_gen_lvx): Likewise.
10514 (rs6000_gen_stvx): Likewise.
10515 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10516 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10517 when memory address is aligned.
10518 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10519 this split to select lvx instruction when memory address is aligned.
10520 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10521 instruction when memory address is aligned.
10522 (*vsx_le_perm_load_v16qi): Likewise.
10523 (four unnamed splitters): Modify to select the stvx instruction
10524 when memory is aligned.
10526 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10528 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10529 which appears in the queue multiple times.
10531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10532 Alan Hayward <alan.hayward@arm.com>
10533 David Sherwood <david.sherwood@arm.com>
10535 * tree-vectorizer.h (vec_lower_bound): New structure.
10536 (_loop_vec_info): Add check_nonzero and lower_bounds.
10537 (LOOP_VINFO_CHECK_NONZERO): New macro.
10538 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10539 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10540 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10541 fields. Make seg_len the distance travelled, not including the
10543 (dr_direction_indicator): Declare.
10544 (dr_zero_step_indicator): Likewise.
10545 (dr_known_forward_stride_p): Likewise.
10546 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10548 (runtime_alias_check_p): Allow runtime alias checks with
10550 (operator ==): Compare access_size and align.
10551 (prune_runtime_alias_test_list): Rework for new distinction between
10552 the access_size and seg_len.
10553 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10555 (get_segment_min_max): New function.
10556 (create_intersect_range_checks): Use it.
10557 (dr_step_indicator): New function.
10558 (dr_direction_indicator): Likewise.
10559 (dr_zero_step_indicator): Likewise.
10560 (dr_known_forward_stride_p): Likewise.
10561 * tree-loop-distribution.c (data_ref_segment_size): Return
10562 DR_STEP * (niters - 1).
10563 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10565 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10566 (vect_preserves_scalar_order_p): New function, split out from...
10567 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10568 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10569 (vect_vfa_access_size): New function.
10570 (vect_vfa_align): Likewise.
10571 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10572 (dump_lower_bound): New function.
10573 (vect_check_lower_bound): Likewise.
10574 (vect_small_gap_p): Likewise.
10575 (vectorizable_with_step_bound_p): Likewise.
10576 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10577 depencies if the vectorization factor is 1. Convert the checks
10578 for nonzero steps into checks on the bounds of DR_STEP. Try using
10579 a bunds check for variable steps if the minimum required step is
10580 relatively small. Update calls to the dr_with_seg_len
10581 constructor and to vect_compile_time_alias.
10582 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10584 (vect_loop_versioning): Call it.
10585 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10587 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10590 Alan Hayward <alan.hayward@arm.com>
10591 David Sherwood <david.sherwood@arm.com>
10593 * doc/sourcebuild.texi (vect_scatter_store): Document.
10594 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10596 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10598 * genopinit.c (main): Add supports_vec_scatter_store and
10599 supports_vec_scatter_store_cached to target_optabs.
10600 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10601 IFN_MASK_SCATTER_STORE.
10602 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10604 * internal-fn.h (internal_store_fn_p): Declare.
10605 (internal_fn_stored_value_index): Likewise.
10606 * internal-fn.c (scatter_store_direct): New macro.
10607 (expand_scatter_store_optab_fn): New function.
10608 (direct_scatter_store_optab_supported_p): New macro.
10609 (internal_store_fn_p): New function.
10610 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10611 IFN_MASK_SCATTER_STORE.
10612 (internal_fn_mask_index): Likewise.
10613 (internal_fn_stored_value_index): New function.
10614 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10615 for scatter stores.
10616 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10617 * optabs-query.c (supports_vec_scatter_store_p): New function.
10618 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10619 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10620 true for scatter stores.
10621 (vect_gather_scatter_fn_p): Handle scatter stores too.
10622 (vect_check_gather_scatter): Consider using scatter stores if
10623 supports_vec_scatter_store_p.
10624 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10625 scatter stores too.
10626 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10627 internal_fn_stored_value_index.
10628 (check_load_store_masking): Handle scatter stores too.
10629 (vect_get_store_rhs): Make public.
10630 (vectorizable_call): Use internal_store_fn_p.
10631 (vectorizable_store): Handle scatter store internal functions.
10632 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10633 when deciding whether the end of the group has been reached.
10634 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10635 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10636 (mask_scatter_store<mode>): New insns.
10638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10639 Alan Hayward <alan.hayward@arm.com>
10640 David Sherwood <david.sherwood@arm.com>
10642 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10643 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10644 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10646 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10647 Use vect_truncate_gather_scatter_offset if we can't treat the
10648 operation as a normal gather load or scatter store.
10649 (get_group_load_store_type): Take the gather_scatter_info
10650 as argument. Try using a gather load or scatter store for
10651 single-element groups.
10652 (get_load_store_type): Update calls to get_group_load_store_type
10653 and vect_use_strided_gather_scatters_p.
10655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10656 Alan Hayward <alan.hayward@arm.com>
10657 David Sherwood <david.sherwood@arm.com>
10659 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10660 optional tree argument.
10661 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10663 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10664 but continue to use the current value as a fallback.
10665 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10666 to compare the updates.
10667 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10668 (get_load_store_type): Use it when handling a strided access.
10669 (vect_get_strided_load_store_ops): New function.
10670 (vect_get_data_ptr_increment): Likewise.
10671 (vectorizable_load): Handle strided gather loads. Always pass
10672 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10675 Alan Hayward <alan.hayward@arm.com>
10676 David Sherwood <david.sherwood@arm.com>
10678 * doc/md.texi (gather_load@var{m}): Document.
10679 (mask_gather_load@var{m}): Likewise.
10680 * genopinit.c (main): Add supports_vec_gather_load and
10681 supports_vec_gather_load_cached to target_optabs.
10682 * optabs-tree.c (init_tree_optimization_optabs): Use
10683 ggc_cleared_alloc to allocate target_optabs.
10684 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10685 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10687 * internal-fn.h (internal_load_fn_p): Declare.
10688 (internal_gather_scatter_fn_p): Likewise.
10689 (internal_fn_mask_index): Likewise.
10690 (internal_gather_scatter_fn_supported_p): Likewise.
10691 * internal-fn.c (gather_load_direct): New macro.
10692 (expand_gather_load_optab_fn): New function.
10693 (direct_gather_load_optab_supported_p): New macro.
10694 (direct_internal_fn_optab): New function.
10695 (internal_load_fn_p): Likewise.
10696 (internal_gather_scatter_fn_p): Likewise.
10697 (internal_fn_mask_index): Likewise.
10698 (internal_gather_scatter_fn_supported_p): Likewise.
10699 * optabs-query.c (supports_at_least_one_mode_p): New function.
10700 (supports_vec_gather_load_p): Likewise.
10701 * optabs-query.h (supports_vec_gather_load_p): Declare.
10702 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10703 and memory_type field.
10704 (NUM_PATTERNS): Bump to 15.
10705 * tree-vect-data-refs.c: Include internal-fn.h.
10706 (vect_gather_scatter_fn_p): New function.
10707 (vect_describe_gather_scatter_call): Likewise.
10708 (vect_check_gather_scatter): Try using internal functions for
10709 gather loads. Recognize existing calls to a gather load function.
10710 (vect_analyze_data_refs): Consider using gather loads if
10711 supports_vec_gather_load_p.
10712 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10713 (vect_get_gather_scatter_offset_type): Likewise.
10714 (vect_convert_mask_for_vectype): Likewise.
10715 (vect_add_conversion_to_patterm): Likewise.
10716 (vect_try_gather_scatter_pattern): Likewise.
10717 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10718 (vect_vect_recog_func_ptrs): Add it.
10719 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10720 internal_fn_mask_index and internal_gather_scatter_fn_p.
10721 (check_load_store_masking): Take the gather_scatter_info as an
10722 argument and handle gather loads.
10723 (vect_get_gather_scatter_ops): New function.
10724 (vectorizable_call): Check internal_load_fn_p.
10725 (vectorizable_load): Likewise. Handle gather load internal
10727 (vectorizable_store): Update call to check_load_store_masking.
10728 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10729 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10730 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10731 (aarch64_gather_scale_operand_d): New predicates.
10732 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10733 (mask_gather_load<mode>): New insns.
10735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10736 Alan Hayward <alan.hayward@arm.com>
10737 David Sherwood <david.sherwood@arm.com>
10739 * optabs.def (fold_left_plus_optab): New optab.
10740 * doc/md.texi (fold_left_plus_@var{m}): Document.
10741 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10742 * internal-fn.c (fold_left_direct): Define.
10743 (expand_fold_left_optab_fn): Likewise.
10744 (direct_fold_left_optab_supported_p): Likewise.
10745 * fold-const-call.c (fold_const_fold_left): New function.
10746 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10747 * tree-parloops.c (valid_reduction_p): New function.
10748 (gather_scalar_reductions): Use it.
10749 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10750 (vect_finish_replace_stmt): Declare.
10751 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10752 (needs_fold_left_reduction_p): New function, split out from...
10753 (vect_is_simple_reduction): ...here. Accept reductions that
10754 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10755 (vect_force_simple_reduction): Also store the reduction type in
10756 the assignment's STMT_VINFO_REDUC_TYPE.
10757 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10758 (merge_with_identity): New function.
10759 (vect_expand_fold_left): Likewise.
10760 (vectorize_fold_left_reduction): Likewise.
10761 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10762 scalar phi in place for it. Check for target support and reject
10763 cases that would reassociate the operation. Defer the transform
10764 phase to vectorize_fold_left_reduction.
10765 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10766 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10767 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10771 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10772 call to ifc_temp_var.
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 * target.def (legitimize_address_displacement): Take the original
10779 offset as a poly_int.
10780 * targhooks.h (default_legitimize_address_displacement): Update
10782 * targhooks.c (default_legitimize_address_displacement): Likewise.
10783 * doc/tm.texi: Regenerate.
10784 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10785 as an argument, moving assert of ad->disp == ad->disp_term to...
10786 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10787 Try calling targetm.legitimize_address_displacement before expanding
10788 the address rather than afterwards, and adjust for the new interface.
10789 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10790 Match the new hook interface. Handle SVE addresses.
10791 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10792 new hook interface.
10794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10796 * Makefile.in (OBJS): Add early-remat.o.
10797 * target.def (select_early_remat_modes): New hook.
10798 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10799 * doc/tm.texi: Regenerate.
10800 * targhooks.h (default_select_early_remat_modes): Declare.
10801 * targhooks.c (default_select_early_remat_modes): New function.
10802 * timevar.def (TV_EARLY_REMAT): New timevar.
10803 * passes.def (pass_early_remat): New pass.
10804 * tree-pass.h (make_pass_early_remat): Declare.
10805 * early-remat.c: New file.
10806 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10808 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10815 vfm1 with a bound_epilog parameter.
10816 (vect_do_peeling): Update calls accordingly, and move the prologue
10817 call earlier in the function. Treat the base bound_epilog as 0 for
10818 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10819 this base when peeling for gaps.
10820 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10821 with fully-masked loops.
10822 (vect_estimate_min_profitable_iters): Handle the single peeled
10823 iteration in that case.
10825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10826 Alan Hayward <alan.hayward@arm.com>
10827 David Sherwood <david.sherwood@arm.com>
10829 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10830 single-element interleaving even if the size is not a power of 2.
10831 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10832 accesses for single-element interleaving if the group size is
10835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10836 Alan Hayward <alan.hayward@arm.com>
10837 David Sherwood <david.sherwood@arm.com>
10839 * doc/md.texi (fold_extract_last_@var{m}): Document.
10840 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10841 * optabs.def (fold_extract_last_optab): New optab.
10842 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10843 * internal-fn.c (fold_extract_direct): New macro.
10844 (expand_fold_extract_optab_fn): Likewise.
10845 (direct_fold_extract_optab_supported_p): Likewise.
10846 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10847 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10848 EXTRACT_LAST_REDUCTION.
10849 (get_initial_def_for_reduction): Do not create an initial vector
10850 for EXTRACT_LAST_REDUCTION reductions.
10851 (vectorizable_reduction): Leave the scalar phi in place for
10852 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10853 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10854 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10855 transform phase to vectorizable_condition.
10856 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10858 (vect_finish_stmt_generation): ...here.
10859 (vect_finish_replace_stmt): New function.
10860 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10861 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10863 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10866 Alan Hayward <alan.hayward@arm.com>
10867 David Sherwood <david.sherwood@arm.com>
10869 * doc/md.texi (extract_last_@var{m}): Document.
10870 * optabs.def (extract_last_optab): New optab.
10871 * internal-fn.def (EXTRACT_LAST): New internal function.
10872 * internal-fn.c (cond_unary_direct): New macro.
10873 (expand_cond_unary_optab_fn): Likewise.
10874 (direct_cond_unary_optab_supported_p): Likewise.
10875 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10876 loops using EXTRACT_LAST.
10877 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10878 (extract_last_<mode>): ...this optab.
10879 (vec_extract<mode><Vel>): Update accordingly.
10881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10882 Alan Hayward <alan.hayward@arm.com>
10883 David Sherwood <david.sherwood@arm.com>
10885 * target.def (empty_mask_is_expensive): New hook.
10886 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10887 * doc/tm.texi: Regenerate.
10888 * targhooks.h (default_empty_mask_is_expensive): Declare.
10889 * targhooks.c (default_empty_mask_is_expensive): New function.
10890 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10891 if the target says that empty masks are expensive.
10892 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10894 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10897 Alan Hayward <alan.hayward@arm.com>
10898 David Sherwood <david.sherwood@arm.com>
10900 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10901 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10902 (vect_use_loop_mask_for_alignment_p): New function.
10903 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10904 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10905 niters_skip argument. Make sure that the first niters_skip elements
10906 of the first iteration are inactive.
10907 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10908 Update call to vect_set_loop_masks_directly.
10909 (get_misalign_in_elems): New function, split out from...
10910 (vect_gen_prolog_loop_niters): ...here.
10911 (vect_update_init_of_dr): Take a code argument that specifies whether
10912 the adjustment should be added or subtracted.
10913 (vect_update_init_of_drs): Likewise.
10914 (vect_prepare_for_masked_peels): New function.
10915 (vect_do_peeling): Skip prologue peeling if we're using a mask
10916 instead. Update call to vect_update_inits_of_drs.
10917 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10919 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10920 alignment. Do not include the number of peeled iterations in
10921 the minimum threshold in that case.
10922 (vectorizable_induction): Adjust the start value down by
10923 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10924 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10925 Take the number of skipped iterations into account when calculating
10927 * tree-vect-stmts.c (vect_gen_while_not): New function.
10929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10930 Alan Hayward <alan.hayward@arm.com>
10931 David Sherwood <david.sherwood@arm.com>
10933 * doc/sourcebuild.texi (vect_fully_masked): Document.
10934 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10935 default value to 0.
10936 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10938 (vect_analyze_loop_2): ...here. Don't check the vectorization
10939 factor against the number of loop iterations if the loop is
10942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10947 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10948 (dump_groups): Update accordingly.
10949 (iv_use::mem_type): New member variable.
10950 (address_p): New function.
10951 (record_use): Add a mem_type argument and initialize the new
10953 (record_group_use): Add a mem_type argument. Use address_p.
10954 Remove obsolete null checks of base_object. Update call to record_use.
10955 (find_interesting_uses_op): Update call to record_group_use.
10956 (find_interesting_uses_cond): Likewise.
10957 (find_interesting_uses_address): Likewise.
10958 (get_mem_type_for_internal_fn): New function.
10959 (find_address_like_use): Likewise.
10960 (find_interesting_uses_stmt): Try find_address_like_use before
10961 calling find_interesting_uses_op.
10962 (addr_offset_valid_p): Use the iv mem_type field as the type
10963 of the addressed memory.
10964 (add_autoinc_candidates): Likewise.
10965 (get_address_cost): Likewise.
10966 (split_small_address_groups_p): Use address_p.
10967 (split_address_groups): Likewise.
10968 (add_iv_candidate_for_use): Likewise.
10969 (autoinc_possible_for_pair): Likewise.
10970 (rewrite_groups): Likewise.
10971 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10972 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10973 (get_alias_ptr_type_for_ptr_address): New function.
10974 (rewrite_use_address): Rewrite address uses in calls that were
10975 identified by find_address_like_use.
10977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10978 Alan Hayward <alan.hayward@arm.com>
10979 David Sherwood <david.sherwood@arm.com>
10981 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10983 * gimple-expr.h (is_gimple_addressable: Likewise.
10984 * gimple-expr.c (is_gimple_address): Likewise.
10985 * internal-fn.c (expand_call_mem_ref): New function.
10986 (expand_mask_load_optab_fn): Use it.
10987 (expand_mask_store_optab_fn): Likewise.
10989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10990 Alan Hayward <alan.hayward@arm.com>
10991 David Sherwood <david.sherwood@arm.com>
10993 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10994 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10995 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10996 (cond_umax@var{mode}): Document.
10997 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10998 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10999 (cond_umin_optab, cond_umax_optab): New optabs.
11000 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11001 (COND_IOR, COND_XOR): New internal functions.
11002 * internal-fn.h (get_conditional_internal_fn): Declare.
11003 * internal-fn.c (cond_binary_direct): New macro.
11004 (expand_cond_binary_optab_fn): Likewise.
11005 (direct_cond_binary_optab_supported_p): Likewise.
11006 (get_conditional_internal_fn): New function.
11007 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11008 Cope with reduction statements that are vectorized as calls rather
11010 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11011 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11012 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11013 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11014 (UNSPEC_COND_EOR): New unspecs.
11015 (optab): Add mappings for them.
11016 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11017 (sve_int_op, sve_fp_op): New int attributes.
11019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11020 Alan Hayward <alan.hayward@arm.com>
11021 David Sherwood <david.sherwood@arm.com>
11023 * optabs.def (while_ult_optab): New optab.
11024 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11025 * internal-fn.def (WHILE_ULT): New internal function.
11026 * internal-fn.h (direct_internal_fn_supported_p): New override
11027 that takes two types as argument.
11028 * internal-fn.c (while_direct): New macro.
11029 (expand_while_optab_fn): New function.
11030 (convert_optab_supported_p): Likewise.
11031 (direct_while_optab_supported_p): New macro.
11032 * wide-int.h (wi::udiv_ceil): New function.
11033 * tree-vectorizer.h (rgroup_masks): New structure.
11034 (vec_loop_masks): New typedef.
11035 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11036 and fully_masked_p.
11037 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11038 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11039 (vect_max_vf): New function.
11040 (slpeel_make_loop_iterate_ntimes): Delete.
11041 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11042 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11043 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11044 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11045 internal-fn.h, stor-layout.h and optabs-query.h.
11046 (vect_set_loop_mask): New function.
11047 (add_preheader_seq): Likewise.
11048 (add_header_seq): Likewise.
11049 (interleave_supported_p): Likewise.
11050 (vect_maybe_permute_loop_masks): Likewise.
11051 (vect_set_loop_masks_directly): Likewise.
11052 (vect_set_loop_condition_masked): Likewise.
11053 (vect_set_loop_condition_unmasked): New function, split out from
11054 slpeel_make_loop_iterate_ntimes.
11055 (slpeel_make_loop_iterate_ntimes): Rename to..
11056 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11057 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11058 (vect_do_peeling): Update call accordingly.
11059 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11061 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11062 mask_compare_type, can_fully_mask_p and fully_masked_p.
11063 (release_vec_loop_masks): New function.
11064 (_loop_vec_info): Use it to free the loop masks.
11065 (can_produce_all_loop_masks_p): New function.
11066 (vect_get_max_nscalars_per_iter): Likewise.
11067 (vect_verify_full_masking): Likewise.
11068 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11069 retries, and free the mask rgroups before retrying. Check loop-wide
11070 reasons for disallowing fully-masked loops. Make the final decision
11071 about whether use a fully-masked loop or not.
11072 (vect_estimate_min_profitable_iters): Do not assume that peeling
11073 for the number of iterations will be needed for fully-masked loops.
11074 (vectorizable_reduction): Disable fully-masked loops.
11075 (vectorizable_live_operation): Likewise.
11076 (vect_halve_mask_nunits): New function.
11077 (vect_double_mask_nunits): Likewise.
11078 (vect_record_loop_mask): Likewise.
11079 (vect_get_loop_mask): Likewise.
11080 (vect_transform_loop): Handle the case in which the final loop
11081 iteration might handle a partial vector. Call vect_set_loop_condition
11082 instead of slpeel_make_loop_iterate_ntimes.
11083 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11084 (check_load_store_masking): New function.
11085 (prepare_load_store_mask): Likewise.
11086 (vectorizable_store): Handle fully-masked loops.
11087 (vectorizable_load): Likewise.
11088 (supportable_widening_operation): Use vect_halve_mask_nunits for
11090 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11091 (vect_gen_while): New function.
11092 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11093 (aarch64_uqdec<mode>): New insn.
11095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11096 Alan Hayward <alan.hayward@arm.com>
11097 David Sherwood <david.sherwood@arm.com>
11099 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11100 (reduc_xor_scal_optab): New optabs.
11101 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11102 (reduc_xor_scal_@var{m}): Document.
11103 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11104 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11105 internal functions.
11106 * fold-const-call.c (fold_const_call): Handle them.
11107 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11108 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11109 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11110 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11111 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11112 (UNSPEC_XORV): New unspecs.
11113 (optab): Add entries for them.
11114 (BITWISEV): New int iterator.
11115 (bit_reduc_op): New int attributes.
11117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11118 Alan Hayward <alan.hayward@arm.com>
11119 David Sherwood <david.sherwood@arm.com>
11121 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11122 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11123 * optabs.def (vec_shl_insert_optab): New optab.
11124 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11125 (duplicate_and_interleave): Likewise.
11126 * tree-vect-loop.c: Include internal-fn.h.
11127 (neutral_op_for_slp_reduction): New function, split out from
11128 get_initial_defs_for_reduction.
11129 (get_initial_def_for_reduction): Handle option 2 for variable-length
11130 vectors by loading the neutral value into a vector and then shifting
11131 the initial value into element 0.
11132 (get_initial_defs_for_reduction): Replace the code argument with
11133 the neutral value calculated by neutral_op_for_slp_reduction.
11134 Use gimple_build_vector for constant-length vectors.
11135 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11136 but the first group_size elements have a neutral value.
11137 Use duplicate_and_interleave otherwise.
11138 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11139 Update call to get_initial_defs_for_reduction. Handle SLP
11140 reductions for variable-length vectors by creating one vector
11141 result for each scalar result, with the elements associated
11142 with other scalar results stubbed out with the neutral value.
11143 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11144 Require IFN_VEC_SHL_INSERT for double reductions on
11145 variable-length vectors, or SLP reductions that have
11146 a neutral value. Require can_duplicate_and_interleave_p
11147 support for variable-length unchained SLP reductions if there
11148 is no neutral value, such as for MIN/MAX reductions. Also require
11149 the number of vector elements to be a multiple of the number of
11150 SLP statements when doing variable-length unchained SLP reductions.
11151 Update call to vect_create_epilog_for_reduction.
11152 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11153 and remove initial values.
11154 (duplicate_and_interleave): Make public.
11155 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11156 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11159 Alan Hayward <alan.hayward@arm.com>
11160 David Sherwood <david.sherwood@arm.com>
11162 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11163 (can_duplicate_and_interleave_p): New function.
11164 (vect_get_and_check_slp_defs): Take the vector of statements
11165 rather than just the current one. Remove excess parentheses.
11166 Restriction rejectinon of vect_constant_def and vect_external_def
11167 for variable-length vectors to boolean types, or types for which
11168 can_duplicate_and_interleave_p is false.
11169 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11170 (duplicate_and_interleave): New function.
11171 (vect_get_constant_vectors): Use gimple_build_vector for
11172 constant-length vectors and suitable variable-length constant
11173 vectors. Use duplicate_and_interleave for other variable-length
11174 vectors. Don't defer the update when inserting new statements.
11176 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11177 Alan Hayward <alan.hayward@arm.com>
11178 David Sherwood <david.sherwood@arm.com>
11180 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11181 min_profitable_iters doesn't go negative.
11183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11184 Alan Hayward <alan.hayward@arm.com>
11185 David Sherwood <david.sherwood@arm.com>
11187 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11188 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11189 * optabs.def (vec_mask_load_lanes_optab): New optab.
11190 (vec_mask_store_lanes_optab): Likewise.
11191 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11192 (MASK_STORE_LANES): Likewise.
11193 * internal-fn.c (mask_load_lanes_direct): New macro.
11194 (mask_store_lanes_direct): Likewise.
11195 (expand_mask_load_optab_fn): Handle masked operations.
11196 (expand_mask_load_lanes_optab_fn): New macro.
11197 (expand_mask_store_optab_fn): Handle masked operations.
11198 (expand_mask_store_lanes_optab_fn): New macro.
11199 (direct_mask_load_lanes_optab_supported_p): Likewise.
11200 (direct_mask_store_lanes_optab_supported_p): Likewise.
11201 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11203 (vect_load_lanes_supported): Likewise.
11204 * tree-vect-data-refs.c (strip_conversion): New function.
11205 (can_group_stmts_p): Likewise.
11206 (vect_analyze_data_ref_accesses): Use it instead of checking
11207 for a pair of assignments.
11208 (vect_store_lanes_supported): Take a masked_p parameter.
11209 (vect_load_lanes_supported): Likewise.
11210 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11211 vect_store_lanes_supported and vect_load_lanes_supported.
11212 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11213 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11214 parameter. Don't allow gaps for masked accesses.
11215 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11216 and vect_load_lanes_supported.
11217 (get_load_store_type): Take a masked_p parameter and update
11218 call to get_group_load_store_type.
11219 (vectorizable_store): Update call to get_load_store_type.
11220 Handle IFN_MASK_STORE_LANES.
11221 (vectorizable_load): Update call to get_load_store_type.
11222 Handle IFN_MASK_LOAD_LANES.
11224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11225 Alan Hayward <alan.hayward@arm.com>
11226 David Sherwood <david.sherwood@arm.com>
11228 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11230 * config/aarch64/aarch64-protos.h
11231 (aarch64_sve_struct_memory_operand_p): Declare.
11232 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11233 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11234 (VPRED, vpred): Handle SVE structure modes.
11235 * config/aarch64/constraints.md (Utx): New constraint.
11236 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11237 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11238 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11239 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11240 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11241 structure modes. Split into pieces after RA.
11242 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11243 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11245 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11246 SVE structure modes.
11247 (aarch64_classify_address): Likewise.
11248 (sizetochar): Move earlier in file.
11249 (aarch64_print_operand): Handle SVE register lists.
11250 (aarch64_array_mode): New function.
11251 (aarch64_sve_struct_memory_operand_p): Likewise.
11252 (TARGET_ARRAY_MODE): Redefine.
11254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11255 Alan Hayward <alan.hayward@arm.com>
11256 David Sherwood <david.sherwood@arm.com>
11258 * target.def (array_mode): New target hook.
11259 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11260 * doc/tm.texi: Regenerate.
11261 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11262 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11263 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11264 targetm.array_mode.
11265 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11269 Alan Hayward <alan.hayward@arm.com>
11270 David Sherwood <david.sherwood@arm.com>
11272 * fold-const.c (fold_binary_loc): Check the argument types
11273 rather than the result type when testing for a vector operation.
11275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11277 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11278 * doc/tm.texi: Regenerate.
11280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11281 Alan Hayward <alan.hayward@arm.com>
11282 David Sherwood <david.sherwood@arm.com>
11284 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11285 (sve): Document new AArch64 extension.
11286 * doc/md.texi (w): Extend the description of the AArch64
11287 constraint to include SVE vectors.
11288 (Upl, Upa): Document new AArch64 predicate constraints.
11289 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11291 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11292 (msve-vector-bits=): New option.
11293 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11294 SVE when these are disabled.
11295 (sve): New extension.
11296 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11297 modes. Adjust their number of units based on aarch64_sve_vg.
11298 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11299 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11300 aarch64_addr_query_type.
11301 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11302 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11303 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11304 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11305 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11306 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11307 (aarch64_simd_imm_zero_p): Delete.
11308 (aarch64_check_zero_based_sve_index_immediate): Declare.
11309 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11310 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11311 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11312 (aarch64_sve_float_mul_immediate_p): Likewise.
11313 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11314 rather than an rtx.
11315 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11316 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11317 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11318 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11319 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11320 (aarch64_regmode_natural_size): Likewise.
11321 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11322 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11324 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11325 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11326 for VG and the SVE predicate registers.
11327 (V_ALIASES): Add a "z"-prefixed alias.
11328 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11329 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11330 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11331 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11332 (REG_CLASS_NAMES): Add entries for them.
11333 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11334 and the predicate registers.
11335 (aarch64_sve_vg): Declare.
11336 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11337 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11338 (REGMODE_NATURAL_SIZE): Define.
11339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11341 * config/aarch64/aarch64.c: Include cfgrtl.h.
11342 (simd_immediate_info): Add a constructor for series vectors,
11343 and an associated step field.
11344 (aarch64_sve_vg): New variable.
11345 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11346 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11347 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11348 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11349 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11350 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11351 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11352 (aarch64_get_mask_mode): New functions.
11353 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11354 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11355 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11356 predicate modes and predicate registers. Explicitly restrict
11357 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11358 to store a vector mode if it is recognized by
11359 aarch64_classify_vector_mode.
11360 (aarch64_regmode_natural_size): New function.
11361 (aarch64_hard_regno_caller_save_mode): Return the original mode
11363 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11364 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11365 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11366 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11368 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11369 does not overlap dest if the function is frame-related. Handle
11371 (aarch64_split_add_offset): New function.
11372 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11373 them aarch64_add_offset.
11374 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11375 and update call to aarch64_sub_sp.
11376 (aarch64_add_cfa_expression): New function.
11377 (aarch64_expand_prologue): Pass extra temporary registers to the
11378 functions above. Handle the case in which we need to emit new
11379 DW_CFA_expressions for registers that were originally saved
11380 relative to the stack pointer, but now have to be expressed
11381 relative to the frame pointer.
11382 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11384 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11385 IP0 and IP1 values for SVE frames.
11386 (aarch64_expand_vec_series): New function.
11387 (aarch64_expand_sve_widened_duplicate): Likewise.
11388 (aarch64_expand_sve_const_vector): Likewise.
11389 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11390 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11391 into the register, rather than emitting a SET directly.
11392 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11393 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11394 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11395 (offset_9bit_signed_scaled_p): New functions.
11396 (aarch64_replicate_bitmask_imm): New function.
11397 (aarch64_bitmask_imm): Use it.
11398 (aarch64_cannot_force_const_mem): Reject expressions involving
11399 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11400 (aarch64_classify_index): Handle SVE indices, by requiring
11401 a plain register index with a scale that matches the element size.
11402 (aarch64_classify_address): Handle SVE addresses. Assert that
11403 the mode of the address is VOIDmode or an integer mode.
11404 Update call to aarch64_classify_symbol.
11405 (aarch64_classify_symbolic_expression): Update call to
11406 aarch64_classify_symbol.
11407 (aarch64_const_vec_all_in_range_p): New function.
11408 (aarch64_print_vector_float_operand): Likewise.
11409 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11410 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11411 and the FP immediates 1.0 and 0.5.
11412 (aarch64_print_address_internal): Handle SVE addresses.
11413 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11414 (aarch64_regno_regclass): Handle predicate registers.
11415 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11417 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11418 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11419 (aarch64_convert_sve_vector_bits): New function.
11420 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11421 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11422 rather than an rtx.
11423 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11424 Handle SVE vector and predicate modes. Accept VL-based constants
11425 that need only one temporary register, and VL offsets that require
11426 no temporary registers.
11427 (aarch64_conditional_register_usage): Mark the predicate registers
11428 as fixed if SVE isn't available.
11429 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11430 Return true for SVE vector and predicate modes.
11431 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11432 rather than an unsigned int. Handle SVE modes.
11433 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11435 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11437 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11438 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11439 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11440 (aarch64_sve_float_mul_immediate_p): New functions.
11441 (aarch64_sve_valid_immediate): New function.
11442 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11443 Explicitly reject structure modes. Check for INDEX constants.
11444 Handle PTRUE and PFALSE constants.
11445 (aarch64_check_zero_based_sve_index_immediate): New function.
11446 (aarch64_simd_imm_zero_p): Delete.
11447 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11448 vector modes. Accept constants in the range of CNT[BHWD].
11449 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11450 ask for an Advanced SIMD mode.
11451 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11452 (aarch64_simd_vector_alignment): Handle SVE predicates.
11453 (aarch64_vectorize_preferred_vector_alignment): New function.
11454 (aarch64_simd_vector_alignment_reachable): Use it instead of
11456 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11457 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11459 (MAX_VECT_LEN): Delete.
11460 (expand_vec_perm_d): Add a vec_flags field.
11461 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11462 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11463 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11465 (aarch64_evpc_rev): Rename to...
11466 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11467 (aarch64_evpc_rev_global): New function.
11468 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11469 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11471 (aarch64_evpc_sve_tbl): New function.
11472 (aarch64_expand_vec_perm_const_1): Update after rename of
11473 aarch64_evpc_rev. Handle SVE permutes too, trying
11474 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11475 than aarch64_evpc_tbl.
11476 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11477 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11478 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11479 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11480 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11481 (aarch64_expand_sve_vcond): New functions.
11482 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11483 of aarch64_vector_mode_p.
11484 (aarch64_dwarf_poly_indeterminate_value): New function.
11485 (aarch64_compute_pressure_classes): Likewise.
11486 (aarch64_can_change_mode_class): Likewise.
11487 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11488 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11489 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11490 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11491 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11492 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11493 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11494 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11496 (Dn, Dl, Dr): Accept const as well as const_vector.
11497 (Dz): Likewise. Compare against CONST0_RTX.
11498 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11499 of "vector" where appropriate.
11500 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11501 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11502 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11503 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11504 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11505 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11506 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11507 (v_int_equiv): Extend to SVE modes.
11508 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11510 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11511 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11512 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11513 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11514 (SVE_COND_FP_CMP): New int iterators.
11515 (perm_hilo): Handle the new unpack unspecs.
11516 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11518 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11519 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11520 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11521 (aarch64_equality_operator, aarch64_constant_vector_operand)
11522 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11523 (aarch64_sve_nonimmediate_operand): Likewise.
11524 (aarch64_sve_general_operand): Likewise.
11525 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11526 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11527 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11528 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11529 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11530 (aarch64_sve_float_arith_immediate): Likewise.
11531 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11532 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11533 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11534 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11535 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11536 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11537 (aarch64_sve_float_arith_operand): Likewise.
11538 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11539 (aarch64_sve_float_mul_operand): Likewise.
11540 (aarch64_sve_vec_perm_operand): Likewise.
11541 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11542 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11543 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11544 as well as const_vector.
11545 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11546 in file. Use CONST0_RTX and CONSTM1_RTX.
11547 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11548 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11549 Use aarch64_simd_imm_zero.
11550 * config/aarch64/aarch64-sve.md: New file.
11551 * config/aarch64/aarch64.md: Include it.
11552 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11553 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11554 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11555 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11556 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11557 (sve): New attribute.
11558 (enabled): Disable instructions with the sve attribute unless
11560 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11561 aarch64_expand_mov_immediate.
11562 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11563 CNT[BHSD] immediates.
11564 (movti): Split CONST_POLY_INT moves into two halves.
11565 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11566 Split additions that need a temporary here if the destination
11567 is the stack pointer.
11568 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11569 (*add<mode>3_poly_1): New instruction.
11570 (set_clobber_cc): New expander.
11572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11574 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11575 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11576 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11577 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11578 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11579 Change innermode from fixed_mode_size to machine_mode.
11580 (simplify_subreg): Update call accordingly. Handle a constant-sized
11581 subreg of a variable-length CONST_VECTOR.
11583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11584 Alan Hayward <alan.hayward@arm.com>
11585 David Sherwood <david.sherwood@arm.com>
11587 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11588 (add_offset_to_base): New function, split out from...
11589 (create_mem_ref): ...here. When handling a scale other than 1,
11590 check first whether the address is valid without the offset.
11591 Add it into the base if so, leaving the index and scale as-is.
11593 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11596 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11597 fold_for_warn before checking if arg2 is INTEGER_CST.
11599 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11601 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11602 (store_multiple_operation): Delete.
11603 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11604 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11605 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11606 guarded by TARGET_STRING.
11607 (rs6000_output_load_multiple): Delete.
11608 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11609 OPTION_MASK_STRING / TARGET_STRING handling.
11610 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11611 (const rs6000_opt_masks) <"string">: Change mask to 0.
11612 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11613 (MASK_STRING): Delete.
11614 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11616 (load_multiple): Delete.
11623 (store_multiple): Delete.
11630 (movmemsi_8reg): Delete.
11631 (corresponding unnamed define_insn): Delete.
11632 (movmemsi_6reg): Delete.
11633 (corresponding unnamed define_insn): Delete.
11634 (movmemsi_4reg): Delete.
11635 (corresponding unnamed define_insn): Delete.
11636 (movmemsi_2reg): Delete.
11637 (corresponding unnamed define_insn): Delete.
11638 (movmemsi_1reg): Delete.
11639 (corresponding unnamed define_insn): Delete.
11640 * config/rs6000/rs6000.opt (mno-string): New.
11641 (mstring): Replace by deprecation warning stub.
11642 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11644 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11646 * regrename.c (regrename_do_replace): If replacing the same
11647 reg multiple times, try to reuse last created gen_raw_REG.
11650 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11651 main to workaround a bug in GDB.
11653 2018-01-12 Tom de Vries <tom@codesourcery.com>
11656 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11658 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11660 PR rtl-optimization/80481
11661 * ira-color.c (get_cap_member): New function.
11662 (allocnos_conflict_by_live_ranges_p): Use it.
11663 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11664 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11666 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11669 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11670 (*saddl_se_1): Ditto.
11671 (*ssubsi_1): Ditto.
11672 (*ssubl_se_1): Ditto.
11674 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11676 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11677 rather than wi::to_widest for DR_INITs.
11678 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11679 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11680 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11682 (vect_analyze_group_access_1): Note that here.
11684 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11686 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11687 polynomial type sizes.
11689 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11691 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11692 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11693 (gimple_add_tmp_var): Likewise.
11695 2018-01-12 Martin Liska <mliska@suse.cz>
11697 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11698 (gimple_alloc_sizes): Likewise.
11699 (dump_gimple_statistics): Use PRIu64 in printf format.
11700 * gimple.h: Change uint64_t to int.
11702 2018-01-12 Martin Liska <mliska@suse.cz>
11704 * tree-core.h: Use uint64_t instead of int.
11705 * tree.c (tree_node_counts): Likewise.
11706 (tree_node_sizes): Likewise.
11707 (dump_tree_statistics): Use PRIu64 in printf format.
11709 2018-01-12 Martin Liska <mliska@suse.cz>
11711 * Makefile.in: As qsort_chk is implemented in vec.c, add
11712 vec.o to linkage of gencfn-macros.
11713 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11714 passing the info to record_node_allocation_statistics.
11715 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11717 * ggc-common.c (struct ggc_usage): Add operator== and use
11718 it in operator< and compare function.
11719 * mem-stats.h (struct mem_usage): Likewise.
11720 * vec.c (struct vec_usage): Remove operator< and compare
11721 function. Can be simply inherited.
11723 2018-01-12 Martin Jambor <mjambor@suse.cz>
11726 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11727 * tree-ssa-math-opts.c: Include domwalk.h.
11728 (convert_mult_to_fma_1): New function.
11729 (fma_transformation_info): New type.
11730 (fma_deferring_state): Likewise.
11731 (cancel_fma_deferring): New function.
11732 (result_of_phi): Likewise.
11733 (last_fma_candidate_feeds_initial_phi): Likewise.
11734 (convert_mult_to_fma): Added deferring logic, split actual
11735 transformation to convert_mult_to_fma_1.
11736 (math_opts_dom_walker): New type.
11737 (math_opts_dom_walker::after_dom_children): New method, body moved
11738 here from pass_optimize_widening_mul::execute, added deferring logic
11740 (pass_optimize_widening_mul::execute): Moved most of code to
11741 math_opts_dom_walker::after_dom_children.
11742 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11743 * config/i386/i386.c (ix86_option_override_internal): Added
11744 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11746 2018-01-12 Richard Biener <rguenther@suse.de>
11749 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11750 inline instance vars.
11752 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11755 * config/rx/rx.c (rx_is_restricted_memory_address):
11756 Handle SUBREG case.
11758 2018-01-12 Richard Biener <rguenther@suse.de>
11760 PR tree-optimization/80846
11761 * target.def (split_reduction): New target hook.
11762 * targhooks.c (default_split_reduction): New function.
11763 * targhooks.h (default_split_reduction): Declare.
11764 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11765 target requests first reduce vectors by combining low and high
11767 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11768 (get_vectype_for_scalar_type_and_size): Export.
11769 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11770 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11771 * doc/tm.texi: Regenerate.
11772 * config/i386/i386.c (ix86_split_reduction): Implement
11773 TARGET_VECTORIZE_SPLIT_REDUCTION.
11775 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11778 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11779 in PIC mode except for TARGET_VXWORKS_RTP.
11780 * config/sparc/sparc.c: Include cfgrtl.h.
11781 (TARGET_INIT_PIC_REG): Define.
11782 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11783 (sparc_pic_register_p): New predicate.
11784 (sparc_legitimate_address_p): Use it.
11785 (sparc_legitimize_pic_address): Likewise.
11786 (sparc_delegitimize_address): Likewise.
11787 (sparc_mode_dependent_address_p): Likewise.
11788 (gen_load_pcrel_sym): Remove 4th parameter.
11789 (load_got_register): Adjust call to above. Remove obsolete stuff.
11790 (sparc_expand_prologue): Do not call load_got_register here.
11791 (sparc_flat_expand_prologue): Likewise.
11792 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11793 (sparc_use_pseudo_pic_reg): New function.
11794 (sparc_init_pic_reg): Likewise.
11795 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11796 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11798 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11800 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11801 Add item for branch_cost.
11803 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11805 PR rtl-optimization/83565
11806 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11807 not extend the result to a larger mode for rotate operations.
11808 (num_sign_bit_copies1): Likewise.
11810 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11813 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11815 Use values-Xc.o for -pedantic.
11816 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11818 2018-01-12 Martin Liska <mliska@suse.cz>
11821 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11823 (possible_polymorphic_call_targets): Use it.
11824 (ipa_devirt): Likewise.
11826 2018-01-12 Martin Liska <mliska@suse.cz>
11828 * profile-count.h (enum profile_quality): Use 0 as invalid
11829 enum value of profile_quality.
11831 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11833 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11834 -mext-string options.
11836 2018-01-12 Richard Biener <rguenther@suse.de>
11838 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11839 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11840 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11842 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11844 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11846 * configure.ac (--with-long-double-format): Add support for the
11847 configuration option to change the default long double format on
11849 * config.gcc (powerpc*-linux*-*): Likewise.
11850 * configure: Regenerate.
11851 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11852 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11853 used without modification.
11855 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11857 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11858 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11859 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11860 MISC_BUILTIN_SPEC_BARRIER.
11861 (rs6000_init_builtins): Likewise.
11862 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11864 (speculation_barrier): New define_insn.
11865 * doc/extend.texi: Document __builtin_speculation_barrier.
11867 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11870 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11871 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11872 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11874 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11875 integral modes instead of "ss" and "sd".
11876 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11877 vectors with 32-bit and 64-bit elements.
11878 (vecdupssescalarmodesuffix): New mode attribute.
11879 (vec_dup<mode>): Use it.
11881 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11884 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11885 frame if argument is passed on stack.
11887 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11890 * ree.c (combine_reaching_defs): Optimize also
11891 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11892 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11894 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11896 PR middle-end/83189
11897 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11899 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11901 PR middle-end/83718
11902 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11903 after they are computed.
11905 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11907 PR tree-optimization/83695
11908 * gimple-loop-linterchange.cc
11909 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11910 reset cached scev information after interchange.
11911 (pass_linterchange::execute): Remove call to scev_reset_htab.
11913 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11915 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11916 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11917 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11918 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11919 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11920 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11921 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11922 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11923 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11924 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11925 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11926 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11927 (V_lane_reg): Likewise.
11928 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11930 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11931 (vfmal_lane_low<mode>_intrinsic,
11932 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11933 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11934 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11935 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11936 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11937 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11939 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11941 * config/arm/arm-cpus.in (fp16fml): New feature.
11942 (ALL_SIMD): Add fp16fml.
11943 (armv8.2-a): Add fp16fml as an option.
11944 (armv8.3-a): Likewise.
11945 (armv8.4-a): Add fp16fml as part of fp16.
11946 * config/arm/arm.h (TARGET_FP16FML): Define.
11947 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11949 * config/arm/arm-modes.def (V2HF): Define.
11950 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11951 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11952 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11953 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11954 vfmsl_low, vfmsl_high): New set of builtins.
11955 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11956 (vfml_op): New code attribute.
11957 (VFMLHALVES): New int iterator.
11958 (VFML, VFMLSEL): New mode attributes.
11959 (V_reg): Define mapping for V2HF.
11960 (V_hi, V_lo): New mode attributes.
11961 (VF_constraint): Likewise.
11962 (vfml_half, vfml_half_selector): New int attributes.
11963 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11965 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11966 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11968 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11969 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11970 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11971 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11973 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11974 Document new effective target and option set.
11976 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11978 * config/arm/arm-cpus.in (armv8_4): New feature.
11979 (ARMv8_4a): New fgroup.
11980 (armv8.4-a): New arch.
11981 * config/arm/arm-tables.opt: Regenerate.
11982 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11983 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11984 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11985 Add matching rules for -march=armv8.4-a and extensions.
11986 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11988 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11991 * config/rx/rx.md (BW): New mode attribute.
11992 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11994 2018-01-11 Richard Biener <rguenther@suse.de>
11996 PR tree-optimization/83435
11997 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11998 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11999 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12001 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12002 Alan Hayward <alan.hayward@arm.com>
12003 David Sherwood <david.sherwood@arm.com>
12005 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12007 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12008 (aarch64_print_address_internal): Use it to check for a zero offset.
12010 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12011 Alan Hayward <alan.hayward@arm.com>
12012 David Sherwood <david.sherwood@arm.com>
12014 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12015 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12016 Return a poly_int64 rather than a HOST_WIDE_INT.
12017 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12018 rather than a HOST_WIDE_INT.
12019 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12020 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12021 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12022 final_offset from HOST_WIDE_INT to poly_int64.
12023 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12024 to_constant when getting the number of units in an Advanced SIMD
12026 (aarch64_builtin_vectorized_function): Check for a constant number
12028 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12030 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12031 attribute instead of GET_MODE_NUNITS.
12032 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12033 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12034 GET_MODE_SIZE for fixed-size registers.
12035 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12036 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12037 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12038 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12039 (aarch64_print_operand, aarch64_print_address_internal)
12040 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12041 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12042 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12043 Handle polynomial GET_MODE_SIZE.
12044 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12045 wider than SImode without modification.
12046 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12047 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12048 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12049 passing and returning SVE modes.
12050 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12051 rather than GEN_INT.
12052 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12053 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12054 (aarch64_allocate_and_probe_stack_space): Likewise.
12055 (aarch64_layout_frame): Cope with polynomial offsets.
12056 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12057 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12058 polynomial offsets.
12059 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12060 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12061 poly_int64 rather than a HOST_WIDE_INT.
12062 (aarch64_get_separate_components, aarch64_process_components)
12063 (aarch64_expand_prologue, aarch64_expand_epilogue)
12064 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12065 (aarch64_anchor_offset): New function, split out from...
12066 (aarch64_legitimize_address): ...here.
12067 (aarch64_builtin_vectorization_cost): Handle polynomial
12068 TYPE_VECTOR_SUBPARTS.
12069 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12071 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12072 number of elements from the PARALLEL rather than the mode.
12073 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12074 rather than GET_MODE_BITSIZE.
12075 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12076 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12077 (aarch64_expand_vec_perm_const_1): Handle polynomial
12078 d->perm.length () and d->perm elements.
12079 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12080 Apply to_constant to d->perm elements.
12081 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12082 polynomial CONST_VECTOR_NUNITS.
12083 (aarch64_move_pointer): Take amount as a poly_int64 rather
12085 (aarch64_progress_pointer): Avoid temporary variable.
12086 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12087 the mode attribute instead of GET_MODE.
12089 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12094 x exists before using it.
12095 (aarch64_add_constant_internal): Rename to...
12096 (aarch64_add_offset_1): ...this. Replace regnum with separate
12097 src and dest rtxes. Handle the case in which they're different,
12098 including when the offset is zero. Replace scratchreg with an rtx.
12099 Use 2 additions if there is no spare register into which we can
12100 move a 16-bit constant.
12101 (aarch64_add_constant): Delete.
12102 (aarch64_add_offset): Replace reg with separate src and dest
12103 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12104 Use aarch64_add_offset_1.
12105 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12106 an rtx rather than an int. Take the delta as a poly_int64
12107 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12108 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12109 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12110 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12111 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12112 and aarch64_add_sp.
12113 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12114 aarch64_add_constant.
12116 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12118 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12119 Use scalar_float_mode.
12121 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12123 * config/aarch64/aarch64-simd.md
12124 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12125 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12126 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12127 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12128 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12129 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12130 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12131 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12132 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12133 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12135 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12138 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12139 targ_options->x_arm_arch_string is non NULL.
12141 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12143 * config/aarch64/aarch64.h
12144 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12146 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12149 * expmed.c (emit_store_flag_force): Swap if const op0
12150 and change VOIDmode to mode of op0.
12152 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12154 PR rtl-optimization/83761
12155 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12156 than bytes to mode_for_size.
12158 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12160 PR middle-end/83189
12161 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12162 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12165 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12167 PR middle-end/83575
12168 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12169 when in layout mode.
12170 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12171 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12174 2018-01-10 Michael Collison <michael.collison@arm.com>
12176 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12177 * config/aarch64/aarch64-option-extension.def: Add
12178 AARCH64_OPT_EXTENSION of 'fp16fml'.
12179 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12180 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12181 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12182 * config/aarch64/constraints.md (Ui7): New constraint.
12183 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12184 (VFMLA_SEL_W): Ditto.
12187 (VFMLA16_LOW): New int iterator.
12188 (VFMLA16_HIGH): Ditto.
12189 (UNSPEC_FMLAL): New unspec.
12190 (UNSPEC_FMLSL): Ditto.
12191 (UNSPEC_FMLAL2): Ditto.
12192 (UNSPEC_FMLSL2): Ditto.
12193 (f16mac): New code attribute.
12194 * config/aarch64/aarch64-simd-builtins.def
12195 (aarch64_fmlal_lowv2sf): Ditto.
12196 (aarch64_fmlsl_lowv2sf): Ditto.
12197 (aarch64_fmlalq_lowv4sf): Ditto.
12198 (aarch64_fmlslq_lowv4sf): Ditto.
12199 (aarch64_fmlal_highv2sf): Ditto.
12200 (aarch64_fmlsl_highv2sf): Ditto.
12201 (aarch64_fmlalq_highv4sf): Ditto.
12202 (aarch64_fmlslq_highv4sf): Ditto.
12203 (aarch64_fmlal_lane_lowv2sf): Ditto.
12204 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12205 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12206 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12207 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12208 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12209 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12210 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12211 (aarch64_fmlal_lane_highv2sf): Ditto.
12212 (aarch64_fmlsl_lane_highv2sf): Ditto.
12213 (aarch64_fmlal_laneq_highv2sf): Ditto.
12214 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12215 (aarch64_fmlalq_lane_highv4sf): Ditto.
12216 (aarch64_fmlsl_lane_highv4sf): Ditto.
12217 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12218 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12219 * config/aarch64/aarch64-simd.md:
12220 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12221 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12222 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12223 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12224 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12225 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12226 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12227 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12228 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12229 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12230 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12231 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12232 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12233 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12234 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12235 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12236 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12237 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12238 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12239 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12240 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12241 (vfmlsl_low_u32): Ditto.
12242 (vfmlalq_low_u32): Ditto.
12243 (vfmlslq_low_u32): Ditto.
12244 (vfmlal_high_u32): Ditto.
12245 (vfmlsl_high_u32): Ditto.
12246 (vfmlalq_high_u32): Ditto.
12247 (vfmlslq_high_u32): Ditto.
12248 (vfmlal_lane_low_u32): Ditto.
12249 (vfmlsl_lane_low_u32): Ditto.
12250 (vfmlal_laneq_low_u32): Ditto.
12251 (vfmlsl_laneq_low_u32): Ditto.
12252 (vfmlalq_lane_low_u32): Ditto.
12253 (vfmlslq_lane_low_u32): Ditto.
12254 (vfmlalq_laneq_low_u32): Ditto.
12255 (vfmlslq_laneq_low_u32): Ditto.
12256 (vfmlal_lane_high_u32): Ditto.
12257 (vfmlsl_lane_high_u32): Ditto.
12258 (vfmlal_laneq_high_u32): Ditto.
12259 (vfmlsl_laneq_high_u32): Ditto.
12260 (vfmlalq_lane_high_u32): Ditto.
12261 (vfmlslq_lane_high_u32): Ditto.
12262 (vfmlalq_laneq_high_u32): Ditto.
12263 (vfmlslq_laneq_high_u32): Ditto.
12264 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12265 (AARCH64_FL_FOR_ARCH8_4): New.
12266 (AARCH64_ISA_F16FML): New ISA flag.
12267 (TARGET_F16FML): New feature flag for fp16fml.
12268 (doc/invoke.texi): Document new fp16fml option.
12270 2018-01-10 Michael Collison <michael.collison@arm.com>
12272 * config/aarch64/aarch64-builtins.c:
12273 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12274 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12275 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12276 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12277 (AARCH64_ISA_SHA3): New ISA flag.
12278 (TARGET_SHA3): New feature flag for sha3.
12279 * config/aarch64/iterators.md (sha512_op): New int attribute.
12280 (CRYPTO_SHA512): New int iterator.
12281 (UNSPEC_SHA512H): New unspec.
12282 (UNSPEC_SHA512H2): Ditto.
12283 (UNSPEC_SHA512SU0): Ditto.
12284 (UNSPEC_SHA512SU1): Ditto.
12285 * config/aarch64/aarch64-simd-builtins.def
12286 (aarch64_crypto_sha512hqv2di): New builtin.
12287 (aarch64_crypto_sha512h2qv2di): Ditto.
12288 (aarch64_crypto_sha512su0qv2di): Ditto.
12289 (aarch64_crypto_sha512su1qv2di): Ditto.
12290 (aarch64_eor3qv8hi): Ditto.
12291 (aarch64_rax1qv2di): Ditto.
12292 (aarch64_xarqv2di): Ditto.
12293 (aarch64_bcaxqv8hi): Ditto.
12294 * config/aarch64/aarch64-simd.md:
12295 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12296 (aarch64_crypto_sha512su0qv2di): Ditto.
12297 (aarch64_crypto_sha512su1qv2di): Ditto.
12298 (aarch64_eor3qv8hi): Ditto.
12299 (aarch64_rax1qv2di): Ditto.
12300 (aarch64_xarqv2di): Ditto.
12301 (aarch64_bcaxqv8hi): Ditto.
12302 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12303 (vsha512h2q_u64): Ditto.
12304 (vsha512su0q_u64): Ditto.
12305 (vsha512su1q_u64): Ditto.
12306 (veor3q_u16): Ditto.
12307 (vrax1q_u64): Ditto.
12308 (vxarq_u64): Ditto.
12309 (vbcaxq_u16): Ditto.
12310 * config/arm/types.md (crypto_sha512): New type attribute.
12311 (crypto_sha3): Ditto.
12312 (doc/invoke.texi): Document new sha3 option.
12314 2018-01-10 Michael Collison <michael.collison@arm.com>
12316 * config/aarch64/aarch64-builtins.c:
12317 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12318 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12319 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12320 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12321 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12322 (AARCH64_ISA_SM4): New ISA flag.
12323 (TARGET_SM4): New feature flag for sm4.
12324 * config/aarch64/aarch64-simd-builtins.def
12325 (aarch64_sm3ss1qv4si): Ditto.
12326 (aarch64_sm3tt1aq4si): Ditto.
12327 (aarch64_sm3tt1bq4si): Ditto.
12328 (aarch64_sm3tt2aq4si): Ditto.
12329 (aarch64_sm3tt2bq4si): Ditto.
12330 (aarch64_sm3partw1qv4si): Ditto.
12331 (aarch64_sm3partw2qv4si): Ditto.
12332 (aarch64_sm4eqv4si): Ditto.
12333 (aarch64_sm4ekeyqv4si): Ditto.
12334 * config/aarch64/aarch64-simd.md:
12335 (aarch64_sm3ss1qv4si): Ditto.
12336 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12337 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12338 (aarch64_sm4eqv4si): Ditto.
12339 (aarch64_sm4ekeyqv4si): Ditto.
12340 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12341 (sm3part_op): Ditto.
12342 (CRYPTO_SM3TT): Ditto.
12343 (CRYPTO_SM3PART): Ditto.
12344 (UNSPEC_SM3SS1): New unspec.
12345 (UNSPEC_SM3TT1A): Ditto.
12346 (UNSPEC_SM3TT1B): Ditto.
12347 (UNSPEC_SM3TT2A): Ditto.
12348 (UNSPEC_SM3TT2B): Ditto.
12349 (UNSPEC_SM3PARTW1): Ditto.
12350 (UNSPEC_SM3PARTW2): Ditto.
12351 (UNSPEC_SM4E): Ditto.
12352 (UNSPEC_SM4EKEY): Ditto.
12353 * config/aarch64/constraints.md (Ui2): New constraint.
12354 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12355 * config/arm/types.md (crypto_sm3): New type attribute.
12356 (crypto_sm4): Ditto.
12357 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12358 (vsm3tt1aq_u32): Ditto.
12359 (vsm3tt1bq_u32): Ditto.
12360 (vsm3tt2aq_u32): Ditto.
12361 (vsm3tt2bq_u32): Ditto.
12362 (vsm3partw1q_u32): Ditto.
12363 (vsm3partw2q_u32): Ditto.
12364 (vsm4eq_u32): Ditto.
12365 (vsm4ekeyq_u32): Ditto.
12366 (doc/invoke.texi): Document new sm4 option.
12368 2018-01-10 Michael Collison <michael.collison@arm.com>
12370 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12371 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12372 (AARCH64_FL_FOR_ARCH8_4): New.
12373 (AARCH64_FL_V8_4): New flag.
12374 (doc/invoke.texi): Document new armv8.4-a option.
12376 2018-01-10 Michael Collison <michael.collison@arm.com>
12378 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12379 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12380 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12381 * config/aarch64/aarch64-option-extension.def: Add
12382 AARCH64_OPT_EXTENSION of 'sha2'.
12383 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12384 (crypto): Disable sha2 and aes if crypto disabled.
12385 (crypto): Enable aes and sha2 if enabled.
12386 (simd): Disable sha2 and aes if simd disabled.
12387 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12389 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12390 (TARGET_SHA2): New feature flag for sha2.
12391 (TARGET_AES): New feature flag for aes.
12392 * config/aarch64/aarch64-simd.md:
12393 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12394 conditional on TARGET_AES.
12395 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12396 (aarch64_crypto_sha1hsi): Make pattern conditional
12398 (aarch64_crypto_sha1hv4si): Ditto.
12399 (aarch64_be_crypto_sha1hv4si): Ditto.
12400 (aarch64_crypto_sha1su1v4si): Ditto.
12401 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12402 (aarch64_crypto_sha1su0v4si): Ditto.
12403 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12404 (aarch64_crypto_sha256su0v4si): Ditto.
12405 (aarch64_crypto_sha256su1v4si): Ditto.
12406 (doc/invoke.texi): Document new aes and sha2 options.
12408 2018-01-10 Martin Sebor <msebor@redhat.com>
12410 PR tree-optimization/83781
12411 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12414 2018-01-11 Martin Sebor <msebor@gmail.com>
12415 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12417 PR tree-optimization/83501
12418 PR tree-optimization/81703
12420 * tree-ssa-strlen.c (get_string_cst): Rename...
12421 (get_string_len): ...to this. Handle global constants.
12422 (handle_char_store): Adjust.
12424 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12425 Jim Wilson <jimw@sifive.com>
12427 * config/riscv/riscv-protos.h (riscv_output_return): New.
12428 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12429 (riscv_attribute_table, riscv_output_return),
12430 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12431 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12432 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12434 (riscv_expand_prologue): Add early return for naked function.
12435 (riscv_expand_epilogue): Likewise.
12436 (riscv_function_ok_for_sibcall): Return false for naked function.
12437 (riscv_set_current_function): New.
12438 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12439 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12440 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12441 * doc/extend.texi (RISC-V Function Attributes): New.
12443 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12445 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12446 check for 128-bit long double before checking TCmode.
12447 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12448 128-bit long doubles before checking TFmode or TCmode.
12449 (FLOAT128_IBM_P): Likewise.
12451 2018-01-10 Martin Sebor <msebor@redhat.com>
12453 PR tree-optimization/83671
12454 * builtins.c (c_strlen): Unconditionally return zero for the empty
12456 Use -Warray-bounds for warnings.
12457 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12458 for non-constant array indices with COMPONENT_REF, arrays of
12459 arrays, and pointers to arrays.
12460 (gimple_fold_builtin_strlen): Determine and set length range for
12461 non-constant character arrays.
12463 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12465 PR middle-end/81897
12466 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12469 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12471 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12473 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12476 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12477 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12478 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12479 indexed_or_indirect_operand predicate.
12480 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12481 (*vsx_le_perm_load_v8hi): Likewise.
12482 (*vsx_le_perm_load_v16qi): Likewise.
12483 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12484 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12485 (*vsx_le_perm_store_v8hi): Likewise.
12486 (*vsx_le_perm_store_v16qi): Likewise.
12487 (eight unnamed splitters): Likewise.
12489 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12491 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12492 * config/rs6000/emmintrin.h: Likewise.
12493 * config/rs6000/mmintrin.h: Likewise.
12494 * config/rs6000/xmmintrin.h: Likewise.
12496 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12499 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12501 * tree.c (tree_nop_conversion): Return true for location wrapper
12503 (maybe_wrap_with_location): New function.
12504 (selftest::check_strip_nops): New function.
12505 (selftest::test_location_wrappers): New function.
12506 (selftest::tree_c_tests): Call it.
12507 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12508 (maybe_wrap_with_location): New decl.
12509 (EXPR_LOCATION_WRAPPER_P): New macro.
12510 (location_wrapper_p): New inline function.
12511 (tree_strip_any_location_wrapper): New inline function.
12513 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12516 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12517 stack_realign_offset for the largest alignment of stack slot
12519 (ix86_find_max_used_stack_alignment): New function.
12520 (ix86_finalize_stack_frame_flags): Use it. Set
12521 max_used_stack_alignment if we don't realign stack.
12522 * config/i386/i386.h (machine_function): Add
12523 max_used_stack_alignment.
12525 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12527 * config/arm/arm.opt (-mbranch-cost): New option.
12528 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12531 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12534 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12535 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12537 2018-01-10 Richard Biener <rguenther@suse.de>
12540 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12541 early out so it also covers the case where we have a non-NULL
12544 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12546 PR tree-optimization/83753
12547 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12548 for non-strided grouped accesses if the number of elements is 1.
12550 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12553 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12554 * i386.h (TARGET_USE_GATHER): Define.
12555 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12557 2018-01-10 Martin Liska <mliska@suse.cz>
12560 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12561 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12563 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12564 CLEANUP_NO_PARTITIONING is not set.
12566 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12568 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12569 for vectors, as a partial revert of r254296.
12570 * rtl.h (const_vec_p): Delete.
12571 (const_vec_duplicate_p): Don't test for vector CONSTs.
12572 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12573 * expmed.c (make_tree): Likewise.
12576 * common.md (E, F): Use CONSTANT_P instead of checking for
12578 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12579 checking for CONST_VECTOR.
12581 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12583 PR middle-end/83575
12584 * predict.c (force_edge_cold): Handle in more sane way edges
12585 with no prediction.
12587 2018-01-09 Carl Love <cel@us.ibm.com>
12589 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12591 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12592 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12593 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12594 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12595 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12596 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12597 * config/rs6000/rs6000-protos.h: Add extern defition for
12598 rs6000_generate_float2_double_code.
12599 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12601 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12602 (float2_v2df): Add define_expand.
12604 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12607 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12608 op_mode in the force_to_mode call.
12610 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12612 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12613 instead of checking each element individually.
12614 (aarch64_evpc_uzp): Likewise.
12615 (aarch64_evpc_zip): Likewise.
12616 (aarch64_evpc_ext): Likewise.
12617 (aarch64_evpc_rev): Likewise.
12618 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12619 instead of checking each element individually. Return true without
12621 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12622 whether all selected elements come from the same input, instead of
12623 checking each element individually. Remove calls to gen_rtx_REG,
12624 start_sequence and end_sequence and instead assert that no rtl is
12627 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12629 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12630 order of HIGH and CONST checks.
12632 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12634 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12635 if the destination isn't an SSA_NAME.
12637 2018-01-09 Richard Biener <rguenther@suse.de>
12639 PR tree-optimization/83668
12640 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12642 (canonicalize_loop_form): ... here, renamed from ...
12643 (canonicalize_loop_closed_ssa_form): ... this and amended to
12644 swap successor edges for loop exit blocks to make us use
12645 the RPO order we need for initial schedule generation.
12647 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12649 PR tree-optimization/64811
12650 * match.pd: When optimizing comparisons with Inf, avoid
12651 introducing or losing exceptions from comparisons with NaN.
12653 2018-01-09 Martin Liska <mliska@suse.cz>
12656 * asan.c (shadow_mem_size): Add gcc_assert.
12658 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12660 Don't save registers in main().
12663 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12664 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12665 * config/avr/avr.c (avr_set_current_function): Don't error if
12666 naked, OS_task or OS_main are specified at the same time.
12667 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12669 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12671 * common/config/avr/avr-common.c (avr_option_optimization_table):
12672 Switch on -mmain-is-OS_task for optimizing compilations.
12674 2018-01-09 Richard Biener <rguenther@suse.de>
12676 PR tree-optimization/83572
12677 * graphite.c: Include cfganal.h.
12678 (graphite_transform_loops): Connect infinite loops to exit
12679 and remove fake edges at the end.
12681 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12683 * ipa-inline.c (edge_badness): Revert accidental checkin.
12685 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12688 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12689 symbols; not inline clones.
12691 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12694 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12695 hard registers. Formatting fixes.
12697 PR preprocessor/83722
12698 * gcc.c (try_generate_repro): Pass
12699 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12700 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12703 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12704 Kito Cheng <kito.cheng@gmail.com>
12706 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12707 (riscv_leaf_function_p): Delete.
12708 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12710 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12712 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12714 (do_ifelse): New function.
12715 (do_isel): New function.
12716 (do_sub3): New function.
12717 (do_add3): New function.
12718 (do_load_mask_compare): New function.
12719 (do_overlap_load_compare): New function.
12720 (expand_compare_loop): New function.
12721 (expand_block_compare): Call expand_compare_loop() when appropriate.
12722 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12723 option description.
12724 (-mblock-compare-inline-loop-limit): New option.
12726 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12729 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12730 Reverse order of second and third operands in first alternative.
12731 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12732 of first and second elements in UNSPEC_VPERMR vector.
12733 (altivec_expand_vec_perm_le): Likewise.
12735 2018-01-08 Jeff Law <law@redhat.com>
12737 PR rtl-optimizatin/81308
12738 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12739 (process_switch): If group_case_labels makes a change, then set
12741 (pass_convert_switch::execute): If a switch is converted, then
12742 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12744 PR rtl-optimization/81308
12745 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12748 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12750 PR target/83663 - Revert r255946
12751 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12752 generation for cases where splatting a value is not useful.
12753 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12754 across a vec_duplicate and a paradoxical subreg forming a vector
12755 mode to a vec_concat.
12757 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12759 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12760 -march=armv8.3-a variants.
12761 * config/arm/t-multilib: Likewise.
12762 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12764 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12766 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12768 (cceq_ior_compare_complement): Give it a name so I can use it, and
12769 change boolean_or_operator predicate to boolean_operator so it can
12770 be used to generate a crand.
12771 (eqne): New code iterator.
12772 (bd/bd_neg): New code_attrs.
12773 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12774 a single define_insn.
12775 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12776 decrement (bdnzt/bdnzf/bdzt/bdzf).
12777 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12778 with the new names of the branch decrement patterns, and added the
12779 names of the branch decrement conditional patterns.
12781 2018-01-08 Richard Biener <rguenther@suse.de>
12783 PR tree-optimization/83563
12784 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12787 2018-01-08 Richard Biener <rguenther@suse.de>
12789 PR middle-end/83713
12790 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12792 2018-01-08 Richard Biener <rguenther@suse.de>
12794 PR tree-optimization/83685
12795 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12796 references to abnormals.
12798 2018-01-08 Richard Biener <rguenther@suse.de>
12801 * dwarf2out.c (output_indirect_strings): Handle empty
12802 skeleton_debug_str_hash.
12803 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12805 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12807 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12808 (emit_store_direct): Likewise.
12809 (arc_trampoline_adjust_address): Likewise.
12810 (arc_asm_trampoline_template): New function.
12811 (arc_initialize_trampoline): Use asm_trampoline_template.
12812 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12813 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12814 * config/arc/arc.md (flush_icache): Delete pattern.
12816 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12818 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12819 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12822 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12825 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12826 by not USED_FOR_TARGET.
12827 (make_pass_resolve_sw_modes): Likewise.
12829 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12831 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12834 2018-01-08 Richard Biener <rguenther@suse.de>
12836 PR middle-end/83580
12837 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12839 2018-01-08 Richard Biener <rguenther@suse.de>
12841 PR middle-end/83517
12842 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12844 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12846 PR middle-end/81897
12847 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12848 basic blocks with a small number of successors.
12849 (convert_control_dep_chain_into_preds): Improve handling of
12851 (dump_predicates): Split apart into...
12852 (dump_pred_chain): ...here...
12853 (dump_pred_info): ...and here.
12854 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12855 (can_chain_union_be_invalidated_p): Improve check for invalidation
12857 (uninit_uses_cannot_happen): Avoid unnecessary if
12858 convert_control_dep_chain_into_preds yielded nothing.
12860 2018-01-06 Martin Sebor <msebor@redhat.com>
12862 PR tree-optimization/83640
12863 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12864 subtracting negative offset from size.
12865 (builtin_access::overlap): Adjust offset bounds of the access to fall
12866 within the size of the object if possible.
12868 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12870 PR rtl-optimization/83699
12871 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12872 extract_bit_field_as_subreg to cases in which the extracted
12873 value is also a vector.
12875 * lra-constraints.c (process_alt_operands): Test for the equivalence
12876 substitutions when detecting a possible reload cycle.
12878 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12881 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12882 by default if flag_selective_schedling{,2}. Formatting fixes.
12884 PR rtl-optimization/83682
12885 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12886 if it has non-VECTOR_MODE element mode.
12887 (vec_duplicate_p): Likewise.
12889 PR middle-end/83694
12890 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12891 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12893 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12896 * config/i386/i386-builtin.def
12897 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12898 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12899 Require also OPTION_MASK_ISA_AVX512F in addition to
12900 OPTION_MASK_ISA_GFNI.
12901 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12902 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12903 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12904 to OPTION_MASK_ISA_GFNI.
12905 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12906 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12907 OPTION_MASK_ISA_AVX512BW.
12908 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12909 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12910 addition to OPTION_MASK_ISA_GFNI.
12911 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12912 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12913 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12914 to OPTION_MASK_ISA_GFNI.
12915 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12916 a requirement for all ISAs rather than any of them with a few
12918 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12920 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12921 bitmasks to be enabled with 3 exceptions, instead of requiring any
12922 enabled ISA with lots of exceptions.
12923 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12924 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12925 Change avx512bw in isa attribute to avx512f.
12926 * config/i386/sgxintrin.h: Add license boilerplate.
12927 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12928 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12929 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12930 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12932 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12933 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12934 temporarily sse2 rather than sse if not enabled already.
12937 * config/i386/sse.md (VI248_VLBW): Rename to ...
12938 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12939 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12940 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12941 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12942 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12943 mode iterator instead of VI248_VLBW.
12945 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12947 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12948 (record_modified): Skip clobbers; add debug output.
12949 (param_change_prob): Use sreal frequencies.
12951 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12953 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12954 punt for user-aligned variables.
12956 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12958 * tree-chrec.c (chrec_contains_symbols): Return true for
12961 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12964 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12965 of (x|y) == x for BICS pattern.
12967 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12969 PR tree-optimization/83605
12970 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12971 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12974 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12976 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12977 * config/epiphany/rtems.h: New file.
12979 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12980 Uros Bizjak <ubizjak@gmail.com>
12983 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12984 QIreg_operand instead of register_operand predicate.
12985 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12986 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12987 comments instead of -fmitigate[-_]rop.
12989 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12992 * cgraphunit.c (symbol_table::compile): Switch to text_section
12993 before calling assembly_start debug hook.
12994 * run-rtl-passes.c (run_rtl_passes): Likewise.
12997 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12999 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13000 range_int_cst_p rather than !symbolic_range_p before calling
13001 extract_range_from_multiplicative_op_1.
13003 2018-01-04 Jeff Law <law@redhat.com>
13005 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13006 redundant test in assertion.
13008 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13010 * doc/rtl.texi: Document machine_mode wrapper classes.
13012 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13014 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13015 using tree_to_uhwi.
13017 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13019 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13020 the VEC_PERM_EXPR fold to fail.
13022 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13025 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13026 to switched_sections.
13028 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13031 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13032 test for d.testing.
13034 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13037 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13038 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13040 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13043 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13044 is BLKmode and bitpos not zero or mode change is needed.
13046 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13049 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13052 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13055 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13056 instead of MULT rtx. Update all corresponding splitters.
13057 (*saddl_se): Ditto.
13058 (*ssub<modesuffix>): Ditto.
13059 (*ssubl_se): Ditto.
13060 (*cmp_sadd_di): Update split patterns.
13061 (*cmp_sadd_si): Ditto.
13062 (*cmp_sadd_sidi): Ditto.
13063 (*cmp_ssub_di): Ditto.
13064 (*cmp_ssub_si): Ditto.
13065 (*cmp_ssub_sidi): Ditto.
13066 * config/alpha/predicates.md (const23_operand): New predicate.
13067 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13068 Look for ASHIFT, not MULT inner operand.
13069 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13071 2018-01-04 Martin Liska <mliska@suse.cz>
13073 PR gcov-profile/83669
13074 * gcov.c (output_intermediate_file): Add version to intermediate
13076 * doc/gcov.texi: Document new field 'version' in intermediate
13077 file format. Fix location of '-k' option of gcov command.
13079 2018-01-04 Martin Liska <mliska@suse.cz>
13082 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13084 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13086 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13088 2018-01-03 Martin Sebor <msebor@redhat.com>
13090 PR tree-optimization/83655
13091 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13092 checking calls with invalid arguments.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13096 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13097 (vectorizable_mask_load_store): Delete.
13098 (vectorizable_call): Return false for masked loads and stores.
13099 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13100 instead of gimple_assign_rhs1.
13101 (vectorizable_load): Handle IFN_MASK_LOAD.
13102 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13106 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13108 (vectorizable_mask_load_store): ...here.
13109 (vectorizable_load): ...and here.
13111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13113 * tree-vect-stmts.c (vect_build_all_ones_mask)
13114 (vect_build_zero_merge_argument): New functions, split out from...
13115 (vectorizable_load): ...here.
13117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13119 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13121 (vectorizable_mask_load_store): ...here.
13122 (vectorizable_store): ...and here.
13124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13126 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13128 (vectorizable_mask_load_store): ...here.
13130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13132 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13133 (vect_model_store_cost): Take a vec_load_store_type instead of a
13135 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13136 (vect_model_store_cost): Take a vec_load_store_type instead of a
13138 (vectorizable_mask_load_store): Update accordingly.
13139 (vectorizable_store): Likewise.
13140 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13144 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13145 IFN_MASK_LOAD calls here rather than...
13146 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13149 Alan Hayward <alan.hayward@arm.com>
13150 David Sherwood <david.sherwood@arm.com>
13152 * expmed.c (extract_bit_field_1): For vector extracts,
13153 fall back to extract_bit_field_as_subreg if vec_extract
13156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13157 Alan Hayward <alan.hayward@arm.com>
13158 David Sherwood <david.sherwood@arm.com>
13160 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13161 they are variable or constant sized.
13162 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13163 slots for constant-sized data.
13165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13166 Alan Hayward <alan.hayward@arm.com>
13167 David Sherwood <david.sherwood@arm.com>
13169 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13170 handling COND_EXPRs with boolean comparisons, try to find a better
13171 basis for the mask type than the boolean itself.
13173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13175 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13176 is calculated and how it can be overridden.
13177 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13178 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13180 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13184 Alan Hayward <alan.hayward@arm.com>
13185 David Sherwood <david.sherwood@arm.com>
13187 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13188 Remove the mode argument.
13189 (aarch64_simd_valid_immediate): Remove the mode and inverse
13191 * config/aarch64/iterators.md (bitsize): New iterator.
13192 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13193 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13194 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13195 aarch64_simd_valid_immediate.
13196 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13197 (aarch64_reg_or_bic_imm): Likewise.
13198 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13199 with an insn_type enum and msl with a modifier_type enum.
13200 Replace element_width with a scalar_mode. Change the shift
13201 to unsigned int. Add constructors for scalar_float_mode and
13202 scalar_int_mode elements.
13203 (aarch64_vect_float_const_representable_p): Delete.
13204 (aarch64_can_const_movi_rtx_p)
13205 (aarch64_simd_scalar_immediate_valid_for_move)
13206 (aarch64_simd_make_constant): Update call to
13207 aarch64_simd_valid_immediate.
13208 (aarch64_advsimd_valid_immediate_hs): New function.
13209 (aarch64_advsimd_valid_immediate): Likewise.
13210 (aarch64_simd_valid_immediate): Remove mode and inverse
13211 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13212 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13213 and aarch64_float_const_representable_p on the result.
13214 (aarch64_output_simd_mov_immediate): Remove mode argument.
13215 Update call to aarch64_simd_valid_immediate and use of
13216 simd_immediate_info.
13217 (aarch64_output_scalar_simd_mov_immediate): Update call
13220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13221 Alan Hayward <alan.hayward@arm.com>
13222 David Sherwood <david.sherwood@arm.com>
13224 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13225 (mode_nunits): Likewise CONST_MODE_NUNITS.
13226 * machmode.def (ADJUST_NUNITS): Document.
13227 * genmodes.c (mode_data::need_nunits_adj): New field.
13228 (blank_mode): Update accordingly.
13229 (adj_nunits): New variable.
13230 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13232 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13233 listed in adj_nunits.
13234 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13235 listed in adj_nunits. Don't emit case statements for such modes.
13236 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13237 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13238 nothing if adj_nunits is nonnull.
13239 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13240 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13241 (emit_mode_fbit): Update use of print_maybe_const_decl.
13242 (emit_move_size): Likewise. Treat the array as non-const
13244 (emit_mode_adjustments): Handle adj_nunits.
13246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13248 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13249 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13250 (VECTOR_MODES): Use it.
13251 (make_vector_modes): Take the prefix as an argument.
13253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13254 Alan Hayward <alan.hayward@arm.com>
13255 David Sherwood <david.sherwood@arm.com>
13257 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13258 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13259 for MODE_VECTOR_BOOL.
13260 * machmode.def (VECTOR_BOOL_MODE): Document.
13261 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13262 (make_vector_bool_mode): New function.
13263 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13265 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13266 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13268 * stor-layout.c (int_mode_for_mode): Likewise.
13269 * tree.c (build_vector_type_for_mode): Likewise.
13270 * varasm.c (output_constant_pool_2): Likewise.
13271 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13272 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13273 for MODE_VECTOR_BOOL.
13274 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13275 of mode class checks.
13276 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13277 instead of a list of mode class checks.
13278 (expand_vector_scalar_condition): Likewise.
13279 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13282 Alan Hayward <alan.hayward@arm.com>
13283 David Sherwood <david.sherwood@arm.com>
13285 * machmode.h (mode_size): Change from unsigned short to
13287 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13288 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13289 or if measurement_type is not polynomial.
13290 (fixed_size_mode::includes_p): Check for constant-sized modes.
13291 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13292 return a poly_uint16 rather than an unsigned short.
13293 (emit_mode_size): Change the type of mode_size from unsigned short
13294 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13295 (emit_mode_adjustments): Cope with polynomial vector sizes.
13296 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13298 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13300 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13301 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13302 * caller-save.c (setup_save_areas): Likewise.
13303 (replace_reg_with_saved_mem): Likewise.
13304 * calls.c (emit_library_call_value_1): Likewise.
13305 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13306 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13307 (gen_lowpart_for_combine): Likewise.
13308 * convert.c (convert_to_integer_1): Likewise.
13309 * cse.c (equiv_constant, cse_insn): Likewise.
13310 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13311 (cselib_subst_to_values): Likewise.
13312 * dce.c (word_dce_process_block): Likewise.
13313 * df-problems.c (df_word_lr_mark_ref): Likewise.
13314 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13315 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13316 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13317 (rtl_for_decl_location): Likewise.
13318 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13319 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13320 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13321 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13322 (expand_expr_real_1): Likewise.
13323 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13324 (pad_below): Likewise.
13325 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13326 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13327 * ira.c (get_subreg_tracking_sizes): Likewise.
13328 * ira-build.c (ira_create_allocno_objects): Likewise.
13329 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13330 (ira_sort_regnos_for_alter_reg): Likewise.
13331 * ira-costs.c (record_operand_costs): Likewise.
13332 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13333 (resolve_simple_move): Likewise.
13334 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13335 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13336 (lra_constraints): Likewise.
13337 (CONST_POOL_OK_P): Reject variable-sized modes.
13338 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13339 (add_pseudo_to_slot, lra_spill): Likewise.
13340 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13341 * optabs-query.c (get_best_extraction_insn): Likewise.
13342 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13343 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13344 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13345 * recog.c (offsettable_address_addr_space_p): Likewise.
13346 * regcprop.c (maybe_mode_change): Likewise.
13347 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13348 * regrename.c (build_def_use): Likewise.
13349 * regstat.c (dump_reg_info): Likewise.
13350 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13351 (find_reloads, find_reloads_subreg_address): Likewise.
13352 * reload1.c (eliminate_regs_1): Likewise.
13353 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13354 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13355 (simplify_binary_operation_1, simplify_subreg): Likewise.
13356 * targhooks.c (default_function_arg_padding): Likewise.
13357 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13358 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13359 (verify_gimple_assign_ternary): Likewise.
13360 * tree-inline.c (estimate_move_cost): Likewise.
13361 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13362 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13363 (get_address_cost_ainc): Likewise.
13364 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13365 (vect_supportable_dr_alignment): Likewise.
13366 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13367 (vectorizable_reduction): Likewise.
13368 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13369 (vectorizable_operation, vectorizable_load): Likewise.
13370 * tree.c (build_same_sized_truth_vector_type): Likewise.
13371 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13372 * var-tracking.c (emit_note_insn_var_location): Likewise.
13373 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13374 (ADDR_VEC_ALIGN): Likewise.
13376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13377 Alan Hayward <alan.hayward@arm.com>
13378 David Sherwood <david.sherwood@arm.com>
13380 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13382 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13383 or if measurement_type is polynomial.
13384 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13385 * combine.c (make_extraction): Likewise.
13386 * dse.c (find_shift_sequence): Likewise.
13387 * dwarf2out.c (mem_loc_descriptor): Likewise.
13388 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13389 (extract_bit_field, extract_low_bits): Likewise.
13390 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13391 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13392 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13393 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13394 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13395 * reload.c (find_reloads): Likewise.
13396 * reload1.c (alter_reg): Likewise.
13397 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13398 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13399 * tree-if-conv.c (predicate_mem_writes): Likewise.
13400 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13401 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13402 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13403 * valtrack.c (dead_debug_insert_temp): Likewise.
13404 * varasm.c (mergeable_constant_section): Likewise.
13405 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13408 Alan Hayward <alan.hayward@arm.com>
13409 David Sherwood <david.sherwood@arm.com>
13411 * expr.c (expand_assignment): Cope with polynomial mode sizes
13412 when assigning to a CONCAT.
13414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13415 Alan Hayward <alan.hayward@arm.com>
13416 David Sherwood <david.sherwood@arm.com>
13418 * machmode.h (mode_precision): Change from unsigned short to
13420 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13422 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13423 or if measurement_type is not polynomial.
13424 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13425 in which the mode is already known to be a scalar_int_mode.
13426 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13427 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13429 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13430 for GET_MODE_PRECISION.
13431 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13432 for GET_MODE_PRECISION.
13433 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13435 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13436 (expand_field_assignment, make_extraction): Likewise.
13437 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13438 (get_last_value): Likewise.
13439 * convert.c (convert_to_integer_1): Likewise.
13440 * cse.c (cse_insn): Likewise.
13441 * expr.c (expand_expr_real_1): Likewise.
13442 * lra-constraints.c (simplify_operand_subreg): Likewise.
13443 * optabs-query.c (can_atomic_load_p): Likewise.
13444 * optabs.c (expand_atomic_load): Likewise.
13445 (expand_atomic_store): Likewise.
13446 * ree.c (combine_reaching_defs): Likewise.
13447 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13448 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13449 * tree.h (type_has_mode_precision_p): Likewise.
13450 * ubsan.c (instrument_si_overflow): Likewise.
13452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13453 Alan Hayward <alan.hayward@arm.com>
13454 David Sherwood <david.sherwood@arm.com>
13456 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13457 polynomial numbers of units.
13458 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13459 (valid_vector_subparts_p): New function.
13460 (build_vector_type): Remove temporary shim and take the number
13461 of units as a poly_uint64 rather than an int.
13462 (build_opaque_vector_type): Take the number of units as a
13463 poly_uint64 rather than an int.
13464 * tree.c (build_vector_from_ctor): Handle polynomial
13465 TYPE_VECTOR_SUBPARTS.
13466 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13467 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13468 (build_vector_from_val): If the number of units is variable,
13469 use build_vec_duplicate_cst for constant operands and
13470 VEC_DUPLICATE_EXPR otherwise.
13471 (make_vector_type): Remove temporary is_constant ().
13472 (build_vector_type, build_opaque_vector_type): Take the number of
13473 units as a poly_uint64 rather than an int.
13474 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13476 * cfgexpand.c (expand_debug_expr): Likewise.
13477 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13478 (store_constructor, expand_expr_real_1): Likewise.
13479 (const_scalar_mask_from_tree): Likewise.
13480 * fold-const-call.c (fold_const_reduction): Likewise.
13481 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13482 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13483 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13484 (fold_relational_const): Likewise.
13485 (native_interpret_vector): Likewise. Change the size from an
13486 int to an unsigned int.
13487 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13488 TYPE_VECTOR_SUBPARTS.
13489 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13490 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13491 duplicating a non-constant operand into a variable-length vector.
13492 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13493 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13494 * ipa-icf.c (sem_variable::equals): Likewise.
13495 * match.pd: Likewise.
13496 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13497 * print-tree.c (print_node): Likewise.
13498 * stor-layout.c (layout_type): Likewise.
13499 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13500 * tree-cfg.c (verify_gimple_comparison): Likewise.
13501 (verify_gimple_assign_binary): Likewise.
13502 (verify_gimple_assign_ternary): Likewise.
13503 (verify_gimple_assign_single): Likewise.
13504 * tree-pretty-print.c (dump_generic_node): Likewise.
13505 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13506 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13507 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13508 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13509 (vect_shift_permute_load_chain): Likewise.
13510 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13511 (expand_vector_condition, optimize_vector_constructor): Likewise.
13512 (lower_vec_perm, get_compute_type): Likewise.
13513 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13514 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13515 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13516 (vect_recog_mask_conversion_pattern): Likewise.
13517 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13518 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13519 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13520 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13521 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13522 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13523 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13524 (supportable_widening_operation): Likewise.
13525 (supportable_narrowing_operation): Likewise.
13526 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13528 * varasm.c (output_constant): Likewise.
13530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13531 Alan Hayward <alan.hayward@arm.com>
13532 David Sherwood <david.sherwood@arm.com>
13534 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13535 so that both the length == 3 and length != 3 cases set up their
13536 own permute vectors. Add comments explaining why we know the
13537 number of elements is constant.
13538 (vect_permute_load_chain): Likewise.
13540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13541 Alan Hayward <alan.hayward@arm.com>
13542 David Sherwood <david.sherwood@arm.com>
13544 * machmode.h (mode_nunits): Change from unsigned char to
13546 (ONLY_FIXED_SIZE_MODES): New macro.
13547 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13548 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13549 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13551 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13552 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13553 or if measurement_type is not polynomial.
13554 * genmodes.c (ZERO_COEFFS): New macro.
13555 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13557 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13558 Use ZERO_COEFFS when emitting initializers.
13559 * data-streamer.h (bp_pack_poly_value): New function.
13560 (bp_unpack_poly_value): Likewise.
13561 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13562 for GET_MODE_NUNITS.
13563 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13564 for GET_MODE_NUNITS.
13565 * tree.c (make_vector_type): Remove temporary shim and make
13566 the real function take the number of units as a poly_uint64
13567 rather than an int.
13568 (build_vector_type_for_mode): Handle polynomial nunits.
13569 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13570 * emit-rtl.c (const_vec_series_p_1): Likewise.
13571 (gen_rtx_CONST_VECTOR): Likewise.
13572 * fold-const.c (test_vec_duplicate_folding): Likewise.
13573 * genrecog.c (validate_pattern): Likewise.
13574 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13575 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13576 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13577 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13578 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13579 * rtlanal.c (subreg_get_info): Likewise.
13580 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13581 (vect_grouped_load_supported): Likewise.
13582 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13583 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13584 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13585 (simplify_const_unary_operation, simplify_binary_operation_1)
13586 (simplify_const_binary_operation, simplify_ternary_operation)
13587 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13588 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13589 instead of CONST_VECTOR_NUNITS.
13590 * varasm.c (output_constant_pool_2): Likewise.
13591 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13592 explicit-encoded elements in the XVEC for variable-length vectors.
13594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13596 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13599 Alan Hayward <alan.hayward@arm.com>
13600 David Sherwood <david.sherwood@arm.com>
13602 * coretypes.h (fixed_size_mode): Declare.
13603 (fixed_size_mode_pod): New typedef.
13604 * builtins.h (target_builtins::x_apply_args_mode)
13605 (target_builtins::x_apply_result_mode): Change type to
13606 fixed_size_mode_pod.
13607 * builtins.c (apply_args_size, apply_result_size, result_vector)
13608 (expand_builtin_apply_args_1, expand_builtin_apply)
13609 (expand_builtin_return): Update accordingly.
13611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13613 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13614 * cselib.c (cselib_hash_rtx): Likewise.
13615 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13616 CONST_VECTOR encoding.
13618 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13619 Jeff Law <law@redhat.com>
13622 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13623 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13624 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13625 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13628 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13629 explicitly probe *sp in a noreturn function if there were any callee
13630 register saves or frame pointer is needed.
13632 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13635 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13636 BLKmode for ternary, binary or unary expressions.
13639 * var-tracking.c (delete_vta_debug_insn): New inline function.
13640 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13641 insns from get_insns () to NULL instead of each bb separately.
13642 Use delete_vta_debug_insn. No longer static.
13643 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13644 delete_vta_debug_insns callers.
13645 * rtl.h (delete_vta_debug_insns): Declare.
13646 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13647 instead of variable_tracking_main.
13649 2018-01-03 Martin Sebor <msebor@redhat.com>
13651 PR tree-optimization/83603
13652 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13653 arguments past the endof the argument list in functions declared
13654 without a prototype.
13655 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13656 Avoid checking when arguments are null.
13658 2018-01-03 Martin Sebor <msebor@redhat.com>
13661 * doc/extend.texi (attribute const): Fix a typo.
13662 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13663 issuing -Wsuggest-attribute for void functions.
13665 2018-01-03 Martin Sebor <msebor@redhat.com>
13667 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13668 offset_int::from instead of wide_int::to_shwi.
13669 (maybe_diag_overlap): Remove assertion.
13670 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13671 * gimple-ssa-sprintf.c (format_directive): Same.
13672 (parse_directive): Same.
13673 (sprintf_dom_walker::compute_format_length): Same.
13674 (try_substitute_return_value): Same.
13676 2018-01-03 Jeff Law <law@redhat.com>
13678 PR middle-end/83654
13679 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13680 non-constant residual for zero at runtime and avoid probing in
13681 that case. Reorganize code for trailing problem to mirror handling
13684 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13686 PR tree-optimization/83501
13687 * tree-ssa-strlen.c (get_string_cst): New.
13688 (handle_char_store): Call get_string_cst.
13690 2018-01-03 Martin Liska <mliska@suse.cz>
13692 PR tree-optimization/83593
13693 * tree-ssa-strlen.c: Include tree-cfg.h.
13694 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13695 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13696 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13698 (strlen_dom_walker::before_dom_children): Call
13699 gimple_purge_dead_eh_edges. Dump tranformation with details
13701 (strlen_dom_walker::before_dom_children): Update call by adding
13702 new argument cleanup_eh.
13703 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13705 2018-01-03 Martin Liska <mliska@suse.cz>
13708 * cif-code.def (VARIADIC_THUNK): New enum value.
13709 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13712 2018-01-03 Jan Beulich <jbeulich@suse.com>
13714 * sse.md (mov<mode>_internal): Tighten condition for when to use
13715 vmovdqu<ssescalarsize> for TI and OI modes.
13717 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13719 Update copyright years.
13721 2018-01-03 Martin Liska <mliska@suse.cz>
13724 * ipa-visibility.c (function_and_variable_visibility): Skip
13725 functions with noipa attribure.
13727 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13729 * gcc.c (process_command): Update copyright notice dates.
13730 * gcov-dump.c (print_version): Ditto.
13731 * gcov.c (print_version): Ditto.
13732 * gcov-tool.c (print_version): Ditto.
13733 * gengtype.c (create_file): Ditto.
13734 * doc/cpp.texi: Bump @copying's copyright year.
13735 * doc/cppinternals.texi: Ditto.
13736 * doc/gcc.texi: Ditto.
13737 * doc/gccint.texi: Ditto.
13738 * doc/gcov.texi: Ditto.
13739 * doc/install.texi: Ditto.
13740 * doc/invoke.texi: Ditto.
13742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13744 * vector-builder.h (vector_builder::m_full_nelts): Change from
13745 unsigned int to poly_uint64.
13746 (vector_builder::full_nelts): Update prototype accordingly.
13747 (vector_builder::new_vector): Likewise.
13748 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13749 (vector_builder::operator ==): Likewise.
13750 (vector_builder::finalize): Likewise.
13751 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13752 Take the number of elements as a poly_uint64 rather than an
13754 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13755 from unsigned int to poly_uint64.
13756 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13757 (vec_perm_indices::new_vector): Likewise.
13758 (vec_perm_indices::length): Likewise.
13759 (vec_perm_indices::nelts_per_input): Likewise.
13760 (vec_perm_indices::input_nelts): Likewise.
13761 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13762 number of elements per input as a poly_uint64 rather than an
13763 unsigned int. Use the original encoding for variable-length
13764 vectors, rather than clamping each individual element.
13765 For the second and subsequent elements in each pattern,
13766 clamp the step and base before clamping their sum.
13767 (vec_perm_indices::series_p): Handle polynomial element counts.
13768 (vec_perm_indices::all_in_range_p): Likewise.
13769 (vec_perm_indices_to_tree): Likewise.
13770 (vec_perm_indices_to_rtx): Likewise.
13771 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13772 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13773 (tree_vector_builder::new_binary_operation): Handle polynomial
13774 element counts. Return false if we need to know the number
13775 of elements at compile time.
13776 * fold-const.c (fold_vec_perm): Punt if the number of elements
13777 isn't known at compile time.
13779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13781 * vec-perm-indices.h (vec_perm_builder): Change element type
13782 from HOST_WIDE_INT to poly_int64.
13783 (vec_perm_indices::element_type): Update accordingly.
13784 (vec_perm_indices::clamp): Handle polynomial element_types.
13785 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13786 (vec_perm_indices::all_in_range_p): Likewise.
13787 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13789 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13790 polynomial vec_perm_indices element types.
13791 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13792 * fold-const.c (fold_vec_perm): Likewise.
13793 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13794 * tree-vect-generic.c (lower_vec_perm): Likewise.
13795 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13796 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13797 element type to HOST_WIDE_INT.
13799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13800 Alan Hayward <alan.hayward@arm.com>
13801 David Sherwood <david.sherwood@arm.com>
13803 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13804 rather than an int. Use plus_constant.
13805 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13806 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13809 Alan Hayward <alan.hayward@arm.com>
13810 David Sherwood <david.sherwood@arm.com>
13812 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13813 a HOST_WIDE_INT to a poly_int64.
13815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13816 Alan Hayward <alan.hayward@arm.com>
13817 David Sherwood <david.sherwood@arm.com>
13819 * calls.c (load_register_parameters): Cope with polynomial
13820 mode sizes. Require a constant size for BLKmode parameters
13821 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13822 forces a parameter to be padded at the lsb end in order to
13823 fill a complete number of words, require the parameter size
13824 to be ordered wrt UNITS_PER_WORD.
13826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13827 Alan Hayward <alan.hayward@arm.com>
13828 David Sherwood <david.sherwood@arm.com>
13830 * reload1.c (spill_stack_slot_width): Change element type
13831 from unsigned int to poly_uint64_pod.
13832 (alter_reg): Treat mode sizes as polynomial.
13834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13835 Alan Hayward <alan.hayward@arm.com>
13836 David Sherwood <david.sherwood@arm.com>
13838 * reload.c (complex_word_subreg_p): New function.
13839 (reload_inner_reg_of_subreg, push_reload): Use it.
13841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13842 Alan Hayward <alan.hayward@arm.com>
13843 David Sherwood <david.sherwood@arm.com>
13845 * lra-constraints.c (process_alt_operands): Reject matched
13846 operands whose sizes aren't ordered.
13847 (match_reload): Refer to this check here.
13849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13850 Alan Hayward <alan.hayward@arm.com>
13851 David Sherwood <david.sherwood@arm.com>
13853 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13854 that the mode size is in the set {1, 2, 4, 8, 16}.
13856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13857 Alan Hayward <alan.hayward@arm.com>
13858 David Sherwood <david.sherwood@arm.com>
13860 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13861 Use plus_constant instead of gen_rtx_PLUS.
13863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13868 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13869 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13870 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13871 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13872 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13873 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13874 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13875 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13876 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13878 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13879 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13880 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13881 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13882 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13883 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13884 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13885 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13886 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13887 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13889 * expr.c (emit_move_resolve_push): Treat the input and result
13890 of PUSH_ROUNDING as a poly_int64.
13891 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13892 (emit_push_insn): Likewise.
13893 * lra-eliminations.c (mark_not_eliminable): Likewise.
13894 * recog.c (push_operand): Likewise.
13895 * reload1.c (elimination_effects): Likewise.
13896 * rtlanal.c (nonzero_bits1): Likewise.
13897 * calls.c (store_one_arg): Likewise. Require the padding to be
13898 known at compile time.
13900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13901 Alan Hayward <alan.hayward@arm.com>
13902 David Sherwood <david.sherwood@arm.com>
13904 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13905 Use plus_constant instead of gen_rtx_PLUS.
13907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13908 Alan Hayward <alan.hayward@arm.com>
13909 David Sherwood <david.sherwood@arm.com>
13911 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13912 rather than an int.
13914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13915 Alan Hayward <alan.hayward@arm.com>
13916 David Sherwood <david.sherwood@arm.com>
13918 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13919 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13920 via stack temporaries. Treat the mode size as polynomial too.
13922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13923 Alan Hayward <alan.hayward@arm.com>
13924 David Sherwood <david.sherwood@arm.com>
13926 * expr.c (expand_expr_real_2): When handling conversions involving
13927 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13928 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13929 as a poly_uint64 too.
13931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13938 Alan Hayward <alan.hayward@arm.com>
13939 David Sherwood <david.sherwood@arm.com>
13941 * combine.c (can_change_dest_mode): Handle polynomial
13942 REGMODE_NATURAL_SIZE.
13943 * expmed.c (store_bit_field_1): Likewise.
13944 * expr.c (store_constructor): Likewise.
13945 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13946 and polynomial REGMODE_NATURAL_SIZE.
13947 (gen_lowpart_common): Likewise.
13948 * reginfo.c (record_subregs_of_mode): Likewise.
13949 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13956 numbers of elements.
13958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13959 Alan Hayward <alan.hayward@arm.com>
13960 David Sherwood <david.sherwood@arm.com>
13962 * match.pd: Cope with polynomial numbers of vector elements.
13964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13965 Alan Hayward <alan.hayward@arm.com>
13966 David Sherwood <david.sherwood@arm.com>
13968 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13969 in a POINTER_PLUS_EXPR.
13971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * omp-simd-clone.c (simd_clone_subparts): New function.
13976 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13977 (ipa_simd_modify_function_body): Likewise.
13979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13984 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13985 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13986 (expand_vector_condition, vector_element): Likewise.
13987 (subparts_gt): New function.
13988 (get_compute_type): Use subparts_gt.
13989 (count_type_subparts): Delete.
13990 (expand_vector_operations_1): Use subparts_gt instead of
13991 count_type_subparts.
13993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13994 Alan Hayward <alan.hayward@arm.com>
13995 David Sherwood <david.sherwood@arm.com>
13997 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13998 (vect_compile_time_alias): ...this new function. Do the calculation
13999 on poly_ints rather than trees.
14000 (vect_prune_runtime_alias_test_list): Update call accordingly.
14002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14003 Alan Hayward <alan.hayward@arm.com>
14004 David Sherwood <david.sherwood@arm.com>
14006 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14008 (vect_schedule_slp_instance): Likewise.
14010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14011 Alan Hayward <alan.hayward@arm.com>
14012 David Sherwood <david.sherwood@arm.com>
14014 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14015 constant and extern definitions for variable-length vectors.
14016 (vect_get_constant_vectors): Note that the number of units
14017 is known to be constant.
14019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14020 Alan Hayward <alan.hayward@arm.com>
14021 David Sherwood <david.sherwood@arm.com>
14023 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14024 of units as polynomial. Choose between WIDE and NARROW based
14027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14028 Alan Hayward <alan.hayward@arm.com>
14029 David Sherwood <david.sherwood@arm.com>
14031 * tree-vect-stmts.c (simd_clone_subparts): New function.
14032 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14035 Alan Hayward <alan.hayward@arm.com>
14036 David Sherwood <david.sherwood@arm.com>
14038 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14039 vectors as polynomial. Use build_index_vector for
14040 IFN_GOMP_SIMD_LANE.
14042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14043 Alan Hayward <alan.hayward@arm.com>
14044 David Sherwood <david.sherwood@arm.com>
14046 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14047 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14048 for variable-length vectors.
14049 (vectorizable_mask_load_store): Treat the number of units as
14050 polynomial, asserting that it is constant if the condition has
14051 already been enforced.
14052 (vectorizable_store, vectorizable_load): Likewise.
14054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14055 Alan Hayward <alan.hayward@arm.com>
14056 David Sherwood <david.sherwood@arm.com>
14058 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14059 of units as polynomial. Punt if we can't tell at compile time
14060 which vector contains the final result.
14062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14063 Alan Hayward <alan.hayward@arm.com>
14064 David Sherwood <david.sherwood@arm.com>
14066 * tree-vect-loop.c (vectorizable_induction): Treat the number
14067 of units as polynomial. Punt on SLP inductions. Use an integer
14068 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14069 cast of such a series for variable-length floating-point
14072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14073 Alan Hayward <alan.hayward@arm.com>
14074 David Sherwood <david.sherwood@arm.com>
14076 * tree.h (build_index_vector): Declare.
14077 * tree.c (build_index_vector): New function.
14078 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14079 of units as polynomial, forcibly converting it to a constant if
14080 vectorizable_reduction has already enforced the condition.
14081 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14082 to create a {1,2,3,...} vector.
14083 (vectorizable_reduction): Treat the number of units as polynomial.
14084 Choose vectype_in based on the largest scalar element size rather
14085 than the smallest number of units. Enforce the restrictions
14088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14089 Alan Hayward <alan.hayward@arm.com>
14090 David Sherwood <david.sherwood@arm.com>
14092 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14093 number of units as polynomial.
14095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14096 Alan Hayward <alan.hayward@arm.com>
14097 David Sherwood <david.sherwood@arm.com>
14099 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14100 * target.def (autovectorize_vector_sizes): Return the vector sizes
14101 by pointer, using vector_sizes rather than a bitmask.
14102 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14103 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14104 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14106 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14107 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14108 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14109 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14110 * omp-general.c (omp_max_vf): Likewise.
14111 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14112 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14113 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14114 * tree-vect-slp.c (vect_slp_bb): Likewise.
14115 * doc/tm.texi: Regenerate.
14116 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14118 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14119 the vector size as a poly_uint64 rather than an unsigned int.
14120 (current_vector_size): Change from an unsigned int to a poly_uint64.
14121 (get_vectype_for_scalar_type): Update accordingly.
14122 * tree.h (build_truth_vector_type): Take the size and number of
14123 units as a poly_uint64 rather than an unsigned int.
14124 (build_vector_type): Add a temporary overload that takes
14125 the number of units as a poly_uint64 rather than an unsigned int.
14126 * tree.c (make_vector_type): Likewise.
14127 (build_truth_vector_type): Take the number of units as a poly_uint64
14128 rather than an unsigned int.
14130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14131 Alan Hayward <alan.hayward@arm.com>
14132 David Sherwood <david.sherwood@arm.com>
14134 * target.def (get_mask_mode): Take the number of units and length
14135 as poly_uint64s rather than unsigned ints.
14136 * targhooks.h (default_get_mask_mode): Update accordingly.
14137 * targhooks.c (default_get_mask_mode): Likewise.
14138 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14139 * doc/tm.texi: Regenerate.
14141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14142 Alan Hayward <alan.hayward@arm.com>
14143 David Sherwood <david.sherwood@arm.com>
14145 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14146 * omp-general.c (omp_max_vf): Likewise.
14147 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14148 (expand_omp_simd): Handle polynomial safelen.
14149 * omp-low.c (omplow_simd_context): Add a default constructor.
14150 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14151 (lower_rec_simd_input_clauses): Update accordingly.
14152 (lower_rec_input_clauses): Likewise.
14154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14155 Alan Hayward <alan.hayward@arm.com>
14156 David Sherwood <david.sherwood@arm.com>
14158 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14159 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14160 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14161 (vect_analyze_slp_cost): Likewise.
14162 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14163 (vect_model_load_cost): Likewise.
14165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14166 Alan Hayward <alan.hayward@arm.com>
14167 David Sherwood <david.sherwood@arm.com>
14169 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14170 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14171 from an unsigned int * to a poly_uint64_pod *.
14172 (calculate_unrolling_factor): New function.
14173 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14176 Alan Hayward <alan.hayward@arm.com>
14177 David Sherwood <david.sherwood@arm.com>
14179 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14180 from an unsigned int to a poly_uint64.
14181 (_loop_vec_info::slp_unrolling_factor): Likewise.
14182 (_loop_vec_info::vectorization_factor): Change from an int
14184 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14185 (vect_get_num_vectors): New function.
14186 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14187 (vect_get_num_copies): Use vect_get_num_vectors.
14188 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14189 to an unsigned int *.
14190 (vect_analyze_data_refs): Change min_vf from an int * to a
14192 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14193 than an unsigned HOST_WIDE_INT.
14194 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14195 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14196 to an unsigned int *.
14197 (vect_analyze_data_ref_dependences): Likewise.
14198 (vect_compute_data_ref_alignment): Handle polynomial vf.
14199 (vect_enhance_data_refs_alignment): Likewise.
14200 (vect_prune_runtime_alias_test_list): Likewise.
14201 (vect_shift_permute_load_chain): Likewise.
14202 (vect_supportable_dr_alignment): Likewise.
14203 (dependence_distance_ge_vf): Take the vectorization factor as a
14204 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14205 (vect_analyze_data_refs): Change min_vf from an int * to a
14207 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14208 vfm1 as a poly_uint64 rather than an int. Make the same change
14209 for the returned bound_scalar.
14210 (vect_gen_vector_loop_niters): Handle polynomial vf.
14211 (vect_do_peeling): Likewise. Update call to
14212 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14213 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14215 * tree-vect-loop.c (vect_determine_vectorization_factor)
14216 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14217 (vect_get_known_peeling_cost): Likewise.
14218 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14219 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14220 (vect_transform_loop): Likewise. Use the lowest possible VF when
14221 updating the upper bounds of the loop.
14222 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14223 rather than an int.
14224 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14225 polynomial unroll factors.
14226 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14227 (vect_make_slp_decision): Likewise.
14228 (vect_supported_load_permutation_p): Likewise, and polynomial
14230 (vect_analyze_slp_cost): Handle polynomial vf.
14231 (vect_slp_analyze_node_operations): Likewise.
14232 (vect_slp_analyze_bb_1): Likewise.
14233 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14234 than an unsigned HOST_WIDE_INT.
14235 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14236 (vectorizable_load): Handle polynomial vf.
14237 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14239 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14242 Alan Hayward <alan.hayward@arm.com>
14243 David Sherwood <david.sherwood@arm.com>
14245 * match.pd: Handle bit operations involving three constants
14246 and try to fold one pair.
14248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14250 * tree-vect-loop-manip.c: Include gimple-fold.h.
14251 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14252 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14253 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14254 Add a path that uses a step of VF instead of 1, but disable it
14256 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14257 and niters_no_overflow parameters. Update calls to
14258 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14259 Create a new SSA name if the latter choses to use a ste other
14260 than zero, and return it via niters_vector_mult_vf_var.
14261 * tree-vect-loop.c (vect_transform_loop): Update calls to
14262 vect_do_peeling, vect_gen_vector_loop_niters and
14263 slpeel_make_loop_iterate_ntimes.
14264 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14265 (vect_gen_vector_loop_niters): Update declarations after above changes.
14267 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14269 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14270 128-bit round to integer instructions.
14271 (ceil<mode>2): Likewise.
14272 (btrunc<mode>2): Likewise.
14273 (round<mode>2): Likewise.
14275 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14277 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14278 unaligned VSX load/store on P8/P9.
14279 (expand_block_clear): Allow the use of unaligned VSX
14280 load/store on P8/P9.
14282 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14284 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14286 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14287 swap associated with both a load and a store.
14289 2018-01-02 Andrew Waterman <andrew@sifive.com>
14291 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14292 * config/riscv/riscv.md (clear_cache): Use it.
14294 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14296 * web.c: Remove out-of-date comment.
14298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14300 * expr.c (fixup_args_size_notes): Check that any existing
14301 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14302 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14303 (emit_single_push_insn): ...here.
14305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14307 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14308 (const_vector_encoded_nelts): New function.
14309 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14310 (const_vector_int_elt, const_vector_elt): Declare.
14311 * emit-rtl.c (const_vector_int_elt_1): New function.
14312 (const_vector_elt): Likewise.
14313 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14314 of CONST_VECTOR_ELT.
14316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14318 * expr.c: Include rtx-vector-builder.h.
14319 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14320 directly on the tree encoding.
14321 (const_vector_from_tree): Likewise.
14322 * optabs.c: Include rtx-vector-builder.h.
14323 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14324 sequence of "u" values.
14325 * vec-perm-indices.c: Include rtx-vector-builder.h.
14326 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14327 directly on the vec_perm_indices encoding.
14329 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14331 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14332 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14333 * rtx-vector-builder.h: New file.
14334 * rtx-vector-builder.c: Likewise.
14335 * rtl.h (rtx_def::u2): Add a const_vector field.
14336 (CONST_VECTOR_NPATTERNS): New macro.
14337 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14338 (CONST_VECTOR_DUPLICATE_P): Likewise.
14339 (CONST_VECTOR_STEPPED_P): Likewise.
14340 (CONST_VECTOR_ENCODED_ELT): Likewise.
14341 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14342 (unwrap_const_vec_duplicate): Likewise.
14343 (const_vec_series_p): Check for a non-duplicated vector encoding.
14344 Say that the function only returns true for integer vectors.
14345 * emit-rtl.c: Include rtx-vector-builder.h.
14346 (gen_const_vec_duplicate_1): Delete.
14347 (gen_const_vector): Call gen_const_vec_duplicate instead of
14348 gen_const_vec_duplicate_1.
14349 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14350 (gen_const_vec_duplicate): Use rtx_vector_builder.
14351 (gen_const_vec_series): Likewise.
14352 (gen_rtx_CONST_VECTOR): Likewise.
14353 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14354 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14355 Build a new vector rather than modifying a CONST_VECTOR in-place.
14356 (handle_special_swappables): Update call accordingly.
14357 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14358 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14359 Build a new vector rather than modifying a CONST_VECTOR in-place.
14360 (handle_special_swappables): Update call accordingly.
14362 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14364 * simplify-rtx.c (simplify_const_binary_operation): Use
14365 CONST_VECTOR_ELT instead of XVECEXP.
14367 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14369 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14370 the selector elements to be different from the data elements
14371 if the selector is a VECTOR_CST.
14372 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14373 ssizetype for the selector.
14375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14377 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14378 before testing each element individually.
14379 * tree-vect-generic.c (lower_vec_perm): Likewise.
14381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14383 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14384 * selftest-run-tests.c (selftest::run_tests): Call it.
14385 * vector-builder.h (vector_builder::operator ==): New function.
14386 (vector_builder::operator !=): Likewise.
14387 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14388 (vec_perm_indices::all_from_input_p): New function.
14389 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14390 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14391 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14392 instead of reading the VECTOR_CST directly. Detect whether both
14393 vector inputs are the same before constructing the vec_perm_indices,
14394 and update the number of inputs argument accordingly. Use the
14395 utility functions added above. Only construct sel2 if we need to.
14397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14399 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14400 the broadcast of the low byte.
14401 (expand_mult_highpart): Use an explicit encoding for the permutes.
14402 * optabs-query.c (can_mult_highpart_p): Likewise.
14403 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14404 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14405 (vectorizable_bswap): Likewise.
14406 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14407 explicit encoding for the power-of-2 permutes.
14408 (vect_permute_store_chain): Likewise.
14409 (vect_grouped_load_supported): Likewise.
14410 (vect_permute_load_chain): Likewise.
14412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14414 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14415 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14416 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14417 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14418 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14419 (vect_gen_perm_mask_any): Likewise.
14421 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14423 * int-vector-builder.h: New file.
14424 * vec-perm-indices.h: Include int-vector-builder.h.
14425 (vec_perm_indices): Redefine as an int_vector_builder.
14426 (auto_vec_perm_indices): Delete.
14427 (vec_perm_builder): Redefine as a stand-alone class.
14428 (vec_perm_indices::vec_perm_indices): New function.
14429 (vec_perm_indices::clamp): Likewise.
14430 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14431 (vec_perm_indices::new_vector): New function.
14432 (vec_perm_indices::new_expanded_vector): Update for new
14433 vec_perm_indices class.
14434 (vec_perm_indices::rotate_inputs): New function.
14435 (vec_perm_indices::all_in_range_p): Operate directly on the
14436 encoded form, without computing elided elements.
14437 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14438 encoding. Update for new vec_perm_indices class.
14439 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14440 the given vec_perm_builder.
14441 (expand_vec_perm_var): Update vec_perm_builder constructor.
14442 (expand_mult_highpart): Use vec_perm_builder instead of
14443 auto_vec_perm_indices.
14444 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14445 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14446 or double series encoding as appropriate.
14447 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14448 vec_perm_indices instead of auto_vec_perm_indices.
14449 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14450 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14451 (vect_permute_store_chain): Likewise.
14452 (vect_grouped_load_supported): Likewise.
14453 (vect_permute_load_chain): Likewise.
14454 (vect_shift_permute_load_chain): Likewise.
14455 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14456 (vect_transform_slp_perm_load): Likewise.
14457 (vect_schedule_slp_instance): Likewise.
14458 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14459 (vectorizable_mask_load_store): Likewise.
14460 (vectorizable_bswap): Likewise.
14461 (vectorizable_store): Likewise.
14462 (vectorizable_load): Likewise.
14463 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14464 vec_perm_indices instead of auto_vec_perm_indices. Use
14465 tree_to_vec_perm_builder to read the vector from a tree.
14466 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14467 vec_perm_builder instead of a vec_perm_indices.
14468 (have_whole_vector_shift): Use vec_perm_builder and
14469 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14470 truncation to calc_vec_perm_mask_for_shift.
14471 (vect_create_epilog_for_reduction): Likewise.
14472 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14473 from auto_vec_perm_indices to vec_perm_indices.
14474 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14475 instead of changing individual elements.
14476 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14477 the vector in d.perm.
14478 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14479 from auto_vec_perm_indices to vec_perm_indices.
14480 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14481 instead of changing individual elements.
14482 (arm_vectorize_vec_perm_const): Use new_vector to install
14483 the vector in d.perm.
14484 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14485 Update vec_perm_builder constructor.
14486 (rs6000_expand_interleave): Likewise.
14487 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14488 (rs6000_expand_interleave): Likewise.
14490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14492 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14493 to qimode could truncate the indices.
14494 * optabs.c (expand_vec_perm_var): Likewise.
14496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14498 * Makefile.in (OBJS): Add vec-perm-indices.o.
14499 * vec-perm-indices.h: New file.
14500 * vec-perm-indices.c: Likewise.
14501 * target.h (vec_perm_indices): Replace with a forward class
14503 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14504 * optabs.h: Include vec-perm-indices.h.
14505 (expand_vec_perm): Delete.
14506 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14507 (expand_vec_perm_const): Declare.
14508 * target.def (vec_perm_const_ok): Replace with...
14509 (vec_perm_const): ...this new hook.
14510 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14511 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14512 * doc/tm.texi: Regenerate.
14513 * optabs.def (vec_perm_const): Delete.
14514 * doc/md.texi (vec_perm_const): Likewise.
14515 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14516 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14517 expand_vec_perm for constant permutation vectors. Assert that
14518 the mode of variable permutation vectors is the integer equivalent
14519 of the mode that is being permuted.
14520 * optabs-query.h (selector_fits_mode_p): Declare.
14521 * optabs-query.c: Include vec-perm-indices.h.
14522 (selector_fits_mode_p): New function.
14523 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14524 is defined, instead of checking whether the vec_perm_const_optab
14525 exists. Use targetm.vectorize.vec_perm_const instead of
14526 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14527 fit in the vector mode before using a variable permute.
14528 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14529 vec_perm_indices instead of an rtx.
14530 (expand_vec_perm): Replace with...
14531 (expand_vec_perm_const): ...this new function. Take the selector
14532 as a vec_perm_indices rather than an rtx. Also take the mode of
14533 the selector. Update call to shift_amt_for_vec_perm_mask.
14534 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14535 Use vec_perm_indices::new_expanded_vector to expand the original
14536 selector into bytes. Check whether the indices fit in the vector
14537 mode before using a variable permute.
14538 (expand_vec_perm_var): Make global.
14539 (expand_mult_highpart): Use expand_vec_perm_const.
14540 * fold-const.c: Includes vec-perm-indices.h.
14541 * tree-ssa-forwprop.c: Likewise.
14542 * tree-vect-data-refs.c: Likewise.
14543 * tree-vect-generic.c: Likewise.
14544 * tree-vect-loop.c: Likewise.
14545 * tree-vect-slp.c: Likewise.
14546 * tree-vect-stmts.c: Likewise.
14547 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14549 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14550 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14551 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14552 (aarch64_vectorize_vec_perm_const): ...this new function.
14553 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14554 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14555 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14556 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14557 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14558 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14559 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14561 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14562 check for NEON modes.
14563 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14564 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14565 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14566 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14568 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14569 the old VEC_PERM_CONST conditions.
14570 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14571 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14572 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14573 (ia64_vectorize_vec_perm_const_ok): Merge into...
14574 (ia64_vectorize_vec_perm_const): ...this new function.
14575 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14576 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14577 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14578 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14579 * config/mips/mips.c (mips_expand_vec_perm_const)
14580 (mips_vectorize_vec_perm_const_ok): Merge into...
14581 (mips_vectorize_vec_perm_const): ...this new function.
14582 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14583 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14584 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14585 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14586 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14587 (rs6000_expand_vec_perm_const): Delete.
14588 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14590 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14591 (altivec_expand_vec_perm_const_le): Take each operand individually.
14592 Operate on constant selectors rather than rtxes.
14593 (altivec_expand_vec_perm_const): Likewise. Update call to
14594 altivec_expand_vec_perm_const_le.
14595 (rs6000_expand_vec_perm_const): Delete.
14596 (rs6000_vectorize_vec_perm_const_ok): Delete.
14597 (rs6000_vectorize_vec_perm_const): New function.
14598 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14599 an element count and rtx array.
14600 (rs6000_expand_extract_even): Update call accordingly.
14601 (rs6000_expand_interleave): Likewise.
14602 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14603 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14604 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14605 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14606 (rs6000_expand_vec_perm_const): Delete.
14607 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14608 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14609 (altivec_expand_vec_perm_const_le): Take each operand individually.
14610 Operate on constant selectors rather than rtxes.
14611 (altivec_expand_vec_perm_const): Likewise. Update call to
14612 altivec_expand_vec_perm_const_le.
14613 (rs6000_expand_vec_perm_const): Delete.
14614 (rs6000_vectorize_vec_perm_const_ok): Delete.
14615 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14616 reference to the SPE evmerge intructions.
14617 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14618 an element count and rtx array.
14619 (rs6000_expand_extract_even): Update call accordingly.
14620 (rs6000_expand_interleave): Likewise.
14621 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14622 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14624 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14628 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14629 vector mode and that that mode matches the mode of the data
14631 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14632 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14633 directly using expand_vec_perm_1 when forcing selectors into
14635 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14639 * optabs-query.h (can_vec_perm_p): Delete.
14640 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14641 * optabs-query.c (can_vec_perm_p): Split into...
14642 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14643 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14644 particular selector is valid.
14645 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14646 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14647 (vect_grouped_load_supported): Likewise.
14648 (vect_shift_permute_load_chain): Likewise.
14649 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14650 (vect_transform_slp_perm_load): Likewise.
14651 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14652 (vectorizable_bswap): Likewise.
14653 (vect_gen_perm_mask_checked): Likewise.
14654 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14655 implementations of variable permutation vectors into account
14656 when deciding which selector to use.
14657 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14658 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14659 with a false third argument.
14660 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14661 to test whether the constant selector is valid and can_vec_perm_var_p
14662 to test whether a variable selector is valid.
14664 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14666 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14667 * optabs-query.c (can_vec_perm_p): Likewise.
14668 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14669 instead of vec_perm_indices.
14670 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14671 (vect_gen_perm_mask_checked): Likewise,
14672 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14673 (vect_gen_perm_mask_checked): Likewise,
14675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14677 * optabs-query.h (qimode_for_vec_perm): Declare.
14678 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14679 (qimode_for_vec_perm): ...this new function.
14680 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14682 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14684 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14685 does not have a conditional at the top.
14687 2018-01-02 Richard Biener <rguenther@suse.de>
14689 * ipa-inline.c (big_speedup_p): Fix expression.
14691 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14694 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14697 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14701 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14702 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14703 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14704 cond_taken_branch_cost 3->4.
14706 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14708 PR tree-optimization/83581
14709 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14710 TODO_cleanup_cfg if any changes have been made.
14712 PR middle-end/83608
14713 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14714 convert_modes if target mode has the right side, but different mode
14717 PR middle-end/83609
14718 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14719 last argument when extracting from CONCAT. If either from_real or
14720 from_imag is NULL, use expansion through memory. If result is not
14721 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14722 the parts directly to inner mode, if even that fails, use expansion
14725 PR middle-end/83623
14726 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14727 check for bswap in mode rather than HImode and use that in expand_unop
14730 Copyright (C) 2018 Free Software Foundation, Inc.
14732 Copying and distribution of this file, with or without modification,
14733 are permitted in any medium without royalty provided the copyright
14734 notice and this notice are preserved.