1 2018-05-30 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85964
4 * tracer.c (better_p): Drop initialized count check, we only
5 call the function with initialized counts now.
6 (find_best_successor): Do find a best edge if one
7 has uninitialized count.
8 (find_best_predecessor): Likewise. Do BB frequency check only
9 if count is initialized.
11 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
13 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14 (aarch64_ldrstr_offset_compare): New.
15 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17 (aarch64_gen_adjusted_ldpstp): Likewise.
19 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
21 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
22 Check for subset of GENERAL_REGS and FP_REGS.
23 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
26 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
28 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
29 and wi::to_poly_offset. Add the current offset and then check
30 whether the sum fits, rather than using an unchecked addition of
31 a checked term. Check for a shwi rather than a uhwi.
32 * expr.c (get_bit_range): Use tree_to_poly_uint64.
33 (store_constructor): Use poly_int_tree_p.
34 (expand_expr_real_1): Likewise.
35 * function.c (assign_temp): Likewise.
36 * fold-const.c (const_binop): Use poly_int_tree_p and
38 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
40 * ipa-icf-gimple.c (func_checker::compare_operand): Use
41 to_poly_offset for MEM offsets.
42 * ipa-icf.c (sem_variable::equals): Likewise.
43 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
44 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
45 wi::to_poly_offset for BIT_FIELD_REF offsets.
46 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
48 * var-tracking.c (emit_note_insn_var_location): Use
51 2018-05-29 Jim Wilson <jimw@sifive.com>
53 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
55 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
58 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
59 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
61 (sse4_1_round<mode>2): Use nonimmediate_operand
62 for operand 1 predicate.
64 2018-05-29 Martin Sebor <msebor@redhat.com>
65 Richard Biener <rguenther@suse.de>
68 * calls.c (get_size_range): Call determine_value_range instead
70 * tree-vrp.h (determine_value_range): Declared new function.
71 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
73 2018-05-29 Richard Biener <rguenther@suse.de>
75 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
76 sure to use non-pattern stmts for get_earlier_stmt arguments.
77 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
78 called on pattern stmts.
79 (get_later_stmt): Likewise.
81 2018-05-29 Martin Liska <mliska@suse.cz>
84 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
87 2018-05-29 Jakub Jelinek <jakub@redhat.com>
89 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
91 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
95 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
96 VEC_PACK_FLOAT_EXPR): New tree codes.
97 * tree-pretty-print.c (op_code_prio): Handle
98 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
99 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
100 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
101 * tree-inline.c (estimate_operator_cost): Likewise.
102 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
103 * fold-const.c (const_binop): Likewise.
104 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
105 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
106 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
107 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
108 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
109 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
110 * expr.c (expand_expr_real_2): Likewise.
111 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
112 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
113 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
115 * optabs.c (expand_widen_pattern_expr): For
116 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
117 sign from result type rather than operand's type.
118 (expand_binop_directly): For vec_packu_float_optab and
119 vec_packs_float_optab allow result type to be different from operand's
121 * optabs-tree.c (optab_for_tree_code): Handle
122 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
123 VEC_PACK_FLOAT_EXPR. Formatting fixes.
124 * tree-vect-generic.c (expand_vector_operations_1): Handle
125 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
127 * tree-vect-stmts.c (supportable_widening_operation): Handle
129 (supportable_narrowing_operation): Handle FLOAT_EXPR.
130 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
131 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
132 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
133 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
135 (vec_pack<floatprefix>_float_<mode>): New expander.
136 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
138 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
139 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
140 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
141 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
142 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
144 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
145 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
146 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
147 VEC_PACK_FLOAT_EXPR): Document.
149 2018-05-29 Richard Biener <rguenther@suse.de>
151 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
153 (stmt_vec_info_vec): Make pointer.
154 (init_stmt_vec_info_vec): Remove.
155 (free_stmt_vec_info_vec): Likewise.
156 (set_stmt_vec_info_vec): New function.
157 (free_stmt_vec_infos): Likewise.
158 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
159 (set_vinfo_for_stmt): Likewise.
160 (get_earlier_stmt): Likewise.
161 (get_later_stmt): Likewise.
162 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
163 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
164 (vec_info::~vec_info): Free stmt_vec_infos.
165 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
166 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
167 (pass_slp_vectorize::execute): Likewise.
168 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
169 (free_stmt_vec_info_vec): Likewise.
170 (set_stmt_vec_info_vec): New function.
171 (free_stmt_vec_infos): Likewise.
172 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
173 the global stmt_vec_info_vec.
174 * tree-parloops.c (gather_scalar_reductions): Use
175 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
178 2018-05-29 Richard Biener <rguenther@suse.de>
180 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
182 2018-05-29 Martin Liska <mliska@suse.cz>
183 David Malcolm <dmalcolm@redhat.com>
185 * vec.c (test_reverse): New.
186 (vec_c_tests): Add new test.
187 * vec.h (vl_ptr>::reverse): New function.
189 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
191 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
193 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
196 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
198 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
200 2018-05-28 Richard Biener <rguenther@suse.de>
202 PR tree-optimization/85933
203 * tree-vect-data-refs.c (vect_record_base_alignments): Only
204 look at stmts marked as vectorizable.
206 2018-05-28 Richard Biener <rguenther@suse.de>
208 PR tree-optimization/85934
209 * tree-vect-generic.c (expand_vector_operations_1): Hoist
210 vector boolean check before scalar optimization.
212 2018-05-28 Jakub Jelinek <jakub@redhat.com>
214 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
217 2018-05-28 Mark Wielaard <mark@klomp.org>
219 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
220 if it is an expression containing a minus sign.
222 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
224 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
226 2018-05-27 Paul Koning <ni1d@arrl.net>
228 * config/pdp11/pdp11.md (truncsihi2): Remove.
230 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
231 Chung-Ju Wu <jasonwucj@gmail.com>
233 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
235 (unaligned_store_dw): Ditto.
236 * config/nds32/nds32-memory-manipulation.c
237 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
238 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
239 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
240 (emit_setmem_word_loop): Rename to ...
241 (emit_setmem_doubleword_loop): ... this.
242 (nds32_gen_dup_4_byte_to_word_value): New function.
243 (nds32_gen_dup_8_byte_to_double_word_value): New function.
244 (nds32_expand_setmem_loop): Refine implementation.
245 (nds32_expand_setmem_loop_v3m): Ditto.
246 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
249 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
251 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
253 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
255 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
256 (nds32_init_machine_status): Initialize machine->attr_naked_p and
257 machine->attr_no_prologue_p.
258 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
259 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
260 (nds32_expand_epilogue): Consider attr_naked_p.
261 (nds32_expand_epilogue_v3pop): Likewise.
262 (nds32_can_use_return_insn): Likewise.
263 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
264 attr_no_prologue_p fields.
265 * config/nds32/nds32.opt (mret-in-naked-func): New option.
267 2018-05-27 Jakub Jelinek <jakub@redhat.com>
270 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
273 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
275 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
277 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
279 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
281 (*<floatsuffix>floatv2div2sf2): Rename to ...
282 (*float<floatunssuffix>v2div2sf2): ... this.
283 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
284 (float<floatunssuffix>v2div2sf2_mask): ... this.
285 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
286 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
287 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
289 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
291 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
293 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
295 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
297 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
299 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
300 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
301 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
302 gen_ufix_truncv8dfv8si2.
303 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
304 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
305 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
306 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
307 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
308 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
309 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
310 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
312 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
316 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
318 2018-05-25 Jim Wilson <jimw@sifive.com>
320 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
321 * config/riscv/riscv.c (struct machine_function): Add
322 interrupt_handler_p and attribute_checked_p fields.
323 (riscv_attribute_table): Add interrupt.
324 (riscv_interrupt_type_p): New.
325 (riscv_save_reg_p): Save extra regs for interrupt handler.
326 (riscv_use_save_libcall): Return false for interrupt handler.
327 (riscv_first_stack_step): Add forward declaration.
328 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
329 for interrupt handler with large frame. Use it for saved reg list.
330 (riscv_expand_prologue): Move flag_stack_usage_info support to
331 eliminate duplication.
332 (riscv_expand_epilogue): Generate mret for interrupt handler.
333 (riscv_epilogue_uses): New.
334 (riscv_can_use_return_insn): Return false for interrupt handler.
335 (riscv_function_ok_for_sibcall): Likewise.
336 (riscv_set_current_function): Add interrupt handler support.
337 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
338 * config/riscv/riscv.md (UNSPECV_MRET): New.
340 (riscv_frflags, riscv_fsflags): Use tab after opcode.
342 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
344 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
346 PR tree-optimization/85712
347 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
348 this candidate has already been replaced in-situ by a copy.
350 2018-05-25 Jason Merrill <jason@redhat.com>
352 PR c++/80485 - inline function non-zero address.
353 * symtab.c (nonzero_address): Check DECL_COMDAT.
355 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
358 * config/alpha/alpha.md (ashlsi3): New insn pattern.
359 (*ashlsi_se): Rename from *ashldi_se. Define as sign
360 extension of SImode operation. Use const123_operand predicate.
362 (*saddl_se_1): Ditto.
364 (*ssubl_se_1): Ditto.
365 * config/alpha/predicates.md (const123_operand): New predicate.
366 * config/alpha/constraints.md (P): Use IN_RANGE.
368 2018-05-25 Richard Biener <rguenther@suse.de>
370 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
372 (ref_maybe_used_by_stmt_p): Likewise.
373 (stmt_may_clobber_ref_p): Likewise.
374 (stmt_may_clobber_ref_p_1): Likewise.
375 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
377 (ref_maybe_used_by_stmt_p): Likewise.
378 (stmt_may_clobber_ref_p): Likewise.
379 (stmt_may_clobber_ref_p_1): Likewise.
380 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
381 the alias oracle to disambiguate DRs with stmts DR analysis
383 (vect_analyze_data_refs): Do not give up on not analyzable
384 DRs for BB vectorization. Remove code truncating the dataref
387 2018-05-25 Jakub Jelinek <jakub@redhat.com>
390 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
391 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
392 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
394 2018-05-25 Richard Biener <rguenther@suse.de>
396 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
397 function, combining stmt data ref gathering and fatal analysis
399 (vect_analyze_data_refs): Remove now redudnant code and simplify.
400 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
401 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
402 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
403 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
405 2018-05-25 Bin Cheng <bin.cheng@arm.com>
407 PR tree-optimization/85720
408 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
409 SCC if all partitions are builtins.
410 (version_loop_by_alias_check): New parameter. Generate cancelable
411 runtime alias check if all partitions are builtins.
412 (distribute_loop): Update call to above function.
414 2018-05-25 Bin Cheng <bin.cheng@arm.com>
416 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
417 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
418 (parm_default_def_partition_arg): Ditto.
419 (set_parm_default_def_partition): Ditto.
420 (get_parm_default_def_partitions): Ditto and make it static.
421 (get_undefined_value_partitions): Ditto and make it static.
422 (remove_ssa_form): Refactor call to init_var_map here.
423 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
424 computation for loop region.
425 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
426 (register_default_def): Delete.
427 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
428 (parm_default_def_partition_arg): Ditto.
429 (set_parm_default_def_partition): Ditto.
430 (get_parm_default_def_partitions): Ditto and make it static.
431 (get_undefined_value_partitions): Ditto and make it static.
432 (coalesce_with_default, coalesce_with_default): Update comment.
433 (create_coalesce_list_for_region): New func factored out from
434 create_outofssa_var_map.
435 (populate_coalesce_list_for_outofssa): New func factored out from
436 create_outofssa_var_map and coalesce_ssa_name.
437 (create_outofssa_var_map): Delete.
438 (coalesce_ssa_name): Refactor to support live range computation.
439 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
440 (get_parm_default_def_partitions): Delete.
441 (get_undefined_value_partitions): Ditto.
442 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
443 computation for loop region.
444 (new_tree_live_info, loe_visit_block): Ditto.
445 (live_worklist, set_var_live_on_entry): Ditto.
446 (calculate_live_on_exit, verify_live_on_entry): Ditto.
447 * tree-ssa-live.h (struct _var_map): New fields.
448 (init_var_map): Change decl.
449 (region_contains_p): New.
451 2018-05-25 Bin Cheng <bin.cheng@arm.com>
453 * tree-ssa-live.h (live_merge_and_clear): Delete.
455 2018-05-25 Richard Biener <rguenther@suse.de>
458 * tree-dump.c (dequeue_and_dump): Remove access to removed
459 operand 2 of a SWITCH_EXPR.
461 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
463 * doc/sourcebuild.texi (vect_double_cond_arith): Include
464 multiplication and division.
465 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
466 (cond_udiv@var{m}, cond_umod@var{m}): Document.
467 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
468 (cond_udiv_optab, cond_umod_optab): New optabs.
469 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
470 (IFN_COND_RDIV): New internal functions.
471 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
472 TRUNC_MOD_EXPR and RDIV_EXPR.
473 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
474 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
476 (SVE_INT_BINARY): Include mult.
477 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
478 (optab, sve_int_op): Handle mult.
479 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
481 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
482 for SVE_INT_BINARY_SD.
484 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
486 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
487 (optab, sve_int_op): Handle div and udiv.
488 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
489 for SVE_INT_BINARY_SD.
490 (*<optab><mode>3): New insn for the same.
492 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
494 * tree-vect-patterns.c: Include predict.h.
495 (vect_recog_divmod_pattern): Restrict check for division support
496 to when optimizing for size.
498 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
500 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
501 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
502 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
503 (gimple_match_op::set_op): Likewise.
504 (gimple_resimplify4): Declare.
505 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
506 (expr::gen_transform): Likewise.
507 (decision_tree::gen): Generate a simplification routine for 4 operands.
508 * gimple-match-head.c (gimple_simplify): Add an overload for
509 4 operands. In the top-level function, handle up to 4 call
510 arguments and call gimple_resimplify4.
511 (gimple_resimplify4): New function.
512 (build_call_internal): Pass a fourth operand.
513 (maybe_push_to_seq): Likewise.
514 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
515 Fold VEC_COND_EXPRs of an operation and a default value into
516 an IFN_COND_* function if possible.
517 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
519 (SVE_COND_FP_BINARY): Include them.
520 (optab, sve_fp_op): Handle them.
521 (SVE_INT_BINARY_REV): New code iterator.
522 (SVE_COND_FP_BINARY_REV): New int iterator.
523 (commutative): New int attribute.
524 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
526 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
528 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
529 (*cond_<optab><mode>): New patterns for reversed operands.
531 2018-05-25 Richard Biener <rguenther@suse.de>
533 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
534 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
535 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
536 (STMT_VINFO_GROUPED_ACCESS): Adjust.
537 * tree-vect-data-refs.c (everywhere): Adjust users.
538 * tree-vect-loop.c (everywhere): Likewise.
539 * tree-vect-slp.c (everywhere): Likewise.
540 * tree-vect-stmts.c (everywhere): Likewise.
541 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
543 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
545 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
547 (gcc_cv_as_section_exclude): ... this.
548 Try Solaris as #exclude syntax.
549 * configure: Regenerate.
550 * config.in: Regenerate.
551 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
553 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
554 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
556 * varasm.c (default_elf_asm_named_section): Don't check if
557 HAVE_GAS_SECTION_EXCLUDE is defined.
559 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
561 * doc/md.texi: Update the documentation of the cond_* optabs
562 to mention the new final operand. Fix GET_MODE_NUNITS call.
563 Describe the scalar case too.
564 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
565 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
567 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
568 (get_conditional_internal_fn): Update comment.
569 * tree-vect-loop.c (vectorizable_reduction): Pass the original
570 accumulator value as a final argument to conditional functions.
571 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
572 a define_expand and add an "else" operand. Assert for now that
573 the else operand is equal to operand 2. Use SVE_INT_BINARY and
574 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
575 (*cond_<optab><mode>): New patterns.
576 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
577 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
578 (UNSPEC_COND_EOR): Delete.
579 (optab): Remove associated mappings.
580 (SVE_INT_BINARY): New code iterator.
581 (sve_int_op): Remove int attribute and add "minus" to the code
583 (SVE_COND_INT_OP): Delete.
584 (SVE_COND_FP_OP): Rename to...
585 (SVE_COND_FP_BINARY): ...this.
587 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
589 * optabs.c (can_reuse_operands_p): New function.
590 (maybe_legitimize_operands): Try to reuse the results for
593 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
595 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
596 Add {q} suffix to insn mnemonic.
598 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
600 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
601 (msp430_warn_func_return): New.
603 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
605 * fold-const.c (tree_nonzero_bits): New function.
606 * fold-const.h (tree_nonzero_bits): Likewise.
607 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
608 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
610 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
614 * varasm.c (assemble_alias): Check ifunc_resolver only on
617 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
620 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
621 when memory input operand is handled.
623 2018-05-24 Luis Machado <luis.machado@linaro.org>
625 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
627 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
629 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
631 * match.pd: Delay FMA folds until after vectorization.
633 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
636 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
637 address check not strict.
639 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
641 * gimple-match.h (gimple_match_op): New class.
642 (mprts_hook): Replace parameters with a gimple_match_op *.
643 (maybe_build_generic_op): Likewise.
644 (gimple_simplified_result_is_gimple_val): Replace parameters with
645 a const gimple_match_op *.
646 (gimple_simplify): Replace code_helper * and tree * parameters with
647 a gimple_match_op * parameter.
648 (gimple_resimplify1): Replace code_helper *, tree and tree *
649 parameters with a gimple_match_op * parameter.
650 (gimple_resimplify2): Likewise.
651 (gimple_resimplify3): Likewise.
652 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
653 parameters with a gimple_match_op * parameter.
654 * gimple-match-head.c (gimple_simplify): Change prototypes of
655 auto-generated functions to take a gimple_match_op * instead of
656 separate code_helper * and tree * parameters. Make the same
657 change in the top-level overload and update calls to the
658 gimple_resimplify routines. Update calls to the auto-generated
659 functions and to maybe_push_res_to_seq in the publicly-facing
660 operation-specific gimple_simplify overloads.
661 (gimple_match_op::MAX_NUM_OPS): Define.
662 (gimple_resimplify1): Replace rcode and ops with a single res_op
663 parameter. Update call to gimple_simplify.
664 (gimple_resimplify2): Likewise.
665 (gimple_resimplify3): Likewise.
666 (mprts_hook): Replace parameters with a gimple_match_op *.
667 (maybe_build_generic_op): Likewise.
668 (build_call_internal): Replace type, nargs and ops with
670 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
671 with a single gimple_match_op *. Update calls to mprts_hook,
672 build_call_internal and gimple_simplified_result_is_gimple_val.
673 Factor out code that is common to the tree_code and combined_fn cases.
674 * genmatch.c (expr::gen_transform): Replace tem_code and
675 tem_ops with a gimple_match_op called tem_op. Update calls
676 to the gimple_resimplify functions and maybe_push_res_to_seq.
677 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
678 res_ops. Update call to the gimple_resimplify functions.
679 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
680 (decision_tree::gen): Make the functions take a gimple_match_op *
681 called res_op instead of separate res_code and res_ops parameters.
682 Update call accordingly.
683 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
684 and ops with a single res_op parameter. Update calls to
685 maybe_build_generic_op and maybe_push_res_to_seq.
686 (fold_stmt_1): Update calls to gimple_simplify and
687 replace_stmt_with_simplification.
688 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
689 and gimple_simplified_result_is_gimple_val.
690 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
692 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
693 with a gimple_match_op *.
694 (vn_nary_build_or_lookup): Likewise. Update call to
695 vn_nary_build_or_lookup_1.
696 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
697 gimple_match_op *. Update calls to the gimple_resimplify routines
698 and to gimple_simplified_result_is_gimple_val.
699 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
700 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
701 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
702 (visit_nary_op): Likewise.
703 (visit_reference_op_load): Likewise.
705 2018-05-23 Luis Machado <luis.machado@linaro.org>
707 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
708 modifier for printing the step amount.
710 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
713 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
716 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
718 * doc/sourcebuild.texi (Endianness): New subsubsection.
720 2018-05-23 Luis Machado <luis.machado@linaro.org>
722 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
723 <prefetch_dynamic_strides>: New const bool field.
724 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
725 prefetch_dynamic_strides.
726 (exynosm1_prefetch_tune): Likewise.
727 (thunderxt88_prefetch_tune): Likewise.
728 (thunderx_prefetch_tune): Likewise.
729 (thunderx2t99_prefetch_tune): Likewise.
730 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
732 (aarch64_override_options_internal): Update to set
733 PARAM_PREFETCH_DYNAMIC_STRIDES.
734 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
735 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
736 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
737 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
738 prefetch-dynamic-strides setting.
740 2018-05-23 Luis Machado <luis.machado@linaro.org>
742 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
743 <minimum_stride>: New const int field.
744 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
745 minimum_stride field defaulting to -1.
746 (exynosm1_prefetch_tune): Likewise.
747 (thunderxt88_prefetch_tune): Likewise.
748 (thunderx_prefetch_tune): Likewise.
749 (thunderx2t99_prefetch_tune): Likewise.
750 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
751 <default_opt_level>: Set to 3.
752 (aarch64_override_options_internal): Update to set
753 PARAM_PREFETCH_MINIMUM_STRIDE.
754 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
755 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
756 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
757 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
758 stride is constant and is below the minimum stride threshold.
760 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
762 * config/arm/arm-cpus.in (mode26): Delete.
763 (armv4): Delete mode26 reference.
764 * config/arm/arm.c (arm_configure_build_target): Delete use of
767 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
769 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
771 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
772 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
774 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
775 (floatunsdidf2): Ditto.
777 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
779 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
780 (fixuns_trunc<mode>si2_avx512f): Ditto.
781 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
782 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
783 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
785 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
787 PR rtl-optimization/79985
788 * df-scan.c (df_insn_refs_collect): Remove special case for
789 global registers and asm statements.
791 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
793 * extend.texi (Global Register Variables): Rewrite the bullet list.
794 Note that the register is available for allocation. Note that access
795 via inline asm must use constraints. Add note about async-signal
796 handlers. Remove paragraph about automagic register selection.
798 2018-05-23 Richard Biener <rguenther@suse.de>
800 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
801 of fixed offset from memset VN.
803 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
805 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
807 (alloc_cand_and_find_basis): Initialize first_interp field.
808 (slsr_process_mul): Modify first_interp field.
809 (slsr_process_add): Likewise.
810 (slsr_process_cast): Modify first_interp field for each new
812 (slsr_process_copy): Likewise.
813 (dump_candidate): Dump first_interp field.
814 (replace_mult_candidate): Process all interpretations, not just
816 (replace_rhs_if_not_dup): Likewise.
817 (replace_one_candidate): Likewise.
819 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
821 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
823 (aarch64_needs_frame_chain): New function.
824 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
826 2018-05-23 Sudakshina Das <sudi.das@arm.com>
829 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
830 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
831 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
832 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
833 as true for strict-align.
834 (aarch64_can_inline_p): Perform checks even when callee has no
835 attributes to check for strict alignment.
836 * doc/extend.texi (AArch64 Function Attributes): Document
838 * doc/invoke.texi: (AArch64 Options): Likewise.
840 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
842 PR tree-optimization/85853
843 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
844 the handling of the root of the node to...
845 (vect_slp_analyze_node_operations_1): ...this new function,
846 and run the whole thing with the child nodes' def types
847 set according to their SLP node's def type.
849 2018-05-23 Richard Biener <rguenther@suse.de>
852 * tree-data-ref.c (create_runtime_alias_checks): Defer
853 and ignore overflow warnings.
855 2018-05-23 Yury Gribov <tetra2005@gmail.com>
857 PR tree-optimization/85822
858 * tree-vrp.c (is_masked_range_test): Fix handling of negative
861 2018-05-23 Richard Biener <rguenther@suse.de>
863 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
864 memset constants via native_interpret_expr.
866 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
869 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
871 (cgraph_node::create_alias): Likewise.
872 (cgraph_node::get_availability): Check ifunc_resolver instead
873 of looking up ifunc attribute.
874 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
875 * varasm.c (do_assemble_alias): Likewise.
876 (assemble_alias): Likewise.
877 (default_binds_local_p_3): Likewise.
878 * cgraph.h (cgraph_node): Add ifunc_resolver.
879 (cgraph_node::only_called_directly_or_aliased_p): Return false
881 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
883 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
884 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
885 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
886 instead of looking up ifunc attribute.
888 2018-05-22 Luis Machado <luis.machado@linaro.org>
890 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
892 2018-05-22 Martin Sebor <msebor@redhat.com>
895 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
896 only when expasion succeeds.
897 (expand_builtin_strcmp): Same.
898 (expand_builtin_strncmp): Same.
900 2018-05-22 Martin Sebor <msebor@redhat.com>
902 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
904 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
905 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
907 * config/aarch64/aarch64-ldpstp.md: Replace uses of
908 aarch64_mem_pair_operand with memory_operand and delete operand swapping
910 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
911 Add check for legitimate_address.
912 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
913 (aarch64_swap_ldrstr_operands): New.
914 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
917 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
918 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
920 * config/aarch64/aarch64.md: New patterns to generate stp
922 (store_pair_sw, store_pair_dw): New patterns to generate stp for
923 single words and double words.
924 (load_pair_sw, load_pair_dw): Likewise.
925 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
927 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
929 * config/aarch64/aarch64-ldpstp.md: Modify peephole
930 for different mode ldpstp and add peephole for merged zero stores.
932 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
934 (aarch64_gen_store_pair): Rename calls to match new patterns.
935 (aarch64_gen_load_pair): Rename calls to match new patterns.
936 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
937 (load_pair<DREG:mode><DREG2:mode>): ... This.
938 (store_pair<mode>): Rename to...
939 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
940 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
942 (V_INT_EQUIV): Handle SImode.
943 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
946 2018-05-22 Martin Sebor <msebor@redhat.com>
949 * calls.c (maybe_warn_nonstring_arg): Use string length to set
950 or ajust the presumed bound on an operation to avoid unnecessary
953 2018-05-22 Martin Sebor <msebor@redhat.com>
955 PR tree-optimization/85826
956 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
957 assuming that a DECL necesarily has a constant size.
959 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
962 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
964 2018-05-22 Richard Biener <rguenther@suse.de>
966 PR tree-optimization/85834
967 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
968 non-constant and non-zero memset arguments.
970 2018-05-22 Martin Liska <mliska@suse.cz>
973 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
975 2018-05-22 Richard Biener <rguenther@suse.de>
977 PR tree-optimization/85863
978 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
979 comparisons when vectype is specified.
980 (vectorizable_condition): Do not specify vectype for
981 vect_is_simple_cond when SLP vectorizing.
983 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
986 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
987 define __ibm128 as long double.
988 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
989 as a distinct type when IEEE 128-bit support is enabled.
990 (init_float128_ieee): Fix up conversions between IFmode and IEEE
991 128-bit types to use the correct functions.
992 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
993 convert between 128-bit floating point types that have different
994 modes but the same representation, instead of using gen_lowpart to
996 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
998 (IFKF_reg): New attributes to give the register constraints for
1000 (extend<mode>tf2_internal): New insns to mark an explicit
1001 conversion between 128-bit floating point types that have a
1002 different mode but share the same representation.
1004 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1006 PR tree-optimization/85814
1007 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1008 a null return from get_strinfo when unsharing the next
1009 strinfo in the chain.
1011 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1014 * varasm.c (weak_finish): Clean up weak_decls.
1016 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1018 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1019 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1020 UNSPEC_UADALP values.
1021 * config/aarch64/iterators.md (ABAL): New int iterator.
1024 (sur): Add mappings for the above.
1025 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1027 (aarch64_<sur>abal<mode>_4): Likewise.
1028 (aarch64_<sur>adalp<mode>_3): Likewise.
1029 (<sur>sadv16qi): New define_expand.
1031 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1033 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1034 (*movdf_internal): Ditto.
1035 (*rcpsf2_sse): Ditto.
1036 (*rsqrtsf2_sse): Ditto.
1037 (*sqrt<mode>2_sse): Ditto.
1039 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1041 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1043 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1044 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1045 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1046 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1048 * config/aarch64/arm_neon.h: Likewise.
1049 * config/aarch64/iterators.md (VQ_I): New.
1051 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1053 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1055 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1058 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1060 * config/nds32/constraints.md (S): New constraint.
1061 * config/nds32/nds32.md (call_internal): Use constraint S.
1062 (call_value_internal): Likewise.
1063 (sibcall_internal): Likewise.
1064 (sibcall_value_internal): Likewise.
1066 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1067 Chung-Ju Wu <jasonwucj@gmail.com>
1069 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1072 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1073 Chung-Ju Wu <jasonwucj@gmail.com>
1075 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1076 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1077 (nds32_rtx_costs_impl): Simplify.
1078 (nds32_address_cost_impl): Simplify.
1079 (nds32_init_rtx_costs): New function.
1080 (nds32_rtx_costs_speed_prefer): Likewise.
1081 (nds32_rtx_costs_size_prefer): Likewise.
1082 (nds32_address_cost_speed_prefer): Likewise.
1083 (nds32_address_cost_speed_fwprop): Likewise.
1084 (nds32_address_cost_size_prefer): Likewise.
1085 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1086 * config/nds32/nds32.c (nds32_option_override): Use
1087 nds32_init_rtx_costs function.
1089 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1091 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1092 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1093 (TARGET_PIPELINE_N8): Likewise.
1094 (TARGET_PIPELINE_N10): Likewise.
1095 (TARGET_PIPELINE_N13): Likewise.
1096 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1098 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1100 * config/nds32/nds32-fpu.md: Update copyright year.
1102 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1104 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1106 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1108 * config/nds32/nds32.c
1109 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1110 * config/nds32/nds32.opt (minline-asm-r15): New option.
1112 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1114 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1116 * config/nds32/nds32.md (abssi2): New pattern.
1118 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1120 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1121 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1122 Merge insn pattern from sse_cvtsi2ss<round_name> and
1123 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1124 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1125 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1126 using SWI48 mode iterator.
1127 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1128 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1129 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1130 pattern from sse_cvttss2si<round_saeonly_name>
1131 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1132 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1133 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1134 using SWI48 mode iterator.
1135 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1136 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1137 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1138 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1139 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1140 using SWI48 mode iterator.
1141 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1142 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1143 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1144 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1145 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1146 SWI48 mode iterator.
1147 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1148 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1149 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1150 pattern from sse_cvttsd2si<round_saeonly_name>
1151 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1153 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1155 * config/nds32/nds32-md-auxiliary.c
1156 (nds32_valid_smw_lwm_base_p): Refine.
1157 (nds32_output_smw_single_word): Refine.
1158 (nds32_output_smw_double_word): New.
1159 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1161 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1163 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1164 (nds32_output_stack_pop): Refine.
1165 (nds32_expand_unaligned_load): Refine.
1166 (nds32_expand_unaligned_store): Refine.
1168 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1169 Chung-Ju Wu <jasonwucj@gmail.com>
1171 * config/nds32/constants.md: Add TP_REGNUM constant.
1172 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1173 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1175 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1176 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1177 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1178 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1179 and PIC code generation.
1180 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1182 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1184 * config/nds32/nds32.md: Support TLS and PIC.
1185 * config/nds32/nds32.c: Support TLS and PIC.
1186 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1187 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1190 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1192 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1193 mode with E_ prefix.
1195 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1196 Chung-Ju Wu <jasonwucj@gmail.com>
1198 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1199 * config/nds32/nds32-md-auxiliary.c
1200 (symbolic_reference_mentioned_p): New.
1201 (nds32_legitimize_ict_address): New.
1202 (nds32_expand_ict_move): New.
1203 (nds32_indirect_call_referenced_p): New.
1204 (nds32_symbol_binds_local_p): Delete.
1205 (nds32_long_call_p): Modify.
1206 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1207 * config/nds32/nds32-protos.h
1208 (symbolic_reference_mentioned_p): Declare.
1209 (nds32_legitimize_ict_address): Declare.
1210 (nds32_expand_ict_move): Declare.
1211 (nds32_indirect_call_referenced_p): Declare.
1212 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1213 (nds32_relax_group): Use nds32_ict_const_p as condition.
1214 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1215 (nds32_asm_file_start): Output ict_model directive in asm code.
1216 (nds32_legitimate_address_p): Consider indirect call.
1217 (nds32_print_operand): Consider indirect call.
1218 (nds32_print_operand_address): Consider indirect call.
1219 (nds32_insert_attributes): Handle "indirect_call" attribute.
1220 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1221 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1222 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1223 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1224 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1225 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1226 (TARGET_ICT_MODEL_SMALL): Define.
1227 (TARGET_ICT_MODEL_LARGE): Define.
1228 * config/nds32/nds32.md (movsi): Consider ict model.
1229 (call, call_value): Consider ict model.
1230 (sibcall, sibcall_value): Consider ict model.
1231 * config/nds32/nds32.opt (mict-model): New option.
1232 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1235 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1236 Monk Chiang <sh.chiang04@gmail.com>
1237 Jim Wilson <jimw@sifive.com>
1239 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1240 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1241 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1242 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1243 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1244 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1245 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1246 compute save_libcall_adjustment properly.
1247 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1248 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1249 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1250 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1251 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1252 (ABI_SPEC): Handle mabi=ilp32e.
1253 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1254 (RVE): Add RVE mask.
1255 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1256 <-march>: Add rv32e as an example.
1258 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1261 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1262 (intra_create_variable_infos): Handle C++ constructors.
1264 2018-05-18 Martin Liska <mliska@suse.cz>
1266 * passes.def: Remove a redundant pass.
1268 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1271 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1273 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1275 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1277 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1279 (armv2, armv2a, armv3, armv3m): Delete architectures.
1280 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1281 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1282 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1284 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1285 (*mulsidi3adddi): Likewise.
1286 (mulsidi3): Likewise.
1287 (*mulsidi3_nov6): Likewise.
1288 (umulsidi3): Likewise.
1289 (umulsidi3_nov6): Likewise.
1290 (umaddsidi4): Likewise.
1291 (*umulsidi3adddi): Likewise.
1292 (smulsi3_highpart): Likewise.
1293 (*smulsi3_highpart_nov6): Likewise.
1294 (umulsi3_highpart): Likewise.
1295 (*umulsi3_highpart_nov6): Likewise.
1296 * config/arm/arm.h (arm_arch3m): Delete.
1297 * config/arm/arm.c (arm_arch3m): Delete.
1298 (arm_option_override_internal): Update armv3-related comment.
1299 (arm_configure_build_target): Delete use of isa_bit_mode32.
1300 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1301 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1302 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1304 (mulusa3): Likewise.
1305 * config/arm/arm-protos.h (arm_arch3m): Delete.
1306 * config/arm/arm-tables.opt: Regenerate.
1307 * config/arm/arm-tune.md: Likewise.
1308 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1309 deleted architectures.
1311 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1313 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1314 (armv5t, armv5te): New features.
1315 (ARMv5, ARMv5e): Delete fgroups.
1316 (ARMv5t, ARMv5te): Adjust for above changes.
1318 (armv5, armv5e): Delete arches.
1319 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1321 (*call_reg_arm): Likewise.
1322 (*call_value_reg_armv5): Likewise.
1323 (*call_value_reg_arm): Likewise.
1324 (*call_symbol): Likewise.
1325 (*call_value_symbol): Likewise.
1326 (*sibcall_insn): Likewise.
1327 (*sibcall_value_insn): Likewise.
1329 (prefetch): Likewise.
1330 (define_split and define_peephole2 dependent on arm_arch5):
1332 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1334 (TARGET_ARM_QBIT): Likewise.
1335 (TARGET_DSP_MULTIPLY): Likewise.
1336 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1337 (arm_arch5, arm_arch5e): Delete.
1338 (arm_arch5t, arm_arch5te): Declare.
1339 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1340 (arm_arch5t): Declare.
1341 (arm_option_reconfigure_globals): Update for the above.
1342 (arm_options_perform_arch_sanity_checks): Update comment, replace
1343 use of arm_arch5 with arm_arch5t.
1344 (use_return_insn): Likewise.
1345 (arm_emit_call_insn): Likewise.
1346 (output_return_instruction): Likewise.
1347 (arm_final_prescan_insn): Likewise.
1348 (arm_coproc_builtin_available): Likewise.
1349 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1350 arm_arch5e with arm_arch5t and arm_arch5te.
1351 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1352 (arm_arch5t, arm_arch5te): Declare.
1353 * config/arm/arm-tables.opt: Regenerate.
1354 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1355 * config/arm/t-multilib: Likewise.
1356 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1357 instead of arm_arch5.
1358 (*call_reg_thumb1): Likewise.
1359 (*call_value_reg_thumb1_v5): Likewise.
1360 (*call_value_reg_thumb1): Likewise.
1361 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1363 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1365 2018-05-18 Martin Liska <mliska@suse.cz>
1367 PR gcov-profile/84846
1368 * doc/gcov.texi: Document -t option of gcov tool.
1370 2018-05-18 Martin Liska <mliska@suse.cz>
1372 PR gcov-profile/84846
1373 * gcov.c (print_usage): Add new -t option.
1374 (process_args): Handle the option.
1375 (generate_results): Use stdout as output when requested by
1378 2018-05-18 Martin Liska <mliska@suse.cz>
1380 PR gcov-profile/84846
1381 * coverage.c (coverage_init): Write PWD to .gcno file.
1382 * doc/gcov.texi: Document how working directory is printed.
1383 * gcov-dump.c (dump_gcov_file): Print PWD.
1384 * gcov.c (output_intermediate_file): Likewise.
1385 (read_graph_file): Read PWD string.
1386 (output_lines): Print PWD.
1388 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1391 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1392 for retval and return false if all args to phi are zero.
1394 2018-05-18 Richard Biener <rguenther@suse.de>
1396 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1398 (evrp_dom_walker::before_dom_children): Call it.
1400 2018-05-18 Richard Biener <rguenther@suse.de>
1402 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1403 results when processing array refs with variable index.
1405 2018-05-18 Toon Moene <toon@moene.org>
1407 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1408 directly after that of -floop-interchange. Indicate that both
1409 options are enabled by default when specifying -O3.
1411 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1413 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1414 iterator. Delete separate integer-mode vec_set<mode> expander.
1415 (aarch64_simd_vec_setv2di): Delete.
1416 (vec_setv2di): Delete.
1417 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1418 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1419 the "w, r" alternative.
1421 2018-05-18 Martin Liska <mliska@suse.cz>
1423 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1424 * tree-pass.h (make_pass_lower_switch_O0): New function.
1425 * tree-switch-conversion.c (node_has_low_bound): Remove.
1426 (node_has_high_bound): Likewise.
1427 (node_is_bounded): Likewise.
1428 (class pass_lower_switch): Make it a template type and create
1430 (pass_lower_switch::execute): Add template argument.
1431 (make_pass_lower_switch): New function.
1432 (make_pass_lower_switch_O0): New function.
1433 (do_jump_if_equal): Remove.
1434 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1435 all the hard work to tree optimization passes.
1437 2018-05-18 Martin Liska <mliska@suse.cz>
1439 * dbgcnt.c (limit_low): Renamed from limit.
1440 (limit_high): New variable.
1441 (dbg_cnt_is_enabled): Check for upper limit.
1442 (dbg_cnt): Adjust dumping.
1443 (dbg_cnt_set_limit_by_index): Add new argument for high
1445 (dbg_cnt_set_limit_by_name): Likewise.
1446 (dbg_cnt_process_single_pair): Parse new format.
1447 (dbg_cnt_process_opt): Use strtok.
1448 (dbg_cnt_list_all_counters): Remove 'value' and add
1450 * doc/invoke.texi: Document changes.
1452 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1454 * doc/sourcebuild.texi (scalar_all_fma): Document.
1455 * tree.def (FMA_EXPR): Delete.
1456 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1457 * internal-fn.c (ternary_direct): New macro.
1458 (expand_ternary_optab_fn): Likewise.
1459 (direct_ternary_optab_supported_p): Likewise.
1460 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1461 * builtins.c (fold_builtin_fma): Delete.
1462 (fold_builtin_3): Don't call it.
1463 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1464 * expr.c (expand_expr_real_2): Likewise.
1465 * fold-const.c (operand_equal_p): Likewise.
1466 (fold_ternary_loc): Likewise.
1467 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1468 * gimple.c (DEFTREECODE): Likewise.
1469 * gimplify.c (gimplify_expr): Likewise.
1470 * optabs-tree.c (optab_for_tree_code): Likewise.
1471 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1472 * tree-eh.c (operation_could_trap_p): Likewise.
1473 (stmt_could_throw_1_p): Likewise.
1474 * tree-inline.c (estimate_operator_cost): Likewise.
1475 * tree-pretty-print.c (dump_generic_node): Likewise.
1476 (op_code_prio): Likewise.
1477 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1478 * tree-ssa-operands.c (get_expr_operands): Likewise.
1479 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1480 * fold-const-call.h (fold_fma): Delete.
1481 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1482 CFN_FNMA and CFN_FNMS.
1484 * genmatch.c (combined_fn): New enum.
1485 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1486 (commutative_op): New function.
1487 (commutate): Use it. Handle more than 2 operands.
1488 (dt_operand::gen_gimple_expr): Use commutative_op.
1489 (parser::parse_expr): Allow :c to be used with non-binary
1490 operators if the commutative operand is known.
1491 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1492 CFN_FMS, CFN_FNMA and CFN_FNMS.
1493 (backprop::process_assign_use): Remove FMA_EXPR handling.
1494 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1495 (gen_hsa_fma): New function.
1496 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1497 IFN_FNMA and IFN_FNMS.
1498 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1499 * gimple-fold.h (follow_all_ssa_edges): Declare.
1500 * gimple-fold.c (follow_all_ssa_edges): New function.
1501 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1502 gimple_build interface and use follow_all_ssa_edges to fold the result.
1503 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1504 instead of checking for optabs directly.
1505 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1506 rather than FMA_EXPRs.
1507 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1508 call to IFN_FMA instead of an FMA_EXPR.
1510 2018-05-17 Jim Wilson <jimw@sifive.com>
1512 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1513 sign extended promoted subreg, and the range does not have the sign bit
1514 set, then do a sign extend.
1516 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1517 test, check for sign extended subreg and/or constant operands, and
1518 do a sign extend in that case.
1520 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1522 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1523 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1525 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1526 Change logics_shift_reg to logics_shift_imm.
1527 (thunderx2t99_fp_loadpair_basic): Delete.
1528 (thunderx2t99_fp_storepair_basic): Delete.
1529 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1530 (thunderx2t99_asimd_polynomial): Delete.
1531 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1532 and neon_fp_mul_d_scalar_q.
1533 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1534 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1535 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1536 (thunderx2t99_asimd_lut): Add missing tbl types.
1537 (thunderx2t99_asimd_ext): Delete.
1538 (thunderx2t99_asimd_load1_1_mult): Delete.
1539 (thunderx2t99_asimd_load1_2_mult): Delete.
1540 (thunderx2t99_asimd_load1_ldp): New.
1541 (thunderx2t99_asimd_load1): New.
1542 (thunderx2t99_asimd_load2): Add missing *load2* types.
1543 (thunderx2t99_asimd_load3): New.
1544 (thunderx2t99_asimd_load4): New.
1545 (thunderx2t99_asimd_store1_1_mult): Delete.
1546 (thunderx2t99_asimd_store1_2_mult): Delete.
1547 (thunderx2t99_asimd_store2_mult): Delete.
1548 (thunderx2t99_asimd_store2_onelane): Delete.
1549 (thunderx2t99_asimd_store_stp): New.
1550 (thunderx2t99_asimd_store1): New.
1551 (thunderx2t99_asimd_store2): New.
1552 (thunderx2t99_asimd_store3): New.
1553 (thunderx2t99_asimd_store4): New.
1555 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1557 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1558 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1560 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1561 Segher Boessenkool <segher@kernel.crashing.org>
1564 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1567 2018-05-17 Richard Biener <rguenther@suse.de>
1569 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1570 for pruning loop and prune defs feeding only already visited PHIs.
1572 2018-05-17 Richard Biener <rguenther@suse.de>
1574 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1576 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1577 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/85793
1580 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1581 for VMAT_ELEMENTWISE.
1583 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1585 * internal-fn.h (lookup_internal_fn): Declare
1586 * internal-fn.c (lookup_internal_fn): New function.
1587 * gimple.c (gimple_build_call_from_tree): Handle calls to
1589 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1590 internal function names.
1591 * tree-pretty-print.c (dump_generic_node): Likewise.
1592 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1594 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1596 * gimple-fold.h (gimple_build): Make the function forms take
1597 combined_fn rather than built_in_function.
1598 (gimple_simplify): Likewise.
1599 * gimple-match-head.c (gimple_simplify): Likewise.
1600 * gimple-fold.c (gimple_build): Likewise.
1601 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1602 rather than gimple_build_call_internal.
1603 (get_initial_defs_for_reduction): Likewise.
1604 (vect_create_epilog_for_reduction): Likewise.
1605 (vectorizable_live_operation): Likewise.
1607 2018-05-17 Martin Liska <mliska@suse.cz>
1609 * gimple-ssa-sprintf.c (format_directive): Do not use
1610 space in between 'G_' and '('.
1612 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1615 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1616 even if the mask is not all ones.
1619 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1621 (ix86_gimple_fold_builtin): Likewise.
1624 * config/i386/i386.c: Include tree-vector-builder.h.
1625 (ix86_vector_shift_count): New function.
1626 (ix86_fold_builtin): Fold shift builtins by scalar count.
1627 (ix86_gimple_fold_builtin): Likewise.
1629 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1630 _mm512_setzero): New intrinsics.
1632 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1633 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1635 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1636 code generation for cases where splatting a value is not useful.
1637 * simplify-rtx.c (simplify_ternary_operation): Simplify
1638 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1639 mode to a vec_concat.
1641 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1643 * config.gcc: Support "goldmont-plus".
1644 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1646 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1647 PROCESSOR_GOLDMONT_PLUS.
1648 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1649 (processor_target_table): Add "goldmont-plus".
1650 (PTA_GOLDMONT_PLUS): Define.
1651 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1652 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1653 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1654 (fold_builtin_cpu): Add "goldmont-plus".
1655 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1656 (ix86_option_override_internal): Add "goldmont-plus".
1657 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1658 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1659 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1660 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1662 2018-05-17 Richard Biener <rguenther@suse.de>
1664 PR tree-optimization/85757
1665 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1666 remove defs that only feed that PHI from further processing.
1668 2018-05-16 Jim Wilson <jimw@sifive.com>
1670 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1672 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1674 2018-05-16 Mark Wielaard <mark@klomp.org>
1676 * dwarf2out.c (count_index_strings): New function.
1677 (output_indirect_strings): Call count_index_strings and generate
1678 header for dwarf_version >= 5.
1680 2018-05-16 Mark Wielaard <mark@klomp.org>
1682 * dwarf2out.c (dwarf_FORM): New function.
1683 (set_indirect_string): Use dwarf_FORM.
1684 (reset_indirect_string): Likewise.
1685 (size_of_die): Likewise.
1686 (value_format): Likewise.
1687 (output_die): Likewise.
1688 (add_skeleton_AT_string): Likewise.
1689 (output_macinfo_op): Likewise.
1690 (index_string): Likewise.
1691 (output_index_string_offset): Likewise.
1692 (output_index_string): Likewise.
1693 (count_index_strings): Likewise.
1695 2018-05-16 Carl Love <cel@us.ibm.com>
1697 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1698 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1700 2018-05-16 Martin Jambor <mjambor@suse.cz>
1702 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1703 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1705 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1707 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1708 (fnma<mode>4): Likewise.
1709 (fms<mode>4): Likewise.
1710 (fnms<mode>4): Likewise.
1711 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1712 (aarch64_fnma<mode>4): Likewise.
1713 (aarch64_fms<mode>4): Likewise.
1714 (aarch64_fnms<mode>4): Likewise.
1715 (aarch64_fnmadd<mode>4): Likewise.
1717 2018-05-16 Jason Merrill <jason@redhat.com>
1719 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1721 2018-05-16 Richard Biener <rguenther@suse.de>
1723 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1724 (dump_stmt_cost): Declare.
1725 (add_stmt_cost): Dump cost we add.
1726 (add_stmt_costs): New function.
1727 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1729 (vect_analyze_stmt): Adjust prototype.
1730 (vectorizable_condition): Likewise.
1731 (vectorizable_live_operation): Likewise.
1732 (vectorizable_reduction): Likewise.
1733 (vectorizable_induction): Likewise.
1734 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1735 cost vector to pass to vectorizable_ and record afterwards.
1736 (vect_model_reduction_cost): Take cost vector argument and adjust.
1737 (vect_model_induction_cost): Likewise.
1738 (vectorizable_reduction): Likewise.
1739 (vectorizable_induction): Likewise.
1740 (vectorizable_live_operation): Likewise.
1741 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1742 SLP_TREE_NUMBER_OF_VEC_STMTS.
1743 (vect_analyze_slp_cost_1): Remove.
1744 (vect_analyze_slp_cost): Likewise.
1745 (vect_slp_analyze_node_operations): Take visited args and
1746 a target cost vector. Avoid processing already visited stmt sets.
1747 (vect_slp_analyze_operations): Use a local cost vector to gather
1748 costs and register those of non-discarded instances.
1749 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1750 (vect_schedule_slp_instance): Remove copying of
1751 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1753 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1754 adding cost. Record cost entry location.
1755 (vect_prologue_cost_for_slp_op): Function to compute cost of
1756 a constant or invariant generated for SLP vect in the prologue,
1757 split out from vect_analyze_slp_cost_1.
1758 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1759 (vect_model_promotion_demotion_cost): Likewise.
1760 (vect_model_store_cost): Likewise, make static.
1761 (vect_model_load_cost): Likewise.
1762 (vectorizable_bswap): Add cost vector arg and adjust.
1763 (vectorizable_call): Likewise.
1764 (vectorizable_simd_clone_call): Likewise.
1765 (vectorizable_conversion): Likewise.
1766 (vectorizable_assignment): Likewise.
1767 (vectorizable_shift): Likewise.
1768 (vectorizable_operation): Likewise.
1769 (vectorizable_store): Likewise.
1770 (vectorizable_load): Likewise.
1771 (vectorizable_condition): Likewise.
1772 (vectorizable_comparison): Likewise.
1773 (can_vectorize_live_stmts): Likewise.
1774 (vect_analyze_stmt): Likewise.
1775 (vect_transform_stmt): Adjust calls to vectorizable_*.
1776 * tree-vectorizer.c: Include gimple-pretty-print.h.
1777 (dump_stmt_cost): New function.
1779 2018-05-16 Richard Biener <rguenther@suse.de>
1781 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1782 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1783 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1784 (check_name): New callback.
1785 (dse_classify_store): Track cycles via a visited bitmap of PHI
1786 defs and simplify handling of in-loop and across loop dead stores
1787 and properly fail for loop-variant refs. Handle byte-tracking with
1788 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1791 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1793 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1794 (vect_get_mask_type_for_stmt): Likewise.
1795 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1797 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1798 to determine the statement's vector type and the vector type that
1799 should be used for calculating nunits. Deal with cases in which
1800 the type has to be deferred.
1801 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1802 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1803 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1804 (vect_determine_vf_for_stmt): New functions, split out from...
1805 (vect_determine_vectorization_factor): ...here.
1806 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1807 (vect_get_mask_type_for_stmt): New functions, split out from
1808 vect_determine_vectorization_factor.
1810 2018-05-16 Richard Biener <rguenther@suse.de>
1812 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1813 verify the [VEC_]COND_EXPR embedded comparison.
1815 2018-05-15 Martin Sebor <msebor@redhat.com>
1817 PR tree-optimization/85753
1818 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1819 RECORD_TYPE in addition to ARRAY_TYPE.
1821 2018-05-15 Martin Sebor <msebor@redhat.com>
1824 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1826 2018-05-15 Richard Biener <rguenther@suse.de>
1828 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1829 add by_clobber_p one. Change algorithm to collect all defs
1830 representing uses we need to walk and try reducing them to
1831 a single one before failing.
1832 (dse_dom_walker::dse_optimize_stmt): Adjust.
1834 2018-05-13 Mark Wielaard <mark@klomp.org>
1836 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1837 (size_of_loc_descr): Likewise.
1838 (output_loc_operands): Likewise.
1839 (output_loc_operands_raw): Likewise.
1840 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1841 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1842 (hash_loc_operands): Likewise.
1843 (compare_loc_operands): Likewise.
1845 2018-05-14 Mark Wielaard <mark@klomp.org>
1847 * dwarf2out.c (count_index_addrs): New function.
1848 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1850 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1852 PR tree-optimization/83648
1853 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1854 return value as malloc candidate.
1856 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1859 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1860 as true in call to suggest_attribute.
1862 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1864 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1865 -mreadonly-in-sdata.
1867 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1869 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1871 (aarch64_crypto_aesd_fused): Likewise.
1873 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1875 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1876 (movsi_aarch64): Likewise.
1877 (load_pairsi): Likewise.
1878 (load_pairdi): Likewise.
1879 (store_pairsi): Likewise.
1880 (store_pairdi): Likewise.
1881 (load_pairsf): Likewise.
1882 (load_pairdf): Likewise.
1883 (store_pairsf): Likewise.
1884 (store_pairdf): Likewise.
1885 (zero_extend): Likewise.
1886 (trunc): Swap alternatives.
1887 (fcvt_target): Add '?' to prefer w over r.
1889 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1892 * config/i386/i386.md: Disallow non-commutative arithmetics in
1893 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1894 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1895 in the peephole2 before it.
1897 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1899 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1900 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1901 (ix86_handle_option): Handle -mcldemote.
1902 * config.gcc: New header.
1903 * config/i386/cldemoteintrin.h: New file.
1904 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1907 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1908 OPTION_MASK_ISA_CLDEMOTE.
1909 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1910 (ix86_valid_target_attribute_inner_p): Ditto.
1911 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1912 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1913 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1914 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1915 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1917 * config/i386/i386.opt: Add -mcldemote.
1918 * config/i386/x86intrin.h: New header.
1919 * doc/invoke.texi: Add -mcldemote.
1921 2018-05-14 Richard Biener <rguenther@suse.de>
1923 * doc/match-and-simplify.texi: Adjust :s documentation.
1925 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1927 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1928 intended memcpy size.
1929 (REORDER_45): Likewise.
1931 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1933 * sort.cc: New file.
1934 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1935 * vec.c (qsort_chk): Use gcc_qsort.
1936 * Makefile.in (OBJS-libcommon): Add sort.o.
1937 (build/sort.o): New target. Use it...
1938 (BUILD_RTL): ... here, and...
1939 (build/gencfn-macros): ... here, and...
1940 (build/genmatch): ... here.
1942 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1943 Chung-Ju Wu <jasonwucj@gmail.com>
1945 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1946 * config/nds32/nds32-graywolf.md: New file.
1947 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1948 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1950 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1951 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1952 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1953 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1954 * config/nds32/pipelines.md: Include n15 settings.
1956 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1957 Chung-Ju Wu <jasonwucj@gmail.com>
1959 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1960 * config/nds32/nds32-n13.md: New file.
1961 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1962 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1964 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1965 * config/nds32/nds32.md (pipeline_model): Add n13.
1966 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1967 * config/nds32/pipelines.md: Include n13 settings.
1969 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1970 Chung-Ju Wu <jasonwucj@gmail.com>
1972 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1973 * config/nds32/nds32-n10.md: New file.
1974 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1975 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1977 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1978 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1979 * config/nds32/nds32.md (pipeline_model): Add n10.
1980 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1981 * config/nds32/pipelines.md: Include n10 settings.
1983 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1984 Kito Cheng <kito.cheng@gmail.com>
1985 Chung-Ju Wu <jasonwucj@gmail.com>
1987 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1988 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1989 Add enum values for DSP extension instructions.
1990 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1992 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1993 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1995 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1996 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1997 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1998 * config/nds32/nds32-intrinsic.md: Likewise.
1999 * config/nds32/nds32_intrinsic.h: Likewise.
2000 * config/nds32/nds32-md-auxiliary.c: Likewise.
2001 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2002 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2003 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2004 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2005 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2006 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2007 TYPE_DMAC in switch statement.
2008 * config/nds32/nds32.c: New checking and implementation for DSP
2009 extension instructions.
2010 * config/nds32/nds32.h: Likewise.
2011 * config/nds32/nds32.md: Likewise.
2012 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2013 * config/nds32/predicates.md: Implement new predicates for DSP
2016 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2018 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2019 Reformat alternatives and attributes so it is easier to identify
2020 which constraints/attributes go with which instruction.
2021 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2022 (mov<mode>_softfloat32, FMOVE64): Likewise.
2023 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2024 (mov<mode>_softfloat64, FMOVE64): Likewise.
2026 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2028 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2030 (Basic PowerPC Built-in Functions): The new name of the
2031 subsection previously known as "PowerPC Built-in Functions".
2032 (Basic PowerPC Built-in Functions Available on all Configurations):
2034 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2035 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2036 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2037 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2039 2018-05-11 Martin Jambor <mjambor@suse.cz>
2042 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2045 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2048 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2050 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2052 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2053 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2054 (ix86_handle_option): Handle -mwaitpkg.
2055 * config.gcc: New header.
2056 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2057 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2058 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2060 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2061 OPTION_MASK_ISA_WAITPKG.
2062 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2063 (ix86_option_override_internal): Add PTA_WAITPKG.
2064 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2065 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2066 IX86_BUILTIN_TPAUSE.
2067 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2068 __builtin_ia32_umwait and __builtin_ia32_tpause.
2069 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2070 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2071 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2072 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2073 UNSPECV_TPAUSE): New.
2074 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2075 * config/i386/i386.opt: Add -mwaitpkg.
2076 * config/i386/waitpkgintrin.h: New file.
2077 * config/i386/x86intrin.h: New header.
2078 * doc/invoke.texi: Add -mwaitpkg.
2080 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2083 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2085 (cortex-m0): Use armv6s-m isa.
2086 (cortex-m0plus): Likewise.
2087 (cortex-m1): Likewise.
2088 (cortex-m0.small-multiply): Likewise.
2089 (cortex-m0plus.small-multiply): Likewise.
2090 (cortex-m1.small-multiply): Likewise.
2092 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2093 Jakub Jelinek <jakub@redhat.com>
2095 PR tree-optimization/85692
2096 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2097 source permute as well.
2099 2018-05-11 Martin Liska <mliska@suse.cz>
2102 * doc/extend.texi: Document LLVM style format for no_sanitize
2105 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2107 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2108 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2109 (mode_supports_vsx_dform_quad): Likewise.
2110 (mode_supports_vmx_dform): Move these functions to be next to the
2111 other mode_supports functions.
2112 (mode_supports_dq_form): Likewise.
2113 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2114 mode_supports_dq_form.
2115 (reg_offset_addressing_ok_p): Likewise.
2116 (offsettable_ok_by_alignment): Likewise.
2117 (rs6000_legitimate_offset_address_p): Likewise.
2118 (legitimate_lo_sum_address_p): Likewise.
2119 (rs6000_legitimize_address): Likewise.
2120 (rs6000_legitimize_reload_address): Likewise.
2121 (rs6000_secondary_reload_inner): Likewise.
2122 (rs6000_preferred_reload_class): Likewise.
2123 (rs6000_output_move_128bit): Likewise.
2125 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2127 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2128 Generate SImode target register for null target.
2129 <case IX86_BUILTIN_XGETBV>: Ditto.
2130 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2131 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2133 2018-05-10 Carl Love <cel@us.ibm.com>
2135 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2136 dcbtt and dcbtstt if operands[2] is 0.
2138 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2141 * config/i386/sse.md (usadv64qi): New expander.
2143 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2145 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2146 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2147 -maltivec=be support.
2148 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2149 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2150 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2151 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2152 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2153 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2154 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2155 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2156 altivec_vsumsws): Adjust.
2157 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2158 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2159 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2161 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2162 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2163 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2164 (altivec_lve<VI_char>x): Delete expand.
2165 (*altivec_lve<VI_char>x_internal): Rename to...
2166 (altivec_lve<VI_char>x): ... this.
2167 (altivec_lvxl_<mode>): Delete expand.
2168 (*altivec_lvxl_<mode>_internal): Rename to ...
2169 (altivec_lvxl_<mode>): ... this.
2170 (altivec_stvxl_<mode>): Delete expand.
2171 (*altivec_stvxl_<mode>_internal): Rename to ...
2172 (altivec_stvxl_<mode>): ... this.
2173 (altivec_stve<VI_char>x): Delete expand.
2174 (*altivec_stve<VI_char>x_internal): Rename to ...
2175 (altivec_stve<VI_char>x): ... this.
2176 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2177 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2178 reduc_plus_scal_<mode>): Adjust.
2179 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2181 (rs6000_cpu_cpp_builtins): Adjust.
2182 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2183 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2184 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2185 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2186 -maltivec=be support.
2187 (rs6000_split_vec_extract_var): Adjust.
2188 (rs6000_split_v4si_init): Adjust.
2189 (swap_selector_for_mode): Delete.
2190 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2191 altivec_expand_stvex_be): Delete.
2192 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2193 -maltivec=be support.
2194 (rs6000_gimple_fold_builtin): Ditto.
2195 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2197 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2198 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2199 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2200 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2201 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2202 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2203 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2204 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2205 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2206 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2207 anonymous split): Adjust.
2208 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2209 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2211 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2213 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2214 when --with-gxx-include-dir is also specified.
2215 * configure: Regenerate.
2217 2018-05-09 Jim Wilson <jimw@sifive.com>
2220 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2221 * config/riscv/t-withmultilib: New.
2222 * config/riscv/withmultilib.h: New.
2223 * doc/install.texi: Document RISC-V --with-multilib-list support.
2225 2018-05-09 Richard Biener <rguenther@suse.de>
2227 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2229 (vect_bb_vectorization_profitable_p): Adjust. Compute
2230 actual scalar cost using the cost vector and the add_stmt_cost
2233 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2235 PR rtl-optimization/85645
2236 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2237 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2239 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2241 PR rtl-optimization/85645
2242 * shrink-wrap.c (spread_components): Return a boolean saying if
2243 anything was changed.
2244 (try_shrink_wrapping_separate): Iterate spread_components until
2245 nothing changes anymore.
2247 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2249 PR rtl-optimization/85645
2250 * regrename.c (build_def_use): Also kill the chains that include the
2251 destination of a REG_CFA_REGISTER note.
2253 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2255 PR rtl-optimization/85645
2256 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2257 insn that has a REG_CFA_REGISTER note.
2259 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2261 * cfgexpand.c (expand_clobber): New function.
2262 (expand_gimple_stmt_1): Use it.
2263 * tree-vect-stmts.c (vect_clobber_variable): New function,
2265 (vectorizable_simd_clone_call): ...here.
2266 (vectorizable_store): Emit a clobber either side of an
2267 IFN_STORE_LANES sequence.
2268 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2270 2018-05-09 Tom de Vries <tom@codesourcery.com>
2273 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2274 (define_insn "trap_if_false"): Add exit after trap.
2276 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2278 PR rtl-optimization/85638
2279 * bb-reorder.c: Include common/common-target.h.
2280 (create_forwarder_block): New function extracted from...
2281 (fix_up_crossing_landing_pad): ...here. Rename into...
2282 (dw2_fix_up_crossing_landing_pad): ...this.
2283 (sjlj_fix_up_crossing_landing_pad): New function.
2284 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2285 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2286 from both partitions and exit the loop after one iteration.
2288 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2291 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2293 (Basic PowerPC Built-in Functions): The new name of the
2294 subsection previously known as "PowerPC Built-in Functions".
2295 (Basic PowerPC Built-in Functions Available on all Configurations):
2297 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2299 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2301 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2303 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2306 2018-05-08 Jim Wilson <jimw@sifive.com>
2308 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2309 (LD_EMUL_SUFFIX): New.
2310 (LINK_SPEC): Use it.
2312 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2314 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2316 (Basic PowerPC Built-in Functions): The new name of the
2317 subsection previously known as "PowerPC Built-in Functions".
2318 (Basic PowerPC Built-in Functions Available on all Configurations):
2320 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2322 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2324 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2326 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2329 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2332 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2333 after cmpelim optimization.
2335 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2337 * config.gcc: Support "goldmont".
2338 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2339 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2341 * config/i386/i386.c (m_GOLDMONT): Define.
2342 (processor_target_table): Add "goldmont".
2343 (PTA_GOLDMONT): Define.
2344 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2345 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2346 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2347 (fold_builtin_cpu): Add "goldmont".
2348 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2349 (ix86_option_override_internal): Add "goldmont".
2350 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2351 (processor_type): Add PROCESSOR_GOLDMONT.
2352 * config/i386/i386.md: Add CPU "glm".
2353 * config/i386/glm.md: New file.
2354 * config/i386/x86-tune.def: Add m_GOLDMONT.
2355 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2357 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2360 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2362 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2363 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2364 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2367 * config/i386/i386.c (ix86_fold_builtin): Handle
2368 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2371 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2372 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2374 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2377 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2378 (check_arch): Likewise.
2379 (check_fpu): Return the result rather than printing it.
2380 (end arch): Fix operator precedence.
2381 (end cpu): Likewise.
2382 (END): Print the result from check_fpu.
2384 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2385 Alan Hayward <alan.hayward@arm.com>
2386 David Sherwood <david.sherwood@arm.com>
2388 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2389 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2390 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2391 (*fcmuo<mode>_and): New patterns.
2393 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2395 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2396 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2397 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2398 (cmp_op, sve_imm_con): New code attributes.
2399 (SVE_COND_INT_CMP, imm_con): Delete.
2400 (cmp_op): Remove above unspecs from int attribute.
2401 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2403 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2404 comparison-specific unspecs.
2405 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2406 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2407 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2408 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2409 (*vec_fcm<cmp_op><mode>): Rename to...
2410 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2411 (*vec_fcmuo<mode>): Rename to...
2412 (*fcmuo<mode>): ...this and adjust likewise.
2413 (*pred_fcm<cmp_op><mode>): New pattern.
2414 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2415 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2417 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2419 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2420 (aarch64_emit_sve_predicated_cond): New function.
2421 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2422 (aarch64_emit_unspec_cond_or): Replace with...
2423 (aarch64_emit_sve_or_conds): ...this new function. Use
2424 aarch64_emit_sve_ptrue_op for the individual comparisons and
2425 aarch64_emit_binop to OR them together.
2426 (aarch64_emit_inverted_unspec_cond): Replace with...
2427 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2428 aarch64_emit_sve_ptrue_op for the comparison and
2429 aarch64_emit_unop to invert the result.
2430 (aarch64_expand_sve_vec_cmp_float): Update after the above
2431 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2433 2018-05-07 Nathan Sidwell <nathan@acm.org>
2435 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2436 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2437 (Backwards Compatibility): Likewise.
2439 2018-05-07 Luis Machado <luis.machado@linaro.org>
2443 2018-05-07 Luis Machado <luis.machado@linaro.org>
2445 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2446 <prefetch_dynamic_strides>: New const bool field.
2447 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2448 prefetch_dynamic_strides.
2449 (exynosm1_prefetch_tune): Likewise.
2450 (thunderxt88_prefetch_tune): Likewise.
2451 (thunderx_prefetch_tune): Likewise.
2452 (thunderx2t99_prefetch_tune): Likewise.
2453 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2454 (aarch64_override_options_internal): Update to set
2455 PARAM_PREFETCH_DYNAMIC_STRIDES.
2456 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2457 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2458 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2459 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2460 prefetch-dynamic-strides setting.
2462 2018-05-07 Luis Machado <luis.machado@linaro.org>
2464 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2465 <minimum_stride>: New const int field.
2466 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2467 minimum_stride field.
2468 (exynosm1_prefetch_tune): Likewise.
2469 (thunderxt88_prefetch_tune): Likewise.
2470 (thunderx_prefetch_tune): Likewise.
2471 (thunderx2t99_prefetch_tune): Likewise.
2472 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2473 (aarch64_override_options_internal): Update to set
2474 PARAM_PREFETCH_MINIMUM_STRIDE.
2475 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2476 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2477 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2478 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2479 stride is constant and is below the minimum stride threshold.
2481 2018-05-07 Luis Machado <luis.machado@linaro.org>
2483 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2486 2018-05-07 Luis Machado <luis.machado@linaro.org>
2488 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2489 <prefetch_dynamic_strides>: New const bool field.
2490 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2491 prefetch_dynamic_strides.
2492 (exynosm1_prefetch_tune): Likewise.
2493 (thunderxt88_prefetch_tune): Likewise.
2494 (thunderx_prefetch_tune): Likewise.
2495 (thunderx2t99_prefetch_tune): Likewise.
2496 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2497 (aarch64_override_options_internal): Update to set
2498 PARAM_PREFETCH_DYNAMIC_STRIDES.
2499 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2500 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2501 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2502 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2503 prefetch-dynamic-strides setting.
2505 2018-05-07 Luis Machado <luis.machado@linaro.org>
2507 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2508 <minimum_stride>: New const int field.
2509 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2510 minimum_stride field.
2511 (exynosm1_prefetch_tune): Likewise.
2512 (thunderxt88_prefetch_tune): Likewise.
2513 (thunderx_prefetch_tune): Likewise.
2514 (thunderx2t99_prefetch_tune): Likewise.
2515 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2516 (aarch64_override_options_internal): Update to set
2517 PARAM_PREFETCH_MINIMUM_STRIDE.
2518 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2519 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2520 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2521 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2522 stride is constant and is below the minimum stride threshold.
2524 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2527 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2528 the type is addressable. Don't force op into register if it has
2531 2018-05-05 Roland McGrath <mcgrathr@google.com>
2534 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2535 any section for which we don't know a specific type it should have,
2536 regardless of name. Previously this was done only for the exact
2537 names ".init_array", ".fini_array", and ".preinit_array".
2538 (default_elf_asm_named_section): Add comment about
2539 relationship with default_section_type_flags and SECTION_NOTYPE.
2540 (get_section): Don't consider it a type conflict if one side has
2541 SECTION_NOTYPE and the other doesn't, as long as neither has the
2542 SECTION_BSS et al used in the default_section_type_flags logic.
2544 2018-05-05 Tom de Vries <tom@codesourcery.com>
2547 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2548 (workaround_barsyncs): New function.
2549 (nvptx_reorg): Use workaround_barsyncs.
2550 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2551 (define_expand "nvptx_membar_cta"): New define_expand.
2552 (define_insn "*nvptx_membar_cta"): New insn.
2554 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2556 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2557 To improve optimization opportunities.
2558 * builtin-types.def: The new needed builtin types for the above.
2560 2018-05-04 Richard Biener <rguenther@suse.de>
2562 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2563 * gimple-ssa-store-merging.c
2564 (imm_store_chain_info::output_merged_store): Remove redundant create,
2565 release split_store vector contents on failure.
2566 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2567 scalar stmt vector on cache hit.
2569 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2571 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2573 * config.gcc (powerpc-xilinx-eabi*): Remove.
2574 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2576 (fusion_addis_mem_combo_load): Ditto.
2577 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2579 (rs6000_cpu_cpp_builtins): Ditto.
2580 * config/rs6000/rs6000-linux.c
2581 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2582 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2583 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2585 (rs6000_setup_reg_addr_masks): Ditto.
2586 (rs6000_init_hard_regno_mode_ok): Ditto.
2587 (rs6000_option_override_internal): Ditto.
2588 (legitimate_lo_sum_address_p): Ditto.
2589 (rs6000_legitimize_address): Ditto.
2590 (rs6000_legitimize_reload_address): Ditto.
2591 (rs6000_legitimate_address_p): Ditto.
2592 (abi_v4_pass_in_fpr): Ditto.
2593 (setup_incoming_varargs): Ditto.
2594 (rs6000_gimplify_va_arg): Ditto.
2595 (rs6000_split_multireg_move): Ditto.
2596 (rs6000_savres_strategy): Ditto.
2597 (rs6000_emit_prologue_components): Ditto.
2598 (rs6000_emit_epilogue_components): Ditto.
2599 (rs6000_emit_prologue): Ditto.
2600 (rs6000_emit_epilogue): Ditto.
2601 (rs6000_elf_file_end): Ditto.
2602 (rs6000_function_value): Ditto.
2603 (rs6000_libcall_value): Ditto.
2604 * config/rs6000/rs6000.h: Ditto.
2605 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2606 (TARGET_MINMAX): ... this. New.
2607 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2608 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2609 (*movsi_internal1_single): Delete.
2610 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2611 mfpu=, mxilinx-fpu): Delete.
2612 * config/rs6000/singlefp.h: Delete.
2613 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2614 * config/rs6000/t-rs6000: Ditto.
2615 * config/rs6000/t-xilinx: Delete.
2616 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2617 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2618 (VStype_simple): Delete.
2619 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2620 * config/rs6000/xfpu.h: Delete.
2621 * config/rs6000/xfpu.md: Delete.
2622 * config/rs6000/xilinx.h: Delete.
2623 * config/rs6000/xilinx.opt: Delete.
2624 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2625 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2627 2018-05-04 Tom de Vries <tom@codesourcery.com>
2630 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2633 2018-05-04 Richard Biener <rguenther@suse.de>
2636 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2637 (expand_complex_div_wide): Likewise.
2638 (expand_complex_operations_1): Likewise.
2639 (expand_complex_libcall): Preserve EH info of the original stmt.
2640 (tree_lower_complex): Handle removed blocks.
2641 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2642 on complex multiplication and division libcall builtins.
2644 2018-05-04 Richard Biener <rguenther@suse.de>
2647 * fold-const.c (negate_expr_p): Restrict negation of operand
2648 zero of a division to when we know that can happen without
2650 (fold_negate_expr_1): Likewise.
2652 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2655 * real.h (real_nextafter): Declare.
2656 * real.c (real_nextafter): New function.
2657 * fold-const-call.c (fold_const_nextafter): New function.
2658 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2659 CASE_CFN_NEXTTOWARD.
2660 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2661 even when arg1_mode is different from arg0_mode.
2663 2018-05-03 Nathan Sidwell <nathan@acm.org>
2665 * doc/extend.texi (Deprecated Features): Remove
2667 (Backwards Compatibility): Likewise.
2668 * doc/invoke.texi (C++ Language Options): Likewise.
2669 (C++ Dialect Options): Likewise.
2671 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2674 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2675 _mm512_mask_mullox_epi64): New intrinsics.
2677 2018-05-03 Tom de Vries <tom@codesourcery.com>
2680 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2681 dump files): Add offload-tree.
2683 2018-05-03 Richard Biener <rguenther@suse.de>
2685 PR tree-optimization/85615
2686 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2687 to loops not nested in BBs loop father to avoid creating multi-entry
2690 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2692 PR tree-optimization/70291
2693 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2694 arguments. Change return type to tree. Emit libcall as a new
2695 statement rather than replacing existing one when inplace_p is true.
2696 (expand_complex_multiplication_components): New function.
2697 (expand_complex_multiplication): Expand floating-point complex
2698 multiplication using the above.
2699 (expand_complex_division): Rename inner_type parameter to type.
2700 Update expand_complex_libcall call-site.
2701 (expand_complex_operations_1): Update expand_complex_multiplication
2702 and expand_complex_division call-sites.
2704 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2707 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2708 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2709 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2710 the highest significant bit of the shift count mask is clear. In
2711 check whether and[sq]i3 is needed verify that all significant bits
2712 of the shift count other than the highest are set.
2714 2018-05-02 Tom de Vries <tom@codesourcery.com>
2717 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2718 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2719 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2720 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2721 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2722 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2723 __builtin_goacc_parlevel_size.
2725 2018-05-02 Richard Biener <rguenther@suse.de>
2727 PR tree-optimization/85597
2728 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2729 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2732 2018-05-02 Tom de Vries <tom@codesourcery.com>
2735 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2736 dump files): Add ltrans-tree.
2738 2018-05-02 Tom de Vries <tom@codesourcery.com>
2741 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2742 dump files): Add wpa-ipa.
2744 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2746 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2747 powerpc*-*-linux*paired* target.
2748 * config/rs6000/750cl.h: Delete.
2749 * config/rs6000/paired.h: Delete.
2750 * config/rs6000/paired.md: Delete.
2751 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2753 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2754 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2755 comment. Remove paired float support.
2756 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2757 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2759 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2760 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2762 * config/rs6000/rs6000.c: Remove paired float support.
2763 (paired_expand_vector_init, paired_expand_vector_move,
2764 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2765 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2766 paired_expand_builtin, paired_expand_predicate_builtin,
2767 paired_init_builtins): Delete.
2768 * config/rs6000/rs6000.h: Remove paired float support.
2769 * config/rs6000/rs6000.md: Remove paired float support.
2770 (move_from_CR_ov_bit): Delete.
2771 * config/rs6000/rs6000.opt (mpaired): Delete.
2772 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2773 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2775 2018-05-02 Richard Biener <rguenther@suse.de>
2778 * gimplify.c (gimplify_save_expr): When in SSA form allow
2779 SAVE_EXPRs to compute to SSA vars.
2781 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2784 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2785 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2786 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2787 clobber operands[2], instead use a new pseudo. Formatting fixes.
2789 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2791 PR tree-optimization/85586
2792 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2793 exit early for statements in the same group if the accesses are
2796 2018-05-02 Tom de Vries <tom@codesourcery.com>
2799 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2802 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2804 PR tree-optimization/85143
2805 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2807 2018-05-01 Tom de Vries <tom@codesourcery.com>
2810 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2811 not found" error message.
2813 2018-05-01 Tom de Vries <tom@codesourcery.com>
2816 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2817 * vec.c (test_ordered_remove_if): New function.
2818 (vec_c_tests): Call test_ordered_remove_if.
2819 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2820 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2821 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2822 VEC_ORDERED_REMOVE_IF.
2824 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2826 PR tree-optimization/82665
2827 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2828 pointer subtraction where arguments come from a memchr call.
2830 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2832 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2833 --push-state --as-needed and --pop-state instead of --as-needed and
2834 --no-as-needed if ld supports it.
2835 * configure: Regenerated.
2838 * doc/install.texi2html: Replace _002d with - and _002a with * in
2839 generated html files using sed.
2841 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2844 * gcc-rich-location.c (blank_line_before_p): New function.
2845 (use_new_line): New function.
2846 (gcc_rich_location::add_fixit_insert_formatted): New function.
2847 * gcc-rich-location.h
2848 (gcc_rich_location::add_fixit_insert_formatted): New function.
2850 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2852 * selftest.c (assert_streq): Rename "expected" and "actual" to
2853 "val1" and "val2". Extend NULL-handling to cover both inputs
2854 symmetrically, while still requiring both to be non-NULL for a pass.
2855 * selftest.h (assert_streq): Rename "expected" and "actual" to
2857 (ASSERT_EQ): Likewise.
2858 (ASSERT_EQ_AT): Likewise.
2859 (ASSERT_KNOWN_EQ): Likewise.
2860 (ASSERT_KNOWN_EQ_AT): Likewise.
2861 (ASSERT_NE): Likewise.
2862 (ASSERT_MAYBE_NE): Likewise.
2863 (ASSERT_MAYBE_NE_AT): Likewise.
2864 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2865 the assertion to pass.
2866 (ASSERT_STREQ_AT): Likewise.
2868 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2870 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2871 interaction with -pie.
2873 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2875 * selftest.h: Fix alphabetization of per-source-file selftest
2878 2018-04-30 Jason Merrill <jason@redhat.com>
2880 PR c++/61982 - dead stores to destroyed objects.
2881 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2884 2018-04-30 Jason Merrill <jason@redhat.com>
2886 * tree.c (build_clobber): New.
2887 * tree.h: Declare it.
2888 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2890 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2892 * diagnostic-show-locus.c (layout::layout): Update for
2893 location_get_source_line returning a char_span.
2894 (struct char_span): Move to input.h.
2895 (struct correction): Update for fields in char_span becoming
2897 (struct source_line): Update for location_get_source_line
2898 returning a char_span.
2899 (layout::print_line): Likewise.
2900 * edit-context.c (edited_file::print_content): Likewise.
2901 (edited_file::print_diff_hunk): Likewise.
2902 (edited_file::print_run_of_changed_lines): Likewise.
2903 (edited_file::get_num_lines): Likewise.
2904 (edited_line::edited_line): Likewise.
2905 * final.c (asm_show_source): Likewise.
2906 * input.c (location_get_source_line): Convert return type
2907 from const char * to char_span, losing the final "line_len"
2909 (dump_location_info): Update for the above.
2910 (get_substring_ranges_for_loc): Likewise. Use a char_span
2911 when handling the literal within the line.
2912 (test_reading_source_line): Update for location_get_source_line
2913 returning a char_span.
2914 * input.h (class char_span): Move here from
2915 diagnostic-show-locus.c, converting from a struct to a class.
2916 Make data members private.
2917 (char_span::operator bool): New.
2918 (char_span::length): New.
2919 (char_span::get_buffer): New.
2920 (char_span::operator[]): New.
2921 (char_span::subspan): Make const.
2922 (char_span::xstrdup): New.
2923 (location_get_source_line): Convert return type from const char *
2924 to char_span, losing the final "line_size" param.
2926 2018-04-30 Jan Hubicka <jh@suse.cz>
2928 * lto-wrapper.c (ltrans_priorities): New static var.
2929 (cmp_priority): New.
2930 (run_gcc): Read priorities and if doing parallel build order
2931 the Makefile by them.
2933 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2935 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2937 2018-04-30 Richard Biener <rguenther@suse.de>
2939 * tree-cfg.c (verify_address): Remove base argument, add
2940 flag whether to check TREE_ADDRESSABLE and do that.
2941 (verify_expr): Remove.
2942 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2943 (verify_gimple_assign_single): Likewise.
2944 (verify_gimple_switch): Likewise.
2945 (verify_expr_location_1): Dereference tp once. Add (disabled)
2946 piece from verify_expr.
2947 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2949 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2951 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2953 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2955 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2956 (small_data_pattern): Likewise.
2957 (arc_rewrite_small_data): Likewise.
2958 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2959 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2960 (get_symbol_alignment): New function.
2961 (legitimate_small_data_address_p): Likewise.
2962 (legitimate_scaled_address): Update, call
2963 legitimate_small_data_address_p.
2964 (output_sdata): New static variable.
2965 (arc_print_operand): Update how we handle small data operands.
2966 (arc_print_operand_address): Likewise.
2967 (arc_legitimate_address_p): Update, use
2968 legitimate_small_data_address_p.
2969 (arc_rewrite_small_data_p): Remove.
2970 (arc_rewrite_small_data_1): Likewise.
2971 (arc_rewrite_small_data): Likewise.
2972 (small_data_pattern): Likewise.
2973 (compact_sda_memory_operand): Update to use
2974 legitimate_small_data_address_p and get_symbol_alignment.
2975 (prepare_move_operands): Don't rewite sdata pattern.
2976 (prepare_extend_operands): Remove.
2977 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2979 (zero_extendqisi2): Likewise.
2980 (zero_extendhisi2): Likewise.
2981 (extendqihi2): Likewise.
2982 (extendqisi2): Likewise.
2983 (extendhisi2): Likewise.
2987 * config/arc/constraints.md (Usd): Change it to memory constraint.
2989 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2991 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2992 as source of std instructions.
2993 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2994 allow 6-bit constants as source for store instructions.
2995 (movdi_insn): Update instruction pattern to allow 6-bit constants
2996 as source for store instructions.
2998 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3000 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3002 2018-04-30 Nathan Sidwell <nathan@acm.org>
3003 Sandra Loosemore <sandra@codesourcery.com>
3005 * dumpfile.c (dump_open): Allow '-' for stdout.
3006 * doc/invoke.texi (Developer Options): Document dump filename
3007 determination early. Document stdin/stdout selection.
3009 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3011 Microblaze Target: PIC data text relative
3013 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3014 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3016 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3017 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3018 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3019 New addressing mode for data-text relative position indepenedent code.
3020 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3021 'ADDRESS_SYMBOLIC_TXT_REL'.
3022 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3023 (microblaze_legitimate_pic_operand): Exclude function calls from
3024 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3025 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3027 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3028 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3029 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3030 for 'address + offset'.
3031 (microblaze_expand_prologue): Add new function prologue call for
3033 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3034 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3035 table in case of TARGET_PIC_DATA_TEXT_REL.
3036 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3037 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3038 Add new macros 'UNSPEC_TEXT',
3039 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3040 + exclude function calls from 'UNSPEC_PLT' in case of data text
3042 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3043 new target hook for generating address diff vector tables in case of
3045 * doc/tm.texi : Regenerate.
3046 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3047 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3048 of addr diff vector generation.
3049 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3050 target hook definition.
3051 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3052 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3053 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3054 option for data text relative.
3056 2018-04-30 Richard Biener <rguenther@suse.de>
3058 * tree-chrec.h (evolution_function_is_constant_p): Remove
3060 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3062 2018-04-30 Richard Biener <rguenther@suse.de>
3065 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3067 2018-04-30 Richard Biener <rguenther@suse.de>
3069 PR tree-optimization/28364
3070 PR tree-optimization/85275
3071 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3072 copying first exit test.
3074 2018-04-28 Mark Wielaard <mark@klomp.org>
3076 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3078 (dwarf_AT): Handle DW_AT_addr_base.
3079 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3081 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3084 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3085 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3086 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3087 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3089 2018-04-28 Richard Biener <rguenther@suse.de>
3091 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3092 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3093 to reflect use. Only add interesting stmts.
3095 2018-04-27 Martin Jambor <mjambor@suse.cz>
3098 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3099 the jump function allows for passing through aggregate values.
3101 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3103 * input.h (in_system_header_at): Convert from macro to inline
3105 (from_macro_expansion_at): Likewise.
3106 (from_macro_definition_at): Likewise.
3108 2018-04-27 Jeff Law <law@redhat.com>
3110 * config.gcc: Mark tile* targets as deprecated/obsolete.
3112 2018-04-27 Richard Biener <rguenther@suse.de>
3114 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3117 2018-04-27 Richard Biener <rguenther@suse.de>
3119 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3121 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3123 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3124 with Yd constraint. Set "preferred_for_speed" attribute from
3125 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3127 (*movdi_internal): Ditto.
3128 (movti_interunit splitters): Remove
3129 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3130 (movdi_interunit splitters): Ditto.
3131 * config/i386/constraints.md (Ye): Remove.
3132 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3134 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3137 * config/aarch64/constraints.md (Usg): Limit to 31.
3140 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3142 PR tree-optimization/85529
3143 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3144 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3145 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3146 zero extension or masking of the MSB bit.
3147 (optimize_range_tests): Add FIRST_BB argument, pass it through
3148 to optimize_range_tests_var_bound.
3149 (maybe_optimize_range_tests, reassociate_bb): Adjust
3150 optimize_range_tests callers.
3152 2018-04-26 Richard Biener <rguenther@suse.de>
3153 Jakub Jelinek <jakub@redhat.com>
3155 * cgraph.h (symbol_table): Just declare debug method here.
3156 * symtab.c (symbol_table::debug): Define.
3158 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3160 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3162 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3164 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3165 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3166 (*movdi_internal): Substitute Yi and Yj constraint with x
3167 and Ym and Yn constraint with y constraint. Update "isa"
3168 attribute and set "preferred_for_speed" attribute from
3169 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3170 (*movsi_internal): Ditto.
3171 (*movdf_internal): Ditto.
3172 (*movsf_internal): Ditto.
3173 (*zero_extendsidi2): Ditto.
3174 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3175 (sse2_loadld): Ditto.
3176 (*vec_extract<ssevecmodelower>_0): Ditto.
3177 (*vec_extractv4si_0_zext_sse4): Ditto.
3178 (vec_concatv2di): Ditto.
3179 (*vec_dup<mode>): Ditto.
3180 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3181 * config/i386/constraints.md (Yi): Remove.
3186 2018-04-26 Nathan Sidwell <nathan@acm.org>
3188 * dumpfile.c (dump_open): New.
3189 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3190 (dump_finish): Detect stdio/stderr by value not name.
3192 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3194 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3196 2018-04-26 Tom de Vries <tom@codesourcery.com>
3199 * config/nvptx/nvptx.c (verify_neutering_jumps)
3200 (verify_neutering_labels): New function
3201 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3203 2018-04-26 Tom de Vries <tom@codesourcery.com>
3206 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3207 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3210 2018-04-26 Richard Biener <rguenther@suse.de>
3211 Tom de Vries <tom@codesourcery.com>
3214 * lto-streamer-out.c (output_function): Fixup loops if required to match
3215 discovery done in the reader.
3217 2018-04-26 Richard Biener <rguenther@suse.de>
3219 PR tree-optimization/85116
3220 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3221 have a loop exit from the single latch predecessor. Remove
3222 case of header with just condition.
3223 (ch_base::copy_headers): Exclude infinite loops from any
3225 (pass_ch::execute): Record exits.
3227 2018-04-26 Richard Biener <rguenther@suse.de>
3229 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3230 prologue cost vector and pass it to vect_get_load_cost.
3231 (vect_get_peeling_costs_all_drs): Likewise.
3232 (vect_peeling_hash_get_lowest_cost): Likewise.
3233 (vect_enhance_data_refs_alignment): Likewise.
3235 2018-04-26 Richard Biener <rguenther@suse.de>
3238 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3239 checking of integer<->pointer conversions.
3240 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3241 sign-/zero-extending pointer types.
3242 (expand_omp_for_static_chunk): Likewise.
3244 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3245 Jean Lee <xiaoyur347@gmail.com>
3247 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3248 (TARGET_ASAN_SHADOW_OFFSET): Define.
3249 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3250 true for -fsanitize=address.
3252 2018-04-25 Mark Wielaard <mark@klomp.org>
3254 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3257 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3259 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3260 than "alu", remove explicit "memory" and "imm_disp" attributes.
3261 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3264 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3265 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3266 gen_lowpart_no_emit.
3268 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3271 * config/i386/i386.c (ix86_expand_builtin): Change memory
3272 operand to XI, extend p0 to Pmode.
3273 * config/i386/i386.md: Change unspec volatile and operand
3274 1 mode to XI, change operand 0 mode to P.
3276 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3278 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3279 GET_MODE_MASK before any checking.
3280 (nds32_can_use_bset_p): Likewise.
3281 (nds32_can_use_btgl_p): Likewise.
3283 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32-doubleword.md: New define_split pattern for
3286 illegal register number.
3288 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3290 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3292 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3296 2018-04-25 Richard Biener <rguenther@suse.de>
3298 * lto-streamer.h (LTO_major_version): Bump to 8.
3300 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3302 * BASE-VER: Set to 9.0.0.
3304 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3306 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3307 in __abskf2 and __powikf2.
3309 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3312 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3313 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3314 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3315 Use the above on operand 2. Reindent.
3316 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3318 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3321 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3323 * config/i386/i386.opt (mcet): Removed.
3324 * doc/install.texi: Remove -mcet documentation.
3325 * doc/invoke.texi: Likewise.
3327 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3330 * doc/install.texi: Remove -mcet from bootstrap-cet.
3332 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3335 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3336 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3340 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3341 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3342 containing a CONST_VECTOR.
3344 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3346 * doc/install.texi: Update newlib dependency for nvptx.
3348 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3351 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3352 instead of INTVAL when shifting x left.
3354 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3356 PR tree-optimization/85478
3357 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3358 vect_grouped_store_supported for single element vectors.
3360 2018-04-24 Richard Biener <rguenther@suse.de>
3363 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3364 load cost increase to the case of non-constant step.
3366 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3369 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3370 destination if any_malformed_asm.
3372 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3375 * expr.c (store_field): In the bitfield case, if the value comes from
3376 a function call and is returned in registers by means of a PARALLEL,
3377 do not change the mode of the temporary unless BLKmode and VOIDmode.
3379 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3381 PR rtl-optimization/85423
3382 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3383 dependencies to debug insns when the previous insn is non-debug.
3385 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3387 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3388 enums into a single definition.
3389 (fls): Fix predicates and printing.
3392 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3394 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3395 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3396 and short u6 immediate.
3397 (check_if_valid_sleep_operand): Remove.
3398 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3400 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3403 flag_always_save_lp condition.
3404 * config/nds32/nds32.opt (malways-save-lp): New option.
3406 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3408 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3409 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3410 * config/nds32/nds32.h
3411 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3412 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3414 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3416 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3417 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3419 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3420 Chung-Ju Wu <jasonwucj@gmail.com>
3422 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3424 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3425 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3427 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3429 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3431 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3433 * config/nds32/nds32-protos.h (nds32_data_alignment,
3434 nds32_local_alignment): Declare.
3435 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3436 nds32_local_alignment): New functions.
3437 (TARGET_CONSTANT_ALIGNMENT): Define.
3438 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3440 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3442 * config/nds32/nds32.c
3443 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3444 (TARGET_MODES_TIEABLE_P): Likewise.
3446 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3448 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3451 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3452 Chung-Ju Wu <jasonwucj@gmail.com>
3454 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3455 for unaligned access.
3456 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3458 * config/nds32/nds32-intrinsic.md: Likewise.
3459 * config/nds32/nds32_intrinsic.h: Likewise.
3460 * config/nds32/nds32.h (nds32_builtins): Likewise.
3461 * config/nds32/nds32.opt (munaligned-access): New option.
3462 * config/nds32/nds32.c (nds32_asm_file_start): Display
3463 flag_unaligned_access status.
3465 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3467 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3468 -mno-relax is present.
3469 * config/riscv/linux.h (LINK_SPEC): Ditto.
3471 2018-04-20 Martin Sebor <msebor@redhat.com>
3474 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3476 (gimple_fold_builtin_stxcpy_chk): Same.
3477 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3479 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3482 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3483 __powikf2 when long double is IEEE 128-bit.
3485 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3487 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3488 step to make sure stack always aligned.
3490 2018-04-20 Carl Love <cel@us.ibm.com>
3493 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3494 size check for arg0.
3496 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3497 Tom de Vries <tom@codesourcery.com>
3500 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3501 Emit insns for calls too.
3502 (nvptx_find_par): Always look for worker-level predecessor insn.
3503 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3505 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3506 (nvptx_process_pars): Propagate frames for calls.
3508 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3511 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3513 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3514 (ix86_handle_option): Don't handle OPT_mibt.
3515 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3517 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3519 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3520 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3521 (ix86_target_macros): Define __CET__ with flag_cf_protection
3522 for -fcf-protection.
3523 * config/i386/i386.c (isa2_opts): Remove -mibt.
3524 * config/i386/i386.h (TARGET_IBT): Removed.
3525 (TARGET_IBT_P): Likewise.
3526 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3527 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3528 * config/i386/i386.opt (mcet): Update help message.
3531 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3532 -mcet as an alias for -mshstk.
3534 2018-04-20 Richard Biener <rguenther@suse.de>
3537 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3538 complexity by forcing a single use of the multiply operand.
3540 2018-04-20 Martin Jambor <mjambor@suse.cz>
3543 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3544 recursion dependency to only apply to non-clones.
3546 2018-04-20 Martin Jambor <mjambor@suse.cz>
3549 * ipa-cp.c (create_specialized_node): Check that clones of
3550 self-recursive edges exist during IPA-CP.
3552 2018-04-19 Toon Moene <toon@moene.org>
3554 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3557 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3559 PR tree-optimization/85467
3560 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3561 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3562 VECTOR_CST element to type.
3564 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3567 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3568 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3569 (builtin_longjmp): Likewise.
3570 (save_stack_nonlocal): New pattern.
3571 (restore_stack_nonlocal): Likewise.
3573 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3576 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3577 Replace ASM_OUTPUT_LABEL with fprintf.
3579 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3582 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3583 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3584 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3585 define __IBT__ and __SHSTK__ for -fcf-protection.
3586 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3588 (ix86_trampoline_init): Likewise.
3589 (x86_output_mi_thunk): Likewise.
3590 (ix86_notrack_prefixed_insn_p): Likewise.
3591 (ix86_option_override_internal): Don't disallow -fcf-protection.
3592 * config/i386/i386.md (rdssp<mode>): Also enable for
3594 (incssp<mode>): Likewise.
3595 (nop_endbr): Likewise.
3596 * config/i386/i386.opt (mcet): Change help message to built-in
3600 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3601 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3602 enable CET built-in functions.
3604 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3606 * common/config/i386/i386-common.c
3607 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3608 OPTION_MASK_ISA_MOVDIRI_UNSET,
3609 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3610 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3611 * config.gcc (movdirintrin.h): New header.
3612 * config/i386/cpuid.h (bit_MOVDIRI,
3613 bit_MOVDIR64B): New bits.
3614 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3616 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3617 (VOID, PVOID, PCVOID)): New function types.
3618 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3619 __builtin_ia32_directstoreu_u64,
3620 __builtin_ia32_movdir64b): New builtins.
3621 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3622 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3624 (ix86_valid_target_attribute_inner_p): Ditto.
3625 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3626 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3627 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3628 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3629 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3630 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3631 (movdiri<mode>, movdir64b_<mode>): New.
3632 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3633 * config/i386/immintrin.h: Include movdirintrin.h.
3634 * config/i386/movdirintrin.h: New file.
3635 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3637 2018-04-19 Richard Biener <rguenther@suse.de>
3640 * cfg.c (clear_bb_flags): When loop state says we have
3641 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3643 2018-04-19 Richard Biener <rguenther@suse.de>
3645 PR tree-optimization/84737
3646 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3647 copying restrict info.
3648 (vect_setup_realignment): Use it.
3649 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3650 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3651 the first DR to all generated stores.
3652 (vectorizable_load): Likewise for loads.
3654 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3656 PR tree-optimization/85446
3657 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3658 the integral and pointer types to have the same precision.
3660 * doc/install.texi: Document --disable-cet being the default and
3663 2018-04-18 Martin Liska <mliska@suse.cz>
3665 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3668 2018-04-18 Martin Liska <mliska@suse.cz>
3671 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3674 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3675 arguments if they are comparable.
3677 2018-04-18 Martin Liska <mliska@suse.cz>
3680 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3683 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3686 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3689 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3690 ENDBR after calling __morestack.
3692 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3695 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3696 by using gcc_base_ver to generate a gcc_driver_version, and use
3697 it when generating GCC_DRIVER_NAME.
3698 * configure: Regenerate.
3700 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3703 * config.gcc: Obsolete powerpc*-*-*spe*.
3705 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3708 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3709 (stabstr_D): Change type of unum from unsigned int to
3710 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3713 2018-04-17 Jim Wilson <jimw@sifive.com>
3716 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3717 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3718 Set arg_pointer_offset after using pretend_args_size.
3720 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3722 PR rtl-optimization/85431
3723 * dse.c (record_store): Ignore zero width stores.
3726 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3727 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3728 __builtin_stack_restore rather than after it.
3729 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3730 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3731 argument instead of virtual_dynamic_stack_rtx.
3733 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3735 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3737 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3738 Add note to error message to explain internal mapping of overloaded
3739 built-in function name to non-overloaded built-in function name.
3740 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3743 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3746 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3747 where the inputs overlap with the output.
3749 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3751 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3752 (=v, v) alternative and explicit "memory" attribute.
3753 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3754 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3756 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3757 "sselog1" type instead of "sselog".
3758 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3759 "sselog". Remove explicit "memory" attribute.
3760 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3761 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3763 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3764 "sselog1" type instead of "sselog". Remove explicit "memory"
3766 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3767 use "sselog1" type instead of "sselog". Remove explicit "memory"
3769 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3770 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3772 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3773 "sselog1" type instead of "sselog". Remove explicit "memory"
3775 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3776 use "sselog1" type instead of "sselog". Remove explicit "memory"
3780 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3783 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3786 2018-04-17 Martin Jambor <mjambor@suse.cz>
3789 * ipa-cp.c (create_specialized_node): Call
3790 expand_all_artificial_thunks if necessary.
3792 2018-04-17 Martin Liska <mliska@suse.cz>
3795 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3796 in message, remote space in between '_G' and '('.
3798 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3801 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3802 avx512f_vmcmp<mode>3<round_saeonly_name>,
3803 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3804 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3805 avx512f_rndscale<mode><round_saeonly_name>,
3806 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3807 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3808 Use %<iptr>2 instead of %2 for -masm=intel.
3809 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3810 avx512f_vcvttss2usi<round_saeonly_name>,
3811 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3813 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3814 avx512f_vcvttsd2usi<round_saeonly_name>,
3815 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3816 Use %q1 instead of %1 for -masm=intel.
3817 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3818 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3819 of %3 for -masm=intel.
3820 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3822 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3824 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3826 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3827 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3829 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3831 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3832 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3833 %g1 and one with %0 and %1.
3834 (avx512er_vmrcp28<mode><round_saeonly_name>,
3835 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3837 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3838 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3839 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3840 of %0 and %{%4%} for -masm=intel.
3841 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3842 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3843 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3844 order of %0 and %{%5%}%{z%} for -masm=intel.
3846 2018-04-17 Jan Hubicka <jh@suse.cz>
3849 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3851 2018-04-17 Martin Liska <mliska@suse.cz>
3854 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3855 for target_clone error message. Make default implementation
3856 clone to be a local declaration.
3857 (separate_attrs): Add new argument and check for an empty
3859 (expand_target_clones): Handle it.
3860 (ipa_target_clone): Make redirection just for target_clones
3863 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3864 Tom de Vries <tom@codesourcery.com>
3867 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3868 tiled basic blocks without omp continue statements.
3870 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3873 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3874 vec_extract expression as having side effects to make sure it gets
3877 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3880 * config/i386/i386.c (get_builtin_code_for_version): Check
3883 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3886 * gcc/config.gcc: Support "skylake".
3887 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3889 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3890 (processor_target_table): Add "skylake".
3891 (ix86_option_override_internal): Add "skylake".
3892 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3893 PROCESSOR_CANNONLAKE.
3894 (get_builtin_code_for_version): Fix priority for
3895 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3896 PROCESSOR_SKYLAKE-AVX512.
3897 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3898 (processor_type): Add PROCESSOR_SKYLAKE.
3900 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3901 Jason Merrill <jason@redhat.com>
3904 * convert.c (convert_to_integer_1): Use direct recursion for
3905 enumeral types and types with a precision less than the number
3906 of bits in their mode.
3908 2018-04-16 Julia Koval <julia.koval@intel.com>
3911 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3912 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3914 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3917 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3918 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3919 and -mno-direct-move.
3921 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3924 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3925 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3926 Ensure negative shifts result in {0}.
3928 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3930 PR rtl-optimization/79916
3931 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3932 regs (if any) to define how to gnerate SD moves when LRA is in
3935 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3937 PR rtl-optimization/85393
3938 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3939 * except.c (expand_dw2_landing_pad_for_region): Make static.
3940 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3941 a label and unconditional jump to old_bb, rather than
3942 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3945 PR rtl-optimization/85376
3946 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3947 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3948 instead of a specific value.
3950 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3951 Bin Cheng <bin.cheng@arm.com>
3953 PR tree-optimization/82965
3954 PR tree-optimization/83991
3955 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3956 by_profile_only parameter.
3957 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3958 information if the loop was predicted to iterate too many times.
3959 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3961 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3964 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3967 2018-04-13 Martin Liska <mliska@suse.cz>
3968 Jakub Jelinek <jakub@redhat.com>
3971 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3972 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3973 * builtins.c (expand_builtin_memory_copy_args): Use
3974 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3975 handle dest_addr == pc_rtx.
3977 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3980 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3981 asked to not generate direct moves.
3982 (fix_trunc<mode>si2_stfiwx): Similar.
3983 (fix_trunc<mode>si2_internal): Similar.
3985 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3988 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3989 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3990 lookup if dest in some wider mode is known to be const0_rtx and
3991 if so, record permanent equivalence for it to be ZERO_EXTEND of
3992 the narrower mode destination.
3994 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3996 * lto-streamer-out.c (output_function): Revert 259346.
3997 * omp-expand.c (expand_oacc_for): Likewise.
3999 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4001 PR rtl-optimization/85354
4002 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4003 * sel-sched.c (sel_global_init): ... here.
4005 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4008 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4009 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4010 mode for PE-COFF targets.
4011 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4012 (i386_pe_asm_lto_end): Likewise.
4013 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4014 (TARGET_ASM_LTO_END): Likewise.
4015 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4016 (i386_pe_asm_lto_start): New function.
4017 (i386_pe_asm_lto_end): Likewise.
4019 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4020 Richard Biener <rguenther@suse.de>
4023 * lto-streamer-out.c (output_function): Fix CFG loop state before
4025 * omp-expand.c (expand_oacc_for): Handle calls to internal
4026 functions like regular functions.
4028 2018-04-12 Richard Biener <rguenther@suse.de>
4031 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4032 for the early LTO debug to properly generate references to it
4033 during DIE emission. Do not re-use that for the skeleton for
4035 (dwarf2out_early_finish): Likewise.
4037 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4040 * config/i386/sse.md
4041 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4042 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4043 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4044 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4045 and output is a reg, avoid creating invalid lowpart subreg, but
4046 instead split into a 512-bit move. Don't split if not AVX512VL,
4047 input is xmm16+ reg and output is a mem.
4048 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4049 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4050 xmm16+ reg and output is a mem.
4052 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4054 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4055 also for flag_dwarf2_cfi_asm.
4057 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4059 PR rtl-optimization/85342
4060 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4061 a bool scalar var inside of the loop instead. Don't try to update
4062 recog_data.operand after failed apply_change_group.
4064 2018-04-12 Tom de Vries <tom@codesourcery.com>
4067 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4068 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4069 array with flexible array member as array without given dimension.
4070 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4071 argument for undefined param to true.
4073 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4076 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4077 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4078 from PowerPC section.
4079 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4080 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4081 help text that is too long.
4082 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4083 help text that is too long.
4084 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4085 help text that is too long.
4087 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4089 * config/alpha/alpha.md (stack_probe_internal): Rename
4090 from "probe_stack". Update all callers.
4092 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4094 PR rtl-optimization/84566
4095 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4096 sched_macro_fuse_insns.
4098 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4101 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4102 (compute_block_dependences): ... from here.
4104 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4106 PR tree-optimization/85331
4107 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4108 from int to HOST_WIDE_INT.
4110 2018-04-11 Martin Jambor <mjambor@suse.cz>
4113 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4114 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4115 not the same as the source val.
4116 (cgraph_edge_brings_value_p): New parameter.
4117 (gather_edges_for_value): Pass destination value to
4118 cgraph_edge_brings_value_p.
4119 (perhaps_add_new_callers): Likewise.
4120 (get_info_about_necessary_edges): Likewise and exclude values brought
4121 only by self-recursive edges.
4122 (create_specialized_node): Redirect only clones of self-calling edges.
4123 (+self_recursive_pass_through_p): New function.
4124 (find_more_scalar_values_for_callers_subset): Use it.
4125 (find_aggregate_values_for_callers_subset): Likewise.
4126 (known_aggs_to_agg_replacement_list): Removed.
4127 (decide_whether_version_node): Re-calculate known constants for all
4128 remaining context clones.
4130 2018-04-11 Richard Biener <rguenther@suse.de>
4133 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4134 from early DWARF output.
4135 (dwarf2out_early_finish): Output line info unconditionally into
4136 early DWARF and add reference to it.
4138 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4141 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4142 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4143 other than V2DFmode using iptr mode attribute.
4144 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4146 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4148 PR rtl-optimization/84659
4149 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4151 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4154 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4156 (output_loc_list): Pass address of a dummy size variable even in the
4157 locview handling loop.
4158 (index_location_lists): Add comment on why skip_loc_list_entry can't
4161 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4164 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4167 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4170 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4171 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4172 and -mstring-compare-inline-limit.
4174 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4177 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4178 for stack clash protection in a register whenever we need it to be in
4181 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4183 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4184 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4186 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4189 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4191 (mlong-double-): Ditto.
4192 * config/rs6000/sysv4.opt (msdata=): Ditto.
4193 (mtls-size=): Ditto.
4195 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4197 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4198 erroneous entries for
4199 "vector int vec_ldl (int, long int *)", and
4200 "vector unsigned int vec_ldl (int, unsigned long int *)".
4201 Add comments and entries for
4202 "vector bool char vec_ldl (int, bool char *)",
4203 "vector bool short vec_ldl (int, bool short *)",
4204 "vector bool int vec_ldl (int, bool int *)",
4205 "vector bool long long vec_ldl (int, bool long long *)",
4206 "vector pixel vec_ldl (int, pixel *)",
4207 "vector long long vec_ldl (int, long long *)",
4208 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4209 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4210 type tree bool_long_long_type_node and correct definition of
4211 bool_V2DI_type_node to make reference to this new type tree.
4212 (rs6000_mangle_type): Replace erroneous reference to
4213 bool_long_type_node with bool_long_long_type_node.
4214 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4215 comments to emphasize sign distinctions for char and int types and
4216 replace RS6000_BTI_bool_long constant with
4217 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4218 use of RS6000_BTI_pixel.
4219 (bool_long_type_node): Remove this macro definition.
4220 (bool_long_long_type_node): New macro definition
4222 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4224 PR rtl-optimization/85300
4225 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4226 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4227 simplify_unary_operation fails.
4229 2018-04-10 Martin Liska <mliska@suse.cz>
4231 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4232 cgraph_edge and ipa_ref.
4234 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4238 * config/i386/sse.md
4239 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4240 computation of the VEC_MERGE selector from mask.
4241 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4242 Fix decoding of the VEC_MERGE selector into mask.
4244 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4246 PR tree-optimization/85286
4247 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4249 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4251 * final.c (final_1): Set insn_last_address as well as
4252 insn_current_address.
4254 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4257 * explow.c (emit_stack_probe): Call validize_mem on memory location
4258 before passing it to gen_probe_stack. Create address operand and
4259 legitimize it for the probe_stack_address case.
4261 2018-04-09 Jan Hubicka <jh@suse.cz>
4264 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4265 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4266 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4268 (free_lang_data): Rebuild type inheritance graph.
4270 2018-04-09 Martin Sebor <msebor@redhat.com>
4272 * invoke.texi (-finline-small-functions): Mention other optimization
4274 (-findirect-inlining, -fpartial-inlining): Same.
4275 (-finline-functions-called-once): Same.
4276 (-freorder-blocks-and-partition): Same.
4278 2018-04-09 Jan Hubicka <jh@suse.cz>
4281 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4282 jumps; choose last target that matches the criteria (i.e.
4283 no partition changes for non-crossing jumps).
4284 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4285 support for redirecting crossing jumps to non-crossing.
4287 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4289 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4290 also for naked functions.
4292 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4294 * config/arc/arc.md (add_shift): New pattern.
4295 (add_shift2): Likewise.
4296 (sub_shift): Likewise.
4297 (sub_shift_cmp0_noout): Likewise.
4298 (compare_si_ashiftsi): Likewise.
4299 (xbfu_cmp0_noout): New combine pattern.
4300 (xbfu_cmp0"): Likewise.
4301 (movsi_set_cc_insn): Place the predicable variant first.
4302 (commutative_binary_cmp0_noout): Remove clobber.
4303 (commutative_binary_cmp0): New pattern.
4304 (noncommutative_binary_cmp0): Likewise.
4305 (noncommutative_binary_cmp0_noout): Likewise.
4306 (noncommutative_binary_comparison_result_used): Removed.
4307 (rsub_cmp0): New pattern.
4308 (rsub_cmp0_noout): Likewise.
4309 (extzvsi): Changed, keep only meaningful variants.
4310 (SQH, SEZ): New iterators.
4311 (SQH_postfix): New mode attribute.
4312 (SEZ_prefix): New code attribute.
4313 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4314 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4315 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4317 (noncommutative_operator): Check the availability of barrel
4320 2018-04-09 Richard Biener <rguenther@suse.de>
4322 PR tree-optimization/85284
4323 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4324 Only use the niter constraining form of simple_iv when the exit
4327 2018-04-09 Tom de Vries <tom@codesourcery.com>
4330 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4331 (define_expand "*memory_barrier"): New define_expand.
4332 (define_insn "memory_barrier"): New insn.
4334 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4336 PR rtl-optimization/80463
4337 PR rtl-optimization/83972
4338 PR rtl-optimization/83480
4340 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4341 correct producer for the insn.
4342 (tidy_control_flow): Fixup seqnos in case of debug insns.
4344 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4346 PR rtl-optimization/83913
4348 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4349 different sched-times when merging exprs.
4351 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4353 PR rtl-optimization/83962
4355 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4356 tidy_fallthru_edge and tidy_control_flow.
4358 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4360 PR rtl-optimization/83530
4362 * sel-sched.c (force_next_insn): New global variable.
4363 (remove_insn_for_debug): When force_next_insn is true, also leave only
4364 next insn in the ready list.
4365 (sel_sched_region): When the region wasn't scheduled, make another pass
4366 over it with force_next_insn set to 1.
4368 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4370 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4372 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4373 for interrupt control.
4374 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4375 functions for interrupt control.
4376 * config/nds32/nds32-intrinsic.md: Likewise.
4377 * config/nds32/nds32_intrinsic.h: Likewise.
4378 * config/nds32/nds32.h (nds32_builtins): Likewise.
4380 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4382 * config/nds32/nds32.c (nds32_init_machine_status,
4383 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4384 strict_aligned_p field.
4385 (nds32_expand_to_rtl_hook): New function.
4386 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4387 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4389 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4390 Chung-Ju Wu <jasonwucj@gmail.com>
4392 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4393 * config/nds32/nds32-n7.md: New file.
4394 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4395 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4397 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4398 * config/nds32/nds32.md (pipeline_model): Add n7.
4399 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4400 * config/nds32/pipelines.md: Include n7 settings.
4402 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4403 Chung-Ju Wu <jasonwucj@gmail.com>
4405 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4406 * config/nds32/nds32-e8.md: New file.
4407 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4408 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4410 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4411 * config/nds32/nds32.md (pipeline_model): Add e8.
4412 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4413 * config/nds32/pipelines.md: Include e8 settings.
4415 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4416 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4419 * config/nds32/nds32-n8.md: New file.
4420 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4421 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4423 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4424 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4425 * config/nds32/nds32.md (pipeline_model): Add n8.
4426 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4427 * config/nds32/pipelines.md: Include n8 settings.
4429 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4430 Chung-Ju Wu <jasonwucj@gmail.com>
4432 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4433 * config/nds32/nds32-n9-2r1w.md: New file.
4434 * config/nds32/nds32-n9-3r2w.md: New file.
4435 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4436 nds32_register_ports): New or modify for cpu n9.
4437 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4439 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4440 * config/nds32/nds32-utils.c: New file.
4441 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4442 TARGET_MUL_SLOW): Define.
4443 * config/nds32/nds32.md (pipeline_model): New attribute.
4444 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4445 New options that support cpu n9.
4446 * config/nds32/pipelines.md: Include n9 settings.
4447 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4449 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4451 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4452 information if necessary.
4453 (output_cond_branch_compare_zero): Likewise.
4454 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4455 (nds32_target_alignment): Refine for alignment.
4456 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4457 (FUNCTION_BOUNDARY): Modify.
4458 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4460 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4462 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4464 * config/nds32/constants.md (unspec_volatile_element): Add values for
4465 TLB operation and data prefetch.
4466 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4467 functions for TLB operation and data prefetch.
4468 * config/nds32/nds32-intrinsic.md: Likewise.
4469 * config/nds32/nds32_intrinsic.h: Likewise.
4470 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4471 (nds32_print_operand): Likewise.
4472 * config/nds32/nds32.h (nds32_builtins): Likewise.
4474 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4475 Andrew Pinski <pinsika@gcc.gnu.org>
4478 * match.pd: Use constant_boolean_node of correct type instead of
4479 boolean_true_node or boolean_false_node for simplifying
4480 pointer comparisons to zero.
4482 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4484 PR tree-optimization/80021
4485 * tree.c (verify_type_variant): Make error call in verify_variant_match
4486 translatable and remove final full stop.
4488 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4490 * config/nds32/constants.md (unspec_volatile_element): Add
4491 UNSPEC_VOLATILE_EH_RETURN.
4492 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4493 nds32_output_stack_pop): Support dwarf exception handling process.
4494 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4495 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4496 exception handling process.
4497 (nds32_compute_stack_frame): Likewise.
4498 (nds32_return_addr_rtx): Likewise.
4499 (nds32_initial_elimination_offset): Likewise.
4500 (nds32_expand_prologue): Likewise.
4501 (nds32_expand_epilogue): Likewise.
4502 (nds32_dynamic_chain_address): New function.
4503 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4505 (DYNAMIC_CHAIN_ADDRESS): Define.
4506 (EH_RETURN_DATA_REGNO): Define.
4507 (EH_RETURN_STACKADJ_RTX): Define.
4508 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4509 patterns for dwarf exception handling.
4511 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4513 * config/nds32/nds32.h: Clean up obsolete macros.
4515 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4517 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4518 Add enum values for particular instructions.
4519 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4520 particular intrinsic functions.
4521 * config/nds32/nds32-intrinsic.md: Likewise.
4522 * config/nds32/nds32_intrinsic.h: Likewise.
4523 * config/nds32/nds32.h (nds32_builtins): Likewise.
4524 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4525 (btst, ave): New patterns for particular instructions.
4527 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4529 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4530 Add enum values for atomic load/store and memory sync.
4531 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4533 * config/nds32/nds32-intrinsic.md: Likewise.
4534 * config/nds32/nds32_intrinsic.h: Likewise.
4535 * config/nds32/nds32.h (nds32_builtins): Likewise.
4537 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4539 PR tree-optimization/85257
4540 * fold-const.c (native_encode_vector): If not all elts could fit
4541 and off is -1, return 0 rather than offset.
4542 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4543 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4544 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4545 adjust buffer in native_interpret_expr call.
4547 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4549 * config/nds32/constants.md (unspec_volatile_element): Add cache
4550 control enum values.
4551 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4552 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4553 * config/nds32/nds32.c (nds32_cctl_names): New.
4554 (nds32_print_operand): Handle cache control register names.
4555 * config/nds32/nds32.h (nds32_builtins): New enum values.
4556 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4558 * config/nds32/nds32.md (type): Add mmu.
4559 * config/nds32/pipelines.md (simple_insn): Add mmu.
4561 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4563 * config/nds32/nds32.md (type): Remove call.
4564 * config/nds32/pipelines.md (simple_insn): Likewise.
4566 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4568 * config/nds32/constants.md (unspec_volatile_element): Add
4569 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4570 UNSPEC_VOLATILE_FMFCFG.
4571 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4572 description for fmfcfg and fmfcsr.
4573 (bdesc_1arg): Add fmtcsr.
4574 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4575 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4576 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4577 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4578 unspec_fmfcfg): New patterns.
4579 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4580 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4581 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4582 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4583 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4584 __nds32__fmfcfg): Define.
4586 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4588 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4589 intrinsic register names.
4590 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4591 intrinsic register enum values and macros.
4593 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4595 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4596 for load/store addressing form.
4597 (nds32_print_operand_address): Likewise.
4599 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4602 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4603 based on LABEL_REF. Remove useless assertion.
4604 (pic_address_needs_scratch): Fix formatting.
4605 (sparc_legitimize_pic_address): Minor tweaks.
4606 (sparc_delegitimize_address): Adjust assertion accordingly.
4607 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4608 into symbolic_operand.
4609 (movsi_high_pic_label_ref): Likewise.
4610 (movsi_lo_sum_pic_label_ref): Likewise.
4611 (movdi_pic_label_ref): Likewise.
4612 (movdi_high_pic_label_ref): Likewise.
4613 (movdi_lo_sum_pic_label_ref): Likewise.
4615 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4617 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4618 custom LIB_SPEC setup.
4620 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4621 Kito Cheng <kito.cheng@gmail.com>
4623 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4624 * config/riscv/freebsd.h: New.
4626 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4628 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4629 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4632 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4633 Kito Cheng <kito.cheng@gmail.com>
4635 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4636 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4637 * config/nds32/nds32-protos.h (nds32_output_call,
4638 nds32_output_return): Declare.
4639 * config/nds32/nds32.md: Refine all the call and return patterns.
4641 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4644 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4645 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4647 PR rtl-optimization/84872
4648 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4649 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4652 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4654 * expr.c (copy_blkmode_to_reg): Revert 254862.
4655 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4657 2018-04-06 Richard Biener <rguenther@suse.de>
4660 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4661 after seeing a component reference with an adjacent field. Treat
4662 refs to arrays at struct end of external decls similar to
4663 refs to unconstrained commons.
4665 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4668 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4669 look through SAVE_EXPRs with non-side-effects argument. Adjust
4671 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4674 2018-04-06 Richard Biener <rguenther@suse.de>
4677 * alias.c (find_base_term): New wrapper around find_base_term
4678 unwinding CSELIB_VAL_PTR changes.
4679 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4682 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4684 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4686 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4687 constant definitions.
4688 ("nop"): lr 0,0 -> nopr r0
4689 ("nop_lr0", "nop_lr1"): New insn definitions.
4691 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4693 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4694 NDS32_V3PUSH_AVAILABLE_P macro.
4696 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4697 Chung-Ju Wu <jasonwucj@gmail.com>
4699 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4700 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4701 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4702 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4703 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4704 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4705 * config/nds32/constraints.md: New constraints and checking for hard
4706 float configuration.
4707 * config/nds32/iterators.md: New mode iterator and attribute for hard
4708 float configuration.
4709 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4711 * config/nds32/nds32-fpu.md: New file.
4712 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4713 deal with hard float code generation.
4714 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4716 (abi_type, float_reg_number): New enum type.
4717 * config/nds32/nds32-predicates.c: New predicates for hard float.
4718 * config/nds32/nds32-protos.h: Declare functions for hard float.
4719 * config/nds32/nds32.c: Implementation for hard float configuration.
4720 * config/nds32/nds32.h: Definitions for hard float configuration.
4721 * config/nds32/nds32.md: Include hard float machine description and
4722 modify patterns for hard float configuration.
4723 * config/nds32/nds32.opt: New options for hard float configuration.
4724 * config/nds32/predicates.md: New predicates for hard float
4727 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4729 * common/config/nds32/nds32-common.c
4730 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4732 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4735 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4736 CONSTRUCTOR_ELT (ctor, ...)->value.
4738 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4741 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4743 2018-04-05 Tom de Vries <tom@codesourcery.com>
4746 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4749 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4750 Kito Cheng <kito.cheng@gmail.com>
4752 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4753 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4754 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4755 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4757 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4758 Kito Cheng <kito.cheng@gmail.com>
4760 * config/nds32/constraints.md (Ufe): New memory constraint.
4761 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4762 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4763 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4765 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4766 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4768 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4770 * config/nds32/nds32.md: Use optimize_size in the condition for
4771 alu-shift instructions.
4773 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4775 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4777 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4779 * config/nds32/nds32.md (negsi2): Refine pattern.
4781 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4782 Chung-Ju Wu <jasonwucj@gmail.com>
4784 * config/nds32/iterators.md (shift_rotate): New code iterator.
4785 (shift): New code attribute.
4786 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4787 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4788 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4789 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4790 bit-wise operations.
4791 (andsi3, *andsi3): Ditto.
4792 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4793 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4794 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4795 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4796 nds32_ior_operand, nds32_xor_operand): New predicates.
4798 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4800 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4801 (addsi3, subsi3): ... this.
4803 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4805 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4807 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4809 * config/nds32/nds32.md: Adjust indention.
4811 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4813 * config/nds32/nds32.md (feature): New attribute.
4815 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4817 * config/nds32/nds32.md (subtype): New attribute.
4819 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4822 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4823 expansion to perform a bitwise AND of the argument followed by a
4824 boolean negation of the result.
4826 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4828 PR rtl-optimization/84878
4829 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4830 the basic block. Assert the use reference is not artificial and that
4831 it has an associated insn.
4833 2018-04-04 Michael Matz <matz@suse.de>
4835 * builtins.c (compute_objsize): Pass correct operand
4836 to array_at_struct_end_p.
4838 2018-04-04 Richard Biener <rguenther@suse.de>
4841 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4842 from contexts for DINFO_LEVEL_TERSE and below.
4844 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4846 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4847 resiter_operand condition.
4848 * config/nds32/nds32.md (*move<mode>): Ditto.
4850 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4851 Monk Chiang <sh.chiang04@gmail.com>
4853 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4855 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4857 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4859 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4860 Kito Cheng <kito.cheng@gmail.com>
4862 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4863 nds32_cond_code_str, output_cond_branch,
4864 output_cond_branch_compare_zero, nds32_expand_cbranch,
4865 nds32_expand_cstore, nds32_expand_movcc,
4866 nds32_output_cbranchsi4_equality_zero,
4867 nds32_output_cbranchsi4_equality_reg,
4868 nds32_output_cbranchsi4_equality_reg_or_const_int,
4869 nds32_output_cbranchsi4_greater_less_zero: New functions.
4870 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4871 nds32_expand_cstore, nds32_expand_movcc,
4872 nds32_output_cbranchsi4_equality_zero,
4873 nds32_output_cbranchsi4_equality_reg,
4874 nds32_output_cbranchsi4_equality_reg_or_const_int,
4875 nds32_output_cbranchsi4_greater_less_zero): Declare.
4876 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4877 nds32_rimm11s_operand): New predicates.
4878 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4879 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4882 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4884 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4885 * config/nds32/nds32.md: Ditto.
4886 * config/nds32/pipelines.md: Ditto.
4888 2018-04-04 Richard Biener <rguenther@suse.de>
4890 PR tree-optimization/85168
4891 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4892 propagating abnormals.
4894 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4896 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4898 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4899 Kito Cheng <kito.cheng@gmail.com>
4901 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4902 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4903 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4904 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4905 * config/nds32/nds32.md (sibcall_internal): New.
4906 (sibcall_register): Remove.
4907 (sibcall_immediate): Remove.
4908 (sibcall_value_internal): New.
4909 (sibcall_value_register): Remove.
4910 (sibcall_value_immediate): Remove.
4911 * config/nds32/predicates.md (nds32_general_register_operand): New.
4912 (nds32_call_address_operand): New.
4914 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4916 PR rtl-optimization/85167
4917 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4918 bb_defs if *split_p, instead preinitialize it to NULL.
4920 PR tree-optimization/85156
4921 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4922 evaluating the argument multiple times.
4924 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4926 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4928 (_mm_cvtpd_ps): Likewise.
4929 (_mm_cvttpd_epi32): Likewise.
4930 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4931 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4932 vector, pixel, and bool following altivec.h include.
4934 2018-04-03 Martin Sebor <msebor@redhat.com>
4936 * doc/extend.texi (Common Function Attributes): Clarify.
4937 (const attribute): Likewise.
4938 (pure attribute): Likewise.
4940 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4943 * config/i386/i386.c (ix86_expand_vector_set): Use
4944 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4946 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4948 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4949 instructions when changing rounding bits to preserve precision bits
4950 in the x87 control word.
4952 2018-04-03 Martin Liska <mliska@suse.cz>
4954 PR tree-optimization/82491
4955 * rtl.h (strip_offset_and_add): Replace += suboffset with
4956 poly_uint64 () + suboffset.
4958 2018-03-29 Martin Liska <mliska@suse.cz>
4959 Martin Jambor <mjambor@suse.cz>
4962 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4963 param_type is not an integral or pointer type.
4965 2018-04-03 Richard Biener <rguenther@suse.de>
4967 * sese.h (recompute_all_dominators): Remove.
4969 2018-04-02 Martin Sebor <msebor@redhat.com>
4971 * doc/invoke.texi (-Wrestrict): Fix typos.
4973 2018-04-02 Jim Wilson <jimw@sifive.com>
4975 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4976 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4977 (<optab>di3, <optab>si3_extend): Likewise.
4978 (<optab>si3_mask, <optab>si3_mask_1): New.
4979 (<optab>di3_mask, <optab>di3_mask_1): New.
4980 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4981 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4982 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4984 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4986 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4989 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4991 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4992 (nds32_canonicalize_comparison): New function.
4994 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4995 Kito Cheng <kito.cheng@gmail.com>
4996 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4998 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4999 * config/nds32/constants.md (unspec_volatile_element): Add
5000 UNSPEC_VOLATILE_RELAX_GROUP.
5001 * config/nds32/nds32-relax-opt.c: New file.
5002 * config/nds32/nds32-predicates.c
5003 (nds32_symbol_load_store_p): New function.
5004 * config/nds32/nds32-protos.h
5005 (nds32_symbol_load_store_p): Declare function.
5006 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5007 * config/nds32/nds32.c
5008 (nds32_register_pass): New function to register pass.
5009 (nds32_register_passes): New function to register passes.
5010 * config/nds32/nds32.md (relax_group): New pattern.
5011 * config/nds32/nds32.opt (mrelax-hint): New option.
5012 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5014 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5016 * config/nds32/t-nds32: Modify files dependency.
5018 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5020 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5021 (PROFILE_HOOK): Define its implementation.
5023 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5025 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5026 type and 32-bit size.
5028 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5031 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5032 (V_128_256): New mode iterator.
5033 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5034 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5035 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5037 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5038 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5040 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5043 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5044 NaN inputs correctly.
5046 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5049 * config/rs6000/vsx.md (??r): New mode attribute.
5050 (*vsx_mov<mode>_64bit): Use it.
5051 (*vsx_mov<mode>_32bit): Likewise.
5053 2018-03-30 Martin Sebor <msebor@redhat.com>
5055 PR tree-optimization/84818
5056 * builtins.c (check_access): Use warning_n.
5058 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5061 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5063 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5066 2018-03-30 Julia Koval <julia.koval@intel.com>
5069 * x86-tune.def (movx, partial_reg_dependency): Enable for
5072 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5075 * lra-constraints.c (process_alt_operands): Move setting
5076 this_alternative_matches below.
5078 2018-03-29 Martin Liska <mliska@suse.cz>
5081 * doc/invoke.texi: Document how LTO works with debug info.
5082 Describe auto-load support of binutils. Mention 'x86-64'
5083 as valid option value of -march option.
5085 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5087 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5090 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5091 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5092 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5095 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5098 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5099 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5100 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5101 for RS6000_BTM_POWERPC64.
5102 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5103 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5104 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5109 2018-03-28 Carl Love <cel@us.ibm.com>
5112 2017-09-27 Carl Love <cel@us.ibm.com>
5114 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5115 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5116 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5119 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5121 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5122 instead of __vector bool.
5123 (_mm_max_pu8): Likewise.
5124 (_mm_min_pi16): Likewise.
5126 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5129 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5130 (DIVWEUO): Likewise.
5132 (DIVDEUO): Likewise.
5133 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5134 DIVWEUO and DIVDEUO.
5135 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5136 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5137 (div_extend): Likewise.
5138 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5140 (__builtin_divweuo): Likewise.
5141 (__builtin_divdeo): Likewise.
5142 (__builtin_divdeuo): Likewise.
5144 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5147 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5148 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5150 PR tree-optimization/82004
5151 * gimple-match-head.c (optimize_pow_to_exp): New function.
5152 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5153 Don't fold to exp if optimize_pow_to_exp is false.
5155 2018-03-28 Martin Liska <mliska@suse.cz>
5158 * calls.c (initialize_argument_information): Fix trailing space.
5159 * common.opt: Fix typo and provide better explanation for
5160 -fsanitize-coverage option.
5161 * config/i386/i386.opt: Fix typo.
5163 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5164 Martin Liska <mliska@suse.cz>
5167 * gimplify.c (asan_poison_variable): Don't do the check for
5168 gimplify_omp_ctxp here.
5169 (gimplify_decl_expr): Do it here.
5170 (gimplify_target_expr): Likewise.
5172 2018-03-28 Martin Liska <mliska@suse.cz>
5175 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5176 chkp_type_bounds_count if MPX is not enabled.
5178 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5180 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5182 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5185 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5186 function to create the function decl for complex long double
5187 multiply and divide for -mabi=ieeelongdouble.
5188 (init_float128_ieee): Call it.
5190 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5193 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5194 -fcf-protection=branch -mibt.
5195 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5197 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5200 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5202 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5205 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5206 extern array declarations.
5208 2018-03-27 Richard Biener <rguenther@suse.de>
5211 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5212 explicit single_use checks.
5214 2018-03-27 Richard Biener <rguenther@suse.de>
5216 PR tree-optimization/85082
5217 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5220 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5222 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5223 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5224 Turn on fasynchronous-unwind-tables and funwind-tables.
5226 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5229 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5230 (*bmi_blsr_<mode>_ccz): Ditto.
5232 2018-03-26 Tom de Vries <tom@codesourcery.com>
5234 PR tree-optimization/85063
5235 * omp-general.c (offloading_function_p): New function. Factor out
5237 * omp-offload.c (pass_omp_target_link::gate): ... here.
5238 * omp-general.h (offloading_function_p): Declare.
5239 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5240 with attribute omp declare target for offloading functions.
5242 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5244 PR tree-optimization/84005
5245 * tree-data-ref.h (get_base_for_alignment): Declare.
5246 * tree-data-ref.c (get_base_for_alignment_1): New function.
5247 (get_base_for_alignment): Likewise.
5248 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5249 get_base_for_alignment to find a suitable base object, instead
5250 of always using drb->base_address.
5252 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5255 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5256 known size by default.
5258 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5261 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5262 and non BLKmode operands.
5264 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5267 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5268 Clean up attributes.
5270 2018-03-23 Richard Biener <rguenther@suse.de>
5273 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5274 we are going to emit early debug for LTO.
5276 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5279 * function.c (match_asm_constraints_1): Don't optimize if input
5280 doesn't satisfy general_operand predicate for output's mode.
5283 * alias.c (write_dependence_p): Don't require for x_canonicalized
5284 non-VOIDmode if x has VOIDmode.
5287 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5288 just don't try to optimize it rather than assert it never happens.
5290 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5292 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5293 macro expansions for definition of ST_INTERNAL_<mode> and
5294 LD_INTERNAL_<mode> builtins.
5295 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5297 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5299 (altivec_expand_st_builtin): Likewise.
5300 (altivec_expand_builtin): Remove calls to deleted functions.
5301 (rs6000_address_for_altivec): Delete this function.
5302 * config/rs6000/vector.md: Remove expands for
5303 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5305 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5308 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5309 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5310 re-computing once computed.
5311 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5312 (arm_init_machine_status): Initialize
5313 machine->static_chain_stack_bytes.
5315 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5318 * doc/extend.texi: Add four new prototypes for vec_ld.
5319 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5320 definitions for more logical presentation.
5321 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5322 entries for V1TI variants of __builtin_altivec_ld builtin.
5323 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5324 handling of V1TI variant of LVX icode pattern.
5325 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5326 (rs6000_gimple_fold_builtin): Likewise.
5327 (altivec_init_builtins): Add code to define
5328 __builtin_altivec_lvx_v1ti function.
5330 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5333 * function.c (match_asm_constraints_1): Don't do the optimization
5334 if input isn't a REG, SUBREG, MEM or constant.
5336 2018-03-22 Tom de Vries <tom@codesourcery.com>
5338 PR tree-optimization/84956
5339 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5340 bb_has_abnormal_pred.
5342 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5345 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5346 DECL_INITIAL (decl) to decl at the end.
5347 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5350 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5352 * doc/extend.texi (__builtin_tgmath): Document when complex
5353 integer types are treated as _Complex _Float64.
5355 2018-03-21 Tom de Vries <tom@codesourcery.com>
5357 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5359 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5361 PR tree-optimization/84960
5362 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5363 if it is ENTRY block, move them into single succ of ENTRY in that case.
5365 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5367 PR tree-optimization/84811
5368 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5369 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5370 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5371 (known_subrange_p): Update accordingly. Cast each value involved
5372 in the size comparison, rather than casting the result of the
5375 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5377 PR tree-optimization/84982
5378 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5379 by flipping the least significant bit rather than all bits from
5380 bitpos to bitpos + bitsize - 1.
5382 2018-03-21 Nathan Sidwell <nathan@acm.org>
5384 * doc/extend.texi (Deprecated Features): Remove mention of
5385 long-deleted deprecations.
5387 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5390 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5391 * configure: Regenerate.
5393 2018-03-21 Tom de Vries <tom@codesourcery.com>
5395 PR tree-optimization/83126
5396 * tree-parloops.c (num_phis): New function.
5397 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5399 2018-03-21 Nathan Sidwell <nathan@acm.org>
5401 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5402 mention anon-struct/union members and trailing attributes.
5404 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5406 PR tree-optimization/84969
5407 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5408 builtin memset partitions if they set different rhs values.
5410 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5412 PR rtl-optimization/84989
5413 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5414 VEC_DUPLICATE with scalar result mode.
5416 2018-03-21 Martin Liska <mliska@suse.cz>
5419 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5420 not intended return statement.
5422 2018-03-21 Martin Liska <mliska@suse.cz>
5425 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5426 (chkp_find_bound_slots_1): Limit number of iterations.
5428 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5431 * Minor grammar fixes for x86 options.
5433 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5436 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5437 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5439 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5442 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5443 (altivec_lvx_<mode>_1op): Likewise.
5444 (altivec_stvx_<mode>_2op): Likewise.
5445 (altivec_stvx_<mode>_1op): Likewise.
5446 (altivec_lvx_<VM2:mode>): New define_expand.
5447 (altivec_stvx_<VM2:mode>): Likewise.
5448 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5449 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5450 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5451 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5452 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5453 (rs6000_gen_lvx): Likewise.
5454 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5455 (altivec_expand_stv_builtin): Likewise.
5456 (altivec_expand_builtin): Likewise.
5457 * config/rs6000/vector.md: Likewise.
5459 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5462 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5465 2018-03-20 Richard Biener <rguenther@suse.de>
5468 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5469 sign-conversions as zero, fall back to standard scalar_stmt
5472 2018-03-20 Martin Liska <mliska@suse.cz>
5475 * predict.c (rebuild_frequencies): Handle case when we have
5476 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5478 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5481 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5482 flag_section_anchors.
5483 * varasm.c (use_blocks_for_decl_p): Remove hack for
5484 dw2_force_const_mem.
5487 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5489 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5490 be created, use lowpart_subreg of operands[0] rather than operands[0]
5492 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5493 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5494 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5495 and n constraint instead of aarch64_shift_imm_di and Usd.
5496 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5497 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5499 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5502 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5503 to favor GPR over NEON registers.
5504 (<shift>di3_neon): Likewise.
5506 2018-03-20 Tom de Vries <tom@codesourcery.com>
5509 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5510 (nvptx_process_pars): Emit bar.sync asap and alap.
5512 2018-03-20 Tom de Vries <tom@codesourcery.com>
5515 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5516 seen_label if seen_label is already set.
5518 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5521 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5522 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5523 Use 1U instead of 1. Formatting fixes.
5526 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5527 instead of TREE_TYPE (s1) for the return value.
5529 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5531 PR tree-optimization/84946
5532 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5533 bitsize + bitsize in poly_uint64 rather than poly_int64.
5536 * dwarf2asm.c: Include fold-const.c.
5537 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5538 of decl rather than decl itself.
5540 PR rtl-optimization/84643
5541 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5543 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5546 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5547 calling assemble_variable.
5549 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5552 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5553 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5555 2018-03-19 Jim Wilson <jimw@sifive.com>
5558 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5559 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5560 (riscv_first_stack_step): Likewise.
5561 (riscv_option_override): Use STACK_BOUNDARY instead of
5563 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5565 (BIGGEST_ALIGNMENT): Set to 128.
5566 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5567 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5570 2018-03-19 Richard Biener <rguenther@suse.de>
5572 PR tree-optimization/84933
5573 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5574 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5576 2018-03-19 Richard Biener <rguenther@suse.de>
5578 PR tree-optimization/84859
5579 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5580 (cond_if_else_store_replacement): Perform sinking operation on
5581 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5582 Generalize what a BB with a single eligible store is.
5584 2018-03-19 Richard Biener <rguenther@suse.de>
5586 PR tree-optimization/84929
5587 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5588 chrec_is_positive against non-chrec arg.
5590 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5593 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5595 2018-03-18 Martin Liska <mliska@suse.cz>
5597 PR rtl-optimization/84635
5598 * regrename.c (build_def_use): Use matches_mode only when
5601 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5603 PR tree-optimization/84913
5604 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5605 vectorize chains of COND_EXPRs.
5607 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5609 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5611 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5613 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5615 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5617 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5619 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5620 Kito Cheng <kito.cheng@gmail.com>
5622 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5623 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5624 (nds32_adjust_reg_alloc_order): New function.
5625 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5627 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5629 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5630 nds32_print_operand, nds32_print_operand_address): Use
5631 HOST_WIDE_INT_PRINT_DEC instead.
5633 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5635 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5637 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5640 * config/i386/i386.c (initial_ix86_tune_features,
5641 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5643 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5644 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5645 rather than 1u << ix86_tune. Formatting fix.
5646 (ix86_option_override_internal): Change ix86_arch_mask from
5647 unsigned int to unsigned HOST_WIDE_INT, initialize to
5648 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5649 (ix86_function_specific_restore): Likewise.
5651 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5654 * postreload.c (reload_combine_recognize_pattern): Perform
5655 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5656 truncate_int_for_mode the result for the destination's mode.
5659 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5660 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5662 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5665 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5668 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5669 regno_allocno_class_array and sorted_pseudos.
5670 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5671 insns where regno is used.
5673 2018-03-16 Martin Liska <mliska@suse.cz>
5676 * multiple_target.c (create_dispatcher_calls): Redirect
5677 reference in the symbol table.
5679 2018-03-16 Martin Liska <mliska@suse.cz>
5682 * multiple_target.c (create_dispatcher_calls): Redirect also
5685 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5689 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5690 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5691 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5693 2018-03-16 Julia Koval <julia.koval@intel.com>
5695 * doc/invoke.texi (Skylake Server): Add CLWB.
5696 Cannonlake): Remove CLWB.
5698 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5700 PR tree-optimization/84841
5701 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5703 (FLOAT_ONE_CONST_TYPE): Define.
5704 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5705 (sort_by_operand_rank): Put entries with higher constant_type last
5706 rather than first to match comments.
5708 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5710 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5713 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5716 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5717 check and use address of target always.
5719 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5722 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5723 (indirect_thunk_bnd_needed): Likewise.
5724 (indirect_thunks_used): Likewise.
5725 (indirect_thunks_bnd_used): Likewise.
5726 (indirect_return_needed): New.
5727 (indirect_return_bnd_needed): Likewise.
5728 (output_indirect_thunk_function): Add a bool argument for
5730 (output_indirect_thunk_function): Don't generate alias for
5731 function return thunk.
5732 (ix86_code_end): Call output_indirect_thunk_function to generate
5733 function return thunks.
5734 (ix86_output_function_return): Set indirect_return_bnd_needed
5735 and indirect_return_needed instead of indirect_thunk_bnd_needed
5736 and indirect_thunk_needed.
5738 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5740 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5741 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5742 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5744 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5745 Paul Hua <paul.hua.gm@gmail.com>
5748 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5750 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5752 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5753 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5756 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5759 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5760 instead of GET_MODE_SIZE when comparing Units.
5762 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5765 * varasm.c (hash_section): Return an unchangeble hash value
5766 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5767 Return !aarch64_can_use_per_function_literal_pools_p ().
5769 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5772 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5773 rather than address of cmode as last argument to prepare_cmp_insn.
5775 2018-03-15 Julia Koval <julia.koval@intel.com>
5777 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5778 F_AVX512VNNI, F_AVX512BITALG): New.
5780 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5783 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5784 insn for floating-point loads and stores.
5786 2018-03-14 Carl Love <cel@us.ibm.com>
5788 * config/rs6000/rs6000-c.c: Add macro definitions for
5789 ALTIVEC_BUILTIN_VEC_PERMXOR.
5790 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5791 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5792 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5793 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5795 * config/doc/extend.texi: Add prototypes for vec_permxor.
5797 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5800 * diagnostic-show-locus.c (class layout_point): Convert m_line
5801 from int to linenum_type.
5802 (line_span::comparator): Use linenum "compare" function when
5803 comparing line numbers.
5804 (test_line_span): New function.
5805 (layout_range::contains_point): Convert param "row" from int to
5807 (layout_range::intersects_line_p): Likewise.
5808 (layout::will_show_line_p): Likewise.
5809 (layout::print_source_line): Likewise.
5810 (layout::should_print_annotation_line_p): Likewise.
5811 (layout::print_annotation_line): Likewise.
5812 (layout::print_leading_fixits): Likewise.
5813 (layout::annotation_line_showed_range_p): Likewise.
5814 (struct line_corrections): Likewise for field m_row.
5815 (line_corrections::line_corrections): Likewise for param "row".
5816 (layout::print_trailing_fixits): Likewise.
5817 (layout::get_state_at_point): Likewise.
5818 (layout::get_x_bound_for_row): Likewise.
5819 (layout::print_line): Likewise.
5820 (diagnostic_show_locus): Likewise for locals "last_line" and
5822 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5823 * input.c (selftest::test_linenum_comparisons): New function.
5824 (selftest::input_c_tests): Call it.
5825 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5826 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5827 * selftest.h (ASSERT_GT): New macro.
5828 (ASSERT_GT_AT): New macro.
5829 (ASSERT_LT): New macro.
5830 (ASSERT_LT_AT): New macro.
5832 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5834 PR rtl-optimization/84780
5835 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5837 2018-03-14 Martin Liska <mliska@suse.cz>
5839 * tree.c (record_node_allocation_statistics): Use
5840 get_stats_node_kind.
5841 (get_stats_node_kind): New function extracted from
5842 record_node_allocation_statistics.
5843 (free_node): Use get_stats_node_kind.
5845 2018-03-14 Richard Biener <rguenther@suse.de>
5847 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5848 that the value-set of ANTIC_IN doesn't grow.
5851 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5853 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5854 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5856 2018-03-14 Julia Koval <julia.koval@intel.com>
5858 * config.gcc (icelake-client, icelake-server): New.
5860 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5861 (initial_ix86_arch_features): Ditto.
5862 (PTA_SKYLAKE): Add SGX.
5863 (PTA_ICELAKE): Remove.
5864 (PTA_ICELAKE_CLIENT): New.
5865 (PTA_ICELAKE_SERVER): New.
5866 (ix86_option_override_internal): Split up icelake on icelake client and
5868 (get_builtin_code_for_version): Ditto.
5869 (fold_builtin_cpu): Ditto.
5870 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5871 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5872 * config/i386/i386.h (processor_type): Ditto.
5873 * doc/invoke.texi: Ditto.
5875 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5878 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5879 INTEGER_CST offset, add it together with bitpos / 8 and
5880 sign extend based on POINTER_SIZE.
5884 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5887 * config/i386/constraints.md (Yc): New register constraint.
5888 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5889 Use Yc constraint for alternative 2 of operand 0. Remove
5890 preferred_for_speed attribute.
5892 2018-03-14 Richard Biener <rguenther@suse.de>
5894 PR tree-optimization/84830
5895 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5896 with the old one to avoid oscillations.
5898 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5901 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5903 (assign_by_spills): Return a flag of reload assignment failure.
5904 Do not process the reload assignment failures. Do not spill other
5905 reload pseudos if they has the same reg class. Update n if
5907 (lra_assign): Add a return arg. Set up from the result of
5908 assign_by_spills call.
5909 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5910 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5911 usage_insns if it is not NULL.
5912 (spill_hard_reg_in_range): New function.
5913 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5914 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5915 function prototypes.
5916 (lra_assign): Change prototype.
5917 * lra.c (lra): Add code to deal with fails by splitting hard reg
5920 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5922 * config/riscv/riscv.opt (mrelax): New option.
5923 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5924 "norelax" when riscv_mrelax is disabled.
5925 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5927 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5930 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5931 reassociation for int modes.
5933 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5935 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5936 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5938 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5939 * config/aarch64/aarch64-sve.md
5940 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5941 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5942 (*extend<mode><Vwide>2): Rename to...
5943 (aarch64_sve_extend<mode><Vwide>2): ...this.
5944 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5945 renaming the old pattern to...
5946 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5948 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5949 define_expand, renaming the old pattern to...
5950 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5951 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5952 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5953 account when deciding which SVE instruction the optab should use.
5954 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5956 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5958 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5959 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5960 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5961 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5962 (tlsdesc_small_<mode>): Turn a define_expand and use
5963 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5964 (tlsdesc_small_advsimd_<mode>): ...this.
5965 (tlsdesc_small_sve_<mode>): New pattern.
5967 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5969 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5970 (UNSPEC_UMUL_HIGHPART): New constants.
5971 (MUL_HIGHPART): New int iteraor.
5972 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5973 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5975 (*<su>mul<mode>3_highpart): New define_insn.
5977 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5980 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5983 2018-03-13 Martin Liska <mliska@suse.cz>
5986 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5988 (sem_item_optimizer::~sem_item_optimizer): Release it.
5989 (sem_item_optimizer::merge_classes): Register variable aliases.
5990 (sem_item_optimizer::fixup_pt_set): New function.
5991 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5992 * ipa-icf.h: Declare new variables and functions.
5994 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5997 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5998 integer_pow2p@2 and test integer_pow2p in condition.
5999 (A < 0 ? C : 0): Similarly for @1.
6002 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6003 characters starting at p contain '\0' character, don't look beyond
6007 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6008 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6011 * reg-stack.c (change_stack): Change update_end var from int to
6012 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6013 also call set_block_for_insn on the newly added insns and rescan.
6016 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6017 on the last operand.
6020 * tree.c (stabilize_reference_1): Return save_expr (e) for
6021 STATEMENT_LIST even if it doesn't have side-effects.
6023 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6025 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6027 2018-03-12 Renlin Li <renlin.li@arm.com>
6029 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6030 aarch64_output_scalar_simd_mov_immediate.
6032 2018-03-12 Martin Sebor <msebor@redhat.com>
6034 PR tree-optimization/83456
6035 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6036 for perfectly overlapping calls to memcpy.
6037 (gimple_fold_builtin_memory_chk): Same.
6038 (gimple_fold_builtin_strcpy): Handle no-warning.
6039 (gimple_fold_builtin_stxcpy_chk): Same.
6040 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6042 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6044 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6045 parameter. Use it for SFmode.
6046 (rs6000_function_arg_advance_1): Adjust.
6047 (rs6000_function_arg): Adjust.
6048 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6050 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6052 PR rtl-optimization/84169
6053 PR rtl-optimization/84780
6054 * combine.c (can_combine_p): Check for a 2-insn combination whether
6055 the destination register is used between the two insns, too.
6057 2018-03-12 Richard Biener <rguenther@suse.de>
6059 PR tree-optimization/84803
6060 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6061 for refs DR analysis didn't process.
6063 2018-03-12 Richard Biener <rguenther@suse.de>
6065 PR tree-optimization/84777
6066 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6067 force-vectorize loops ignore whether we are optimizing for size.
6069 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6071 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6072 (TARGET_MD_ASM_ADJUST): Define.
6074 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6075 Kito Cheng <kito.cheng@gmail.com>
6076 Chung-Ju Wu <jasonwucj@gmail.com>
6078 * config/nds32/nds32.c (nds32_compute_stack_frame,
6079 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6080 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6081 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6082 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6083 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6084 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6085 * config/nds32/nds32.md (prologue, epilogue): Use macro
6086 NDS32_V3PUSH_AVAILABLE_P to do checking.
6088 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6091 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6092 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6093 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6094 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6095 addition of most attributes on !orig_type_die or the attribute not
6096 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6098 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6099 Chung-Ju Wu <jasonwucj@gmail.com>
6101 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6103 * config/nds32/nds32.opt (mvh): New option.
6105 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6106 Chung-Ju Wu <jasonwucj@gmail.com>
6108 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6110 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6111 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6114 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6115 Chung-Ju Wu <jasonwucj@gmail.com>
6117 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6119 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6120 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6122 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6123 Kito Cheng <kito.cheng@gmail.com>
6124 Chung-Ju Wu <jasonwucj@gmail.com>
6126 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6127 UNSPEC_FFMISM and UNSPEC_FLMISM.
6128 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6129 for ffb, ffmism and flmism.
6130 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6131 (unspec_ffmism): Ditto.
6132 (unspec_flmism): Ditto.
6133 (nds32_expand_builtin_impl): Check if string extension is available.
6134 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6135 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6137 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6140 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6143 * lra-assigns.c (assign_by_spills): Return a flag of reload
6144 assignment failure. Do not process the reload assignment
6145 failures. Do not spill other reload pseudos if they has the same
6147 (lra_assign): Add a return arg. Set up from the result of
6148 assign_by_spills call.
6149 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6150 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6151 usage_insns if it is not NULL.
6152 (spill_hard_reg_in_range): New function.
6153 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6154 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6155 function prototypes.
6156 (lra_assign): Change prototype.
6157 * lra.c (lra): Add code to deal with fails by splitting hard reg
6160 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6163 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6165 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6168 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6169 (dw_val_node): Add val_symbolic_view.
6170 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6171 (symview_upper_bound): New.
6172 (new_line_info_table): Initialize symviews_since_reset.
6173 (dwarf2out_source_line): Count symviews_since_reset and set
6174 symview_upper_bound.
6175 (dw_val_equal_p): Handle symview.
6176 (add_AT_symview): New.
6177 (print_dw_val): Handle symview.
6178 (attr_checksum, attr_checksum_ordered): Likewise.
6179 (same_dw_val_p, size_of_die): Likewise.
6180 (value_format, output_die): Likewise.
6181 (add_high_low_attributes): Use add_AT_symview for entry_view.
6182 (dwarf2out_finish): Reset symview_upper_bound, clear
6185 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6188 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6189 Add strict argument and use it.
6190 (rs6000_split_multireg_move): Update for new strict argument.
6191 (mem_operand_gpr): Disallow all non-offsettable addresses.
6192 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6194 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6197 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6198 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6199 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6202 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6203 decl, use remap_type if we want to use the type.
6205 2018-03-09 Martin Sebor <msebor@redhat.com>
6207 PR tree-optimization/84526
6208 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6210 (builtin_access::generic_overlap): Be prepared to handle non-array
6213 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6215 PR rtl-optimization/84682
6216 * lra-constraints.c (process_address_1): Check is_address flag
6217 for address constraints.
6218 (process_alt_operands): Likewise.
6219 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6220 preprocess_constraints.
6221 * recog.h (preprocess_constraints): Add oploc parameter.
6223 * recog.c (preprocess_constraints): Test address_operand for
6224 CT_ADDRESS constraints.
6226 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6229 * lra-assigns.c (assign_by_spills): Return a flag of reload
6230 assignment failure. Do not process the reload assignment
6231 failures. Do not spill other reload pseudos if they has the same
6233 (lra_assign): Add a return arg. Set up from the result of
6234 assign_by_spills call.
6235 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6236 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6237 usage_insns if it is not NULL.
6238 (spill_hard_reg_in_range): New function.
6239 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6240 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6241 function prototypes.
6242 (lra_assign): Change prototype.
6243 * lra.c (lra): Add code to deal with fails by splitting hard reg
6246 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6249 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6250 Accept complain bool parameter. Only emit errors if it is true.
6251 (arm_parse_cpu_option_name): Likewise.
6252 (arm_target_thumb_only): Adjust callers of the above.
6253 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6254 prototype to take a default true bool parameter.
6255 (arm_parse_arch_option_name): Likewise.
6257 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6258 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6262 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6263 * configure: Regenerate.
6264 * configure.ac ("linker --version-script option"): New.
6265 ("linker soname option"): New.
6267 2018-03-09 Richard Biener <rguenther@suse.de>
6269 PR tree-optimization/84775
6270 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6271 immediate uses of predicate stmts and mark them modified.
6274 PR tree-optimization/84178
6275 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6277 (version_loop_for_if_conversion): Delay update_ssa call.
6278 (tree_if_conversion): Delay update_ssa until after predicate
6281 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6284 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6285 when the function accesses prior frames.
6287 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6290 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6291 gen_llsym, otherwise call maybe_gen_llsym.
6294 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6295 has ',' character inside of it.
6297 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6300 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6301 as clobbering CC_REGNUM.
6303 2018-03-08 Richard Biener <rguenther@suse.de>
6306 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6307 (follow_copies_to_constant): Do not follow SSA names registered
6310 2018-03-08 Richard Biener <rguenther@suse.de>
6312 PR tree-optimization/84178
6313 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6315 (version_loop_for_if_conversion): Delay update_ssa call.
6316 (tree_if_conversion): Delay update_ssa until after predicate
6319 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6321 PR tree-optimization/84178
6322 * tree-if-conv.c (release_bb_predicate): Remove the
6323 the assertion that the stmts have NULL use_ops.
6324 Discard the statements, asserting that they haven't
6325 yet been added to a BB.
6327 2018-03-08 Richard Biener <rguenther@suse.de>
6329 PR tree-optimization/84746
6330 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6331 (phi_translate): Pass in destination ANTIC_OUT set.
6332 (phi_translate_1): Likewise. For a simplified result lookup
6333 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6334 (phi_translate_set): Adjust.
6335 (do_pre_regular_insertion): Likewise.
6336 (do_pre_partial_partial_insertion): Likewise.
6338 2018-03-08 Martin Liska <mliska@suse.cz>
6340 PR gcov-profile/84735
6341 * doc/gcov.texi: Document usage of profile files.
6342 * gcov-io.h: Document changes in the format.
6344 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6348 * dwarf2out.c (struct dw_line_info_table): Update comments for
6350 (FORCE_RESET_NEXT_VIEW): New.
6351 (FORCE_RESETTING_VIEW_P): New.
6352 (RESETTING_VIEW_P): Check for -1 too.
6353 (ZERO_VIEW_P): Likewise.
6354 (new_line_info_table): Force-reset next view.
6355 (dwarf2out_begin_function): Likewise.
6356 (dwarf2out_source_line): Simplify zero_view_p initialization.
6357 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6358 view directly. Omit view when omitting .loc at line 0.
6360 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6362 PR tree-optimization/84740
6363 * tree-switch-conversion.c (process_switch): Call build_constructors
6364 only if info.phi_count is non-zero.
6366 PR tree-optimization/84739
6367 * tree-tailcall.c (find_tail_calls): Check call arguments against
6368 DECL_ARGUMENTS (current_function_decl) rather than
6369 DECL_ARGUMENTS (func) when checking for tail recursion.
6371 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6373 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6374 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6375 Volker Reichelt's entry and add entries for people that perform
6376 GCC fuzzy testing and report numerous bugs.
6378 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6381 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6382 readonly data in sdata, if that is disabled.
6383 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6384 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6385 -mreadonly-in-sdata option.
6387 2018-03-07 Martin Sebor <msebor@redhat.com>
6389 PR tree-optimization/84468
6390 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6391 basic block when looking for nul assignment.
6393 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6396 * except.h (output_function_exception_table): Adjust prototype.
6397 * except.c (output_function_exception_table): Remove FNNAME parameter
6398 and add SECTION parameter. Ouput one part of the table at a time.
6399 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6400 the first part of the exception table and emit unwind directives.
6401 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6402 (i386_pe_seh_cold_init): Likewise.
6403 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6404 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6405 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6406 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6407 * config/i386/winnt.c: Include except.h.
6408 (struct seh_frame_state): Add reg_offset, after_prologue and
6409 in_cold_section fields.
6410 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6411 (i386_pe_seh_cold_init): New function.
6412 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6413 to seh->in_cold_section.
6414 (seh_emit_push): Record the offset of the push.
6415 (seh_emit_save): Record the offet of the save.
6416 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6417 Test seh->after_prologue to disregard the epilogue.
6418 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6419 (i386_pe_end_cold_function): New function.
6421 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6424 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6425 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6428 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6430 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6431 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6434 * multiple_target.c: Include tree-inline.h and intl.h.
6435 (expand_target_clones): Diagnose and fail if node->definition and
6436 !tree_versionable_function_p (node->decl).
6438 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6440 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6442 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6443 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6444 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6446 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6449 * combine.c (try_combine): Use reg_or_subregno instead of handling
6450 just paradoxical SUBREGs and REGs.
6452 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6454 * config/arc/arc.c (arc_finalize_pic): Remove function.
6455 (arc_must_save_register): We use single base PIC register, remove
6456 checks to save/restore the PIC register.
6457 (arc_expand_prologue): Likewise.
6458 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6460 (arc_verify_short): Likewise.
6461 (arc_attr_type): Likewise.
6462 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6463 (walk_stores): Likewise.
6464 (arc_address_cost): Make it static.
6465 (arc_verify_short): Likewise.
6466 (branch_dest): Likewise.
6467 (arc_attr_type): Likewise.
6468 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6469 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6470 (arc_final_prescan_insn): Remove inserting the nops due to
6471 hardware hazards. It is done in reorg step.
6472 (insn_length_variant_t): Remove.
6473 (insn_length_parameters_t): Likewise.
6474 (arc_insn_length_parameters): Likewise.
6475 (arc_get_insn_variants): Likewise.
6476 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6478 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6481 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6484 PR tree-optimization/84687
6485 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6487 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6489 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6491 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6492 Rename to ppc_speculation_barrier.
6493 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6494 __builtin_ppc_speculation_barrier.
6496 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6499 * combine.c (combine_simplify_rtx): Don't try to simplify if
6500 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6503 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6505 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6506 to 32 bytes when compiling for POWER9.
6508 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6511 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6512 regparm >= 3 with no arg reg available also for calls with
6513 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6516 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6518 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6520 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6523 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6525 2018-03-05 Richard Biener <rguenther@suse.de>
6527 PR tree-optimization/84486
6528 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6529 When inserting a __builtin_assume_aligned call set the LHS
6530 SSA name alignment info accordingly.
6532 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6534 PR tree-optimization/84114
6535 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6536 Avoid reassociation of FLOAT_MODE addition.
6538 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6540 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6541 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6542 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6543 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6544 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6545 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6546 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6548 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6549 __builtin_ia32_wbinvd): New builtins.
6550 (SPECIAL_ARGS2): New.
6551 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6552 (SPECIAL_ARGS2): New.
6553 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6554 (ix86_valid_target_attribute_inner_p): Ditto.
6555 (ix86_init_mmx_sse_builtins): Add special_args2.
6556 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6557 TARGET_WBNOINVD_P): New.
6558 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6559 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6560 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6561 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6562 * config/i386/pconfigintrin.h: New file.
6563 * config/i386/wbnoinvdintrin.h: Ditto.
6564 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6565 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6567 2018-03-05 Richard Biener <rguenther@suse.de>
6569 PR tree-optimization/84670
6570 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6572 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6573 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6574 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6575 doesn't grow if all successors (recursively) were visited at least
6578 2018-03-05 Richard Biener <rguenther@suse.de>
6580 PR tree-optimization/84650
6581 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6582 if executed in the loop pipeline.
6584 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6586 * doc/configfiles.texi (Configuration Files): Move info about
6587 conditionalizing $target-protos.h to...
6588 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6589 differs from $target-protos.h.
6591 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6592 Chung-Ju Wu <jasonwucj@gmail.com>
6594 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6595 * config/nds32/nds32-multiple.md (setmemsi): Define.
6596 * config/nds32/nds32-memory-manipulation.c
6597 (nds32_gen_dup_4_byte_to_word_value): New.
6598 (emit_setmem_word_loop): New.
6599 (emit_setmem_byte_loop): New.
6600 (nds32_expand_setmem_loop): New.
6601 (nds32_expand_setmem_loop_v3m): New.
6602 (nds32_expand_setmem_unroll): New.
6603 (nds32_expand_setmem): New.
6605 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6606 Chung-Ju Wu <jasonwucj@gmail.com>
6608 * config/nds32/nds32-memory-manipulation.c
6609 (nds32_emit_load_store): New.
6610 (nds32_emit_post_inc_load_store): New.
6611 (nds32_emit_mem_move): New.
6612 (nds32_emit_mem_move_block): New.
6613 (nds32_expand_movmemsi_loop_unknown_size): New.
6614 (nds32_expand_movmemsi_loop_known_size): New.
6615 (nds32_expand_movmemsi_loop): New.
6616 (nds32_expand_movmemsi_unroll): New.
6617 (nds32_expand_movmemqi): Rename ...
6618 (nds32_expand_movmemsi): ... to this.
6619 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6620 (movmemsi): ... to this.
6621 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6622 (nds32_expand_movmemsi): ... to this.
6624 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6625 Monk Chiang <sh.chiang04@gmail.com>
6626 Chung-Ju Wu <jasonwucj@gmail.com>
6628 * config/nds32/nds32-protos.h
6629 (nds32_expand_load_multiple): New arguments.
6630 (nds32_expand_store_multiple): Ditto.
6631 (nds32_valid_multiple_load_store): Rename ...
6632 (nds32_valid_multiple_load_store_p): ... to this.
6633 * config/nds32/nds32-memory-manipulation.c
6634 (nds32_expand_load_multiple): Refine implementation.
6635 (nds32_expand_store_multiple): Ditto.
6636 * config/nds32/nds32-multiple.md
6637 (load_multiple): Update nds32_expand_load_multiple interface.
6638 (store_multiple): Update nds32_expand_store_multiple interface.
6639 * config/nds32/nds32-predicates.c
6640 (nds32_valid_multiple_load_store): Rename ...
6641 (nds32_valid_multiple_load_store_p): ... to this and refine
6643 * config/nds32/predicates.md
6644 (nds32_load_multiple_and_update_address_operation): New predicate.
6645 (nds32_store_multiple_and_update_address_operation): New predicate.
6647 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6648 Chung-Ju Wu <jasonwucj@gmail.com>
6650 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6651 (combo): New attribute.
6652 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6654 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6656 * config/nds32/nds32.opt: Change -mcmodel= default value.
6658 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6659 Monk Chiang <sh.chiang04@gmail.com>
6660 Chung-Ju Wu <jasonwucj@gmail.com>
6662 * config/nds32/constants.md (unspec_element): New enum.
6663 * config/nds32/constraints.md (Umw): New constraint.
6664 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6665 * config/nds32/nds32-intrinsic.md: Likewise.
6666 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6667 (nds32_valid_smw_lwm_base_p): New.
6668 (nds32_output_smw_single_word): New.
6669 (nds32_output_lmw_single_word): New.
6670 (nds32_expand_unaligned_load): New.
6671 (nds32_expand_unaligned_store): New.
6672 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6673 (nds32_output_smw_single_word): Declare.
6674 (nds32_output_lmw_single_word): Declare.
6675 (nds32_expand_unaligned_load): Declare.
6676 (nds32_expand_unaligned_store): Declare.
6677 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6678 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6679 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6680 NDS32_BUILTIN_UASTORE_DW.
6681 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6684 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6685 Kito Cheng <kito.cheng@gmail.com>
6686 Chung-Ju Wu <jasonwucj@gmail.com>
6688 * config/nds32/nds32-intrinsic.c
6689 (nds32_expand_builtin_null_ftype_reg): Delete.
6690 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6691 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6692 (nds32_read_argument): New.
6693 (nds32_legitimize_target): Ditto.
6694 (nds32_legitimize_argument): Ditto.
6695 (nds32_check_constant_argument): Ditto.
6696 (nds32_expand_unop_builtin): Ditto.
6697 (nds32_expand_unopimm_builtin): Ditto.
6698 (nds32_expand_binop_builtin): Ditto.
6699 (nds32_builtin_decl_impl): Ditto.
6700 (builtin_description): Ditto.
6701 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6702 (nds32_init_builtins_impl): Ditto.
6703 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6704 (nds32_builtin_decl): New.
6705 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6706 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6708 2018-03-02 Jeff Law <law@redhat.com>
6710 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6711 (redundant_insn, fill_simple_delay_slots): Likewise.
6712 (fill_slots_from_thread): Likewise.
6713 * resource.c (mark_referenced_resources): Likewise.
6714 (mark_set_resources, find_dead_or_set_registers): Likewise.
6716 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6718 * substring-locations.h (format_warning_va): Formatting fix for
6720 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6722 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6723 * substring-locations.c: Include intl.h.
6724 (format_warning_va): Turned into small wrapper around
6725 format_warning_n_va, renamed to ...
6726 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6727 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6729 (format_warning_at_substring_n): New function.
6730 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6731 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6732 format_warning_at_substring with just a shorter name instead of
6733 const function pointer.
6734 (fmtwarn_n): New function.
6735 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6736 appropriate, get rid of all the fmtstr temporaries, move conditionals
6737 with G_() wrapped string literals directly into fmtwarn arguments,
6738 cast dir.len to (int), formatting fixes.
6740 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6742 * doc/invoke.texi: Remove "Cilk Plus" references.
6744 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6745 Richard Biener <rguenther@suse.de>
6748 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6749 for error or warning attributes if CALL_FROM_THUNK_P is set.
6752 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6755 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6756 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6759 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6760 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6762 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6763 instead of -1U in last predictors element's probability member.
6765 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6768 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6769 arguments if they are comparable.
6771 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6773 PR tree-optimization/84634
6774 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6775 masks and masked_loop_p with a single loop_masks, making sure it's
6776 null for bb vectorization.
6778 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6780 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6781 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6782 loop->force_vectorize to check whether there is no alias.
6784 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6787 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6789 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6791 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6792 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6793 instead of a loop around prev_real_insn.
6794 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6798 * config/i386/i386.c (ix86_print_operand): Use conditional
6799 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6802 2018-03-02 Richard Biener <rguenther@suse.de>
6804 PR tree-optimization/84427
6805 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6806 (bitmap_set_subtract_values): Rewrite to handle multiple
6809 (prune_clobbered_mems): Likewise.
6810 (phi_translate): Take edge instead of pred/phiblock.
6811 (phi_translate_1): Likewise.
6812 (phi_translate_set): Likewise. Insert all translated
6813 exprs for a value into the set, keeping possibly multiple
6814 expressions per value.
6815 (compute_antic_aux): Adjust for phi_translate changes.
6816 When intersecting union the expressions and prune those
6817 not in the final value set, keeping possibly multiple
6818 expressions per value. Do not use value-insertion
6819 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6820 all expressions. Add verification that the value-sets
6821 only shrink during iteration.
6822 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6823 (do_pre_regular_insertion): Likewise.
6824 (do_pre_partial_partial_insertion): Likewise.
6826 2018-03-02 Richard Biener <rguenther@suse.de>
6829 * config/darwin.c (saved_debug_info_level): New static global.
6830 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6831 (darwin_asm_lto_end): Restore debug info generation settings.
6833 2018-03-01 Martin Liska <mliska@suse.cz>
6836 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6839 2018-03-01 Richard Biener <rguenther@suse.de>
6842 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6843 type annotation with LTO when debug was disabled at compile-time.
6845 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6847 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6849 (mips_final_postscan_insn): Likewise.
6851 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6853 PR rtl-optimization/84528
6854 * alias.c (init_alias_target): Add commentary.
6855 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6856 a unique base value if the frame pointer is not eliminated
6857 to the stack pointer.
6859 2018-03-01 Tom de Vries <tom@codesourcery.com>
6861 PR rtl-optimization/83327
6862 * lra-int.h (hard_regs_spilled_into): Declare.
6863 * lra.c (hard_regs_spilled_into): Define.
6864 (init_reg_info): Init hard_regs_spilled_into.
6865 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6866 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6867 (process_bb_lives): Handle hard_regs_spilled_into.
6868 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6869 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6871 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6873 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6874 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6875 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6876 * config/rs6000/aix72.h: New file.
6878 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6880 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6881 instead of warning_at with conditional singular and plural messages
6885 * stor-layout.c (update_alignment_for_field): For
6886 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6887 && !DECL_PACKED (field), do the alignment update, just use
6888 only desired_align instead of MAX (type_align, desired_align)
6890 (place_field): Don't do known_align < desired_align handling
6891 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6892 is non-NULL, instead do it after rli->prev_field handling and
6893 only if not within a bitfield word. For DECL_PACKED (field)
6894 use type_align of BITS_PER_UNIT.
6896 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6898 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6899 superfluous parentheses and trailing spaces.
6901 2018-02-28 Richard Biener <rguenther@suse.de>
6903 PR tree-optimization/84584
6904 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6905 SCoPs with fake exit edge.
6907 2018-02-28 Martin Liska <mliska@suse.cz>
6910 * timevar.c (timer::print): Fix format to properly print 100%
6913 2018-02-28 Richard Biener <rguenther@suse.de>
6916 * genmatch.c (capture_info::walk_match): Do not mark
6917 captured expressions without operands as expr_p given
6918 they act more like predicates and should be subject to
6919 "lost tail" side-effect preserving.
6921 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6923 PR rtl-optimization/81611
6924 * auto-inc-dec.c (attempt_change): Move dead note from
6925 mem_insn if it's the next use of regno
6926 (find_address): Take address use of reg holding
6927 non-incremented value. Add parm to limit search to the named
6929 (merge_in_block): Attempt to use a mem insn that is the next
6930 use of the original regno.
6932 2018-02-27 Martin Sebor <msebor@redhat.com>
6935 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6936 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6938 2018-02-27 Martin Sebor <msebor@redhat.com>
6940 PR translation/84207
6941 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6942 n argument to unsigned HOST_WIDE_INT.
6943 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6944 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6945 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6946 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6948 2018-02-27 Richard Biener <rguenther@suse.de>
6950 PR tree-optimization/84512
6951 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6952 Do not use the estimate returned from record_stmt_cost for
6953 the scalar iteration cost but sum properly using add_stmt_cost.
6955 2018-02-27 Richard Biener <rguenther@suse.de>
6957 PR tree-optimization/84466
6958 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6959 Adjust last change to less strictly validate use operands.
6961 2018-02-27 Martin Liska <mliska@suse.cz>
6963 PR gcov-profile/84548
6964 * gcov.c (process_file): Allow partial overlap and consider it
6965 also as group functions.
6966 (output_lines): Properly calculate range of lines for a group.
6968 2018-02-27 Martin Liska <mliska@suse.cz>
6970 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6971 'ggc' suffixes. Change first column width.
6972 (timer::print): Fix formatting of the column.
6974 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6976 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6977 preserve inline entry blocks for the sake of debug inline
6978 entry point markers alone.
6979 (remove_unused_locals): Suggest in comments a better place to
6980 force the preservation of inline entry blocks that are
6981 otherwise unused, but do not preserve them.
6983 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6985 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6987 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6990 * config/i386/constraints.md (Bs): Replace
6991 ix86_indirect_branch_register with
6992 TARGET_INDIRECT_BRANCH_REGISTER.
6994 * config/i386/i386.md (indirect_jump): Likewise.
6995 (tablejump): Likewise.
6996 (*sibcall_memory): Likewise.
6997 (*sibcall_value_memory): Likewise.
6998 Peepholes of indirect call and jump via memory: Likewise.
6999 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7000 (*sibcall_value_GOT_32): Likewise.
7001 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7002 (GOT_memory_operand): Likewise.
7003 (call_insn_operand): Likewise.
7004 (sibcall_insn_operand): Likewise.
7005 (GOT32_symbol_operand): Likewise.
7006 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7008 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7010 PR rtl-optimization/83496
7011 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7012 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7013 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7014 redundant insn, if any.
7015 (relax_delay_slots): Likewise.
7016 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7018 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7020 PR tree-optimization/83965
7021 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7022 that grouped statements are part of a reduction chain. Return
7023 true if the statement is not marked as a reduction itself but
7025 (vect_recog_dot_prod_pattern): Don't check whether the statement
7026 is part of a group here.
7027 (vect_recog_sad_pattern): Likewise.
7028 (vect_recog_widen_sum_pattern): Likewise.
7030 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7033 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7035 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7038 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7040 (ix86_output_indirect_function_return): New prototype.
7041 (ix86_split_simple_return_pop_internal): Likewise.
7042 * config/i386/i386.c (indirect_return_via_cx): New.
7043 (indirect_return_via_cx_bnd): Likewise.
7044 (indirect_thunk_name): Handle return va CX_REG.
7045 (output_indirect_thunk_function): Create alias for
7046 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7047 (ix86_output_indirect_jmp): Remove the bool argument.
7048 (ix86_output_indirect_function_return): New function.
7049 (ix86_split_simple_return_pop_internal): Likewise.
7050 * config/i386/i386.md (*indirect_jump): Don't pass false
7051 to ix86_output_indirect_jmp.
7052 (*tablejump_1): Likewise.
7053 (simple_return_pop_internal): Change it to define_insn_and_split.
7054 Call ix86_split_simple_return_pop_internal to split it for
7056 (simple_return_indirect_internal): Call
7057 ix86_output_indirect_function_return instead of
7058 ix86_output_indirect_jmp.
7060 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7063 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7064 memset and value initialization afterwards.
7066 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7068 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7070 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7073 * common/config/aarch64/aarch64-common.c
7074 (aarch_option_optimization_table[]): Switch
7075 off fomit-frame-pointer
7077 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7078 Chung-Ju Wu <jasonwucj@gmail.com>
7080 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7082 (store_multiple): Ditto.
7084 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7086 * config.gcc: Add --with-cpu support for nds32 target.
7087 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7088 * config/nds32/nds32.opt: Add -mcpu= option.
7090 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7092 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7093 isel=yes): Warn for these deprecated options.
7095 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7097 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7098 ISA_2_5_MASKS_EMBEDDED.
7100 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7102 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7103 p->max as pointers rather than using iterative_hash_expr.
7105 2018-02-23 Carl Love <cel@us.ibm.com>
7107 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7108 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7110 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7111 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7112 P8V_BUILTIN_VEC_VUNSIGNED2.
7114 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7117 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7118 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7120 * lra-constraints.c (curr_insn_transform): Set up
7121 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7123 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7124 * lra-eliminations.c (spill_pseudos): Ditto.
7125 (process_insn_for_elimination): Ditto.
7126 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7127 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7128 LRA_NON_CLOBBERED_ALT.
7130 2018-02-22 Martin Sebor <msebor@redhat.com>
7132 PR tree-optimization/84480
7133 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7134 to maybe_diag_stxncpy_trunc. Call it.
7135 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7136 from gimple_fold_builtin_strcpy. Print inlining stack.
7137 (handle_builtin_stxncpy): Print inlining stack.
7138 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7140 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7143 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7144 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7145 and -fcheck-pointer-bounds are used together.
7146 (indirect_thunk_prefix): New enum.
7147 (indirect_thunk_need_prefix): New function.
7148 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7149 "_nt" instead of "_bnd" for NOTRACK prefix.
7150 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7151 (output_indirect_thunk_function): Likewise.
7153 (ix86_code_end): Update output_indirect_thunk_function calls.
7154 (ix86_output_indirect_branch_via_reg): Replace
7155 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7156 (ix86_output_indirect_branch_via_push): Likewise.
7157 (ix86_output_function_return): Likewise.
7158 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7159 incompatible with -fcf-protection=branch and
7160 -fcheck-pointer-bounds.
7162 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7165 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7166 Change gcc_assert call to output_operand_lossage.
7168 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7170 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7172 2018-02-22 DJ Delorie <dj@redhat.com>
7173 Sebastian Perta <sebastian.perta@renesas.com>
7174 Oleg Endo <olegendo@gcc.gnu.org>
7176 * config/rx/rx.c (rx_rtx_costs): New function.
7177 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7179 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7181 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7183 2018-02-22 Martin Liska <mliska@suse.cz>
7186 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7187 Add "native" as a possible value.
7189 2018-02-22 Martin Liska <mliska@suse.cz>
7192 * config/i386/i386.c (ix86_option_override_internal):
7193 Add "native" as a possible value for -march and -mtune.
7195 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7198 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7199 to all type variants.
7201 PR tree-optimization/84503
7202 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7203 width as info->bitpos + info->bitsize - start.
7204 (merged_store_group::merge_overlapping): Simplify width computation.
7205 (check_no_overlap): New function.
7206 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7207 start + width and last_order of the group, fail if check_no_overlap
7209 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7210 to group if check_no_overlap fails.
7212 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7214 * config/rs6000/altivec.md: Delete contraint arguments to
7215 define_expand, define_split, and define_peephole2, and in
7216 define_insn_and_split if always unused.
7217 * config/rs6000/darwin.md: Ditto.
7218 * config/rs6000/dfp.md: Ditto.
7219 * config/rs6000/rs6000.md: Ditto.
7220 * config/rs6000/sync.md: Ditto.
7221 * config/rs6000/vector.md: Ditto.
7222 * config/rs6000/vsx.md: Ditto.
7224 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7226 * config/rs6000/altivec.md: Write output control strings as braced
7227 blocks instead of double-quoted strings.
7228 * config/rs6000/darwin.md: Ditto.
7229 * config/rs6000/rs6000.md: Ditto.
7230 * config/rs6000/vector.md: Ditto.
7231 * config/rs6000/vsx.md: Ditto.
7233 2018-02-21 Jason Merrill <jason@redhat.com>
7235 PR c++/84314 - ICE with templates and fastcall attribute.
7236 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7238 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7240 * ipa-cp.c (determine_versionability): Fix comment typos.
7242 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7245 * ipa-cp.c (determine_versionability): Do not version functions caling
7248 2018-02-21 Martin Liska <mliska@suse.cz>
7251 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7252 Add "native" as a possible value.
7253 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7254 the macro when native cpu detection is available.
7256 2018-02-21 Martin Liska <mliska@suse.cz>
7259 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7260 Add "native" as a possible value.
7261 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7262 when native cpu detection is available.
7264 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7265 Martin Sebor <msebor@redhat.com>
7267 PR tree-optimization/84478
7268 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7270 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7271 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7272 support which is conservatively correct, for 2 only stay conservative
7273 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7274 argument to the 2 argument get_range_strlen, adjust 6 arg
7275 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7277 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7278 (gimple_fold_builtin_strlen): Pass true as last argument to
7281 2018-02-20 Martin Sebor <msebor@redhat.com>
7284 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7285 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7286 (builtin_memref::builtin_memref): Factor out parts into
7287 set_base_and_offset and call it.
7289 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7292 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7293 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7294 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7295 search at the associated MODE_INT.
7297 2018-02-20 Jeff Law <law@redhat.com>
7300 PR tree-optimization/81592
7302 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7303 for range data rather than using global data.
7304 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7305 range data rather than using global data.
7306 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7307 pass it to children as needed.
7308 (struct directive::fmtresult): Similarly.
7309 (struct directive::set_width): Similarly.
7310 (struct directive::set_precision): Similarly.
7311 (format_integer, format_directive, parse_directive): Similarly.
7312 (format_none): Accept unnamed vr_values parameter.
7313 (format_percent, format_floating, format_character): Similarly.
7314 (format_string, format_plain): Similarly.
7315 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7316 the EVRP range analyzer for range data rather than using global data.
7317 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7318 gimple-ssa-evrp-analyze.h
7319 (class sprintf_dom_walker): Add after_dom_children member function.
7320 Add evrp_range_analyzer member.
7321 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7322 range analyzer as needed.
7323 (sprintf_dom_walker::after_dom_children): New member function.
7324 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7326 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7327 (evrp_range_analyzer::pop_to_marker): Likewise.
7329 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7331 PR tree-optimization/84419
7332 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7333 with the required type if its current type is compatible but
7336 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7339 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7340 after vectorization.
7342 2018-02-20 Martin Liska <mliska@suse.cz>
7345 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7346 possible values if we don't have a hint.
7348 2018-02-20 Martin Liska <mliska@suse.cz>
7352 * final.c (shorten_branches): Build align_tab array with one
7354 * opts.c (finish_options): Add alignment option limit check.
7355 (MAX_CODE_ALIGN): Likewise.
7356 (MAX_CODE_ALIGN_VALUE): Likewise.
7357 * doc/invoke.texi: Document maximum allowed option value for
7358 all -falign-* options.
7360 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7363 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7364 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7365 * var-tracking.c (emit_note_insn_var_location): Remove all references
7366 to NOTE_INSN_CALL_ARG_LOCATION.
7367 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7368 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7369 Use copy_rtx_if_shared.
7370 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7371 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7372 (dwarf2out_var_location): Remove handling of
7373 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7375 * final.c (final_scan_insn): Remove all references to
7376 NOTE_INSN_CALL_ARG_LOCATION.
7377 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7378 before dumping final insns.
7379 * except.c (emit_note_eh_region_end): Remove all references to
7380 NOTE_INSN_CALL_ARG_LOCATION.
7381 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7382 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7383 * config/arc/arc.c (hwloop_optimize): Likewise.
7384 * config/arm/arm.c (create_fix_barrier): Likewise.
7385 * config/s390/s390.c (s390_chunkify_start): Likewise.
7386 * config/sh/sh.c (find_barrier): Likewise.
7387 * config/i386/i386.c (rest_of_insert_endbranch,
7388 ix86_seh_fixup_eh_fallthru): Likewise.
7389 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7390 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7391 * config/frv/frv.c (frv_function_prologue): Likewise.
7392 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7394 (note_outside_basic_block_p): Remove all references to
7395 NOTE_INSN_CALL_ARG_LOCATION.
7396 * gengtype.c (adjust_field_rtx_def): Likewise.
7397 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7399 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7400 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7403 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7406 PR tree-optimization/84452
7407 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7408 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7411 2018-02-19 Martin Liska <mliska@suse.cz>
7414 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7416 2018-02-19 Martin Liska <mliska@suse.cz>
7417 Richard Sandiford <richard.sandiford@linaro.org>
7419 PR tree-optimization/82491
7420 * gimple-fold.c (get_base_constructor): Make earlier bail out
7423 2018-02-19 Carl Love <cel@us.ibm.com>
7425 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7426 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7428 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7429 P8V_BUILTIN_VEC_NEG.
7431 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7433 * config/rl78/rl78.md (movdf): New define expand.
7435 2018-02-19 Martin Liska <mliska@suse.cz>
7438 * doc/invoke.texi: Fix typo.
7439 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7441 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7443 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7444 handle rs6000_single_float and rs6000_double_float specially for
7447 2018-02-16 Jeff Law <law@redhat.com>
7449 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7452 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7455 * ipa-inline.c (inline_small_functions): Fix a typo.
7457 2018-02-16 Nathan Sidwell <nathan@acm.org>
7459 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7461 2018-02-16 Carl Love <cel@us.ibm.com>
7463 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7464 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7465 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7466 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7467 expansion to P8V_BUILTIN_VEC_FLOAT2.
7469 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7471 PR rtl-optimization/70023
7472 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7473 src_regno into account.
7475 2018-02-16 Carl Love <cel@us.ibm.com>
7477 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7478 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7479 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7480 * config/rs6000/rs6000.c: Remove case statements for
7481 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7482 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7483 and P9V_BUILTIN_VEC_VINSERT4B.
7484 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7485 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7486 * config/rs6000/vsx.md:
7487 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7490 2018-02-16 Carl Love <cel@us.ibm.com>
7492 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7494 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7496 * config/rs6000/rs6000-c.c: Add the definitions for
7497 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7498 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7499 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7500 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7501 definition for insert4b and define insn *insert3b_internal.
7502 * doc/extend.texi: Add documentation for vec_extract4b.
7504 2018-02-16 Nathan Sidwell <nathan@acm.org>
7506 * doc/extend.texi (Backwards Compatibility): Mention friend
7507 injection. Note for-scope is deprecated.
7508 * doc/invoke.texi (-ffriend-injection): Deprecate.
7510 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7512 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7513 that moved to I2, also allow destinations that are a paradoxical
7514 subreg (instead of a normal reg).
7516 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7519 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7522 2018-02-16 Richard Biener <rguenther@suse.de>
7524 PR tree-optimization/84037
7525 PR tree-optimization/84016
7527 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7528 Adjust vec_construct for the fact we need additional higher latency
7529 128bit inserts for AVX256 and AVX512 vector builds.
7530 (ix86_add_stmt_cost): Scale vector construction cost for
7533 2018-02-16 Richard Biener <rguenther@suse.de>
7535 PR tree-optimization/84417
7536 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7537 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7538 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7540 2018-02-16 Martin Liska <mliska@suse.cz>
7543 * internal-fn.def (ASAN_CHECK): Set proper flags.
7544 (ASAN_MARK): Likewise.
7546 2018-02-16 Julia Koval <julia.koval@intel.com>
7548 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7549 from PTA_CANNONLAKE.
7551 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7554 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7555 Use ++iter rather than iter++ for std::list iterators.
7556 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7557 defer deleting them until all nodes in the forest are processed. Do
7558 free even leaf nodes. Change to_process into auto_vec.
7561 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7562 * vec.h (vec_default_construct): Use memset instead of placement new
7563 if BROKEN_VALUE_INITIALIZATION is defined.
7564 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7565 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7568 PR rtl-optimization/83723
7569 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7570 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7571 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7573 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7575 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7577 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7579 PR rtl-optimization/81443
7580 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7581 from inner REGs to paradoxical SUBREGs.
7583 2018-02-16 Richard Biener <rguenther@suse.de>
7585 PR tree-optimization/84399
7586 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7587 For operands we can analyze at their definition make sure we can
7588 analyze them at each use as well.
7590 2018-02-16 Richard Biener <rguenther@suse.de>
7592 PR tree-optimization/84190
7593 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7594 volatile accesses if the decl isn't volatile.
7596 2018-02-15 Jason Merrill <jason@redhat.com>
7598 PR c++/84314 - ICE with templates and fastcall attribute.
7599 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7600 TYPE_CANONICAL on an existing type.
7602 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7604 PR tree-optimization/84383
7605 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7606 dstoff nor call operand_equal_p if dstbase is NULL.
7608 PR tree-optimization/84334
7609 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7610 also a CONSTANT_CLASS_P, punt.
7612 2018-02-14 Jim Wilson <jimw@sifive.com>
7614 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7615 first SMALL_OPERAND check. New local min_second_step. Move assert
7616 to where locals are set. Add TARGET_RVC support.
7617 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7619 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7621 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7623 2018-02-14 Martin Sebor <msebor@redhat.com>
7625 PR tree-optimization/83698
7626 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7627 arrays constrain the offset range to their bounds.
7628 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7629 (builtin_access::overlap): Avoid setting the size of overlap if it's
7631 (maybe_diag_overlap): Also consider arrays when deciding what values
7632 of offsets to include in diagnostics.
7634 2018-02-14 Martin Sebor <msebor@redhat.com>
7637 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7638 that correspond to the kind of a declaration.
7640 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7643 * config/pa/pa.md: Load address of PIC label using the linkage table
7644 if the label is nonlocal.
7646 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7648 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7649 warning message if user requests -maltivec=be.
7650 * doc/invoke.texi: Document deprecation of -maltivec=be.
7652 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7655 * config/rs6000/rs6000-c.c: Update definitions for
7656 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7657 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7659 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7662 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7663 add _get_ssp intrinsics. Remove argument from
7664 __builtin_ia32_rdssp[d|q].
7665 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7666 * config/i386/i386-builtin.def: Remove argument from
7667 __builtin_ia32_rdssp[d|q].
7668 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7669 ix86_expand_special_args_builtin for _rdssp[d|q].
7670 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7671 Clear register before usage.
7672 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7673 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7675 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7677 PR tree-optimization/84357
7678 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7679 operand 1 of an ARRAY_REF too.
7681 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7684 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7685 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7687 (set_of_reg): New struct.
7688 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7689 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7690 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7692 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7693 Split into bitclr, bitset, bitinvert patterns if appropriate.
7694 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7695 use rx_fuse_in_memory_bitop.
7696 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7697 to named insn, correct maximum insn length.
7699 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7702 * machmode.def: Define a complex mode for PARTIAL_INT.
7703 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7705 * doc/rtl.texi: Document CSPImode.
7706 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7708 (msp430_hard_regno_nregs_with_padding): Likewise.
7710 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7713 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7715 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7717 PR rtl-optimization/84169
7718 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7719 we generated a parallel as new i3 and we split that to new i2 and i3
7720 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7721 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7722 those to i2, not i1. Partially rewrite this scan code.
7724 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7727 * stor-layout.c (place_field): For variable length fields, adjust
7728 offset_align afterwards not just based on the field's alignment,
7729 but also on the size.
7732 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7733 of exps and logs in the use_exp2 case.
7735 2018-02-13 Jeff Law <law@redhat.com>
7737 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7740 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7743 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7747 * common.opt (gas-loc-support, gas-locview-support): New.
7748 (ginline-points, ginternal-reset-location-views): New.
7749 * doc/invoke.texi: Document them. Use @itemx where intended.
7750 (gvariable-location-views): Adjust.
7751 * target.def (reset_location_view): New.
7752 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7753 (TARGET_RESET_LOCATION_VIEW): New.
7754 * doc/tm.texi: Rebuilt.
7755 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7756 (dwarf2out_default_as_locview_support): New.
7757 (output_asm_line_debug_info): Use option variables.
7758 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7759 (output_loc_list): Likewise.
7760 (add_high_low_attributes): Check option variables.
7761 Don't output entry view attribute in strict mode.
7762 (gen_inlined_subroutine_die): Check option variables.
7763 (dwarf2out_inline_entry): Likewise.
7764 (init_sections_and_labels): Likewise.
7765 (dwarf2out_early_finish): Likewise.
7766 (maybe_reset_location_view): New, from...
7767 (dwarf2out_var_location): ... here. Call it.
7768 * debug.h (dwarf2out_default_as_loc_support): Declare.
7769 (dwarf2out_default_as_locview_support): Declare.
7770 * hooks.c (hook_int_rtx_insn_0): New.
7771 * hooks.h (hook_int_rtx_insn_0): Declare.
7772 * toplev.c (process_options): Take -gas-loc-support and
7773 -gas-locview-support from dwarf2out. Enable
7774 -gvariable-location-views by default only with locview
7775 assembler support. Enable -ginternal-reset-location-views by
7776 default only if the target defines the corresponding hook.
7777 Enable -ginline-points by default if location views are
7778 enabled; force it disabled if statement frontiers are
7780 * tree-inline.c (expand_call_inline): Check option variables.
7781 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7783 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7785 PR tree-optimization/84321
7786 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7787 handling. Also check whether the anti-range contains any values
7788 that satisfy the mask; switch to a VR_RANGE if not.
7790 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7793 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7795 2018-02-13 Martin Jambor <mjambor@suse.cz>
7798 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7799 of call statements, also set location of a load to a temporary.
7801 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7803 * config/rl78/rl78.c (add_vector_labels): New function.
7804 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7805 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7806 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7807 which checks that no arguments are passed.
7808 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7809 * doc/extend.texi: Documentation for the new attribute.
7811 2018-02-13 Andreas Schwab <schwab@suse.de>
7813 * config/riscv/linux.h (CPP_SPEC): Define.
7815 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7818 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7819 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7820 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7821 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7822 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7823 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7824 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7825 temporarily for AES and PCLMUL builtins.
7827 PR tree-optimization/84339
7828 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7829 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7833 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7834 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7835 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7837 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7839 * omp-simd-clone.h: New file.
7840 * omp-simd-clone.c: Include omp-simd-clone.h.
7841 (expand_simd_clones): No longer static.
7842 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7843 cgraph.h and omp-simd-clone.h.
7844 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7845 (vect_recog_widen_shift_pattern): Formatting fix.
7846 (vect_pattern_recog_1): Don't check optab for calls.
7849 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7850 operands[2] into a REG before using gen_lowpart on it.
7852 2018-02-12 Jeff Law <law@redhat.com>
7855 * config/sh/sh.c (find_barrier): Consider a sibling call
7858 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7859 successfully back substituting a reg.
7861 2018-02-12 Richard Biener <rguenther@suse.de>
7863 PR tree-optimization/84037
7864 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7865 parameter, move visited init to caller.
7866 (vect_slp_analyze_operations): Separate cost from validity
7867 check, initialize visited once for all instances.
7868 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7870 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7872 (vect_model_promotion_demotion_cost): Likewise.
7873 (vectorizable_bswap): Guard cost modeling with !slp_node
7874 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7876 (vectorizable_call): Likewise.
7877 (vectorizable_conversion): Likewise.
7878 (vectorizable_assignment): Likewise.
7879 (vectorizable_shift): Likewise.
7880 (vectorizable_operation): Likewise.
7881 (vectorizable_store): Likewise.
7882 (vectorizable_load): Likewise.
7883 (vectorizable_condition): Likewise.
7884 (vectorizable_comparison): Likewise.
7886 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7889 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7890 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7891 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7893 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7896 * params.def (inline-min-speedup): Increase from 8 to 15.
7897 (max-inline-insns-auto): Decrease from 40 to 30.
7898 * ipa-split.c (consider_split): Add some buffer for function to
7899 be considered inlining candidate.
7900 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7903 2018-02-12 Richard Biener <rguenther@suse.de>
7905 PR tree-optimization/84037
7906 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7907 matched stmts if we cannot swap the non-matched ones.
7909 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7911 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7912 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7913 _mm_maskz_scalef_round_ss): New intrinsics.
7914 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7915 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7916 __builtin_ia32_scalefss_round): Remove.
7917 (__builtin_ia32_scalefsd_mask_round,
7918 __builtin_ia32_scalefss_mask_round): New intrinsics.
7919 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7920 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7921 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7922 "<round_constraint>")): Changed to ...
7923 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7924 "<round_scalar_constraint>")): ... this.
7925 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7926 %0, %1, %2<round_op3>}"): Changed to ...
7927 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7928 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7929 %2<round_scalar_mask_op3>}"): ... this.
7930 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7932 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7934 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7935 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7936 (_mm_maskz_sqrt_round_ss): New intrinsics.
7937 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7938 (__builtin_ia32_sqrtsd_mask_round)
7939 (__builtin_ia32_sqrtss_mask_round): New builtins.
7940 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7941 (__builtin_ia32_sqrtss_round): Remove.
7942 (__builtin_ia32_sqrtsd_mask_round)
7943 (__builtin_ia32_sqrtss_mask_round): New builtins.
7944 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7945 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7946 ((match_operand:VF_128 1 "vector_operand"
7947 "xBm,<round_constraint>")): Changed to ...
7948 ((match_operand:VF_128 1 "vector_operand"
7949 "xBm,<round_scalar_constraint>")): ... this.
7950 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7951 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7952 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7953 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7954 %<iptr>1<round_scalar_mask_op3>}): ... this.
7955 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7956 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7958 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7961 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7962 Cast vec_cmpeq result to correct type.
7963 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7964 Cast vec_cmpgt result to correct type.
7966 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7968 * final.c (final_scan_insn_1): Renamed from...
7969 (final_scan_insn): ... this. New wrapper, to recover
7970 seen from the outermost call in recursive ones.
7971 * config/sparc/sparc.c (output_return): Drop seen from call.
7972 (output_sibcall): Likewise.
7973 * config/visium/visium.c (output_branch): Likewise.
7975 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7977 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7980 2018-02-10 Alan Modra <amodra@gmail.com>
7983 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7984 Specify LR as an input.
7986 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7989 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7990 remove_member_access_dummy_vars): New functions.
7991 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7992 lower_omp_1, execute_lower_omp): Use them.
7994 PR rtl-optimization/84308
7995 * shrink-wrap.c (spread_components): Release todo vector.
7997 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7999 PR rtl-optimization/57193
8000 * ira-color.c (struct allocno_color_data): Add member
8001 conflict_allocno_hard_prefs.
8002 (update_conflict_allocno_hard_prefs): New.
8003 (bucket_allocno_compare_func): Add a preference based on
8004 conflict_allocno_hard_prefs.
8005 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8006 (color_allocnos): Remove a dead code. Initiate
8007 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8009 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8012 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8013 constraint from =wa to wa. Avoid a subreg on the output operand,
8014 instead use a pseudo and subreg it in a move.
8015 (p9_xxbrd_<mode>): Changed to ...
8016 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8017 (p9_xxbrd_v2df): New expander.
8018 (p9_xxbrw_<mode>): Changed to ...
8019 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8020 (p9_xxbrw_v4sf): New expander.
8022 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8024 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8026 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8029 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8030 multiply in 32-bit mode.
8031 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8032 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8035 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8037 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8038 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8039 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8040 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8042 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8045 * dwarf2out.c (is_trivial_indirect_ref): New function.
8046 (dwarf2out_late_global_decl): Do not generate a location
8047 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8048 and that are not defined in the current unit.
8050 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8052 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8053 instead of a libcall for UNORDERED.
8055 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8058 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8059 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8061 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8064 * config/s390/s390.c (s390_set_current_function): Invoke
8065 s390_indirect_branch_settings also if fndecl didn't change.
8067 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8069 * config/rs6000/rs6000.md (blockage): Set length to zero.
8071 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8073 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8075 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8078 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8079 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8083 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8084 PARALLEL incoming that failed vt_get_decl_and_offset check.
8087 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8088 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8090 (get_variable_section): For decls in named .bss* sections pass true as
8091 second argument to bss_initializer_p.
8093 2018-02-09 Marek Polacek <polacek@redhat.com>
8094 Jakub Jelinek <jakub@redhat.com>
8097 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8098 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8099 Sync some changes from cxx_fold_indirect_ref.
8101 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8103 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8105 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8106 (BLOCK_INLINE_ENTRY_LABEL): New.
8107 (dwarf2out_var_location): Disregard inline entry markers.
8108 (inline_entry_data): New struct.
8109 (inline_entry_data_hasher): New hashtable type.
8110 (inline_entry_data_hasher::hash): New.
8111 (inline_entry_data_hasher::equal): New.
8112 (inline_entry_data_table): New variable.
8113 (add_high_low_attributes): Add DW_AT_entry_pc and
8114 DW_AT_GNU_entry_view attributes if a pending entry is found
8115 in inline_entry_data_table. Add old entry_pc attribute only
8116 if debug nonbinding markers are disabled.
8117 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8118 markers are enabled.
8119 (block_within_block_p, dwarf2out_inline_entry): New.
8120 (dwarf2out_finish): Check that no entries remained in
8121 inline_entry_data_table.
8122 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8123 (final_scan_insn, notice_source_line): Likewise.
8124 (rest_of_clean_state): Skip inline entry markers.
8125 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8127 * gimple.c (gimple_build_debug_inline_entry): New.
8128 * gimple.h (enum gimple_debug_subcode): Add
8129 GIMPLE_DEBUG_INLINE_ENTRY.
8130 (gimple_build_debug_inline_entry): Declare.
8131 (gimple_debug_inline_entry_p): New.
8132 (gimple_debug_nonbind_marker_p): Adjust.
8133 * insn-notes.def (INLINE_ENTRY): New.
8134 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8135 inline entry marker notes.
8136 (print_insn): Likewise.
8137 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8138 (INSN_DEBUG_MARKER_KIND): Likewise.
8139 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8140 * tree-inline.c (expand_call_inline): Build and insert
8141 debug_inline_entry stmt.
8142 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8143 inline entry blocks early, if nonbind markers are enabled.
8144 (dump_scope_block): Dump fragment info.
8145 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8146 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8147 (gimple_build_debug_inline_entry): New.
8148 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8149 Enable/disable inline entry points too.
8150 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8151 (DEBUG_INSN): Describe inline entry markers.
8153 * common.opt (gvariable-location-views): New.
8154 (gvariable-location-views=incompat5): New.
8155 * config.in: Rebuilt.
8156 * configure: Rebuilt.
8157 * configure.ac: Test assembler for view support.
8158 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8159 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8160 * dwarf2out.c (var_loc_view): New typedef.
8161 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8162 (dwarf2out_locviews_in_attribute): New.
8163 (dwarf2out_locviews_in_loclist): New.
8164 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8165 (enum dw_line_info_opcode): Add LI_adv_address.
8166 (struct dw_line_info_table): Add view.
8167 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8168 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8169 (zero_view_p): New variable.
8170 (ZERO_VIEW_P): New macro.
8171 (output_asm_line_debug_info): New.
8172 (struct var_loc_node): Add view.
8173 (add_AT_view_list, AT_loc_list): New.
8174 (add_var_loc_to_decl): Add view param. Test it against last.
8175 (new_loc_list): Add view params. Record them.
8176 (AT_loc_list_ptr): Handle loc and view lists.
8177 (view_list_to_loc_list_val_node): New.
8178 (print_dw_val): Handle dw_val_class_view_list.
8179 (size_of_die): Likewise.
8180 (value_format): Likewise.
8181 (loc_list_has_views): New.
8182 (gen_llsym): Set vl_symbol too.
8183 (maybe_gen_llsym, skip_loc_list_entry): New.
8184 (dwarf2out_maybe_output_loclist_view_pair): New.
8185 (output_loc_list): Output view list or entries too.
8186 (output_view_list_offset): New.
8187 (output_die): Handle dw_val_class_view_list.
8188 (output_dwarf_version): New.
8189 (output_compilation_unit_header): Use it.
8190 (output_skeleton_debug_sections): Likewise.
8191 (output_rnglists, output_line_info): Likewise.
8192 (output_pubnames, output_aranges): Update version comments.
8193 (output_one_line_info_table): Output view numbers in asm comments.
8194 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8195 Call maybe_gen_llsym.
8196 (loc_list_from_tree_1): Adjust.
8197 (add_AT_location_description): Create view list attribute if
8198 needed, check it's absent otherwise.
8199 (convert_cfa_to_fb_loc_list): Adjust.
8200 (maybe_emit_file): Call output_asm_line_debug_info for test.
8201 (dwarf2out_var_location): Reset views as needed. Precompute
8202 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8203 attribute. Set view.
8204 (new_line_info_table): Reset next view.
8205 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8206 (dwarf2out_source_line): Likewise. Output view resets and labels to
8207 the assembler, or select appropriate line info opcodes.
8208 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8209 (optimize_string_length): Catch it. Adjust.
8210 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8211 dw_val_class_view_list, and remove it if no longer needed.
8212 (hash_loc_list): Hash view numbers.
8213 (loc_list_hasher::equal): Compare them.
8214 (optimize_location_lists): Check whether a view list symbol is
8215 needed, and whether the locview attribute is present, and
8216 whether they match. Remove the locview attribute if no longer
8218 (index_location_lists): Call skip_loc_list_entry for test.
8219 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8220 Use output_dwarf_version.
8221 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8222 (struct dw_val_node): Add val_view_list.
8223 * final.c (SEEN_NEXT_VIEW): New.
8224 (set_next_view_needed): New.
8225 (clear_next_view_needed): New.
8226 (maybe_output_next_view): New.
8227 (final_start_function): Rename to...
8228 (final_start_function_1): ... this. Take pointer to FIRST,
8229 add SEEN parameter. Emit param bindings in the initial view.
8230 (final_start_function): Reintroduce SEEN-less interface.
8231 (final): Rename to...
8232 (final_1): ... this. Take SEEN parameter. Output final pending
8233 next view at the end.
8234 (final): Reintroduce seen-less interface.
8235 (final_scan_insn): Output pending next view before switching
8236 sections or ending a block. Mark the next view as needed when
8237 outputting variable locations. Notify debug backend of section
8238 changes, and of location view changes.
8239 (rest_of_handle_final): Adjust.
8240 * toplev.c (process_options): Autodetect value for debug variable
8241 location views option. Warn on incompat5 without -gdwarf-5.
8242 * doc/invoke.texi (gvariable-location-views): New.
8243 (gvariable-location-views=incompat5): New.
8244 (gno-variable-location-views): New.
8246 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8248 PR tree-optimization/84136
8249 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8250 that the result of find_edge is non-NULL.
8252 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8255 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8256 storing integer register in SImode. Fix cost of 256 and 512
8257 byte aligned SSE register store.
8259 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8261 * config/i386/i386.c (ix86_multiplication_cost): Fix
8262 multiplication cost for TARGET_AVX512DQ.
8264 2018-02-08 Marek Polacek <polacek@redhat.com>
8266 PR tree-optimization/84238
8267 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8270 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8272 PR tree-optimization/84265
8273 * tree-vect-stmts.c (vectorizable_store): Don't treat
8274 VMAT_CONTIGUOUS accesses as grouped.
8275 (vectorizable_load): Likewise.
8277 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8279 PR tree-optimization/81635
8280 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8281 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8282 (test_round_for_mask): New functions.
8283 (wide_int_cc_tests): Call test_round_for_mask.
8284 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8285 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8286 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8287 range returned by get_range_info.
8289 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8292 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8293 * symtab.c: Include builtins.h
8294 (symtab_node::output_to_lto_symbol_table_p): Move here
8295 from lto-streamer-out.c:output_symbol_p.
8296 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8297 (output_symbol_p): Move all logic to symtab.c
8298 (produce_symtab): Update.
8300 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8302 * config/s390/s390-opts.h (enum indirect_branch): Define.
8303 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8304 (s390_indirect_branch_via_thunk)
8305 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8306 (enum s390_indirect_branch_type): Define.
8307 * config/s390/s390.c (struct s390_frame_layout, struct
8308 machine_function): Remove.
8309 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8310 (indirect_branch_table_label_no, indirect_branch_table_name):
8312 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8313 (enum s390_indirect_branch_option): Define.
8314 (s390_return_addr_from_memory): New function.
8315 (s390_handle_string_attribute): New function.
8316 (s390_attribute_table): Add new attribute handler.
8317 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8318 (s390_indirect_branch_via_thunk): New function.
8319 (s390_indirect_branch_via_inline_thunk): New function.
8320 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8321 sibling call optimization for non z10 compiles.
8322 (s390_emit_call): Force indirect branch target to be a single
8323 register. Add r1 clobber for non-z10 compiles.
8324 (s390_emit_epilogue): Emit return jump via return_use expander.
8325 (s390_reorg): Handle JUMP_INSNs as execute targets.
8326 (s390_option_override_internal): Perform validity checks for the
8327 new command line options.
8328 (s390_indirect_branch_attrvalue): New function.
8329 (s390_indirect_branch_settings): New function.
8330 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8331 (s390_output_indirect_thunk_function): New function.
8332 (s390_code_end): Implement target hook.
8333 (s390_case_values_threshold): Implement target hook.
8334 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8336 * config/s390/s390.h (struct s390_frame_layout)
8337 (struct machine_function): Move here from s390.c.
8338 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8339 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8340 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8341 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8342 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8343 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8344 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8345 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8346 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8347 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8348 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8349 (mnemonic attribute): Add values which aren't recognized
8351 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8352 pattern for branch conversion. Fix mnemonic attribute.
8353 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8354 indirect branch via thunk if requested.
8355 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8356 ("*indirect_jump"): Disable for branch conversion using out of
8358 ("indirect_jump_via_thunk<mode>_z10")
8359 ("indirect_jump_via_thunk<mode>")
8360 ("indirect_jump_via_inlinethunk<mode>_z10")
8361 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8362 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8363 ("casesi_jump_via_inlinethunk<mode>_z10")
8364 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8365 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8366 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8367 ("*indirect2_jump"): Disable for branch conversion.
8368 ("casesi_jump"): Turn into expander and expand patterns for branch
8370 ("return_use"): New expander.
8371 ("*return"): Emit return via thunk and rename it to ...
8372 ("*return<mode>"): ... this one.
8373 * config/s390/s390.opt: Add new options and and enum for the
8376 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8378 * lra-constraints.c (match_reload): Unconditionally use
8379 gen_lowpart_SUBREG, rather than selecting between that
8380 and equivalent gen_rtx_SUBREG code.
8382 2018-02-08 Richard Biener <rguenther@suse.de>
8384 PR tree-optimization/84233
8385 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8386 changed flag instead of boguously re-using phi_inserted.
8388 2018-02-08 Martin Jambor <mjambor@suse.cz>
8390 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8391 static local variables.
8393 2018-02-08 Richard Biener <rguenther@suse.de>
8395 PR tree-optimization/84278
8396 * tree-vect-stmts.c (vectorizable_store): When looking for
8397 smaller vector types to perform grouped strided loads/stores
8398 make sure the mode is supported by the target.
8399 (vectorizable_load): Likewise.
8401 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8403 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8404 Increase LDP/STP opportunities by adding adjacent callee-saves.
8406 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8408 PR rtl-optimization/84068
8409 PR rtl-optimization/83459
8410 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8412 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8414 PR tree-optimization/84224
8415 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8416 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8419 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8422 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8423 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8424 count, remove one USE.
8426 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8428 * doc/install.texi (Configuration): Document the
8429 --with-long-double-format={ibm,ieee} PowerPC configuration
8433 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8434 Convert from define_expand to be define_insn_and_split. Rework
8435 float/double/_Float128 conversions to QI/HI/SImode to work with
8436 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8437 conversions to QI/HImode types did a store and then a load to
8438 truncate the value. For conversions to VSX registers, don't split
8439 the insn, instead emit the code directly. Use the code iterator
8440 any_fix to combine signed and unsigned conversions.
8441 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8442 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8443 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8444 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8445 (fix_<mode>di2_hw): Likewise.
8446 (fixuns_<mode>di2_hw): Likewise.
8447 (fix_<mode>si2_hw): Likewise.
8448 (fixuns_<mode>si2_hw): Likewise.
8449 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8450 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8451 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8452 fix<uns>_trunc<SFDF:mode>si2_p8.
8453 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8455 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8456 (fix<uns>_<mode>_mem): Likewise.
8457 (fctiw<u>z_<mode>_mem): Likewise.
8458 (fix<uns>_<mode>_mem): Likewise.
8459 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8460 the register allocator from doing a direct move to the GPRs to do
8461 a store, and instead use the ISA 3.0 store byte/half-word from
8462 vector register instruction. For IEEE 128-bit floating point,
8463 also optimize stores of 32-bit ints.
8464 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8466 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8468 * genextract.c (push_pathstr_operand): New function to support
8470 (walk_rtx): Call push_pathstr_operand.
8471 (print_path): Support [a-zA-Z].
8473 2018-02-07 Richard Biener <rguenther@suse.de>
8475 PR tree-optimization/84037
8476 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8477 (cse_and_gimplify_to_preheader): Declare.
8478 (vect_get_place_in_interleaving_chain): Likewise.
8479 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8481 (_loop_vec_info::~_loop_vec_info): Delete it.
8482 (cse_and_gimplify_to_preheader): New function.
8483 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8484 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8485 (vectorizable_load): Likewise. For grouped stores always base
8486 the IV on the first element.
8487 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8488 condition before gimplifying.
8490 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8492 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8493 *DIV_EXPR and *MOD_EXPR.
8495 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8498 * config/i386/i386.c (ix86_option_override_internal): Mask out
8499 the CF_SET bit when checking -fcf-protection.
8501 2018-02-07 Tom de Vries <tom@codesourcery.com>
8504 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8507 2018-02-07 Richard Biener <rguenther@suse.de>
8509 PR tree-optimization/84204
8510 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8513 PR tree-optimization/84205
8514 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8515 special-case isl_ast_op_zdiv_r.
8517 PR tree-optimization/84223
8518 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8519 Only add conditions from within the region.
8520 (gather_bbs::after_dom_children): Adjust.
8522 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8525 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8526 * config/avr/avr.md: Only post-reload split REG-REG moves if
8527 either register is GENERAL_REG_P.
8529 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8531 PR tree-optimization/84235
8532 * tree-ssa-scopedtables.c
8533 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8534 if the subtraction is performed in floating point type where NaNs are
8535 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8536 build 1. Formatting fix.
8538 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8541 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8542 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8543 and skip it regardless of bb boundaries. Use CALL_P macro,
8544 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8545 unnecessarily, formatting fix.
8547 2018-02-06 Michael Collison <michael.collison@arm.com>
8549 * config/arm/thumb2.md:
8550 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8551 (*thumb_mov_notscc): Ditto.
8553 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8556 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8557 unsigned_fix, not "s".
8559 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8561 * configure.ac (gcc_fn_eh_frame_ro): New function.
8562 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8563 correct .eh_frame permissions.
8564 * configure: Regenerate.
8566 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8568 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8571 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8573 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8574 Display warning message for -mno-speculate-indirect-jumps.
8576 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8578 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8580 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8582 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8584 PR tree-optimization/84225
8585 * tree-eh.c (find_trapping_overflow): Only call
8586 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8588 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8591 * config/i386/i386.c: Reimplement the check of possible options
8592 -mibt/-mshstk conbination. Change error messages.
8593 * doc/invoke.texi: Fix a typo: remove extra '='.
8595 2018-02-06 Marek Polacek <polacek@redhat.com>
8597 PR tree-optimization/84228
8598 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8600 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8603 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8604 emitted arch directives.
8605 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8606 __ARM_FEATURE_COPROC before changing architectures.
8608 2018-02-06 Richard Biener <rguenther@suse.de>
8610 * config/i386/i386.c (print_reg): Fix typo.
8611 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8613 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8615 * configure: Regenerate.
8617 2018-02-05 Martin Sebor <msebor@redhat.com>
8619 PR tree-optimization/83369
8620 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8623 2018-02-05 Martin Liska <mliska@suse.cz>
8625 * doc/invoke.texi: Cherry-pick upstream r323995.
8627 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8629 * ira.c (ira_init_register_move_cost): Adjust comment.
8631 2018-02-05 Martin Liska <mliska@suse.cz>
8633 PR gcov-profile/84137
8634 * doc/gcov.texi: Fix typo in documentation.
8636 2018-02-05 Martin Liska <mliska@suse.cz>
8638 PR gcov-profile/83879
8639 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8640 using dlopen functionality.
8642 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8644 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8645 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8646 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8647 _mm_maskz_range_round_ss): New intrinsics.
8648 (__builtin_ia32_rangesd128_round)
8649 (__builtin_ia32_rangess128_round): Remove.
8650 (__builtin_ia32_rangesd128_mask_round,
8651 __builtin_ia32_rangess128_mask_round): New builtins.
8652 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8653 __builtin_ia32_rangess128_round): Remove.
8654 (__builtin_ia32_rangesd128_mask_round,
8655 __builtin_ia32_rangess128_mask_round): New builtins.
8656 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8657 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8658 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8659 "<round_saeonly_constraint>")): Changed to ...
8660 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8661 "<round_saeonly_scalar_constraint>")): ... this.
8662 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8663 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8664 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8665 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8666 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8668 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8670 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8672 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8673 Remove all values except native, 8540 and 8548.
8675 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8677 * config/i386/i386.c (ix86_output_function_return): Pass
8678 INVALID_REGNUM, instead of -1, as invalid register number to
8679 indirect_thunk_name and output_indirect_thunk.
8681 2018-02-02 Julia Koval <julia.koval@intel.com>
8683 * config.gcc: Add -march=icelake.
8684 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8685 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8686 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8687 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8688 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8689 (processor_target_table): Add icelake.
8690 (ix86_option_override_internal): Handle new PTAs.
8691 (get_builtin_code_for_version): Handle icelake.
8692 (M_INTEL_COREI7_ICELAKE): New.
8693 (fold_builtin_cpu): Handle icelake.
8694 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8695 * doc/invoke.texi: Add -march=icelake.
8697 2018-02-02 Julia Koval <julia.koval@intel.com>
8699 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8700 to wide_int_bitmask.
8701 * wide-int-bitmask.h: New.
8703 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8706 * config/i386/i386.md: Replace Pmode with word_mode in
8707 builtin_setjmp_setup and builtin_longjmp to support x32.
8709 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8713 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8715 (rs6000_supported_cpu_names): New static variable.
8716 (linux_cpu_translation_table): Likewise.
8717 (elf_platform) <cpu>: Define new static variable and use it.
8718 Translate kernel AT_PLATFORM name to canonical name if needed.
8719 Error if platform name is unknown.
8721 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8724 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8726 2018-02-01 Jeff Law <law@redhat.com>
8729 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8730 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8731 the scratch if RELEASE_VIA_POP is false.
8732 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8733 If we have to save a temporary register, decrement SIZE appropriately.
8734 Pass new arguments to release_scratch_register_on_entry.
8735 (ix86_adjust_stack_and_probe): Likewise.
8736 (ix86_emit_probe_stack_range): Pass new arguments to
8737 release_scratch_register_on_entry.
8739 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8741 PR rtl-optimization/84157
8742 * combine.c (change_zero_ext): Use REG_P predicate in
8743 front of HARD_REGISTER_P predicate.
8745 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8747 * config/avr/avr.c (avr_option_override): Move disabling of
8748 -fdelete-null-pointer-checks to...
8749 * common/config/avr/avr-common.c (avr_option_optimization_table):
8752 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8754 PR tree-optimization/81635
8755 * tree-data-ref.c (split_constant_offset_1): For types that
8756 wrap on overflow, try to use range info to prove that wrapping
8759 2018-02-01 Renlin Li <renlin.li@arm.com>
8762 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8764 (aarch64_register_move_cost): Likewise.
8765 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8767 (REG_CLASS_NAMES): Likewise.
8768 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8769 TAILCALL_ADDR_REGS. Remove IP registers.
8770 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8772 2018-02-01 Richard Biener <rguenther@suse.de>
8774 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8775 for specifying RPO order and allow NULL for that.
8776 * domwalk.c (dom_walker::dom_walker): Likewise.
8777 (dom_walker::walk): Handle NULL RPO order.
8778 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8780 (rewrite_update_dom_walker): Likewise.
8781 (mark_def_dom_walker): Likewise.
8783 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8785 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8786 (aarch64_maybe_expand_sve_subreg_move): Declare.
8787 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8788 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8790 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8791 that are semantically a reverse operation.
8792 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8793 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8794 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8796 (aarch64_can_change_mode_class): For big-endian, forbid changes
8797 between two SVE modes if they have different element sizes.
8799 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8801 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8802 the TImode handling for big-endian targets.
8804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8806 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8807 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8809 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8810 Remove BSWAP handing for big-endian targets and use the form of
8811 LD1RQ appropariate for the mode.
8813 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8815 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8816 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8819 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8822 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8823 check for operands that need to go through aarch64_sve_reload_be.
8825 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8827 PR tree-optimization/81661
8828 PR tree-optimization/84117
8829 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8830 * tree-eh.c: Include gimplify.h.
8831 (find_trapping_overflow, replace_trapping_overflow,
8832 rewrite_to_non_trapping_overflow): New functions.
8833 * tree-vect-loop.c: Include tree-eh.h.
8834 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8835 * tree-data-ref.c: Include tree-eh.h.
8836 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8838 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8840 PR rtl-optimization/84123
8841 * combine.c (change_zero_ext): Check if hard register satisfies
8842 can_change_dest_mode before calling gen_lowpart_SUBREG.
8844 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8847 * ira.c (ira_init_register_move_cost): Remove assert.
8849 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8851 PR rtl-optimization/84071
8852 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8853 * doc/tm.texi: Regenerate.
8855 2018-01-31 Richard Biener <rguenther@suse.de>
8857 PR tree-optimization/84132
8858 * tree-data-ref.c (analyze_miv_subscript): Properly
8859 check whether evolution_function_is_affine_multivariate_p
8860 before calling gcd_of_steps_may_divide_p.
8862 2018-01-31 Julia Koval <julia.koval@intel.com>
8865 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8866 * config/i386/i386.md (rdpid_rex64) New.
8867 (rdpid): Make 32bit only.
8869 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8872 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8873 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8875 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8878 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8880 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8882 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8884 PR rtl-optimization/84071
8885 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8886 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8888 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8890 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8891 (arc_attribute_table): Add 'aux' attribute.
8892 (arc_in_small_data_p): Consider aux like variables.
8893 (arc_is_aux_reg_p): New function.
8894 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8895 (arc_get_aux_arg): New function.
8896 (prepare_move_operands): Handle aux-register access.
8897 (arc_handle_aux_attribute): New function.
8898 * doc/extend.texi (ARC Variable attributes): Add subsection.
8900 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8902 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8903 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8904 (arc_attribute_table): Add 'uncached' attribute.
8905 (arc_print_operand): Print '.di' flag for uncached memory
8907 (arc_in_small_data_p): Do not consider for small data the uncached
8909 (arc_is_uncached_mem_p): New function.
8910 * config/arc/predicates.md (compact_store_memory_operand): Check
8911 for uncached memory accesses.
8912 (nonvol_nonimm_operand): Likewise.
8913 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8915 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8918 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8919 falign-loops=): Add Optimization flag.
8921 2018-01-30 Jeff Law <law@redhat.com>
8924 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8925 INT_REGISTERS_SAVED. Check it prior to calling
8926 get_scratch_register_on_entry.
8927 (ix86_adjust_stack_and_probe): Similarly.
8928 (ix86_emit_probe_stack_range): Similarly.
8929 (ix86_expand_prologue): Corresponding changes.
8931 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8934 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8935 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8937 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8940 * lra-constraints.c (curr_insn_transform): Process AND in the
8943 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8945 PR rtl-optimization/83986
8946 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8947 dependence against last_pending_memory_flush in addition to
8950 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8952 PR tree-optimization/81611
8953 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8956 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8959 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8962 2018-01-30 Richard Biener <rguenther@suse.de>
8963 Jakub Jelinek <jakub@redhat.com>
8965 PR tree-optimization/84111
8966 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8967 inner loops added during recursion, as they don't have up-to-date
8970 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8973 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8974 (can_inline_edge_by_limits_p): ... here.
8975 (can_early_inline_edge_p, check_callers,
8976 update_caller_keys, update_callee_keys, recursive_inlining,
8977 add_new_edges_to_heap, speculation_useful_p,
8978 inline_small_functions,
8979 inline_small_functions, flatten_function,
8980 inline_to_all_callers_1): Update.
8982 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8984 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8987 2018-01-30 Richard Biener <rguenther@suse.de>
8989 PR tree-optimization/83008
8990 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8991 invariant and constant vector uses in stmts when they need
8994 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8997 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8998 * configure: Regenerate.
9000 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9002 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9004 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9005 Use gen_rtx_REG rather than gen_lowpart.
9007 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9009 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9010 rather than 0 when creating partial subregs.
9012 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9014 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9017 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9020 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9021 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9022 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9023 flags. This restores the settings used before the 2017-07-24.
9024 Turning off pre increment/decrement/modify allows IVOPTS to
9025 optimize DF/SF loops where the index is an int.
9027 2018-01-29 Richard Biener <rguenther@suse.de>
9028 Kelvin Nilsen <kelvin@gcc.gnu.org>
9031 * tree-vect-stmts.c (vectorizable_call): Don't call
9032 targetm.vectorize_builtin_md_vectorized_function if callee is
9035 2018-01-22 Carl Love <cel@us.ibm.com>
9037 * doc/extend.tex: Fix typo in second arg in
9038 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9040 2018-01-29 Richard Biener <rguenther@suse.de>
9042 PR tree-optimization/84086
9043 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9044 (flush_ssaname_freelist): When SSA names were released reset
9045 the SCEV hash table.
9047 2018-01-29 Richard Biener <rguenther@suse.de>
9049 PR tree-optimization/84057
9050 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9051 removed paths when removing edges.
9053 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9055 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9056 -mfunction-return=@var{choice}.
9058 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9061 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9063 (layout::print_source_line): Likewise.
9064 (test_get_line_width_without_trailing_whitespace): Add test cases.
9066 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9069 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9072 2018-01-26 Jim Wilson <jimw@sifive.com>
9074 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9076 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9079 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9081 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9082 and CMP + SUB-immediate -> SUBS.
9084 2018-01-26 Martin Sebor <msebor@redhat.com>
9086 PR tree-optimization/83896
9087 * tree-ssa-strlen.c (get_string_len): Rename...
9088 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9089 Avoid assuming length is constant.
9090 (handle_char_store): Use HOST_WIDE_INT for string length.
9092 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9095 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9096 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9098 2018-01-26 Richard Biener <rguenther@suse.de>
9100 PR rtl-optimization/84003
9101 * dse.c (record_store): Only record redundant stores when
9102 the earlier store aliases at least all accesses the later one does.
9104 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9106 PR rtl-optimization/83985
9107 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9108 REG_CFA_RESTORE insns.
9109 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9110 REG_CFA_RESTORE insns here.
9113 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9114 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9116 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9118 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9119 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9120 (arc_init): Likewise.
9121 (arc_override_options): Likewise.
9122 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9124 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9126 * config/arc/arc.h (TARGET_DBNZ): Define.
9127 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9128 properly set the tune attribute.
9129 (dbnz): Use TARGET_DBNZ guard.
9130 * config/arc/arc.opt (mtune): Add core3 option.
9132 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9134 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9135 recognize new pic like addresses.
9136 (arc_delegitimize_address): Clean up.
9138 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9140 * config/arc/arc-arches.def: Option mrf16 valid for all
9142 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9143 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9144 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9145 * config/arc/arc-tables.opt: Regenerate.
9146 * config/arc/arc.c (arc_conditional_register_usage): Handle
9147 reduced register file case.
9148 (arc_file_start): Set must have build attributes.
9149 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9151 * config/arc/arc.opt (mrf16): Add new option.
9152 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9153 * config/arc/genmultilib.awk: Handle new mrf16 option.
9154 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9155 * config/arc/t-multilib: Regenerate.
9156 * doc/invoke.texi (ARC Options): Document mrf16 option.
9158 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9160 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9161 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9162 (arc_attribute_table): Add 'secure_call' attribute.
9163 (arc_print_operand): Print secure call operand.
9164 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9166 (arc_is_secure_call_p): New function. * config/arc/arc.md
9167 (call_i): Add support for sjli instruction.
9168 (call_value_i): Likewise.
9169 * config/arc/constraints.md (Csc): New constraint.
9171 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9172 John Eric Martin <John.Martin@emmicro-us.com>
9174 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9175 * config/arc/arc.c (_arc_jli_section): New struct.
9176 (arc_jli_section): New type.
9177 (rc_jli_sections): New static variable.
9178 (arc_handle_jli_attribute): New function.
9179 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9180 (arc_file_end): New function.
9181 (TARGET_ASM_FILE_END): Define.
9182 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9183 (arc_add_jli_section): New function.
9184 (jli_call_scan): Likewise.
9185 (arc_reorg): Call jli_call_scan.
9186 (arc_output_addsi): Remove 'S' from printing asm operand.
9187 (arc_is_jli_call_p): New function.
9188 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9190 (movhi_insn): Likewise.
9191 (movsi_insn): Likewise.
9192 (movsi_set_cc_insn): Likewise.
9193 (loadqi_update): Likewise.
9194 (load_zeroextendqisi_update): Likewise.
9195 (load_signextendqisi_update): Likewise.
9196 (loadhi_update): Likewise.
9197 (load_zeroextendhisi_update): Likewise.
9198 (load_signextendhisi_update): Likewise.
9199 (loadsi_update): Likewise.
9200 (loadsf_update): Likewise.
9201 (movsicc_insn): Likewise.
9202 (bset_insn): Likewise.
9203 (bxor_insn): Likewise.
9204 (bclr_insn): Likewise.
9205 (bmsk_insn): Likewise.
9206 (bicsi3_insn): Likewise.
9207 (cmpsi_cc_c_insn): Likewise.
9208 (movsi_ne): Likewise.
9209 (movsi_cond_exec): Likewise.
9210 (clrsbsi2): Likewise.
9221 (call_i): Remove 'S' asm letter, add jli instruction.
9222 (call_value_i): Likewise.
9223 * config/arc/arc.op (mjli-always): New option.
9224 * config/arc/constraints.md (Cji): New constraint.
9225 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9227 (subsf3_fpx): Likewise.
9228 (mulsf3_fpx): Likewise.
9229 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9231 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9233 * doc/invoke.texi (ARC): Document mjli-always option.
9235 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9237 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9238 avoid addition with 0 and use incw and decw where possible.
9240 2018-01-26 Richard Biener <rguenther@suse.de>
9242 PR tree-optimization/81082
9243 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9244 association if it requires casting to unsigned.
9245 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9246 from fold_plusminus_mult_expr to catch important cases late when
9247 range info is available.
9249 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9251 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9252 * configure.ac (hidden_linkonce): New test.
9253 * configure: Regenerate.
9254 * config.in: Regenerate.
9256 2018-01-26 Julia Koval <julia.koval@intel.com>
9258 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9259 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9260 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9261 _mm_mask_bitshuffle_epi64_mask): Fix type.
9262 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9263 USI_FTYPE_V4DI_V4DI_USI): Remove.
9264 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9265 __builtin_ia32_vpshufbitqmb256_mask,
9266 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9267 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9268 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9270 2018-01-26 Alan Modra <amodra@gmail.com>
9273 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9274 UNSPEC_VBPERMQ. Sort other unspecs.
9276 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9278 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9280 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9283 * predict.c (drop_profile): Do not push/pop cfun; update also
9285 (handle_missing_profiles): Fix logic looking for zero profiles.
9287 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9290 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9291 on functions with #pragma omp declare simd or functions with simd
9293 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9294 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9295 Remove trailing \n from warning_at calls.
9297 2018-01-25 Tom de Vries <tom@codesourcery.com>
9300 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9301 for neutered workers.
9303 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9306 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9307 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9309 2018-01-24 Jeff Law <law@redhat.com>
9312 * i386.c (get_probe_interval): Move to earlier point.
9313 (ix86_compute_frame_layout): If -fstack-clash-protection and
9314 the frame is larger than the probe interval, then use pushes
9315 to save registers rather than reg->mem moves.
9316 (ix86_expand_prologue): Remove conditional for int_registers_saved
9319 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9322 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9323 min/max for never referenced object.
9325 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9328 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9330 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9331 attributes from DECL_ATTRIBUTES (decl) without affecting
9332 DECL_ATTRIBUTES (current_function_decl).
9333 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9334 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9336 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9338 PR tree-optimization/83979
9339 * fold-const.c (fold_comparison): Use constant_boolean_node
9340 instead of boolean_{true,false}_node.
9342 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9344 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9347 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9349 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9350 Simplify the clause that sets the length attribute.
9351 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9352 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9353 clause that sets the length attribute.
9354 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9356 2018-01-24 Tom de Vries <tom@codesourcery.com>
9359 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9360 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9361 Add strict parameter.
9362 (prevent_branch_around_nothing): Insert dummy insn between branch to
9363 label and label with no ptx insn inbetween.
9364 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9366 2018-01-24 Tom de Vries <tom@codesourcery.com>
9369 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9370 for neutered threads in warp.
9371 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9373 2018-01-24 Richard Biener <rguenther@suse.de>
9375 PR tree-optimization/83176
9376 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9379 2018-01-24 Richard Biener <rguenther@suse.de>
9381 PR tree-optimization/82819
9382 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9383 code generating pluses that are no-ops in the target precision.
9385 2018-01-24 Richard Biener <rguenther@suse.de>
9388 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9390 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9392 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9393 to merge probabilities.
9394 * predict.c (probably_never_executed): Also mark as cold functions
9395 with global 0 profile and guessed local profile.
9396 * profile-count.c (profile_probability::combine_with_count): New
9398 * profile-count.h (profile_probability::operator*,
9399 profile_probability::operator*=, profile_probability::operator/,
9400 profile_probability::operator/=): Reduce precision to adjusted
9401 and set value to guessed on contradictory divisions.
9402 (profile_probability::combine_with_freq): Remove.
9403 (profile_probability::combine_wiht_count): Declare.
9404 (profile_count::force_nonzero):: Set to adjusted.
9405 (profile_count::probability_in):: Set quality to adjusted.
9406 * tree-ssa-tail-merge.c (replace_block_by): Use
9409 2018-01-23 Andrew Waterman <andrew@sifive.com>
9410 Jim Wilson <jimw@sifive.com>
9412 * config/riscv/riscv.c (riscv_stack_boundary): New.
9413 (riscv_option_override): Set riscv_stack_boundary. Handle
9414 riscv_preferred_stack_boundary_arg.
9415 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9416 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9417 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9418 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9419 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9420 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9422 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9425 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9426 of struct ix86_frame.
9427 (ix86_expand_epilogue): Likewise. Add a local variable for
9428 the reg_save_offset field in struct ix86_frame.
9430 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9432 PR tree-optimization/82604
9433 * tree-loop-distribution.c (enum partition_kind): New enum item
9434 PKIND_PARTIAL_MEMSET.
9435 (partition_builtin_p): Support above new enum item.
9436 (generate_code_for_partition): Ditto.
9437 (compute_access_range): Differentiate cases that equality can be
9438 proven at all loops, the innermost loops or no loops.
9439 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9440 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9441 (finalize_partitions, distribute_loop): Don't fuse partition of
9442 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9443 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9446 2018-01-23 Martin Liska <mliska@suse.cz>
9448 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9449 order to ignore the predictor.
9450 (PRED_POLYMORPHIC_CALL): Likewise.
9451 (PRED_RECURSIVE_CALL): Likewise.
9453 2018-01-23 Martin Liska <mliska@suse.cz>
9455 * tree-profile.c (tree_profiling): Print function header to
9456 aware reader which function we are working on.
9457 * value-prof.c (gimple_find_values_to_profile): Do not print
9458 not interesting value histograms.
9460 2018-01-23 Martin Liska <mliska@suse.cz>
9462 * profile-count.h (enum profile_quality): Add
9463 profile_uninitialized as the first value. Do not number values
9464 as they are zero based.
9465 (profile_count::verify): Update sanity check.
9466 (profile_probability::verify): Likewise.
9468 2018-01-23 Nathan Sidwell <nathan@acm.org>
9470 * doc/invoke.texi (ffor-scope): Deprecate.
9472 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9474 PR tree-optimization/83510
9475 * domwalk.c (set_all_edges_as_executable): New function.
9476 (dom_walker::dom_walker): Convert bool param
9477 "skip_unreachable_blocks" to enum reachability. Move setup of
9478 edge flags to set_all_edges_as_executable and only do it when
9479 reachability is REACHABLE_BLOCKS.
9480 * domwalk.h (enum dom_walker::reachability): New enum.
9481 (dom_walker::dom_walker): Convert bool param
9482 "skip_unreachable_blocks" to enum reachability.
9483 (set_all_edges_as_executable): New decl.
9484 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9485 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9487 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9488 but converting true to REACHABLE_BLOCKS.
9489 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9491 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9492 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9493 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9495 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9496 if check_all_array_refs will be called.
9498 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9500 * tree.c (selftest::test_location_wrappers): Add more test
9503 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9505 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9506 (selftest::test_bit_in_range): Likewise.
9508 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9511 * doc/sourcebuild.texi (vect_float): Say that the selector
9512 only describes the situation when -funsafe-math-optimizations is on.
9513 (vect_float_strict): Document.
9515 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9517 PR tree-optimization/83965
9518 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9519 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9520 instead of checking only for a reduction.
9521 (vect_recog_widen_sum_pattern): Likewise.
9523 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9525 * predict.c (probably_never_executed): Only use precise profile info.
9526 (compute_function_frequency): Skip after inlining hack since we now
9527 have quality checking.
9529 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9531 * profile-count.h (profile_probability::very_unlikely,
9532 profile_probability::unlikely, profile_probability::even): Set
9533 precision to guessed.
9535 2018-01-23 Richard Biener <rguenther@suse.de>
9537 PR tree-optimization/83963
9538 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9539 Properly terminate dominator walk when crossing the exit edge not
9540 when visiting its source block.
9542 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9545 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9546 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9548 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9550 PR tree-optimization/83957
9551 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9552 semicolon after for body surrounded by braces.
9554 PR tree-optimization/83081
9555 * profile-count.h (profile_probability::split): New method.
9556 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9557 Use profile_probability::split.
9558 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9559 when splitting a single conditional jump into 2.
9561 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9563 PR tree-optimization/69452
9564 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9567 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9569 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9570 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9571 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9573 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9575 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9576 * config/rl78/rl78.md (movdi): New define_expand.
9577 * config/rl78/rl78.c (rl78_split_movdi): New function.
9579 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9582 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9584 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9585 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9586 128-bit to produce an UNSPEC move to get the double word with the
9587 signbit and then a shift directly to do signbit.
9588 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9589 implementation with a new version that just does either a direct
9590 move or a regular move. Move memory interface to separate insns.
9591 Move insns so they are next to the expander.
9592 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9593 with signbit move. Split big and little endian case.
9594 (signbit<mode>2_dm_mem_le): Likewise.
9595 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9596 (signbit<mode>2_dm2): Likewise.
9598 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9600 * config/rl78/rl78.md (anddi3): New define_expand.
9602 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9604 * config/rl78/rl78.md (umindi3): New define_expand.
9606 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9608 * config/rl78/rl78.md (smindi3): New define_expand.
9610 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9612 * config/rl78/rl78.md (smaxdi3): New define_expand.
9614 2018-01-22 Carl Love <cel@us.ibm.com>
9616 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9617 LVX_V1TI): Add macro expansion.
9618 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9619 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9620 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9621 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9622 Change check to determine if the instruction is a byte reversing
9623 entry. Fix typo in comment.
9624 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9625 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9626 Add def_builtin calls for new builtins.
9627 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9628 Add define_insn expansion.
9630 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9632 * config/rl78/rl78.md (umaxdi3): New define_expand.
9634 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9636 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9637 for non-QImode registers.
9639 2018-01-22 Richard Biener <rguenther@suse.de>
9641 PR tree-optimization/83963
9642 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9643 including the loop exit block.
9644 (scop_detection::merge_sese): Likewise.
9645 (scop_detection::add_scop): Do it here instead.
9647 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9649 * doc/sourcebuild.texi (arm_softfloat): Document.
9651 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9654 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9655 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9656 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9658 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9659 David Edelsohn <dje.gcc@gmail.com>
9662 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9663 Change "crset eq" to "crset 2".
9664 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9665 (*call_indirect_aix<mode>_nospec): Likewise.
9666 (*call_value_indirect_aix<mode>_nospec): Likewise.
9667 (*call_indirect_elfv2<mode>_nospec): Likewise.
9668 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9669 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9670 change assembly output from . to $.
9671 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9672 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9673 (*tablejump<mode>_internal1_nospec): Likewise.
9675 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9678 * config/sh/sh_optimize_sett_clrt.cc:
9679 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9681 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9683 PR tree-optimization/83940
9684 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9685 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9686 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9687 use it to pass back the definition type.
9688 (vect_check_store_rhs): Likewise rhs_dt_out.
9689 (vect_build_gather_load_calls): Add a mask_dt argument and use
9690 it instead of a call to vect_is_simple_use.
9691 (vectorizable_store): Update calls to vect_check_load_store_mask
9692 and vect_check_store_rhs. Use the dt returned by the latter instead
9693 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9694 instead of calls to vect_is_simple_use. Pass the scalar rather
9695 than the vector operand to vect_is_simple_use when handling
9696 second and subsequent copies of an rhs value.
9697 (vectorizable_load): Update calls to vect_check_load_store_mask
9698 and vect_build_gather_load_calls. Use the cached mask_dt and
9699 gs_info.offset_dt instead of calls to vect_is_simple_use.
9701 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9704 * tree-emutls.c: Include gimplify.h.
9705 (lower_emutls_2): New function.
9706 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9707 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9708 it before further processing.
9711 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9712 UINTVAL (trueop1) instead of INTVAL (op1).
9714 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9718 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9719 INCOMING_FRAME_SP_OFFSET if not defined.
9720 (scan_trace): Add ENTRY argument. If true and
9721 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9722 emit a note to adjust the CFA offset.
9723 (create_cfi_notes): Adjust scan_trace callers.
9724 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9725 INCOMING_FRAME_SP_OFFSET in the CIE.
9726 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9727 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9729 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9730 * doc/tm.texi: Regenerated.
9732 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9734 PR rtl-optimization/83147
9735 * lra-constraints.c (remove_inheritance_pseudos): Use
9736 lra_substitute_pseudo_within_insn.
9738 2018-01-19 Tom de Vries <tom@codesourcery.com>
9739 Cesar Philippidis <cesar@codesourcery.com>
9742 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9744 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9747 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9748 spaces for function labels.
9750 2018-01-19 Martin Liska <mliska@suse.cz>
9752 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9753 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9754 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9755 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9756 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9757 (PRED_CONST_RETURN): Change from 69 to 65.
9758 (PRED_NULL_RETURN): Change from 91 to 71.
9759 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9760 (PRED_LOOP_GUARD): Change from 66 to 73.
9762 2018-01-19 Martin Liska <mliska@suse.cz>
9764 * predict.c (predict_insn_def): Add new assert.
9765 (struct branch_predictor): Change type to signed integer.
9766 (test_prediction_value_range): Amend test to cover
9768 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9769 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9770 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9771 (PRED_LOOP_IV_COMPARE): Likewise.
9772 * predict.h (PROB_UNINITIALIZED): Define new constant.
9774 2018-01-19 Martin Liska <mliska@suse.cz>
9776 * predict.c (dump_prediction): Add new format for
9777 analyze_brprob.py script which is enabled with -details
9779 * profile-count.h (precise_p): New function.
9781 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9783 PR tree-optimization/83922
9784 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9785 there are no statements that need masking.
9786 (vect_active_double_reduction_p): New function.
9787 (vect_analyze_loop_operations): Use it when handling phis that
9788 are not in the loop header.
9790 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9792 PR tree-optimization/83914
9793 * tree-vect-loop.c (vectorizable_induction): Don't convert
9794 init_expr or apply the peeling adjustment for inductions
9795 that are nested within the vectorized loop.
9797 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9799 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9802 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9806 * function.h (gimplify_parameters): Add gimple_seq * argument.
9807 * function.c: Include gimple.h and options.h.
9808 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9809 for the added local temporaries if needed.
9810 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9811 if there are any parameter cleanups, wrap whole body into a
9812 try/finally with the cleanups.
9814 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9817 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9818 Use GET_MODE_CLASS for scalar floating point.
9820 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9824 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9825 Fix call of call_cgraph_insertion_hooks.
9827 2018-01-18 Martin Sebor <msebor@redhat.com>
9829 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9831 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9834 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9837 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9840 * common.opt: (-ffile-prefix-map): New option.
9841 * opts.c (common_handle_option): Defer it.
9842 * opts-global.c (handle_common_deferred_options): Handle it.
9843 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9844 * file-prefix-map.h: New file.
9845 (remap_debug_filename, add_debug_prefix_map): ...here.
9846 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9847 * final.c (debug_prefix_map, add_debug_prefix_map
9848 remap_debug_filename): Move to...
9849 * file-prefix-map.c: New file.
9850 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9851 generalize, get rid of alloca(), use strrchr() instead of strchr().
9852 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9853 Implement in terms of add_prefix_map().
9854 (remap_macro_filename, remap_debug_filename): Implement in term of
9856 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9857 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9858 * dbxout.c: Include file-prefix-map.h.
9859 * varasm.c: Likewise.
9860 * vmsdbgout.c: Likewise.
9861 * xcoffout.c: Likewise.
9862 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9863 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9864 * doc/invoke.texi (-ffile-prefix-map): Document.
9865 (-fdebug-prefix-map): Update description.
9867 2018-01-18 Martin Liska <mliska@suse.cz>
9869 * config/i386/i386.c (indirect_thunk_name): Document that also
9871 (output_indirect_thunk): Document why both instructions
9872 (pause and lfence) are generated.
9874 2018-01-18 Richard Biener <rguenther@suse.de>
9876 PR tree-optimization/83887
9877 * graphite-scop-detection.c
9878 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9879 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9880 (scop_detection::merge_sese): Re-implement with a flood-fill
9881 algorithm that properly finds a SESE region if it exists.
9883 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9886 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9887 pointer_diff optimizations use view_convert instead of convert.
9889 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9891 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9892 Generate different code for -mno-speculate-indirect-jumps.
9893 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9894 (*call_indirect_aix<mode>): Disable for
9895 -mno-speculate-indirect-jumps.
9896 (*call_indirect_aix<mode>_nospec): New define_insn.
9897 (*call_value_indirect_aix<mode>): Disable for
9898 -mno-speculate-indirect-jumps.
9899 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9900 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9901 -mno-speculate-indirect-jumps.
9902 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9904 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9906 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9907 long double type, set the flags for noting the default long double
9908 type, even if we don't pass or return a long double type.
9910 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9913 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9916 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9918 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9919 support for merge[hl].
9920 (fold_mergehl_helper): New helper function.
9921 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9922 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9923 (altivec_vmrglw_direct): Add xxmrglw insn.
9925 2018-01-17 Andrew Waterman <andrew@sifive.com>
9927 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9928 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9930 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9933 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9934 call the lto_location_cache before reading the
9935 DECL_SOURCE_LOCATION of the types.
9937 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9938 Richard Sandiford <richard.sandiford@linaro.org>
9940 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9941 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9942 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9943 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9944 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9946 * config/aarch64/constraints.md (aarch64_movti_operand):
9948 * config/aarch64/predicates.md (Uti): Add new constraint.
9950 2018-01-17 Carl Love <cel@us.ibm.com>
9952 * config/rs6000/vsx.md (define_expand xl_len_r,
9953 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9954 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9956 (define_expand, define_insn): Move the shift left from the
9957 define_insn to the define_expand for lxvl and stxvl instructions.
9958 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9959 and XL_LEN_R definitions to PURE.
9961 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9963 * config/i386/i386.c (indirect_thunk_name): Declare regno
9964 as unsigned int. Compare regno with INVALID_REGNUM.
9965 (output_indirect_thunk): Ditto.
9966 (output_indirect_thunk_function): Ditto.
9967 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9968 in the call to output_indirect_thunk_function.
9970 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9973 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9974 rather than the size of inner_type to determine the stack slot size
9975 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9977 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9980 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9983 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9985 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9986 endian Linux systems to optionally enable multilibs for selecting
9987 the long double type if the user configured an explicit type.
9988 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9989 have no long double multilibs if not defined.
9990 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9991 warn if the user used -mabi={ieee,ibm}longdouble and we built
9992 multilibs for long double.
9993 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9994 appropriate multilib option.
9995 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9997 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9998 for building long double multilibs.
9999 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10001 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10003 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10006 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10008 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10011 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10014 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10017 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10019 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10020 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10022 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10024 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10025 different rtl trees depending on TARGET_64BIT.
10026 (rs6000_gen_lvx): Likewise.
10028 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10030 * config/visium/visium.md (nop): Tweak comment.
10031 (hazard_nop): Likewise.
10033 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10035 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10036 -mspeculate-indirect-jumps.
10037 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10038 for -mno-speculate-indirect-jumps.
10039 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10040 (*call_value_indirect_elfv2<mode>): Disable for
10041 -mno-speculate-indirect-jumps.
10042 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10043 (indirect_jump): Emit different RTL for
10044 -mno-speculate-indirect-jumps.
10045 (*indirect_jump<mode>): Disable for
10046 -mno-speculate-indirect-jumps.
10047 (*indirect_jump<mode>_nospec): New define_insn.
10048 (tablejump): Emit different RTL for
10049 -mno-speculate-indirect-jumps.
10050 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10051 (tablejumpsi_nospec): New define_expand.
10052 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10053 (tablejumpdi_nospec): New define_expand.
10054 (*tablejump<mode>_internal1): Disable for
10055 -mno-speculate-indirect-jumps.
10056 (*tablejump<mode>_internal1_nospec): New define_insn.
10057 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10060 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10062 * caller-save.c (insert_save): Drop unnecessary parameter. All
10065 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10066 Richard Biener <rguenth@suse.de>
10069 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10070 return early, inline manually is_gimple_sizepos. Make sure if we
10071 call gimplify_expr we don't end up with a gimple constant.
10072 * tree.c (variably_modified_type_p): Don't return true for
10073 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10074 * gimplify.h (is_gimple_sizepos): Remove.
10076 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10078 PR tree-optimization/83857
10079 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10080 vectorizable_live_operation for pure SLP statements.
10081 (vectorizable_live_operation): Handle PHIs.
10083 2018-01-16 Richard Biener <rguenther@suse.de>
10085 PR tree-optimization/83867
10086 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10087 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10089 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10092 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10093 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10094 If off is not INTEGER_CST, issue a may not be aligned warning
10095 rather than isn't aligned. Use isn%'t rather than isn't.
10096 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10098 <case MULT_EXPR>: Improve the case when bottom and one of the
10099 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10100 operand, in that case check if the other operand is multiple of
10101 bottom divided by the INTEGER_CST operand.
10103 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10106 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10107 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10108 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10109 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10110 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10111 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10112 (pa_function_arg_size): New function.
10114 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10116 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10117 in a separate statement.
10119 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10121 PR tree-optimization/83847
10122 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10123 group gathers and scatters.
10125 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10127 PR rtl-optimization/86620
10128 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10130 PR rtl-optimization/83213
10131 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10132 to last if both are JUMP_INSNs.
10134 PR tree-optimization/83843
10135 * gimple-ssa-store-merging.c
10136 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10137 store_immediate_info for bswap/nop orig_stores.
10139 2018-01-15 Andrew Waterman <andrew@sifive.com>
10141 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10143 <UDIV>: Increase cost if !TARGET_DIV.
10145 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10147 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10148 (define_attr "cr_logical_3op"): New.
10149 (cceq_ior_compare): Adjust.
10150 (cceq_ior_compare_complement): Adjust.
10151 (*cceq_rev_compare): Adjust.
10152 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10153 (is_cracked_insn): Adjust.
10154 (insn_must_be_first_in_group): Adjust.
10155 * config/rs6000/40x.md: Adjust.
10156 * config/rs6000/440.md: Adjust.
10157 * config/rs6000/476.md: Adjust.
10158 * config/rs6000/601.md: Adjust.
10159 * config/rs6000/603.md: Adjust.
10160 * config/rs6000/6xx.md: Adjust.
10161 * config/rs6000/7450.md: Adjust.
10162 * config/rs6000/7xx.md: Adjust.
10163 * config/rs6000/8540.md: Adjust.
10164 * config/rs6000/cell.md: Adjust.
10165 * config/rs6000/e300c2c3.md: Adjust.
10166 * config/rs6000/e500mc.md: Adjust.
10167 * config/rs6000/e500mc64.md: Adjust.
10168 * config/rs6000/e5500.md: Adjust.
10169 * config/rs6000/e6500.md: Adjust.
10170 * config/rs6000/mpc.md: Adjust.
10171 * config/rs6000/power4.md: Adjust.
10172 * config/rs6000/power5.md: Adjust.
10173 * config/rs6000/power6.md: Adjust.
10174 * config/rs6000/power7.md: Adjust.
10175 * config/rs6000/power8.md: Adjust.
10176 * config/rs6000/power9.md: Adjust.
10177 * config/rs6000/rs64.md: Adjust.
10178 * config/rs6000/titan.md: Adjust.
10180 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10182 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10183 ix86_indirect_branch_register logic.
10185 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10187 * config/i386/constraints.md (Bs): Update
10188 ix86_indirect_branch_register check. Don't check
10189 ix86_indirect_branch_register with GOT_memory_operand.
10191 * config/i386/predicates.md (GOT_memory_operand): Don't check
10192 ix86_indirect_branch_register here.
10193 (GOT32_symbol_operand): Likewise.
10195 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10197 * config/i386/predicates.md (constant_call_address_operand):
10198 Rewrite ix86_indirect_branch_register logic.
10199 (sibcall_insn_operand): Likewise.
10201 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10203 * config/i386/constraints.md (Bs): Replace
10204 ix86_indirect_branch_thunk_register with
10205 ix86_indirect_branch_register.
10207 * config/i386/i386.md (indirect_jump): Likewise.
10208 (tablejump): Likewise.
10209 (*sibcall_memory): Likewise.
10210 (*sibcall_value_memory): Likewise.
10211 Peepholes of indirect call and jump via memory: Likewise.
10212 * config/i386/i386.opt: Likewise.
10213 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10214 (GOT_memory_operand): Likewise.
10215 (call_insn_operand): Likewise.
10216 (sibcall_insn_operand): Likewise.
10217 (GOT32_symbol_operand): Likewise.
10219 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10221 PR middle-end/83837
10222 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10223 type rather than type addr's type points to.
10224 (expand_omp_atomic_mutex): Likewise.
10225 (expand_omp_atomic): Likewise.
10227 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10230 * config/i386/i386.c (output_indirect_thunk_function): Use
10231 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10232 for __x86_return_thunk.
10234 2018-01-15 Richard Biener <rguenther@suse.de>
10236 PR middle-end/83850
10237 * expmed.c (extract_bit_field_1): Fix typo.
10239 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10242 * config/arm/iterators.md (VF): New mode iterator.
10243 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10244 Remove integer-related logic from pattern.
10245 (neon_vabd<mode>_3): Likewise.
10247 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10249 PR middle-end/82694
10250 * common.opt (fstrict-overflow): No longer an alias.
10251 (fwrapv-pointer): New option.
10252 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10253 also for pointer types based on flag_wrapv_pointer.
10254 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10255 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10256 opts->x_flag_wrapv got set.
10257 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10258 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10259 POINTER_TYPE_OVERFLOW_UNDEFINED.
10260 * match.pd: Likewise in address comparison pattern.
10261 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10263 2018-01-15 Richard Biener <rguenther@suse.de>
10266 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10267 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10268 Reset type names to their identifier if their TYPE_DECL doesn't
10269 have linkage (and thus is used for ODR and devirt).
10270 (save_debug_info_for_decl): Remove.
10271 (save_debug_info_for_type): Likewise.
10272 (add_tree_to_fld_list): Adjust.
10273 * tree-pretty-print.c (dump_generic_node): Make dumping of
10274 type names more robust.
10276 2018-01-15 Richard Biener <rguenther@suse.de>
10278 * BASE-VER: Bump to 8.0.1.
10280 2018-01-14 Martin Sebor <msebor@redhat.com>
10283 * builtins.c (check_access): Avoid warning when the no-warning bit
10286 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10288 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10289 * ira-color (allocno_hard_regs_compare): Likewise.
10291 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10294 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10295 Use .pushsection/.popsection.
10297 2018-01-14 Martin Sebor <msebor@redhat.com>
10300 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10302 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10304 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10305 entry from extra_headers.
10306 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10307 extra_headers, make the list bitwise identical to the i?86-*-* one.
10309 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10311 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10312 -mcmodel=large with -mindirect-branch=thunk,
10313 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10314 -mfunction-return=thunk-extern.
10315 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10316 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10317 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10319 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10321 * config/i386/i386.c (print_reg): Print the name of the full
10322 integer register without '%'.
10323 (ix86_print_operand): Handle 'V'.
10324 * doc/extend.texi: Document 'V' modifier.
10326 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10328 * config/i386/constraints.md (Bs): Disallow memory operand for
10329 -mindirect-branch-register.
10331 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10332 (GOT_memory_operand): Likewise.
10333 (call_insn_operand): Likewise.
10334 (sibcall_insn_operand): Likewise.
10335 (GOT32_symbol_operand): Likewise.
10336 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10337 for -mindirect-branch-register.
10338 (tablejump): Likewise.
10339 (*sibcall_memory): Likewise.
10340 (*sibcall_value_memory): Likewise.
10341 Disallow peepholes of indirect call and jump via memory for
10342 -mindirect-branch-register.
10343 (*call_pop): Replace m with Bw.
10344 (*call_value_pop): Likewise.
10345 (*sibcall_pop_memory): Replace m with Bs.
10346 * config/i386/i386.opt (mindirect-branch-register): New option.
10347 * doc/invoke.texi: Document -mindirect-branch-register option.
10349 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10351 * config/i386/i386-protos.h (ix86_output_function_return): New.
10352 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10353 set function_return_type.
10354 (indirect_thunk_name): Add ret_p to indicate thunk for function
10356 (output_indirect_thunk_function): Pass false to
10357 indirect_thunk_name.
10358 (ix86_output_indirect_branch_via_reg): Likewise.
10359 (ix86_output_indirect_branch_via_push): Likewise.
10360 (output_indirect_thunk_function): Create alias for function
10361 return thunk if regno < 0.
10362 (ix86_output_function_return): New function.
10363 (ix86_handle_fndecl_attribute): Handle function_return.
10364 (ix86_attribute_table): Add function_return.
10365 * config/i386/i386.h (machine_function): Add
10366 function_return_type.
10367 * config/i386/i386.md (simple_return_internal): Use
10368 ix86_output_function_return.
10369 (simple_return_internal_long): Likewise.
10370 * config/i386/i386.opt (mfunction-return=): New option.
10371 (indirect_branch): Mention -mfunction-return=.
10372 * doc/extend.texi: Document function_return function attribute.
10373 * doc/invoke.texi: Document -mfunction-return= option.
10375 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10377 * config/i386/i386-opts.h (indirect_branch): New.
10378 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10379 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10380 with local indirect jump when converting indirect call and jump.
10381 (ix86_set_indirect_branch_type): New.
10382 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10383 (indirectlabelno): New.
10384 (indirect_thunk_needed): Likewise.
10385 (indirect_thunk_bnd_needed): Likewise.
10386 (indirect_thunks_used): Likewise.
10387 (indirect_thunks_bnd_used): Likewise.
10388 (INDIRECT_LABEL): Likewise.
10389 (indirect_thunk_name): Likewise.
10390 (output_indirect_thunk): Likewise.
10391 (output_indirect_thunk_function): Likewise.
10392 (ix86_output_indirect_branch_via_reg): Likewise.
10393 (ix86_output_indirect_branch_via_push): Likewise.
10394 (ix86_output_indirect_branch): Likewise.
10395 (ix86_output_indirect_jmp): Likewise.
10396 (ix86_code_end): Call output_indirect_thunk_function if needed.
10397 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10399 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10400 (ix86_attribute_table): Add indirect_branch.
10401 * config/i386/i386.h (machine_function): Add indirect_branch_type
10402 and has_local_indirect_jump.
10403 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10405 (tablejump): Likewise.
10406 (*indirect_jump): Use ix86_output_indirect_jmp.
10407 (*tablejump_1): Likewise.
10408 (simple_return_indirect_internal): Likewise.
10409 * config/i386/i386.opt (mindirect-branch=): New option.
10410 (indirect_branch): New.
10413 (thunk-inline): Likewise.
10414 (thunk-extern): Likewise.
10415 * doc/extend.texi: Document indirect_branch function attribute.
10416 * doc/invoke.texi: Document -mindirect-branch= option.
10418 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10421 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10423 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10425 * ipa-inline.c (want_inline_small_function_p): Return false if
10426 inlining has already failed with CIF_FINAL_ERROR.
10427 (update_caller_keys): Call want_inline_small_function_p before
10429 (update_callee_keys): Likewise.
10431 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10433 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10435 (rs6000_quadword_masked_address_p): Likewise.
10436 (quad_aligned_load_p): Likewise.
10437 (quad_aligned_store_p): Likewise.
10438 (const_load_sequence_p): Add comment to describe the outer-most loop.
10439 (mimic_memory_attributes_and_flags): New function.
10440 (rs6000_gen_stvx): Likewise.
10441 (replace_swapped_aligned_store): Likewise.
10442 (rs6000_gen_lvx): Likewise.
10443 (replace_swapped_aligned_load): Likewise.
10444 (replace_swapped_load_constant): Capitalize argument name in
10445 comment describing this function.
10446 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10447 and stores that access quad-word aligned addresses and replace
10448 with stvx or lvx instructions when appropriate.
10449 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10450 New function prototype.
10451 (rs6000_quadword_masked_address_p): Likewise.
10452 (rs6000_gen_lvx): Likewise.
10453 (rs6000_gen_stvx): Likewise.
10454 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10455 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10456 when memory address is aligned.
10457 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10458 this split to select lvx instruction when memory address is aligned.
10459 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10460 instruction when memory address is aligned.
10461 (*vsx_le_perm_load_v16qi): Likewise.
10462 (four unnamed splitters): Modify to select the stvx instruction
10463 when memory is aligned.
10465 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10467 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10468 which appears in the queue multiple times.
10470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10471 Alan Hayward <alan.hayward@arm.com>
10472 David Sherwood <david.sherwood@arm.com>
10474 * tree-vectorizer.h (vec_lower_bound): New structure.
10475 (_loop_vec_info): Add check_nonzero and lower_bounds.
10476 (LOOP_VINFO_CHECK_NONZERO): New macro.
10477 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10478 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10479 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10480 fields. Make seg_len the distance travelled, not including the
10482 (dr_direction_indicator): Declare.
10483 (dr_zero_step_indicator): Likewise.
10484 (dr_known_forward_stride_p): Likewise.
10485 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10487 (runtime_alias_check_p): Allow runtime alias checks with
10489 (operator ==): Compare access_size and align.
10490 (prune_runtime_alias_test_list): Rework for new distinction between
10491 the access_size and seg_len.
10492 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10494 (get_segment_min_max): New function.
10495 (create_intersect_range_checks): Use it.
10496 (dr_step_indicator): New function.
10497 (dr_direction_indicator): Likewise.
10498 (dr_zero_step_indicator): Likewise.
10499 (dr_known_forward_stride_p): Likewise.
10500 * tree-loop-distribution.c (data_ref_segment_size): Return
10501 DR_STEP * (niters - 1).
10502 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10504 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10505 (vect_preserves_scalar_order_p): New function, split out from...
10506 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10507 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10508 (vect_vfa_access_size): New function.
10509 (vect_vfa_align): Likewise.
10510 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10511 (dump_lower_bound): New function.
10512 (vect_check_lower_bound): Likewise.
10513 (vect_small_gap_p): Likewise.
10514 (vectorizable_with_step_bound_p): Likewise.
10515 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10516 depencies if the vectorization factor is 1. Convert the checks
10517 for nonzero steps into checks on the bounds of DR_STEP. Try using
10518 a bunds check for variable steps if the minimum required step is
10519 relatively small. Update calls to the dr_with_seg_len
10520 constructor and to vect_compile_time_alias.
10521 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10523 (vect_loop_versioning): Call it.
10524 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10526 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10529 Alan Hayward <alan.hayward@arm.com>
10530 David Sherwood <david.sherwood@arm.com>
10532 * doc/sourcebuild.texi (vect_scatter_store): Document.
10533 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10535 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10537 * genopinit.c (main): Add supports_vec_scatter_store and
10538 supports_vec_scatter_store_cached to target_optabs.
10539 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10540 IFN_MASK_SCATTER_STORE.
10541 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10543 * internal-fn.h (internal_store_fn_p): Declare.
10544 (internal_fn_stored_value_index): Likewise.
10545 * internal-fn.c (scatter_store_direct): New macro.
10546 (expand_scatter_store_optab_fn): New function.
10547 (direct_scatter_store_optab_supported_p): New macro.
10548 (internal_store_fn_p): New function.
10549 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10550 IFN_MASK_SCATTER_STORE.
10551 (internal_fn_mask_index): Likewise.
10552 (internal_fn_stored_value_index): New function.
10553 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10554 for scatter stores.
10555 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10556 * optabs-query.c (supports_vec_scatter_store_p): New function.
10557 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10558 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10559 true for scatter stores.
10560 (vect_gather_scatter_fn_p): Handle scatter stores too.
10561 (vect_check_gather_scatter): Consider using scatter stores if
10562 supports_vec_scatter_store_p.
10563 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10564 scatter stores too.
10565 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10566 internal_fn_stored_value_index.
10567 (check_load_store_masking): Handle scatter stores too.
10568 (vect_get_store_rhs): Make public.
10569 (vectorizable_call): Use internal_store_fn_p.
10570 (vectorizable_store): Handle scatter store internal functions.
10571 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10572 when deciding whether the end of the group has been reached.
10573 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10574 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10575 (mask_scatter_store<mode>): New insns.
10577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10578 Alan Hayward <alan.hayward@arm.com>
10579 David Sherwood <david.sherwood@arm.com>
10581 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10582 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10583 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10585 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10586 Use vect_truncate_gather_scatter_offset if we can't treat the
10587 operation as a normal gather load or scatter store.
10588 (get_group_load_store_type): Take the gather_scatter_info
10589 as argument. Try using a gather load or scatter store for
10590 single-element groups.
10591 (get_load_store_type): Update calls to get_group_load_store_type
10592 and vect_use_strided_gather_scatters_p.
10594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10595 Alan Hayward <alan.hayward@arm.com>
10596 David Sherwood <david.sherwood@arm.com>
10598 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10599 optional tree argument.
10600 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10602 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10603 but continue to use the current value as a fallback.
10604 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10605 to compare the updates.
10606 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10607 (get_load_store_type): Use it when handling a strided access.
10608 (vect_get_strided_load_store_ops): New function.
10609 (vect_get_data_ptr_increment): Likewise.
10610 (vectorizable_load): Handle strided gather loads. Always pass
10611 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10614 Alan Hayward <alan.hayward@arm.com>
10615 David Sherwood <david.sherwood@arm.com>
10617 * doc/md.texi (gather_load@var{m}): Document.
10618 (mask_gather_load@var{m}): Likewise.
10619 * genopinit.c (main): Add supports_vec_gather_load and
10620 supports_vec_gather_load_cached to target_optabs.
10621 * optabs-tree.c (init_tree_optimization_optabs): Use
10622 ggc_cleared_alloc to allocate target_optabs.
10623 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10624 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10626 * internal-fn.h (internal_load_fn_p): Declare.
10627 (internal_gather_scatter_fn_p): Likewise.
10628 (internal_fn_mask_index): Likewise.
10629 (internal_gather_scatter_fn_supported_p): Likewise.
10630 * internal-fn.c (gather_load_direct): New macro.
10631 (expand_gather_load_optab_fn): New function.
10632 (direct_gather_load_optab_supported_p): New macro.
10633 (direct_internal_fn_optab): New function.
10634 (internal_load_fn_p): Likewise.
10635 (internal_gather_scatter_fn_p): Likewise.
10636 (internal_fn_mask_index): Likewise.
10637 (internal_gather_scatter_fn_supported_p): Likewise.
10638 * optabs-query.c (supports_at_least_one_mode_p): New function.
10639 (supports_vec_gather_load_p): Likewise.
10640 * optabs-query.h (supports_vec_gather_load_p): Declare.
10641 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10642 and memory_type field.
10643 (NUM_PATTERNS): Bump to 15.
10644 * tree-vect-data-refs.c: Include internal-fn.h.
10645 (vect_gather_scatter_fn_p): New function.
10646 (vect_describe_gather_scatter_call): Likewise.
10647 (vect_check_gather_scatter): Try using internal functions for
10648 gather loads. Recognize existing calls to a gather load function.
10649 (vect_analyze_data_refs): Consider using gather loads if
10650 supports_vec_gather_load_p.
10651 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10652 (vect_get_gather_scatter_offset_type): Likewise.
10653 (vect_convert_mask_for_vectype): Likewise.
10654 (vect_add_conversion_to_patterm): Likewise.
10655 (vect_try_gather_scatter_pattern): Likewise.
10656 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10657 (vect_vect_recog_func_ptrs): Add it.
10658 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10659 internal_fn_mask_index and internal_gather_scatter_fn_p.
10660 (check_load_store_masking): Take the gather_scatter_info as an
10661 argument and handle gather loads.
10662 (vect_get_gather_scatter_ops): New function.
10663 (vectorizable_call): Check internal_load_fn_p.
10664 (vectorizable_load): Likewise. Handle gather load internal
10666 (vectorizable_store): Update call to check_load_store_masking.
10667 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10668 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10669 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10670 (aarch64_gather_scale_operand_d): New predicates.
10671 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10672 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
10679 * doc/md.texi (fold_left_plus_@var{m}): Document.
10680 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10681 * internal-fn.c (fold_left_direct): Define.
10682 (expand_fold_left_optab_fn): Likewise.
10683 (direct_fold_left_optab_supported_p): Likewise.
10684 * fold-const-call.c (fold_const_fold_left): New function.
10685 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10686 * tree-parloops.c (valid_reduction_p): New function.
10687 (gather_scalar_reductions): Use it.
10688 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10689 (vect_finish_replace_stmt): Declare.
10690 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10691 (needs_fold_left_reduction_p): New function, split out from...
10692 (vect_is_simple_reduction): ...here. Accept reductions that
10693 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10694 (vect_force_simple_reduction): Also store the reduction type in
10695 the assignment's STMT_VINFO_REDUC_TYPE.
10696 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10697 (merge_with_identity): New function.
10698 (vect_expand_fold_left): Likewise.
10699 (vectorize_fold_left_reduction): Likewise.
10700 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10701 scalar phi in place for it. Check for target support and reject
10702 cases that would reassociate the operation. Defer the transform
10703 phase to vectorize_fold_left_reduction.
10704 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10705 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10706 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10710 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10711 call to ifc_temp_var.
10713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10714 Alan Hayward <alan.hayward@arm.com>
10715 David Sherwood <david.sherwood@arm.com>
10717 * target.def (legitimize_address_displacement): Take the original
10718 offset as a poly_int.
10719 * targhooks.h (default_legitimize_address_displacement): Update
10721 * targhooks.c (default_legitimize_address_displacement): Likewise.
10722 * doc/tm.texi: Regenerate.
10723 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10724 as an argument, moving assert of ad->disp == ad->disp_term to...
10725 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10726 Try calling targetm.legitimize_address_displacement before expanding
10727 the address rather than afterwards, and adjust for the new interface.
10728 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10729 Match the new hook interface. Handle SVE addresses.
10730 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10731 new hook interface.
10733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10735 * Makefile.in (OBJS): Add early-remat.o.
10736 * target.def (select_early_remat_modes): New hook.
10737 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10738 * doc/tm.texi: Regenerate.
10739 * targhooks.h (default_select_early_remat_modes): Declare.
10740 * targhooks.c (default_select_early_remat_modes): New function.
10741 * timevar.def (TV_EARLY_REMAT): New timevar.
10742 * passes.def (pass_early_remat): New pass.
10743 * tree-pass.h (make_pass_early_remat): Declare.
10744 * early-remat.c: New file.
10745 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10747 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10750 Alan Hayward <alan.hayward@arm.com>
10751 David Sherwood <david.sherwood@arm.com>
10753 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10754 vfm1 with a bound_epilog parameter.
10755 (vect_do_peeling): Update calls accordingly, and move the prologue
10756 call earlier in the function. Treat the base bound_epilog as 0 for
10757 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10758 this base when peeling for gaps.
10759 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10760 with fully-masked loops.
10761 (vect_estimate_min_profitable_iters): Handle the single peeled
10762 iteration in that case.
10764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10765 Alan Hayward <alan.hayward@arm.com>
10766 David Sherwood <david.sherwood@arm.com>
10768 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10769 single-element interleaving even if the size is not a power of 2.
10770 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10771 accesses for single-element interleaving if the group size is
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
10779 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10780 * optabs.def (fold_extract_last_optab): New optab.
10781 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10782 * internal-fn.c (fold_extract_direct): New macro.
10783 (expand_fold_extract_optab_fn): Likewise.
10784 (direct_fold_extract_optab_supported_p): Likewise.
10785 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10786 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10787 EXTRACT_LAST_REDUCTION.
10788 (get_initial_def_for_reduction): Do not create an initial vector
10789 for EXTRACT_LAST_REDUCTION reductions.
10790 (vectorizable_reduction): Leave the scalar phi in place for
10791 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10792 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10793 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10794 transform phase to vectorizable_condition.
10795 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10797 (vect_finish_stmt_generation): ...here.
10798 (vect_finish_replace_stmt): New function.
10799 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10800 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10802 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10805 Alan Hayward <alan.hayward@arm.com>
10806 David Sherwood <david.sherwood@arm.com>
10808 * doc/md.texi (extract_last_@var{m}): Document.
10809 * optabs.def (extract_last_optab): New optab.
10810 * internal-fn.def (EXTRACT_LAST): New internal function.
10811 * internal-fn.c (cond_unary_direct): New macro.
10812 (expand_cond_unary_optab_fn): Likewise.
10813 (direct_cond_unary_optab_supported_p): Likewise.
10814 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10815 loops using EXTRACT_LAST.
10816 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10817 (extract_last_<mode>): ...this optab.
10818 (vec_extract<mode><Vel>): Update accordingly.
10820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10821 Alan Hayward <alan.hayward@arm.com>
10822 David Sherwood <david.sherwood@arm.com>
10824 * target.def (empty_mask_is_expensive): New hook.
10825 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10826 * doc/tm.texi: Regenerate.
10827 * targhooks.h (default_empty_mask_is_expensive): Declare.
10828 * targhooks.c (default_empty_mask_is_expensive): New function.
10829 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10830 if the target says that empty masks are expensive.
10831 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10833 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10840 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10841 (vect_use_loop_mask_for_alignment_p): New function.
10842 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10843 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10844 niters_skip argument. Make sure that the first niters_skip elements
10845 of the first iteration are inactive.
10846 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10847 Update call to vect_set_loop_masks_directly.
10848 (get_misalign_in_elems): New function, split out from...
10849 (vect_gen_prolog_loop_niters): ...here.
10850 (vect_update_init_of_dr): Take a code argument that specifies whether
10851 the adjustment should be added or subtracted.
10852 (vect_update_init_of_drs): Likewise.
10853 (vect_prepare_for_masked_peels): New function.
10854 (vect_do_peeling): Skip prologue peeling if we're using a mask
10855 instead. Update call to vect_update_inits_of_drs.
10856 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10858 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10859 alignment. Do not include the number of peeled iterations in
10860 the minimum threshold in that case.
10861 (vectorizable_induction): Adjust the start value down by
10862 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10863 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10864 Take the number of skipped iterations into account when calculating
10866 * tree-vect-stmts.c (vect_gen_while_not): New function.
10868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10869 Alan Hayward <alan.hayward@arm.com>
10870 David Sherwood <david.sherwood@arm.com>
10872 * doc/sourcebuild.texi (vect_fully_masked): Document.
10873 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10874 default value to 0.
10875 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10877 (vect_analyze_loop_2): ...here. Don't check the vectorization
10878 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10886 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10887 (dump_groups): Update accordingly.
10888 (iv_use::mem_type): New member variable.
10889 (address_p): New function.
10890 (record_use): Add a mem_type argument and initialize the new
10892 (record_group_use): Add a mem_type argument. Use address_p.
10893 Remove obsolete null checks of base_object. Update call to record_use.
10894 (find_interesting_uses_op): Update call to record_group_use.
10895 (find_interesting_uses_cond): Likewise.
10896 (find_interesting_uses_address): Likewise.
10897 (get_mem_type_for_internal_fn): New function.
10898 (find_address_like_use): Likewise.
10899 (find_interesting_uses_stmt): Try find_address_like_use before
10900 calling find_interesting_uses_op.
10901 (addr_offset_valid_p): Use the iv mem_type field as the type
10902 of the addressed memory.
10903 (add_autoinc_candidates): Likewise.
10904 (get_address_cost): Likewise.
10905 (split_small_address_groups_p): Use address_p.
10906 (split_address_groups): Likewise.
10907 (add_iv_candidate_for_use): Likewise.
10908 (autoinc_possible_for_pair): Likewise.
10909 (rewrite_groups): Likewise.
10910 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10911 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10912 (get_alias_ptr_type_for_ptr_address): New function.
10913 (rewrite_use_address): Rewrite address uses in calls that were
10914 identified by find_address_like_use.
10916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10917 Alan Hayward <alan.hayward@arm.com>
10918 David Sherwood <david.sherwood@arm.com>
10920 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10922 * gimple-expr.h (is_gimple_addressable: Likewise.
10923 * gimple-expr.c (is_gimple_address): Likewise.
10924 * internal-fn.c (expand_call_mem_ref): New function.
10925 (expand_mask_load_optab_fn): Use it.
10926 (expand_mask_store_optab_fn): Likewise.
10928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10929 Alan Hayward <alan.hayward@arm.com>
10930 David Sherwood <david.sherwood@arm.com>
10932 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10933 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10934 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10935 (cond_umax@var{mode}): Document.
10936 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10937 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10938 (cond_umin_optab, cond_umax_optab): New optabs.
10939 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10940 (COND_IOR, COND_XOR): New internal functions.
10941 * internal-fn.h (get_conditional_internal_fn): Declare.
10942 * internal-fn.c (cond_binary_direct): New macro.
10943 (expand_cond_binary_optab_fn): Likewise.
10944 (direct_cond_binary_optab_supported_p): Likewise.
10945 (get_conditional_internal_fn): New function.
10946 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10947 Cope with reduction statements that are vectorized as calls rather
10949 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10950 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10951 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10952 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10953 (UNSPEC_COND_EOR): New unspecs.
10954 (optab): Add mappings for them.
10955 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10956 (sve_int_op, sve_fp_op): New int attributes.
10958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10959 Alan Hayward <alan.hayward@arm.com>
10960 David Sherwood <david.sherwood@arm.com>
10962 * optabs.def (while_ult_optab): New optab.
10963 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10964 * internal-fn.def (WHILE_ULT): New internal function.
10965 * internal-fn.h (direct_internal_fn_supported_p): New override
10966 that takes two types as argument.
10967 * internal-fn.c (while_direct): New macro.
10968 (expand_while_optab_fn): New function.
10969 (convert_optab_supported_p): Likewise.
10970 (direct_while_optab_supported_p): New macro.
10971 * wide-int.h (wi::udiv_ceil): New function.
10972 * tree-vectorizer.h (rgroup_masks): New structure.
10973 (vec_loop_masks): New typedef.
10974 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10975 and fully_masked_p.
10976 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10977 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10978 (vect_max_vf): New function.
10979 (slpeel_make_loop_iterate_ntimes): Delete.
10980 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10981 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10982 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10983 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10984 internal-fn.h, stor-layout.h and optabs-query.h.
10985 (vect_set_loop_mask): New function.
10986 (add_preheader_seq): Likewise.
10987 (add_header_seq): Likewise.
10988 (interleave_supported_p): Likewise.
10989 (vect_maybe_permute_loop_masks): Likewise.
10990 (vect_set_loop_masks_directly): Likewise.
10991 (vect_set_loop_condition_masked): Likewise.
10992 (vect_set_loop_condition_unmasked): New function, split out from
10993 slpeel_make_loop_iterate_ntimes.
10994 (slpeel_make_loop_iterate_ntimes): Rename to..
10995 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10996 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10997 (vect_do_peeling): Update call accordingly.
10998 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11000 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11001 mask_compare_type, can_fully_mask_p and fully_masked_p.
11002 (release_vec_loop_masks): New function.
11003 (_loop_vec_info): Use it to free the loop masks.
11004 (can_produce_all_loop_masks_p): New function.
11005 (vect_get_max_nscalars_per_iter): Likewise.
11006 (vect_verify_full_masking): Likewise.
11007 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11008 retries, and free the mask rgroups before retrying. Check loop-wide
11009 reasons for disallowing fully-masked loops. Make the final decision
11010 about whether use a fully-masked loop or not.
11011 (vect_estimate_min_profitable_iters): Do not assume that peeling
11012 for the number of iterations will be needed for fully-masked loops.
11013 (vectorizable_reduction): Disable fully-masked loops.
11014 (vectorizable_live_operation): Likewise.
11015 (vect_halve_mask_nunits): New function.
11016 (vect_double_mask_nunits): Likewise.
11017 (vect_record_loop_mask): Likewise.
11018 (vect_get_loop_mask): Likewise.
11019 (vect_transform_loop): Handle the case in which the final loop
11020 iteration might handle a partial vector. Call vect_set_loop_condition
11021 instead of slpeel_make_loop_iterate_ntimes.
11022 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11023 (check_load_store_masking): New function.
11024 (prepare_load_store_mask): Likewise.
11025 (vectorizable_store): Handle fully-masked loops.
11026 (vectorizable_load): Likewise.
11027 (supportable_widening_operation): Use vect_halve_mask_nunits for
11029 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11030 (vect_gen_while): New function.
11031 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11032 (aarch64_uqdec<mode>): New insn.
11034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11035 Alan Hayward <alan.hayward@arm.com>
11036 David Sherwood <david.sherwood@arm.com>
11038 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11039 (reduc_xor_scal_optab): New optabs.
11040 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11041 (reduc_xor_scal_@var{m}): Document.
11042 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11043 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11044 internal functions.
11045 * fold-const-call.c (fold_const_call): Handle them.
11046 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11047 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11048 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11049 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11050 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11051 (UNSPEC_XORV): New unspecs.
11052 (optab): Add entries for them.
11053 (BITWISEV): New int iterator.
11054 (bit_reduc_op): New int attributes.
11056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11057 Alan Hayward <alan.hayward@arm.com>
11058 David Sherwood <david.sherwood@arm.com>
11060 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11061 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11062 * optabs.def (vec_shl_insert_optab): New optab.
11063 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11064 (duplicate_and_interleave): Likewise.
11065 * tree-vect-loop.c: Include internal-fn.h.
11066 (neutral_op_for_slp_reduction): New function, split out from
11067 get_initial_defs_for_reduction.
11068 (get_initial_def_for_reduction): Handle option 2 for variable-length
11069 vectors by loading the neutral value into a vector and then shifting
11070 the initial value into element 0.
11071 (get_initial_defs_for_reduction): Replace the code argument with
11072 the neutral value calculated by neutral_op_for_slp_reduction.
11073 Use gimple_build_vector for constant-length vectors.
11074 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11075 but the first group_size elements have a neutral value.
11076 Use duplicate_and_interleave otherwise.
11077 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11078 Update call to get_initial_defs_for_reduction. Handle SLP
11079 reductions for variable-length vectors by creating one vector
11080 result for each scalar result, with the elements associated
11081 with other scalar results stubbed out with the neutral value.
11082 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11083 Require IFN_VEC_SHL_INSERT for double reductions on
11084 variable-length vectors, or SLP reductions that have
11085 a neutral value. Require can_duplicate_and_interleave_p
11086 support for variable-length unchained SLP reductions if there
11087 is no neutral value, such as for MIN/MAX reductions. Also require
11088 the number of vector elements to be a multiple of the number of
11089 SLP statements when doing variable-length unchained SLP reductions.
11090 Update call to vect_create_epilog_for_reduction.
11091 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11092 and remove initial values.
11093 (duplicate_and_interleave): Make public.
11094 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11095 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11098 Alan Hayward <alan.hayward@arm.com>
11099 David Sherwood <david.sherwood@arm.com>
11101 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11102 (can_duplicate_and_interleave_p): New function.
11103 (vect_get_and_check_slp_defs): Take the vector of statements
11104 rather than just the current one. Remove excess parentheses.
11105 Restriction rejectinon of vect_constant_def and vect_external_def
11106 for variable-length vectors to boolean types, or types for which
11107 can_duplicate_and_interleave_p is false.
11108 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11109 (duplicate_and_interleave): New function.
11110 (vect_get_constant_vectors): Use gimple_build_vector for
11111 constant-length vectors and suitable variable-length constant
11112 vectors. Use duplicate_and_interleave for other variable-length
11113 vectors. Don't defer the update when inserting new statements.
11115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11116 Alan Hayward <alan.hayward@arm.com>
11117 David Sherwood <david.sherwood@arm.com>
11119 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11120 min_profitable_iters doesn't go negative.
11122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11123 Alan Hayward <alan.hayward@arm.com>
11124 David Sherwood <david.sherwood@arm.com>
11126 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11127 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11128 * optabs.def (vec_mask_load_lanes_optab): New optab.
11129 (vec_mask_store_lanes_optab): Likewise.
11130 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11131 (MASK_STORE_LANES): Likewise.
11132 * internal-fn.c (mask_load_lanes_direct): New macro.
11133 (mask_store_lanes_direct): Likewise.
11134 (expand_mask_load_optab_fn): Handle masked operations.
11135 (expand_mask_load_lanes_optab_fn): New macro.
11136 (expand_mask_store_optab_fn): Handle masked operations.
11137 (expand_mask_store_lanes_optab_fn): New macro.
11138 (direct_mask_load_lanes_optab_supported_p): Likewise.
11139 (direct_mask_store_lanes_optab_supported_p): Likewise.
11140 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11142 (vect_load_lanes_supported): Likewise.
11143 * tree-vect-data-refs.c (strip_conversion): New function.
11144 (can_group_stmts_p): Likewise.
11145 (vect_analyze_data_ref_accesses): Use it instead of checking
11146 for a pair of assignments.
11147 (vect_store_lanes_supported): Take a masked_p parameter.
11148 (vect_load_lanes_supported): Likewise.
11149 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11150 vect_store_lanes_supported and vect_load_lanes_supported.
11151 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11152 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11153 parameter. Don't allow gaps for masked accesses.
11154 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11155 and vect_load_lanes_supported.
11156 (get_load_store_type): Take a masked_p parameter and update
11157 call to get_group_load_store_type.
11158 (vectorizable_store): Update call to get_load_store_type.
11159 Handle IFN_MASK_STORE_LANES.
11160 (vectorizable_load): Update call to get_load_store_type.
11161 Handle IFN_MASK_LOAD_LANES.
11163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11164 Alan Hayward <alan.hayward@arm.com>
11165 David Sherwood <david.sherwood@arm.com>
11167 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11169 * config/aarch64/aarch64-protos.h
11170 (aarch64_sve_struct_memory_operand_p): Declare.
11171 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11172 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11173 (VPRED, vpred): Handle SVE structure modes.
11174 * config/aarch64/constraints.md (Utx): New constraint.
11175 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11176 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11177 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11178 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11179 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11180 structure modes. Split into pieces after RA.
11181 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11182 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11184 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11185 SVE structure modes.
11186 (aarch64_classify_address): Likewise.
11187 (sizetochar): Move earlier in file.
11188 (aarch64_print_operand): Handle SVE register lists.
11189 (aarch64_array_mode): New function.
11190 (aarch64_sve_struct_memory_operand_p): Likewise.
11191 (TARGET_ARRAY_MODE): Redefine.
11193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11194 Alan Hayward <alan.hayward@arm.com>
11195 David Sherwood <david.sherwood@arm.com>
11197 * target.def (array_mode): New target hook.
11198 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11199 * doc/tm.texi: Regenerate.
11200 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11201 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11202 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11203 targetm.array_mode.
11204 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11208 Alan Hayward <alan.hayward@arm.com>
11209 David Sherwood <david.sherwood@arm.com>
11211 * fold-const.c (fold_binary_loc): Check the argument types
11212 rather than the result type when testing for a vector operation.
11214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11216 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11217 * doc/tm.texi: Regenerate.
11219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11220 Alan Hayward <alan.hayward@arm.com>
11221 David Sherwood <david.sherwood@arm.com>
11223 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11224 (sve): Document new AArch64 extension.
11225 * doc/md.texi (w): Extend the description of the AArch64
11226 constraint to include SVE vectors.
11227 (Upl, Upa): Document new AArch64 predicate constraints.
11228 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11230 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11231 (msve-vector-bits=): New option.
11232 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11233 SVE when these are disabled.
11234 (sve): New extension.
11235 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11236 modes. Adjust their number of units based on aarch64_sve_vg.
11237 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11238 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11239 aarch64_addr_query_type.
11240 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11241 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11242 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11243 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11244 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11245 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11246 (aarch64_simd_imm_zero_p): Delete.
11247 (aarch64_check_zero_based_sve_index_immediate): Declare.
11248 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11249 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11250 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11251 (aarch64_sve_float_mul_immediate_p): Likewise.
11252 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11253 rather than an rtx.
11254 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11255 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11256 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11257 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11258 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11259 (aarch64_regmode_natural_size): Likewise.
11260 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11261 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11263 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11264 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11265 for VG and the SVE predicate registers.
11266 (V_ALIASES): Add a "z"-prefixed alias.
11267 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11268 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11269 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11270 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11271 (REG_CLASS_NAMES): Add entries for them.
11272 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11273 and the predicate registers.
11274 (aarch64_sve_vg): Declare.
11275 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11276 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11277 (REGMODE_NATURAL_SIZE): Define.
11278 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11280 * config/aarch64/aarch64.c: Include cfgrtl.h.
11281 (simd_immediate_info): Add a constructor for series vectors,
11282 and an associated step field.
11283 (aarch64_sve_vg): New variable.
11284 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11285 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11286 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11287 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11288 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11289 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11290 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11291 (aarch64_get_mask_mode): New functions.
11292 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11293 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11294 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11295 predicate modes and predicate registers. Explicitly restrict
11296 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11297 to store a vector mode if it is recognized by
11298 aarch64_classify_vector_mode.
11299 (aarch64_regmode_natural_size): New function.
11300 (aarch64_hard_regno_caller_save_mode): Return the original mode
11302 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11303 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11304 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11305 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11307 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11308 does not overlap dest if the function is frame-related. Handle
11310 (aarch64_split_add_offset): New function.
11311 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11312 them aarch64_add_offset.
11313 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11314 and update call to aarch64_sub_sp.
11315 (aarch64_add_cfa_expression): New function.
11316 (aarch64_expand_prologue): Pass extra temporary registers to the
11317 functions above. Handle the case in which we need to emit new
11318 DW_CFA_expressions for registers that were originally saved
11319 relative to the stack pointer, but now have to be expressed
11320 relative to the frame pointer.
11321 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11323 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11324 IP0 and IP1 values for SVE frames.
11325 (aarch64_expand_vec_series): New function.
11326 (aarch64_expand_sve_widened_duplicate): Likewise.
11327 (aarch64_expand_sve_const_vector): Likewise.
11328 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11329 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11330 into the register, rather than emitting a SET directly.
11331 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11332 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11333 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11334 (offset_9bit_signed_scaled_p): New functions.
11335 (aarch64_replicate_bitmask_imm): New function.
11336 (aarch64_bitmask_imm): Use it.
11337 (aarch64_cannot_force_const_mem): Reject expressions involving
11338 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11339 (aarch64_classify_index): Handle SVE indices, by requiring
11340 a plain register index with a scale that matches the element size.
11341 (aarch64_classify_address): Handle SVE addresses. Assert that
11342 the mode of the address is VOIDmode or an integer mode.
11343 Update call to aarch64_classify_symbol.
11344 (aarch64_classify_symbolic_expression): Update call to
11345 aarch64_classify_symbol.
11346 (aarch64_const_vec_all_in_range_p): New function.
11347 (aarch64_print_vector_float_operand): Likewise.
11348 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11349 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11350 and the FP immediates 1.0 and 0.5.
11351 (aarch64_print_address_internal): Handle SVE addresses.
11352 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11353 (aarch64_regno_regclass): Handle predicate registers.
11354 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11356 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11357 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11358 (aarch64_convert_sve_vector_bits): New function.
11359 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11360 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11361 rather than an rtx.
11362 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11363 Handle SVE vector and predicate modes. Accept VL-based constants
11364 that need only one temporary register, and VL offsets that require
11365 no temporary registers.
11366 (aarch64_conditional_register_usage): Mark the predicate registers
11367 as fixed if SVE isn't available.
11368 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11369 Return true for SVE vector and predicate modes.
11370 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11371 rather than an unsigned int. Handle SVE modes.
11372 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11374 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11376 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11377 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11378 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11379 (aarch64_sve_float_mul_immediate_p): New functions.
11380 (aarch64_sve_valid_immediate): New function.
11381 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11382 Explicitly reject structure modes. Check for INDEX constants.
11383 Handle PTRUE and PFALSE constants.
11384 (aarch64_check_zero_based_sve_index_immediate): New function.
11385 (aarch64_simd_imm_zero_p): Delete.
11386 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11387 vector modes. Accept constants in the range of CNT[BHWD].
11388 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11389 ask for an Advanced SIMD mode.
11390 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11391 (aarch64_simd_vector_alignment): Handle SVE predicates.
11392 (aarch64_vectorize_preferred_vector_alignment): New function.
11393 (aarch64_simd_vector_alignment_reachable): Use it instead of
11395 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11396 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11398 (MAX_VECT_LEN): Delete.
11399 (expand_vec_perm_d): Add a vec_flags field.
11400 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11401 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11402 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11404 (aarch64_evpc_rev): Rename to...
11405 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11406 (aarch64_evpc_rev_global): New function.
11407 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11408 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11410 (aarch64_evpc_sve_tbl): New function.
11411 (aarch64_expand_vec_perm_const_1): Update after rename of
11412 aarch64_evpc_rev. Handle SVE permutes too, trying
11413 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11414 than aarch64_evpc_tbl.
11415 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11416 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11417 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11418 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11419 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11420 (aarch64_expand_sve_vcond): New functions.
11421 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11422 of aarch64_vector_mode_p.
11423 (aarch64_dwarf_poly_indeterminate_value): New function.
11424 (aarch64_compute_pressure_classes): Likewise.
11425 (aarch64_can_change_mode_class): Likewise.
11426 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11427 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11428 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11429 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11430 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11431 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11432 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11433 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11435 (Dn, Dl, Dr): Accept const as well as const_vector.
11436 (Dz): Likewise. Compare against CONST0_RTX.
11437 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11438 of "vector" where appropriate.
11439 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11440 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11441 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11442 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11443 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11444 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11445 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11446 (v_int_equiv): Extend to SVE modes.
11447 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11449 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11450 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11451 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11452 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11453 (SVE_COND_FP_CMP): New int iterators.
11454 (perm_hilo): Handle the new unpack unspecs.
11455 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11457 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11458 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11459 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11460 (aarch64_equality_operator, aarch64_constant_vector_operand)
11461 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11462 (aarch64_sve_nonimmediate_operand): Likewise.
11463 (aarch64_sve_general_operand): Likewise.
11464 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11465 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11466 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11467 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11468 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11469 (aarch64_sve_float_arith_immediate): Likewise.
11470 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11471 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11472 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11473 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11474 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11475 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11476 (aarch64_sve_float_arith_operand): Likewise.
11477 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11478 (aarch64_sve_float_mul_operand): Likewise.
11479 (aarch64_sve_vec_perm_operand): Likewise.
11480 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11481 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11482 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11483 as well as const_vector.
11484 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11485 in file. Use CONST0_RTX and CONSTM1_RTX.
11486 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11487 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11488 Use aarch64_simd_imm_zero.
11489 * config/aarch64/aarch64-sve.md: New file.
11490 * config/aarch64/aarch64.md: Include it.
11491 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11492 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11493 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11494 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11495 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11496 (sve): New attribute.
11497 (enabled): Disable instructions with the sve attribute unless
11499 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11500 aarch64_expand_mov_immediate.
11501 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11502 CNT[BHSD] immediates.
11503 (movti): Split CONST_POLY_INT moves into two halves.
11504 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11505 Split additions that need a temporary here if the destination
11506 is the stack pointer.
11507 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11508 (*add<mode>3_poly_1): New instruction.
11509 (set_clobber_cc): New expander.
11511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11513 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11514 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11515 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11516 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11517 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11518 Change innermode from fixed_mode_size to machine_mode.
11519 (simplify_subreg): Update call accordingly. Handle a constant-sized
11520 subreg of a variable-length CONST_VECTOR.
11522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11523 Alan Hayward <alan.hayward@arm.com>
11524 David Sherwood <david.sherwood@arm.com>
11526 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11527 (add_offset_to_base): New function, split out from...
11528 (create_mem_ref): ...here. When handling a scale other than 1,
11529 check first whether the address is valid without the offset.
11530 Add it into the base if so, leaving the index and scale as-is.
11532 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11535 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11536 fold_for_warn before checking if arg2 is INTEGER_CST.
11538 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11540 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11541 (store_multiple_operation): Delete.
11542 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11543 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11544 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11545 guarded by TARGET_STRING.
11546 (rs6000_output_load_multiple): Delete.
11547 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11548 OPTION_MASK_STRING / TARGET_STRING handling.
11549 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11550 (const rs6000_opt_masks) <"string">: Change mask to 0.
11551 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11552 (MASK_STRING): Delete.
11553 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11555 (load_multiple): Delete.
11562 (store_multiple): Delete.
11569 (movmemsi_8reg): Delete.
11570 (corresponding unnamed define_insn): Delete.
11571 (movmemsi_6reg): Delete.
11572 (corresponding unnamed define_insn): Delete.
11573 (movmemsi_4reg): Delete.
11574 (corresponding unnamed define_insn): Delete.
11575 (movmemsi_2reg): Delete.
11576 (corresponding unnamed define_insn): Delete.
11577 (movmemsi_1reg): Delete.
11578 (corresponding unnamed define_insn): Delete.
11579 * config/rs6000/rs6000.opt (mno-string): New.
11580 (mstring): Replace by deprecation warning stub.
11581 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11583 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11585 * regrename.c (regrename_do_replace): If replacing the same
11586 reg multiple times, try to reuse last created gen_raw_REG.
11589 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11590 main to workaround a bug in GDB.
11592 2018-01-12 Tom de Vries <tom@codesourcery.com>
11595 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11597 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11599 PR rtl-optimization/80481
11600 * ira-color.c (get_cap_member): New function.
11601 (allocnos_conflict_by_live_ranges_p): Use it.
11602 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11603 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11605 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11608 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11609 (*saddl_se_1): Ditto.
11610 (*ssubsi_1): Ditto.
11611 (*ssubl_se_1): Ditto.
11613 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11615 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11616 rather than wi::to_widest for DR_INITs.
11617 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11618 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11619 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11621 (vect_analyze_group_access_1): Note that here.
11623 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11625 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11626 polynomial type sizes.
11628 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11630 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11631 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11632 (gimple_add_tmp_var): Likewise.
11634 2018-01-12 Martin Liska <mliska@suse.cz>
11636 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11637 (gimple_alloc_sizes): Likewise.
11638 (dump_gimple_statistics): Use PRIu64 in printf format.
11639 * gimple.h: Change uint64_t to int.
11641 2018-01-12 Martin Liska <mliska@suse.cz>
11643 * tree-core.h: Use uint64_t instead of int.
11644 * tree.c (tree_node_counts): Likewise.
11645 (tree_node_sizes): Likewise.
11646 (dump_tree_statistics): Use PRIu64 in printf format.
11648 2018-01-12 Martin Liska <mliska@suse.cz>
11650 * Makefile.in: As qsort_chk is implemented in vec.c, add
11651 vec.o to linkage of gencfn-macros.
11652 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11653 passing the info to record_node_allocation_statistics.
11654 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11656 * ggc-common.c (struct ggc_usage): Add operator== and use
11657 it in operator< and compare function.
11658 * mem-stats.h (struct mem_usage): Likewise.
11659 * vec.c (struct vec_usage): Remove operator< and compare
11660 function. Can be simply inherited.
11662 2018-01-12 Martin Jambor <mjambor@suse.cz>
11665 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11666 * tree-ssa-math-opts.c: Include domwalk.h.
11667 (convert_mult_to_fma_1): New function.
11668 (fma_transformation_info): New type.
11669 (fma_deferring_state): Likewise.
11670 (cancel_fma_deferring): New function.
11671 (result_of_phi): Likewise.
11672 (last_fma_candidate_feeds_initial_phi): Likewise.
11673 (convert_mult_to_fma): Added deferring logic, split actual
11674 transformation to convert_mult_to_fma_1.
11675 (math_opts_dom_walker): New type.
11676 (math_opts_dom_walker::after_dom_children): New method, body moved
11677 here from pass_optimize_widening_mul::execute, added deferring logic
11679 (pass_optimize_widening_mul::execute): Moved most of code to
11680 math_opts_dom_walker::after_dom_children.
11681 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11682 * config/i386/i386.c (ix86_option_override_internal): Added
11683 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11685 2018-01-12 Richard Biener <rguenther@suse.de>
11688 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11689 inline instance vars.
11691 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11694 * config/rx/rx.c (rx_is_restricted_memory_address):
11695 Handle SUBREG case.
11697 2018-01-12 Richard Biener <rguenther@suse.de>
11699 PR tree-optimization/80846
11700 * target.def (split_reduction): New target hook.
11701 * targhooks.c (default_split_reduction): New function.
11702 * targhooks.h (default_split_reduction): Declare.
11703 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11704 target requests first reduce vectors by combining low and high
11706 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11707 (get_vectype_for_scalar_type_and_size): Export.
11708 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11709 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11710 * doc/tm.texi: Regenerate.
11711 * config/i386/i386.c (ix86_split_reduction): Implement
11712 TARGET_VECTORIZE_SPLIT_REDUCTION.
11714 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11717 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11718 in PIC mode except for TARGET_VXWORKS_RTP.
11719 * config/sparc/sparc.c: Include cfgrtl.h.
11720 (TARGET_INIT_PIC_REG): Define.
11721 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11722 (sparc_pic_register_p): New predicate.
11723 (sparc_legitimate_address_p): Use it.
11724 (sparc_legitimize_pic_address): Likewise.
11725 (sparc_delegitimize_address): Likewise.
11726 (sparc_mode_dependent_address_p): Likewise.
11727 (gen_load_pcrel_sym): Remove 4th parameter.
11728 (load_got_register): Adjust call to above. Remove obsolete stuff.
11729 (sparc_expand_prologue): Do not call load_got_register here.
11730 (sparc_flat_expand_prologue): Likewise.
11731 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11732 (sparc_use_pseudo_pic_reg): New function.
11733 (sparc_init_pic_reg): Likewise.
11734 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11735 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11737 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11739 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11740 Add item for branch_cost.
11742 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11744 PR rtl-optimization/83565
11745 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11746 not extend the result to a larger mode for rotate operations.
11747 (num_sign_bit_copies1): Likewise.
11749 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11752 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11754 Use values-Xc.o for -pedantic.
11755 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11757 2018-01-12 Martin Liska <mliska@suse.cz>
11760 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11762 (possible_polymorphic_call_targets): Use it.
11763 (ipa_devirt): Likewise.
11765 2018-01-12 Martin Liska <mliska@suse.cz>
11767 * profile-count.h (enum profile_quality): Use 0 as invalid
11768 enum value of profile_quality.
11770 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11772 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11773 -mext-string options.
11775 2018-01-12 Richard Biener <rguenther@suse.de>
11777 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11778 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11779 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11781 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11783 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11785 * configure.ac (--with-long-double-format): Add support for the
11786 configuration option to change the default long double format on
11788 * config.gcc (powerpc*-linux*-*): Likewise.
11789 * configure: Regenerate.
11790 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11791 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11792 used without modification.
11794 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11796 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11797 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11798 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11799 MISC_BUILTIN_SPEC_BARRIER.
11800 (rs6000_init_builtins): Likewise.
11801 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11803 (speculation_barrier): New define_insn.
11804 * doc/extend.texi: Document __builtin_speculation_barrier.
11806 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11809 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11810 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11811 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11813 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11814 integral modes instead of "ss" and "sd".
11815 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11816 vectors with 32-bit and 64-bit elements.
11817 (vecdupssescalarmodesuffix): New mode attribute.
11818 (vec_dup<mode>): Use it.
11820 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11823 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11824 frame if argument is passed on stack.
11826 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11829 * ree.c (combine_reaching_defs): Optimize also
11830 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11831 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11833 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11835 PR middle-end/83189
11836 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11838 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11840 PR middle-end/83718
11841 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11842 after they are computed.
11844 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11846 PR tree-optimization/83695
11847 * gimple-loop-linterchange.cc
11848 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11849 reset cached scev information after interchange.
11850 (pass_linterchange::execute): Remove call to scev_reset_htab.
11852 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11854 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11855 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11856 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11857 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11858 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11859 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11860 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11861 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11862 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11863 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11864 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11865 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11866 (V_lane_reg): Likewise.
11867 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11869 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11870 (vfmal_lane_low<mode>_intrinsic,
11871 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11872 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11873 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11874 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11875 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11876 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11878 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11880 * config/arm/arm-cpus.in (fp16fml): New feature.
11881 (ALL_SIMD): Add fp16fml.
11882 (armv8.2-a): Add fp16fml as an option.
11883 (armv8.3-a): Likewise.
11884 (armv8.4-a): Add fp16fml as part of fp16.
11885 * config/arm/arm.h (TARGET_FP16FML): Define.
11886 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11888 * config/arm/arm-modes.def (V2HF): Define.
11889 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11890 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11891 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11892 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11893 vfmsl_low, vfmsl_high): New set of builtins.
11894 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11895 (vfml_op): New code attribute.
11896 (VFMLHALVES): New int iterator.
11897 (VFML, VFMLSEL): New mode attributes.
11898 (V_reg): Define mapping for V2HF.
11899 (V_hi, V_lo): New mode attributes.
11900 (VF_constraint): Likewise.
11901 (vfml_half, vfml_half_selector): New int attributes.
11902 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11904 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11905 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11907 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11908 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11909 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11910 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11912 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11913 Document new effective target and option set.
11915 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11917 * config/arm/arm-cpus.in (armv8_4): New feature.
11918 (ARMv8_4a): New fgroup.
11919 (armv8.4-a): New arch.
11920 * config/arm/arm-tables.opt: Regenerate.
11921 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11922 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11923 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11924 Add matching rules for -march=armv8.4-a and extensions.
11925 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11927 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11930 * config/rx/rx.md (BW): New mode attribute.
11931 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11933 2018-01-11 Richard Biener <rguenther@suse.de>
11935 PR tree-optimization/83435
11936 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11937 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11938 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11940 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11941 Alan Hayward <alan.hayward@arm.com>
11942 David Sherwood <david.sherwood@arm.com>
11944 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11946 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11947 (aarch64_print_address_internal): Use it to check for a zero offset.
11949 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11950 Alan Hayward <alan.hayward@arm.com>
11951 David Sherwood <david.sherwood@arm.com>
11953 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11954 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11955 Return a poly_int64 rather than a HOST_WIDE_INT.
11956 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11957 rather than a HOST_WIDE_INT.
11958 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11959 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11960 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11961 final_offset from HOST_WIDE_INT to poly_int64.
11962 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11963 to_constant when getting the number of units in an Advanced SIMD
11965 (aarch64_builtin_vectorized_function): Check for a constant number
11967 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11969 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11970 attribute instead of GET_MODE_NUNITS.
11971 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11972 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11973 GET_MODE_SIZE for fixed-size registers.
11974 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11975 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11976 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11977 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11978 (aarch64_print_operand, aarch64_print_address_internal)
11979 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11980 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11981 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11982 Handle polynomial GET_MODE_SIZE.
11983 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11984 wider than SImode without modification.
11985 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11986 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11987 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11988 passing and returning SVE modes.
11989 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11990 rather than GEN_INT.
11991 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11992 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11993 (aarch64_allocate_and_probe_stack_space): Likewise.
11994 (aarch64_layout_frame): Cope with polynomial offsets.
11995 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11996 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11997 polynomial offsets.
11998 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11999 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12000 poly_int64 rather than a HOST_WIDE_INT.
12001 (aarch64_get_separate_components, aarch64_process_components)
12002 (aarch64_expand_prologue, aarch64_expand_epilogue)
12003 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12004 (aarch64_anchor_offset): New function, split out from...
12005 (aarch64_legitimize_address): ...here.
12006 (aarch64_builtin_vectorization_cost): Handle polynomial
12007 TYPE_VECTOR_SUBPARTS.
12008 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12010 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12011 number of elements from the PARALLEL rather than the mode.
12012 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12013 rather than GET_MODE_BITSIZE.
12014 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12015 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12016 (aarch64_expand_vec_perm_const_1): Handle polynomial
12017 d->perm.length () and d->perm elements.
12018 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12019 Apply to_constant to d->perm elements.
12020 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12021 polynomial CONST_VECTOR_NUNITS.
12022 (aarch64_move_pointer): Take amount as a poly_int64 rather
12024 (aarch64_progress_pointer): Avoid temporary variable.
12025 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12026 the mode attribute instead of GET_MODE.
12028 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12029 Alan Hayward <alan.hayward@arm.com>
12030 David Sherwood <david.sherwood@arm.com>
12032 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12033 x exists before using it.
12034 (aarch64_add_constant_internal): Rename to...
12035 (aarch64_add_offset_1): ...this. Replace regnum with separate
12036 src and dest rtxes. Handle the case in which they're different,
12037 including when the offset is zero. Replace scratchreg with an rtx.
12038 Use 2 additions if there is no spare register into which we can
12039 move a 16-bit constant.
12040 (aarch64_add_constant): Delete.
12041 (aarch64_add_offset): Replace reg with separate src and dest
12042 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12043 Use aarch64_add_offset_1.
12044 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12045 an rtx rather than an int. Take the delta as a poly_int64
12046 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12047 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12048 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12049 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12050 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12051 and aarch64_add_sp.
12052 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12053 aarch64_add_constant.
12055 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12057 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12058 Use scalar_float_mode.
12060 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12062 * config/aarch64/aarch64-simd.md
12063 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12064 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12065 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12066 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12067 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12068 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12069 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12070 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12071 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12072 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12074 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12077 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12078 targ_options->x_arm_arch_string is non NULL.
12080 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12082 * config/aarch64/aarch64.h
12083 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12085 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12088 * expmed.c (emit_store_flag_force): Swap if const op0
12089 and change VOIDmode to mode of op0.
12091 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12093 PR rtl-optimization/83761
12094 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12095 than bytes to mode_for_size.
12097 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12099 PR middle-end/83189
12100 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12101 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12104 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12106 PR middle-end/83575
12107 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12108 when in layout mode.
12109 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12110 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12113 2018-01-10 Michael Collison <michael.collison@arm.com>
12115 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12116 * config/aarch64/aarch64-option-extension.def: Add
12117 AARCH64_OPT_EXTENSION of 'fp16fml'.
12118 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12119 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12120 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12121 * config/aarch64/constraints.md (Ui7): New constraint.
12122 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12123 (VFMLA_SEL_W): Ditto.
12126 (VFMLA16_LOW): New int iterator.
12127 (VFMLA16_HIGH): Ditto.
12128 (UNSPEC_FMLAL): New unspec.
12129 (UNSPEC_FMLSL): Ditto.
12130 (UNSPEC_FMLAL2): Ditto.
12131 (UNSPEC_FMLSL2): Ditto.
12132 (f16mac): New code attribute.
12133 * config/aarch64/aarch64-simd-builtins.def
12134 (aarch64_fmlal_lowv2sf): Ditto.
12135 (aarch64_fmlsl_lowv2sf): Ditto.
12136 (aarch64_fmlalq_lowv4sf): Ditto.
12137 (aarch64_fmlslq_lowv4sf): Ditto.
12138 (aarch64_fmlal_highv2sf): Ditto.
12139 (aarch64_fmlsl_highv2sf): Ditto.
12140 (aarch64_fmlalq_highv4sf): Ditto.
12141 (aarch64_fmlslq_highv4sf): Ditto.
12142 (aarch64_fmlal_lane_lowv2sf): Ditto.
12143 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12144 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12145 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12146 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12147 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12148 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12149 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12150 (aarch64_fmlal_lane_highv2sf): Ditto.
12151 (aarch64_fmlsl_lane_highv2sf): Ditto.
12152 (aarch64_fmlal_laneq_highv2sf): Ditto.
12153 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12154 (aarch64_fmlalq_lane_highv4sf): Ditto.
12155 (aarch64_fmlsl_lane_highv4sf): Ditto.
12156 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12157 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12158 * config/aarch64/aarch64-simd.md:
12159 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12160 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12161 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12162 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12163 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12164 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12165 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12166 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12167 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12168 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12169 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12170 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12171 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12172 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12173 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12174 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12175 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12176 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12177 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12178 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12179 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12180 (vfmlsl_low_u32): Ditto.
12181 (vfmlalq_low_u32): Ditto.
12182 (vfmlslq_low_u32): Ditto.
12183 (vfmlal_high_u32): Ditto.
12184 (vfmlsl_high_u32): Ditto.
12185 (vfmlalq_high_u32): Ditto.
12186 (vfmlslq_high_u32): Ditto.
12187 (vfmlal_lane_low_u32): Ditto.
12188 (vfmlsl_lane_low_u32): Ditto.
12189 (vfmlal_laneq_low_u32): Ditto.
12190 (vfmlsl_laneq_low_u32): Ditto.
12191 (vfmlalq_lane_low_u32): Ditto.
12192 (vfmlslq_lane_low_u32): Ditto.
12193 (vfmlalq_laneq_low_u32): Ditto.
12194 (vfmlslq_laneq_low_u32): Ditto.
12195 (vfmlal_lane_high_u32): Ditto.
12196 (vfmlsl_lane_high_u32): Ditto.
12197 (vfmlal_laneq_high_u32): Ditto.
12198 (vfmlsl_laneq_high_u32): Ditto.
12199 (vfmlalq_lane_high_u32): Ditto.
12200 (vfmlslq_lane_high_u32): Ditto.
12201 (vfmlalq_laneq_high_u32): Ditto.
12202 (vfmlslq_laneq_high_u32): Ditto.
12203 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12204 (AARCH64_FL_FOR_ARCH8_4): New.
12205 (AARCH64_ISA_F16FML): New ISA flag.
12206 (TARGET_F16FML): New feature flag for fp16fml.
12207 (doc/invoke.texi): Document new fp16fml option.
12209 2018-01-10 Michael Collison <michael.collison@arm.com>
12211 * config/aarch64/aarch64-builtins.c:
12212 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12213 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12214 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12215 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12216 (AARCH64_ISA_SHA3): New ISA flag.
12217 (TARGET_SHA3): New feature flag for sha3.
12218 * config/aarch64/iterators.md (sha512_op): New int attribute.
12219 (CRYPTO_SHA512): New int iterator.
12220 (UNSPEC_SHA512H): New unspec.
12221 (UNSPEC_SHA512H2): Ditto.
12222 (UNSPEC_SHA512SU0): Ditto.
12223 (UNSPEC_SHA512SU1): Ditto.
12224 * config/aarch64/aarch64-simd-builtins.def
12225 (aarch64_crypto_sha512hqv2di): New builtin.
12226 (aarch64_crypto_sha512h2qv2di): Ditto.
12227 (aarch64_crypto_sha512su0qv2di): Ditto.
12228 (aarch64_crypto_sha512su1qv2di): Ditto.
12229 (aarch64_eor3qv8hi): Ditto.
12230 (aarch64_rax1qv2di): Ditto.
12231 (aarch64_xarqv2di): Ditto.
12232 (aarch64_bcaxqv8hi): Ditto.
12233 * config/aarch64/aarch64-simd.md:
12234 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12235 (aarch64_crypto_sha512su0qv2di): Ditto.
12236 (aarch64_crypto_sha512su1qv2di): Ditto.
12237 (aarch64_eor3qv8hi): Ditto.
12238 (aarch64_rax1qv2di): Ditto.
12239 (aarch64_xarqv2di): Ditto.
12240 (aarch64_bcaxqv8hi): Ditto.
12241 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12242 (vsha512h2q_u64): Ditto.
12243 (vsha512su0q_u64): Ditto.
12244 (vsha512su1q_u64): Ditto.
12245 (veor3q_u16): Ditto.
12246 (vrax1q_u64): Ditto.
12247 (vxarq_u64): Ditto.
12248 (vbcaxq_u16): Ditto.
12249 * config/arm/types.md (crypto_sha512): New type attribute.
12250 (crypto_sha3): Ditto.
12251 (doc/invoke.texi): Document new sha3 option.
12253 2018-01-10 Michael Collison <michael.collison@arm.com>
12255 * config/aarch64/aarch64-builtins.c:
12256 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12257 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12258 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12259 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12260 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12261 (AARCH64_ISA_SM4): New ISA flag.
12262 (TARGET_SM4): New feature flag for sm4.
12263 * config/aarch64/aarch64-simd-builtins.def
12264 (aarch64_sm3ss1qv4si): Ditto.
12265 (aarch64_sm3tt1aq4si): Ditto.
12266 (aarch64_sm3tt1bq4si): Ditto.
12267 (aarch64_sm3tt2aq4si): Ditto.
12268 (aarch64_sm3tt2bq4si): Ditto.
12269 (aarch64_sm3partw1qv4si): Ditto.
12270 (aarch64_sm3partw2qv4si): Ditto.
12271 (aarch64_sm4eqv4si): Ditto.
12272 (aarch64_sm4ekeyqv4si): Ditto.
12273 * config/aarch64/aarch64-simd.md:
12274 (aarch64_sm3ss1qv4si): Ditto.
12275 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12276 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12277 (aarch64_sm4eqv4si): Ditto.
12278 (aarch64_sm4ekeyqv4si): Ditto.
12279 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12280 (sm3part_op): Ditto.
12281 (CRYPTO_SM3TT): Ditto.
12282 (CRYPTO_SM3PART): Ditto.
12283 (UNSPEC_SM3SS1): New unspec.
12284 (UNSPEC_SM3TT1A): Ditto.
12285 (UNSPEC_SM3TT1B): Ditto.
12286 (UNSPEC_SM3TT2A): Ditto.
12287 (UNSPEC_SM3TT2B): Ditto.
12288 (UNSPEC_SM3PARTW1): Ditto.
12289 (UNSPEC_SM3PARTW2): Ditto.
12290 (UNSPEC_SM4E): Ditto.
12291 (UNSPEC_SM4EKEY): Ditto.
12292 * config/aarch64/constraints.md (Ui2): New constraint.
12293 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12294 * config/arm/types.md (crypto_sm3): New type attribute.
12295 (crypto_sm4): Ditto.
12296 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12297 (vsm3tt1aq_u32): Ditto.
12298 (vsm3tt1bq_u32): Ditto.
12299 (vsm3tt2aq_u32): Ditto.
12300 (vsm3tt2bq_u32): Ditto.
12301 (vsm3partw1q_u32): Ditto.
12302 (vsm3partw2q_u32): Ditto.
12303 (vsm4eq_u32): Ditto.
12304 (vsm4ekeyq_u32): Ditto.
12305 (doc/invoke.texi): Document new sm4 option.
12307 2018-01-10 Michael Collison <michael.collison@arm.com>
12309 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12310 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12311 (AARCH64_FL_FOR_ARCH8_4): New.
12312 (AARCH64_FL_V8_4): New flag.
12313 (doc/invoke.texi): Document new armv8.4-a option.
12315 2018-01-10 Michael Collison <michael.collison@arm.com>
12317 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12318 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12319 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12320 * config/aarch64/aarch64-option-extension.def: Add
12321 AARCH64_OPT_EXTENSION of 'sha2'.
12322 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12323 (crypto): Disable sha2 and aes if crypto disabled.
12324 (crypto): Enable aes and sha2 if enabled.
12325 (simd): Disable sha2 and aes if simd disabled.
12326 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12328 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12329 (TARGET_SHA2): New feature flag for sha2.
12330 (TARGET_AES): New feature flag for aes.
12331 * config/aarch64/aarch64-simd.md:
12332 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12333 conditional on TARGET_AES.
12334 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12335 (aarch64_crypto_sha1hsi): Make pattern conditional
12337 (aarch64_crypto_sha1hv4si): Ditto.
12338 (aarch64_be_crypto_sha1hv4si): Ditto.
12339 (aarch64_crypto_sha1su1v4si): Ditto.
12340 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12341 (aarch64_crypto_sha1su0v4si): Ditto.
12342 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12343 (aarch64_crypto_sha256su0v4si): Ditto.
12344 (aarch64_crypto_sha256su1v4si): Ditto.
12345 (doc/invoke.texi): Document new aes and sha2 options.
12347 2018-01-10 Martin Sebor <msebor@redhat.com>
12349 PR tree-optimization/83781
12350 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12353 2018-01-11 Martin Sebor <msebor@gmail.com>
12354 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12356 PR tree-optimization/83501
12357 PR tree-optimization/81703
12359 * tree-ssa-strlen.c (get_string_cst): Rename...
12360 (get_string_len): ...to this. Handle global constants.
12361 (handle_char_store): Adjust.
12363 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12364 Jim Wilson <jimw@sifive.com>
12366 * config/riscv/riscv-protos.h (riscv_output_return): New.
12367 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12368 (riscv_attribute_table, riscv_output_return),
12369 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12370 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12371 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12373 (riscv_expand_prologue): Add early return for naked function.
12374 (riscv_expand_epilogue): Likewise.
12375 (riscv_function_ok_for_sibcall): Return false for naked function.
12376 (riscv_set_current_function): New.
12377 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12378 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12379 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12380 * doc/extend.texi (RISC-V Function Attributes): New.
12382 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12384 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12385 check for 128-bit long double before checking TCmode.
12386 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12387 128-bit long doubles before checking TFmode or TCmode.
12388 (FLOAT128_IBM_P): Likewise.
12390 2018-01-10 Martin Sebor <msebor@redhat.com>
12392 PR tree-optimization/83671
12393 * builtins.c (c_strlen): Unconditionally return zero for the empty
12395 Use -Warray-bounds for warnings.
12396 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12397 for non-constant array indices with COMPONENT_REF, arrays of
12398 arrays, and pointers to arrays.
12399 (gimple_fold_builtin_strlen): Determine and set length range for
12400 non-constant character arrays.
12402 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12404 PR middle-end/81897
12405 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12408 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12410 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12412 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12415 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12416 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12417 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12418 indexed_or_indirect_operand predicate.
12419 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12420 (*vsx_le_perm_load_v8hi): Likewise.
12421 (*vsx_le_perm_load_v16qi): Likewise.
12422 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12423 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12424 (*vsx_le_perm_store_v8hi): Likewise.
12425 (*vsx_le_perm_store_v16qi): Likewise.
12426 (eight unnamed splitters): Likewise.
12428 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12430 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12431 * config/rs6000/emmintrin.h: Likewise.
12432 * config/rs6000/mmintrin.h: Likewise.
12433 * config/rs6000/xmmintrin.h: Likewise.
12435 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12438 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12440 * tree.c (tree_nop_conversion): Return true for location wrapper
12442 (maybe_wrap_with_location): New function.
12443 (selftest::check_strip_nops): New function.
12444 (selftest::test_location_wrappers): New function.
12445 (selftest::tree_c_tests): Call it.
12446 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12447 (maybe_wrap_with_location): New decl.
12448 (EXPR_LOCATION_WRAPPER_P): New macro.
12449 (location_wrapper_p): New inline function.
12450 (tree_strip_any_location_wrapper): New inline function.
12452 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12455 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12456 stack_realign_offset for the largest alignment of stack slot
12458 (ix86_find_max_used_stack_alignment): New function.
12459 (ix86_finalize_stack_frame_flags): Use it. Set
12460 max_used_stack_alignment if we don't realign stack.
12461 * config/i386/i386.h (machine_function): Add
12462 max_used_stack_alignment.
12464 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12466 * config/arm/arm.opt (-mbranch-cost): New option.
12467 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12470 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12473 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12474 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12476 2018-01-10 Richard Biener <rguenther@suse.de>
12479 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12480 early out so it also covers the case where we have a non-NULL
12483 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12485 PR tree-optimization/83753
12486 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12487 for non-strided grouped accesses if the number of elements is 1.
12489 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12492 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12493 * i386.h (TARGET_USE_GATHER): Define.
12494 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12496 2018-01-10 Martin Liska <mliska@suse.cz>
12499 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12500 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12502 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12503 CLEANUP_NO_PARTITIONING is not set.
12505 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12507 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12508 for vectors, as a partial revert of r254296.
12509 * rtl.h (const_vec_p): Delete.
12510 (const_vec_duplicate_p): Don't test for vector CONSTs.
12511 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12512 * expmed.c (make_tree): Likewise.
12515 * common.md (E, F): Use CONSTANT_P instead of checking for
12517 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12518 checking for CONST_VECTOR.
12520 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12522 PR middle-end/83575
12523 * predict.c (force_edge_cold): Handle in more sane way edges
12524 with no prediction.
12526 2018-01-09 Carl Love <cel@us.ibm.com>
12528 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12530 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12531 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12532 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12533 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12534 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12535 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12536 * config/rs6000/rs6000-protos.h: Add extern defition for
12537 rs6000_generate_float2_double_code.
12538 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12540 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12541 (float2_v2df): Add define_expand.
12543 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12546 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12547 op_mode in the force_to_mode call.
12549 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12551 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12552 instead of checking each element individually.
12553 (aarch64_evpc_uzp): Likewise.
12554 (aarch64_evpc_zip): Likewise.
12555 (aarch64_evpc_ext): Likewise.
12556 (aarch64_evpc_rev): Likewise.
12557 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12558 instead of checking each element individually. Return true without
12560 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12561 whether all selected elements come from the same input, instead of
12562 checking each element individually. Remove calls to gen_rtx_REG,
12563 start_sequence and end_sequence and instead assert that no rtl is
12566 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12568 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12569 order of HIGH and CONST checks.
12571 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12573 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12574 if the destination isn't an SSA_NAME.
12576 2018-01-09 Richard Biener <rguenther@suse.de>
12578 PR tree-optimization/83668
12579 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12581 (canonicalize_loop_form): ... here, renamed from ...
12582 (canonicalize_loop_closed_ssa_form): ... this and amended to
12583 swap successor edges for loop exit blocks to make us use
12584 the RPO order we need for initial schedule generation.
12586 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12588 PR tree-optimization/64811
12589 * match.pd: When optimizing comparisons with Inf, avoid
12590 introducing or losing exceptions from comparisons with NaN.
12592 2018-01-09 Martin Liska <mliska@suse.cz>
12595 * asan.c (shadow_mem_size): Add gcc_assert.
12597 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12599 Don't save registers in main().
12602 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12603 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12604 * config/avr/avr.c (avr_set_current_function): Don't error if
12605 naked, OS_task or OS_main are specified at the same time.
12606 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12608 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12610 * common/config/avr/avr-common.c (avr_option_optimization_table):
12611 Switch on -mmain-is-OS_task for optimizing compilations.
12613 2018-01-09 Richard Biener <rguenther@suse.de>
12615 PR tree-optimization/83572
12616 * graphite.c: Include cfganal.h.
12617 (graphite_transform_loops): Connect infinite loops to exit
12618 and remove fake edges at the end.
12620 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12622 * ipa-inline.c (edge_badness): Revert accidental checkin.
12624 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12627 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12628 symbols; not inline clones.
12630 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12633 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12634 hard registers. Formatting fixes.
12636 PR preprocessor/83722
12637 * gcc.c (try_generate_repro): Pass
12638 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12639 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12642 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12643 Kito Cheng <kito.cheng@gmail.com>
12645 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12646 (riscv_leaf_function_p): Delete.
12647 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12649 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12651 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12653 (do_ifelse): New function.
12654 (do_isel): New function.
12655 (do_sub3): New function.
12656 (do_add3): New function.
12657 (do_load_mask_compare): New function.
12658 (do_overlap_load_compare): New function.
12659 (expand_compare_loop): New function.
12660 (expand_block_compare): Call expand_compare_loop() when appropriate.
12661 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12662 option description.
12663 (-mblock-compare-inline-loop-limit): New option.
12665 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12668 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12669 Reverse order of second and third operands in first alternative.
12670 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12671 of first and second elements in UNSPEC_VPERMR vector.
12672 (altivec_expand_vec_perm_le): Likewise.
12674 2018-01-08 Jeff Law <law@redhat.com>
12676 PR rtl-optimizatin/81308
12677 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12678 (process_switch): If group_case_labels makes a change, then set
12680 (pass_convert_switch::execute): If a switch is converted, then
12681 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12683 PR rtl-optimization/81308
12684 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12687 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12689 PR target/83663 - Revert r255946
12690 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12691 generation for cases where splatting a value is not useful.
12692 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12693 across a vec_duplicate and a paradoxical subreg forming a vector
12694 mode to a vec_concat.
12696 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12698 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12699 -march=armv8.3-a variants.
12700 * config/arm/t-multilib: Likewise.
12701 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12703 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12705 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12707 (cceq_ior_compare_complement): Give it a name so I can use it, and
12708 change boolean_or_operator predicate to boolean_operator so it can
12709 be used to generate a crand.
12710 (eqne): New code iterator.
12711 (bd/bd_neg): New code_attrs.
12712 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12713 a single define_insn.
12714 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12715 decrement (bdnzt/bdnzf/bdzt/bdzf).
12716 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12717 with the new names of the branch decrement patterns, and added the
12718 names of the branch decrement conditional patterns.
12720 2018-01-08 Richard Biener <rguenther@suse.de>
12722 PR tree-optimization/83563
12723 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12726 2018-01-08 Richard Biener <rguenther@suse.de>
12728 PR middle-end/83713
12729 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12731 2018-01-08 Richard Biener <rguenther@suse.de>
12733 PR tree-optimization/83685
12734 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12735 references to abnormals.
12737 2018-01-08 Richard Biener <rguenther@suse.de>
12740 * dwarf2out.c (output_indirect_strings): Handle empty
12741 skeleton_debug_str_hash.
12742 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12744 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12746 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12747 (emit_store_direct): Likewise.
12748 (arc_trampoline_adjust_address): Likewise.
12749 (arc_asm_trampoline_template): New function.
12750 (arc_initialize_trampoline): Use asm_trampoline_template.
12751 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12752 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12753 * config/arc/arc.md (flush_icache): Delete pattern.
12755 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12757 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12758 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12761 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12764 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12765 by not USED_FOR_TARGET.
12766 (make_pass_resolve_sw_modes): Likewise.
12768 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12770 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12773 2018-01-08 Richard Biener <rguenther@suse.de>
12775 PR middle-end/83580
12776 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12778 2018-01-08 Richard Biener <rguenther@suse.de>
12780 PR middle-end/83517
12781 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12783 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12785 PR middle-end/81897
12786 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12787 basic blocks with a small number of successors.
12788 (convert_control_dep_chain_into_preds): Improve handling of
12790 (dump_predicates): Split apart into...
12791 (dump_pred_chain): ...here...
12792 (dump_pred_info): ...and here.
12793 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12794 (can_chain_union_be_invalidated_p): Improve check for invalidation
12796 (uninit_uses_cannot_happen): Avoid unnecessary if
12797 convert_control_dep_chain_into_preds yielded nothing.
12799 2018-01-06 Martin Sebor <msebor@redhat.com>
12801 PR tree-optimization/83640
12802 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12803 subtracting negative offset from size.
12804 (builtin_access::overlap): Adjust offset bounds of the access to fall
12805 within the size of the object if possible.
12807 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12809 PR rtl-optimization/83699
12810 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12811 extract_bit_field_as_subreg to cases in which the extracted
12812 value is also a vector.
12814 * lra-constraints.c (process_alt_operands): Test for the equivalence
12815 substitutions when detecting a possible reload cycle.
12817 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12820 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12821 by default if flag_selective_schedling{,2}. Formatting fixes.
12823 PR rtl-optimization/83682
12824 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12825 if it has non-VECTOR_MODE element mode.
12826 (vec_duplicate_p): Likewise.
12828 PR middle-end/83694
12829 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12830 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12832 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12835 * config/i386/i386-builtin.def
12836 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12837 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12838 Require also OPTION_MASK_ISA_AVX512F in addition to
12839 OPTION_MASK_ISA_GFNI.
12840 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12841 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12842 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12843 to OPTION_MASK_ISA_GFNI.
12844 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12845 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12846 OPTION_MASK_ISA_AVX512BW.
12847 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12848 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12849 addition to OPTION_MASK_ISA_GFNI.
12850 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12851 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12852 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12853 to OPTION_MASK_ISA_GFNI.
12854 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12855 a requirement for all ISAs rather than any of them with a few
12857 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12859 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12860 bitmasks to be enabled with 3 exceptions, instead of requiring any
12861 enabled ISA with lots of exceptions.
12862 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12863 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12864 Change avx512bw in isa attribute to avx512f.
12865 * config/i386/sgxintrin.h: Add license boilerplate.
12866 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12867 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12868 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12869 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12871 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12872 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12873 temporarily sse2 rather than sse if not enabled already.
12876 * config/i386/sse.md (VI248_VLBW): Rename to ...
12877 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12878 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12879 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12880 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12881 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12882 mode iterator instead of VI248_VLBW.
12884 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12886 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12887 (record_modified): Skip clobbers; add debug output.
12888 (param_change_prob): Use sreal frequencies.
12890 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12892 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12893 punt for user-aligned variables.
12895 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12897 * tree-chrec.c (chrec_contains_symbols): Return true for
12900 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12903 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12904 of (x|y) == x for BICS pattern.
12906 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12908 PR tree-optimization/83605
12909 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12910 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12913 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12915 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12916 * config/epiphany/rtems.h: New file.
12918 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12919 Uros Bizjak <ubizjak@gmail.com>
12922 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12923 QIreg_operand instead of register_operand predicate.
12924 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12925 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12926 comments instead of -fmitigate[-_]rop.
12928 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12931 * cgraphunit.c (symbol_table::compile): Switch to text_section
12932 before calling assembly_start debug hook.
12933 * run-rtl-passes.c (run_rtl_passes): Likewise.
12936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12938 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12939 range_int_cst_p rather than !symbolic_range_p before calling
12940 extract_range_from_multiplicative_op_1.
12942 2018-01-04 Jeff Law <law@redhat.com>
12944 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12945 redundant test in assertion.
12947 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12949 * doc/rtl.texi: Document machine_mode wrapper classes.
12951 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12953 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12954 using tree_to_uhwi.
12956 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12958 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12959 the VEC_PERM_EXPR fold to fail.
12961 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12964 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12965 to switched_sections.
12967 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12970 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12971 test for d.testing.
12973 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12976 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12977 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12979 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12982 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12983 is BLKmode and bitpos not zero or mode change is needed.
12985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12988 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12991 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12994 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12995 instead of MULT rtx. Update all corresponding splitters.
12996 (*saddl_se): Ditto.
12997 (*ssub<modesuffix>): Ditto.
12998 (*ssubl_se): Ditto.
12999 (*cmp_sadd_di): Update split patterns.
13000 (*cmp_sadd_si): Ditto.
13001 (*cmp_sadd_sidi): Ditto.
13002 (*cmp_ssub_di): Ditto.
13003 (*cmp_ssub_si): Ditto.
13004 (*cmp_ssub_sidi): Ditto.
13005 * config/alpha/predicates.md (const23_operand): New predicate.
13006 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13007 Look for ASHIFT, not MULT inner operand.
13008 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13010 2018-01-04 Martin Liska <mliska@suse.cz>
13012 PR gcov-profile/83669
13013 * gcov.c (output_intermediate_file): Add version to intermediate
13015 * doc/gcov.texi: Document new field 'version' in intermediate
13016 file format. Fix location of '-k' option of gcov command.
13018 2018-01-04 Martin Liska <mliska@suse.cz>
13021 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13023 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13025 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13027 2018-01-03 Martin Sebor <msebor@redhat.com>
13029 PR tree-optimization/83655
13030 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13031 checking calls with invalid arguments.
13033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13035 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13036 (vectorizable_mask_load_store): Delete.
13037 (vectorizable_call): Return false for masked loads and stores.
13038 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13039 instead of gimple_assign_rhs1.
13040 (vectorizable_load): Handle IFN_MASK_LOAD.
13041 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13045 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13047 (vectorizable_mask_load_store): ...here.
13048 (vectorizable_load): ...and here.
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13052 * tree-vect-stmts.c (vect_build_all_ones_mask)
13053 (vect_build_zero_merge_argument): New functions, split out from...
13054 (vectorizable_load): ...here.
13056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13058 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13060 (vectorizable_mask_load_store): ...here.
13061 (vectorizable_store): ...and here.
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13065 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13067 (vectorizable_mask_load_store): ...here.
13069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13071 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13072 (vect_model_store_cost): Take a vec_load_store_type instead of a
13074 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13075 (vect_model_store_cost): Take a vec_load_store_type instead of a
13077 (vectorizable_mask_load_store): Update accordingly.
13078 (vectorizable_store): Likewise.
13079 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13083 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13084 IFN_MASK_LOAD calls here rather than...
13085 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13088 Alan Hayward <alan.hayward@arm.com>
13089 David Sherwood <david.sherwood@arm.com>
13091 * expmed.c (extract_bit_field_1): For vector extracts,
13092 fall back to extract_bit_field_as_subreg if vec_extract
13095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13096 Alan Hayward <alan.hayward@arm.com>
13097 David Sherwood <david.sherwood@arm.com>
13099 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13100 they are variable or constant sized.
13101 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13102 slots for constant-sized data.
13104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13105 Alan Hayward <alan.hayward@arm.com>
13106 David Sherwood <david.sherwood@arm.com>
13108 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13109 handling COND_EXPRs with boolean comparisons, try to find a better
13110 basis for the mask type than the boolean itself.
13112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13114 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13115 is calculated and how it can be overridden.
13116 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13117 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13119 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13123 Alan Hayward <alan.hayward@arm.com>
13124 David Sherwood <david.sherwood@arm.com>
13126 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13127 Remove the mode argument.
13128 (aarch64_simd_valid_immediate): Remove the mode and inverse
13130 * config/aarch64/iterators.md (bitsize): New iterator.
13131 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13132 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13133 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13134 aarch64_simd_valid_immediate.
13135 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13136 (aarch64_reg_or_bic_imm): Likewise.
13137 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13138 with an insn_type enum and msl with a modifier_type enum.
13139 Replace element_width with a scalar_mode. Change the shift
13140 to unsigned int. Add constructors for scalar_float_mode and
13141 scalar_int_mode elements.
13142 (aarch64_vect_float_const_representable_p): Delete.
13143 (aarch64_can_const_movi_rtx_p)
13144 (aarch64_simd_scalar_immediate_valid_for_move)
13145 (aarch64_simd_make_constant): Update call to
13146 aarch64_simd_valid_immediate.
13147 (aarch64_advsimd_valid_immediate_hs): New function.
13148 (aarch64_advsimd_valid_immediate): Likewise.
13149 (aarch64_simd_valid_immediate): Remove mode and inverse
13150 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13151 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13152 and aarch64_float_const_representable_p on the result.
13153 (aarch64_output_simd_mov_immediate): Remove mode argument.
13154 Update call to aarch64_simd_valid_immediate and use of
13155 simd_immediate_info.
13156 (aarch64_output_scalar_simd_mov_immediate): Update call
13159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13160 Alan Hayward <alan.hayward@arm.com>
13161 David Sherwood <david.sherwood@arm.com>
13163 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13164 (mode_nunits): Likewise CONST_MODE_NUNITS.
13165 * machmode.def (ADJUST_NUNITS): Document.
13166 * genmodes.c (mode_data::need_nunits_adj): New field.
13167 (blank_mode): Update accordingly.
13168 (adj_nunits): New variable.
13169 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13171 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13172 listed in adj_nunits.
13173 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13174 listed in adj_nunits. Don't emit case statements for such modes.
13175 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13176 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13177 nothing if adj_nunits is nonnull.
13178 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13179 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13180 (emit_mode_fbit): Update use of print_maybe_const_decl.
13181 (emit_move_size): Likewise. Treat the array as non-const
13183 (emit_mode_adjustments): Handle adj_nunits.
13185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13187 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13188 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13189 (VECTOR_MODES): Use it.
13190 (make_vector_modes): Take the prefix as an argument.
13192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13193 Alan Hayward <alan.hayward@arm.com>
13194 David Sherwood <david.sherwood@arm.com>
13196 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13197 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13198 for MODE_VECTOR_BOOL.
13199 * machmode.def (VECTOR_BOOL_MODE): Document.
13200 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13201 (make_vector_bool_mode): New function.
13202 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13204 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13205 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13207 * stor-layout.c (int_mode_for_mode): Likewise.
13208 * tree.c (build_vector_type_for_mode): Likewise.
13209 * varasm.c (output_constant_pool_2): Likewise.
13210 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13211 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13212 for MODE_VECTOR_BOOL.
13213 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13214 of mode class checks.
13215 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13216 instead of a list of mode class checks.
13217 (expand_vector_scalar_condition): Likewise.
13218 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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_size): Change from unsigned short to
13226 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13227 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13228 or if measurement_type is not polynomial.
13229 (fixed_size_mode::includes_p): Check for constant-sized modes.
13230 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13231 return a poly_uint16 rather than an unsigned short.
13232 (emit_mode_size): Change the type of mode_size from unsigned short
13233 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13234 (emit_mode_adjustments): Cope with polynomial vector sizes.
13235 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13237 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13239 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13240 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13241 * caller-save.c (setup_save_areas): Likewise.
13242 (replace_reg_with_saved_mem): Likewise.
13243 * calls.c (emit_library_call_value_1): Likewise.
13244 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13245 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13246 (gen_lowpart_for_combine): Likewise.
13247 * convert.c (convert_to_integer_1): Likewise.
13248 * cse.c (equiv_constant, cse_insn): Likewise.
13249 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13250 (cselib_subst_to_values): Likewise.
13251 * dce.c (word_dce_process_block): Likewise.
13252 * df-problems.c (df_word_lr_mark_ref): Likewise.
13253 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13254 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13255 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13256 (rtl_for_decl_location): Likewise.
13257 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13258 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13259 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13260 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13261 (expand_expr_real_1): Likewise.
13262 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13263 (pad_below): Likewise.
13264 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13265 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13266 * ira.c (get_subreg_tracking_sizes): Likewise.
13267 * ira-build.c (ira_create_allocno_objects): Likewise.
13268 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13269 (ira_sort_regnos_for_alter_reg): Likewise.
13270 * ira-costs.c (record_operand_costs): Likewise.
13271 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13272 (resolve_simple_move): Likewise.
13273 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13274 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13275 (lra_constraints): Likewise.
13276 (CONST_POOL_OK_P): Reject variable-sized modes.
13277 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13278 (add_pseudo_to_slot, lra_spill): Likewise.
13279 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13280 * optabs-query.c (get_best_extraction_insn): Likewise.
13281 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13282 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13283 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13284 * recog.c (offsettable_address_addr_space_p): Likewise.
13285 * regcprop.c (maybe_mode_change): Likewise.
13286 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13287 * regrename.c (build_def_use): Likewise.
13288 * regstat.c (dump_reg_info): Likewise.
13289 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13290 (find_reloads, find_reloads_subreg_address): Likewise.
13291 * reload1.c (eliminate_regs_1): Likewise.
13292 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13293 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13294 (simplify_binary_operation_1, simplify_subreg): Likewise.
13295 * targhooks.c (default_function_arg_padding): Likewise.
13296 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13297 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13298 (verify_gimple_assign_ternary): Likewise.
13299 * tree-inline.c (estimate_move_cost): Likewise.
13300 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13301 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13302 (get_address_cost_ainc): Likewise.
13303 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13304 (vect_supportable_dr_alignment): Likewise.
13305 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13306 (vectorizable_reduction): Likewise.
13307 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13308 (vectorizable_operation, vectorizable_load): Likewise.
13309 * tree.c (build_same_sized_truth_vector_type): Likewise.
13310 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13311 * var-tracking.c (emit_note_insn_var_location): Likewise.
13312 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13313 (ADDR_VEC_ALIGN): Likewise.
13315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13316 Alan Hayward <alan.hayward@arm.com>
13317 David Sherwood <david.sherwood@arm.com>
13319 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13321 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13322 or if measurement_type is polynomial.
13323 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13324 * combine.c (make_extraction): Likewise.
13325 * dse.c (find_shift_sequence): Likewise.
13326 * dwarf2out.c (mem_loc_descriptor): Likewise.
13327 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13328 (extract_bit_field, extract_low_bits): Likewise.
13329 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13330 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13331 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13332 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13333 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13334 * reload.c (find_reloads): Likewise.
13335 * reload1.c (alter_reg): Likewise.
13336 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13337 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13338 * tree-if-conv.c (predicate_mem_writes): Likewise.
13339 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13340 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13341 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13342 * valtrack.c (dead_debug_insert_temp): Likewise.
13343 * varasm.c (mergeable_constant_section): Likewise.
13344 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13347 Alan Hayward <alan.hayward@arm.com>
13348 David Sherwood <david.sherwood@arm.com>
13350 * expr.c (expand_assignment): Cope with polynomial mode sizes
13351 when assigning to a CONCAT.
13353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13354 Alan Hayward <alan.hayward@arm.com>
13355 David Sherwood <david.sherwood@arm.com>
13357 * machmode.h (mode_precision): Change from unsigned short to
13359 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13361 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13362 or if measurement_type is not polynomial.
13363 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13364 in which the mode is already known to be a scalar_int_mode.
13365 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13366 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13368 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13369 for GET_MODE_PRECISION.
13370 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13371 for GET_MODE_PRECISION.
13372 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13374 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13375 (expand_field_assignment, make_extraction): Likewise.
13376 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13377 (get_last_value): Likewise.
13378 * convert.c (convert_to_integer_1): Likewise.
13379 * cse.c (cse_insn): Likewise.
13380 * expr.c (expand_expr_real_1): Likewise.
13381 * lra-constraints.c (simplify_operand_subreg): Likewise.
13382 * optabs-query.c (can_atomic_load_p): Likewise.
13383 * optabs.c (expand_atomic_load): Likewise.
13384 (expand_atomic_store): Likewise.
13385 * ree.c (combine_reaching_defs): Likewise.
13386 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13387 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13388 * tree.h (type_has_mode_precision_p): Likewise.
13389 * ubsan.c (instrument_si_overflow): Likewise.
13391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13392 Alan Hayward <alan.hayward@arm.com>
13393 David Sherwood <david.sherwood@arm.com>
13395 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13396 polynomial numbers of units.
13397 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13398 (valid_vector_subparts_p): New function.
13399 (build_vector_type): Remove temporary shim and take the number
13400 of units as a poly_uint64 rather than an int.
13401 (build_opaque_vector_type): Take the number of units as a
13402 poly_uint64 rather than an int.
13403 * tree.c (build_vector_from_ctor): Handle polynomial
13404 TYPE_VECTOR_SUBPARTS.
13405 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13406 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13407 (build_vector_from_val): If the number of units is variable,
13408 use build_vec_duplicate_cst for constant operands and
13409 VEC_DUPLICATE_EXPR otherwise.
13410 (make_vector_type): Remove temporary is_constant ().
13411 (build_vector_type, build_opaque_vector_type): Take the number of
13412 units as a poly_uint64 rather than an int.
13413 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13415 * cfgexpand.c (expand_debug_expr): Likewise.
13416 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13417 (store_constructor, expand_expr_real_1): Likewise.
13418 (const_scalar_mask_from_tree): Likewise.
13419 * fold-const-call.c (fold_const_reduction): Likewise.
13420 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13421 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13422 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13423 (fold_relational_const): Likewise.
13424 (native_interpret_vector): Likewise. Change the size from an
13425 int to an unsigned int.
13426 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13427 TYPE_VECTOR_SUBPARTS.
13428 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13429 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13430 duplicating a non-constant operand into a variable-length vector.
13431 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13432 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13433 * ipa-icf.c (sem_variable::equals): Likewise.
13434 * match.pd: Likewise.
13435 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13436 * print-tree.c (print_node): Likewise.
13437 * stor-layout.c (layout_type): Likewise.
13438 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13439 * tree-cfg.c (verify_gimple_comparison): Likewise.
13440 (verify_gimple_assign_binary): Likewise.
13441 (verify_gimple_assign_ternary): Likewise.
13442 (verify_gimple_assign_single): Likewise.
13443 * tree-pretty-print.c (dump_generic_node): Likewise.
13444 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13445 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13446 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13447 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13448 (vect_shift_permute_load_chain): Likewise.
13449 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13450 (expand_vector_condition, optimize_vector_constructor): Likewise.
13451 (lower_vec_perm, get_compute_type): Likewise.
13452 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13453 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13454 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13455 (vect_recog_mask_conversion_pattern): Likewise.
13456 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13457 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13458 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13459 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13460 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13461 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13462 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13463 (supportable_widening_operation): Likewise.
13464 (supportable_narrowing_operation): Likewise.
13465 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13467 * varasm.c (output_constant): Likewise.
13469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13470 Alan Hayward <alan.hayward@arm.com>
13471 David Sherwood <david.sherwood@arm.com>
13473 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13474 so that both the length == 3 and length != 3 cases set up their
13475 own permute vectors. Add comments explaining why we know the
13476 number of elements is constant.
13477 (vect_permute_load_chain): Likewise.
13479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13480 Alan Hayward <alan.hayward@arm.com>
13481 David Sherwood <david.sherwood@arm.com>
13483 * machmode.h (mode_nunits): Change from unsigned char to
13485 (ONLY_FIXED_SIZE_MODES): New macro.
13486 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13487 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13488 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13490 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13491 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13492 or if measurement_type is not polynomial.
13493 * genmodes.c (ZERO_COEFFS): New macro.
13494 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13496 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13497 Use ZERO_COEFFS when emitting initializers.
13498 * data-streamer.h (bp_pack_poly_value): New function.
13499 (bp_unpack_poly_value): Likewise.
13500 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13501 for GET_MODE_NUNITS.
13502 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13503 for GET_MODE_NUNITS.
13504 * tree.c (make_vector_type): Remove temporary shim and make
13505 the real function take the number of units as a poly_uint64
13506 rather than an int.
13507 (build_vector_type_for_mode): Handle polynomial nunits.
13508 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13509 * emit-rtl.c (const_vec_series_p_1): Likewise.
13510 (gen_rtx_CONST_VECTOR): Likewise.
13511 * fold-const.c (test_vec_duplicate_folding): Likewise.
13512 * genrecog.c (validate_pattern): Likewise.
13513 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13514 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13515 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13516 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13517 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13518 * rtlanal.c (subreg_get_info): Likewise.
13519 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13520 (vect_grouped_load_supported): Likewise.
13521 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13522 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13523 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13524 (simplify_const_unary_operation, simplify_binary_operation_1)
13525 (simplify_const_binary_operation, simplify_ternary_operation)
13526 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13527 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13528 instead of CONST_VECTOR_NUNITS.
13529 * varasm.c (output_constant_pool_2): Likewise.
13530 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13531 explicit-encoded elements in the XVEC for variable-length vectors.
13533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13535 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13538 Alan Hayward <alan.hayward@arm.com>
13539 David Sherwood <david.sherwood@arm.com>
13541 * coretypes.h (fixed_size_mode): Declare.
13542 (fixed_size_mode_pod): New typedef.
13543 * builtins.h (target_builtins::x_apply_args_mode)
13544 (target_builtins::x_apply_result_mode): Change type to
13545 fixed_size_mode_pod.
13546 * builtins.c (apply_args_size, apply_result_size, result_vector)
13547 (expand_builtin_apply_args_1, expand_builtin_apply)
13548 (expand_builtin_return): Update accordingly.
13550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13552 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13553 * cselib.c (cselib_hash_rtx): Likewise.
13554 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13555 CONST_VECTOR encoding.
13557 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13558 Jeff Law <law@redhat.com>
13561 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13562 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13563 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13564 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13567 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13568 explicitly probe *sp in a noreturn function if there were any callee
13569 register saves or frame pointer is needed.
13571 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13574 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13575 BLKmode for ternary, binary or unary expressions.
13578 * var-tracking.c (delete_vta_debug_insn): New inline function.
13579 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13580 insns from get_insns () to NULL instead of each bb separately.
13581 Use delete_vta_debug_insn. No longer static.
13582 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13583 delete_vta_debug_insns callers.
13584 * rtl.h (delete_vta_debug_insns): Declare.
13585 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13586 instead of variable_tracking_main.
13588 2018-01-03 Martin Sebor <msebor@redhat.com>
13590 PR tree-optimization/83603
13591 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13592 arguments past the endof the argument list in functions declared
13593 without a prototype.
13594 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13595 Avoid checking when arguments are null.
13597 2018-01-03 Martin Sebor <msebor@redhat.com>
13600 * doc/extend.texi (attribute const): Fix a typo.
13601 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13602 issuing -Wsuggest-attribute for void functions.
13604 2018-01-03 Martin Sebor <msebor@redhat.com>
13606 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13607 offset_int::from instead of wide_int::to_shwi.
13608 (maybe_diag_overlap): Remove assertion.
13609 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13610 * gimple-ssa-sprintf.c (format_directive): Same.
13611 (parse_directive): Same.
13612 (sprintf_dom_walker::compute_format_length): Same.
13613 (try_substitute_return_value): Same.
13615 2018-01-03 Jeff Law <law@redhat.com>
13617 PR middle-end/83654
13618 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13619 non-constant residual for zero at runtime and avoid probing in
13620 that case. Reorganize code for trailing problem to mirror handling
13623 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13625 PR tree-optimization/83501
13626 * tree-ssa-strlen.c (get_string_cst): New.
13627 (handle_char_store): Call get_string_cst.
13629 2018-01-03 Martin Liska <mliska@suse.cz>
13631 PR tree-optimization/83593
13632 * tree-ssa-strlen.c: Include tree-cfg.h.
13633 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13634 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13635 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13637 (strlen_dom_walker::before_dom_children): Call
13638 gimple_purge_dead_eh_edges. Dump tranformation with details
13640 (strlen_dom_walker::before_dom_children): Update call by adding
13641 new argument cleanup_eh.
13642 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13644 2018-01-03 Martin Liska <mliska@suse.cz>
13647 * cif-code.def (VARIADIC_THUNK): New enum value.
13648 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13651 2018-01-03 Jan Beulich <jbeulich@suse.com>
13653 * sse.md (mov<mode>_internal): Tighten condition for when to use
13654 vmovdqu<ssescalarsize> for TI and OI modes.
13656 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13658 Update copyright years.
13660 2018-01-03 Martin Liska <mliska@suse.cz>
13663 * ipa-visibility.c (function_and_variable_visibility): Skip
13664 functions with noipa attribure.
13666 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13668 * gcc.c (process_command): Update copyright notice dates.
13669 * gcov-dump.c (print_version): Ditto.
13670 * gcov.c (print_version): Ditto.
13671 * gcov-tool.c (print_version): Ditto.
13672 * gengtype.c (create_file): Ditto.
13673 * doc/cpp.texi: Bump @copying's copyright year.
13674 * doc/cppinternals.texi: Ditto.
13675 * doc/gcc.texi: Ditto.
13676 * doc/gccint.texi: Ditto.
13677 * doc/gcov.texi: Ditto.
13678 * doc/install.texi: Ditto.
13679 * doc/invoke.texi: Ditto.
13681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13683 * vector-builder.h (vector_builder::m_full_nelts): Change from
13684 unsigned int to poly_uint64.
13685 (vector_builder::full_nelts): Update prototype accordingly.
13686 (vector_builder::new_vector): Likewise.
13687 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13688 (vector_builder::operator ==): Likewise.
13689 (vector_builder::finalize): Likewise.
13690 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13691 Take the number of elements as a poly_uint64 rather than an
13693 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13694 from unsigned int to poly_uint64.
13695 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13696 (vec_perm_indices::new_vector): Likewise.
13697 (vec_perm_indices::length): Likewise.
13698 (vec_perm_indices::nelts_per_input): Likewise.
13699 (vec_perm_indices::input_nelts): Likewise.
13700 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13701 number of elements per input as a poly_uint64 rather than an
13702 unsigned int. Use the original encoding for variable-length
13703 vectors, rather than clamping each individual element.
13704 For the second and subsequent elements in each pattern,
13705 clamp the step and base before clamping their sum.
13706 (vec_perm_indices::series_p): Handle polynomial element counts.
13707 (vec_perm_indices::all_in_range_p): Likewise.
13708 (vec_perm_indices_to_tree): Likewise.
13709 (vec_perm_indices_to_rtx): Likewise.
13710 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13711 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13712 (tree_vector_builder::new_binary_operation): Handle polynomial
13713 element counts. Return false if we need to know the number
13714 of elements at compile time.
13715 * fold-const.c (fold_vec_perm): Punt if the number of elements
13716 isn't known at compile time.
13718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13720 * vec-perm-indices.h (vec_perm_builder): Change element type
13721 from HOST_WIDE_INT to poly_int64.
13722 (vec_perm_indices::element_type): Update accordingly.
13723 (vec_perm_indices::clamp): Handle polynomial element_types.
13724 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13725 (vec_perm_indices::all_in_range_p): Likewise.
13726 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13728 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13729 polynomial vec_perm_indices element types.
13730 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13731 * fold-const.c (fold_vec_perm): Likewise.
13732 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13733 * tree-vect-generic.c (lower_vec_perm): Likewise.
13734 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13735 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13736 element type to HOST_WIDE_INT.
13738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 Alan Hayward <alan.hayward@arm.com>
13740 David Sherwood <david.sherwood@arm.com>
13742 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13743 rather than an int. Use plus_constant.
13744 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13745 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13752 a HOST_WIDE_INT to a poly_int64.
13754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13755 Alan Hayward <alan.hayward@arm.com>
13756 David Sherwood <david.sherwood@arm.com>
13758 * calls.c (load_register_parameters): Cope with polynomial
13759 mode sizes. Require a constant size for BLKmode parameters
13760 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13761 forces a parameter to be padded at the lsb end in order to
13762 fill a complete number of words, require the parameter size
13763 to be ordered wrt UNITS_PER_WORD.
13765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13766 Alan Hayward <alan.hayward@arm.com>
13767 David Sherwood <david.sherwood@arm.com>
13769 * reload1.c (spill_stack_slot_width): Change element type
13770 from unsigned int to poly_uint64_pod.
13771 (alter_reg): Treat mode sizes as polynomial.
13773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13774 Alan Hayward <alan.hayward@arm.com>
13775 David Sherwood <david.sherwood@arm.com>
13777 * reload.c (complex_word_subreg_p): New function.
13778 (reload_inner_reg_of_subreg, push_reload): Use it.
13780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13781 Alan Hayward <alan.hayward@arm.com>
13782 David Sherwood <david.sherwood@arm.com>
13784 * lra-constraints.c (process_alt_operands): Reject matched
13785 operands whose sizes aren't ordered.
13786 (match_reload): Refer to this check here.
13788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13789 Alan Hayward <alan.hayward@arm.com>
13790 David Sherwood <david.sherwood@arm.com>
13792 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13793 that the mode size is in the set {1, 2, 4, 8, 16}.
13795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13796 Alan Hayward <alan.hayward@arm.com>
13797 David Sherwood <david.sherwood@arm.com>
13799 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13800 Use plus_constant instead of gen_rtx_PLUS.
13802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13803 Alan Hayward <alan.hayward@arm.com>
13804 David Sherwood <david.sherwood@arm.com>
13806 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13807 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13808 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13809 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13810 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13811 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13812 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13813 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13814 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13815 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13817 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13818 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13819 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13820 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13821 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13822 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13823 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13824 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13825 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13826 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13828 * expr.c (emit_move_resolve_push): Treat the input and result
13829 of PUSH_ROUNDING as a poly_int64.
13830 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13831 (emit_push_insn): Likewise.
13832 * lra-eliminations.c (mark_not_eliminable): Likewise.
13833 * recog.c (push_operand): Likewise.
13834 * reload1.c (elimination_effects): Likewise.
13835 * rtlanal.c (nonzero_bits1): Likewise.
13836 * calls.c (store_one_arg): Likewise. Require the padding to be
13837 known at compile time.
13839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13840 Alan Hayward <alan.hayward@arm.com>
13841 David Sherwood <david.sherwood@arm.com>
13843 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13844 Use plus_constant instead of gen_rtx_PLUS.
13846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13847 Alan Hayward <alan.hayward@arm.com>
13848 David Sherwood <david.sherwood@arm.com>
13850 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13851 rather than an int.
13853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13854 Alan Hayward <alan.hayward@arm.com>
13855 David Sherwood <david.sherwood@arm.com>
13857 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13858 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13859 via stack temporaries. Treat the mode size as polynomial too.
13861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13862 Alan Hayward <alan.hayward@arm.com>
13863 David Sherwood <david.sherwood@arm.com>
13865 * expr.c (expand_expr_real_2): When handling conversions involving
13866 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13867 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13868 as a poly_uint64 too.
13870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13871 Alan Hayward <alan.hayward@arm.com>
13872 David Sherwood <david.sherwood@arm.com>
13874 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13877 Alan Hayward <alan.hayward@arm.com>
13878 David Sherwood <david.sherwood@arm.com>
13880 * combine.c (can_change_dest_mode): Handle polynomial
13881 REGMODE_NATURAL_SIZE.
13882 * expmed.c (store_bit_field_1): Likewise.
13883 * expr.c (store_constructor): Likewise.
13884 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13885 and polynomial REGMODE_NATURAL_SIZE.
13886 (gen_lowpart_common): Likewise.
13887 * reginfo.c (record_subregs_of_mode): Likewise.
13888 * rtlanal.c (read_modify_subreg_p): Likewise.
13890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13895 numbers of elements.
13897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13898 Alan Hayward <alan.hayward@arm.com>
13899 David Sherwood <david.sherwood@arm.com>
13901 * match.pd: Cope with polynomial numbers of vector elements.
13903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13904 Alan Hayward <alan.hayward@arm.com>
13905 David Sherwood <david.sherwood@arm.com>
13907 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13908 in a POINTER_PLUS_EXPR.
13910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13911 Alan Hayward <alan.hayward@arm.com>
13912 David Sherwood <david.sherwood@arm.com>
13914 * omp-simd-clone.c (simd_clone_subparts): New function.
13915 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13916 (ipa_simd_modify_function_body): Likewise.
13918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13919 Alan Hayward <alan.hayward@arm.com>
13920 David Sherwood <david.sherwood@arm.com>
13922 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13923 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13924 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13925 (expand_vector_condition, vector_element): Likewise.
13926 (subparts_gt): New function.
13927 (get_compute_type): Use subparts_gt.
13928 (count_type_subparts): Delete.
13929 (expand_vector_operations_1): Use subparts_gt instead of
13930 count_type_subparts.
13932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13937 (vect_compile_time_alias): ...this new function. Do the calculation
13938 on poly_ints rather than trees.
13939 (vect_prune_runtime_alias_test_list): Update call accordingly.
13941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13942 Alan Hayward <alan.hayward@arm.com>
13943 David Sherwood <david.sherwood@arm.com>
13945 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13947 (vect_schedule_slp_instance): Likewise.
13949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13950 Alan Hayward <alan.hayward@arm.com>
13951 David Sherwood <david.sherwood@arm.com>
13953 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13954 constant and extern definitions for variable-length vectors.
13955 (vect_get_constant_vectors): Note that the number of units
13956 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13963 of units as polynomial. Choose between WIDE and NARROW based
13966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13967 Alan Hayward <alan.hayward@arm.com>
13968 David Sherwood <david.sherwood@arm.com>
13970 * tree-vect-stmts.c (simd_clone_subparts): New function.
13971 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13974 Alan Hayward <alan.hayward@arm.com>
13975 David Sherwood <david.sherwood@arm.com>
13977 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13978 vectors as polynomial. Use build_index_vector for
13979 IFN_GOMP_SIMD_LANE.
13981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13982 Alan Hayward <alan.hayward@arm.com>
13983 David Sherwood <david.sherwood@arm.com>
13985 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13986 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13987 for variable-length vectors.
13988 (vectorizable_mask_load_store): Treat the number of units as
13989 polynomial, asserting that it is constant if the condition has
13990 already been enforced.
13991 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
13998 of units as polynomial. Punt if we can't tell at compile time
13999 which vector contains the final result.
14001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14002 Alan Hayward <alan.hayward@arm.com>
14003 David Sherwood <david.sherwood@arm.com>
14005 * tree-vect-loop.c (vectorizable_induction): Treat the number
14006 of units as polynomial. Punt on SLP inductions. Use an integer
14007 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14008 cast of such a series for variable-length floating-point
14011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14012 Alan Hayward <alan.hayward@arm.com>
14013 David Sherwood <david.sherwood@arm.com>
14015 * tree.h (build_index_vector): Declare.
14016 * tree.c (build_index_vector): New function.
14017 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14018 of units as polynomial, forcibly converting it to a constant if
14019 vectorizable_reduction has already enforced the condition.
14020 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14021 to create a {1,2,3,...} vector.
14022 (vectorizable_reduction): Treat the number of units as polynomial.
14023 Choose vectype_in based on the largest scalar element size rather
14024 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
14032 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14039 * target.def (autovectorize_vector_sizes): Return the vector sizes
14040 by pointer, using vector_sizes rather than a bitmask.
14041 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14042 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14043 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14045 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14046 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14047 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14048 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14049 * omp-general.c (omp_max_vf): Likewise.
14050 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14051 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14052 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14053 * tree-vect-slp.c (vect_slp_bb): Likewise.
14054 * doc/tm.texi: Regenerate.
14055 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14057 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14058 the vector size as a poly_uint64 rather than an unsigned int.
14059 (current_vector_size): Change from an unsigned int to a poly_uint64.
14060 (get_vectype_for_scalar_type): Update accordingly.
14061 * tree.h (build_truth_vector_type): Take the size and number of
14062 units as a poly_uint64 rather than an unsigned int.
14063 (build_vector_type): Add a temporary overload that takes
14064 the number of units as a poly_uint64 rather than an unsigned int.
14065 * tree.c (make_vector_type): Likewise.
14066 (build_truth_vector_type): Take the number of units as a poly_uint64
14067 rather than an unsigned int.
14069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14070 Alan Hayward <alan.hayward@arm.com>
14071 David Sherwood <david.sherwood@arm.com>
14073 * target.def (get_mask_mode): Take the number of units and length
14074 as poly_uint64s rather than unsigned ints.
14075 * targhooks.h (default_get_mask_mode): Update accordingly.
14076 * targhooks.c (default_get_mask_mode): Likewise.
14077 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14078 * doc/tm.texi: Regenerate.
14080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14081 Alan Hayward <alan.hayward@arm.com>
14082 David Sherwood <david.sherwood@arm.com>
14084 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14085 * omp-general.c (omp_max_vf): Likewise.
14086 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14087 (expand_omp_simd): Handle polynomial safelen.
14088 * omp-low.c (omplow_simd_context): Add a default constructor.
14089 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14090 (lower_rec_simd_input_clauses): Update accordingly.
14091 (lower_rec_input_clauses): Likewise.
14093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14094 Alan Hayward <alan.hayward@arm.com>
14095 David Sherwood <david.sherwood@arm.com>
14097 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14098 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14099 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14100 (vect_analyze_slp_cost): Likewise.
14101 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14102 (vect_model_load_cost): Likewise.
14104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14105 Alan Hayward <alan.hayward@arm.com>
14106 David Sherwood <david.sherwood@arm.com>
14108 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14109 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14110 from an unsigned int * to a poly_uint64_pod *.
14111 (calculate_unrolling_factor): New function.
14112 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14115 Alan Hayward <alan.hayward@arm.com>
14116 David Sherwood <david.sherwood@arm.com>
14118 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14119 from an unsigned int to a poly_uint64.
14120 (_loop_vec_info::slp_unrolling_factor): Likewise.
14121 (_loop_vec_info::vectorization_factor): Change from an int
14123 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14124 (vect_get_num_vectors): New function.
14125 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14126 (vect_get_num_copies): Use vect_get_num_vectors.
14127 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14128 to an unsigned int *.
14129 (vect_analyze_data_refs): Change min_vf from an int * to a
14131 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14132 than an unsigned HOST_WIDE_INT.
14133 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14134 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14135 to an unsigned int *.
14136 (vect_analyze_data_ref_dependences): Likewise.
14137 (vect_compute_data_ref_alignment): Handle polynomial vf.
14138 (vect_enhance_data_refs_alignment): Likewise.
14139 (vect_prune_runtime_alias_test_list): Likewise.
14140 (vect_shift_permute_load_chain): Likewise.
14141 (vect_supportable_dr_alignment): Likewise.
14142 (dependence_distance_ge_vf): Take the vectorization factor as a
14143 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14144 (vect_analyze_data_refs): Change min_vf from an int * to a
14146 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14147 vfm1 as a poly_uint64 rather than an int. Make the same change
14148 for the returned bound_scalar.
14149 (vect_gen_vector_loop_niters): Handle polynomial vf.
14150 (vect_do_peeling): Likewise. Update call to
14151 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14152 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14154 * tree-vect-loop.c (vect_determine_vectorization_factor)
14155 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14156 (vect_get_known_peeling_cost): Likewise.
14157 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14158 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14159 (vect_transform_loop): Likewise. Use the lowest possible VF when
14160 updating the upper bounds of the loop.
14161 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14162 rather than an int.
14163 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14164 polynomial unroll factors.
14165 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14166 (vect_make_slp_decision): Likewise.
14167 (vect_supported_load_permutation_p): Likewise, and polynomial
14169 (vect_analyze_slp_cost): Handle polynomial vf.
14170 (vect_slp_analyze_node_operations): Likewise.
14171 (vect_slp_analyze_bb_1): Likewise.
14172 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14173 than an unsigned HOST_WIDE_INT.
14174 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14175 (vectorizable_load): Handle polynomial vf.
14176 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14178 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14181 Alan Hayward <alan.hayward@arm.com>
14182 David Sherwood <david.sherwood@arm.com>
14184 * match.pd: Handle bit operations involving three constants
14185 and try to fold one pair.
14187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14189 * tree-vect-loop-manip.c: Include gimple-fold.h.
14190 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14191 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14192 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14193 Add a path that uses a step of VF instead of 1, but disable it
14195 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14196 and niters_no_overflow parameters. Update calls to
14197 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14198 Create a new SSA name if the latter choses to use a ste other
14199 than zero, and return it via niters_vector_mult_vf_var.
14200 * tree-vect-loop.c (vect_transform_loop): Update calls to
14201 vect_do_peeling, vect_gen_vector_loop_niters and
14202 slpeel_make_loop_iterate_ntimes.
14203 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14204 (vect_gen_vector_loop_niters): Update declarations after above changes.
14206 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14208 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14209 128-bit round to integer instructions.
14210 (ceil<mode>2): Likewise.
14211 (btrunc<mode>2): Likewise.
14212 (round<mode>2): Likewise.
14214 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14216 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14217 unaligned VSX load/store on P8/P9.
14218 (expand_block_clear): Allow the use of unaligned VSX
14219 load/store on P8/P9.
14221 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14223 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14225 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14226 swap associated with both a load and a store.
14228 2018-01-02 Andrew Waterman <andrew@sifive.com>
14230 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14231 * config/riscv/riscv.md (clear_cache): Use it.
14233 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14235 * web.c: Remove out-of-date comment.
14237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14239 * expr.c (fixup_args_size_notes): Check that any existing
14240 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14241 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14242 (emit_single_push_insn): ...here.
14244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14246 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14247 (const_vector_encoded_nelts): New function.
14248 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14249 (const_vector_int_elt, const_vector_elt): Declare.
14250 * emit-rtl.c (const_vector_int_elt_1): New function.
14251 (const_vector_elt): Likewise.
14252 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14253 of CONST_VECTOR_ELT.
14255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14257 * expr.c: Include rtx-vector-builder.h.
14258 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14259 directly on the tree encoding.
14260 (const_vector_from_tree): Likewise.
14261 * optabs.c: Include rtx-vector-builder.h.
14262 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14263 sequence of "u" values.
14264 * vec-perm-indices.c: Include rtx-vector-builder.h.
14265 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14266 directly on the vec_perm_indices encoding.
14268 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14270 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14271 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14272 * rtx-vector-builder.h: New file.
14273 * rtx-vector-builder.c: Likewise.
14274 * rtl.h (rtx_def::u2): Add a const_vector field.
14275 (CONST_VECTOR_NPATTERNS): New macro.
14276 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14277 (CONST_VECTOR_DUPLICATE_P): Likewise.
14278 (CONST_VECTOR_STEPPED_P): Likewise.
14279 (CONST_VECTOR_ENCODED_ELT): Likewise.
14280 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14281 (unwrap_const_vec_duplicate): Likewise.
14282 (const_vec_series_p): Check for a non-duplicated vector encoding.
14283 Say that the function only returns true for integer vectors.
14284 * emit-rtl.c: Include rtx-vector-builder.h.
14285 (gen_const_vec_duplicate_1): Delete.
14286 (gen_const_vector): Call gen_const_vec_duplicate instead of
14287 gen_const_vec_duplicate_1.
14288 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14289 (gen_const_vec_duplicate): Use rtx_vector_builder.
14290 (gen_const_vec_series): Likewise.
14291 (gen_rtx_CONST_VECTOR): Likewise.
14292 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14293 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14294 Build a new vector rather than modifying a CONST_VECTOR in-place.
14295 (handle_special_swappables): Update call accordingly.
14296 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14297 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14298 Build a new vector rather than modifying a CONST_VECTOR in-place.
14299 (handle_special_swappables): Update call accordingly.
14301 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14303 * simplify-rtx.c (simplify_const_binary_operation): Use
14304 CONST_VECTOR_ELT instead of XVECEXP.
14306 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14308 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14309 the selector elements to be different from the data elements
14310 if the selector is a VECTOR_CST.
14311 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14312 ssizetype for the selector.
14314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14316 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14317 before testing each element individually.
14318 * tree-vect-generic.c (lower_vec_perm): Likewise.
14320 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14322 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14323 * selftest-run-tests.c (selftest::run_tests): Call it.
14324 * vector-builder.h (vector_builder::operator ==): New function.
14325 (vector_builder::operator !=): Likewise.
14326 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14327 (vec_perm_indices::all_from_input_p): New function.
14328 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14329 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14330 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14331 instead of reading the VECTOR_CST directly. Detect whether both
14332 vector inputs are the same before constructing the vec_perm_indices,
14333 and update the number of inputs argument accordingly. Use the
14334 utility functions added above. Only construct sel2 if we need to.
14336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14338 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14339 the broadcast of the low byte.
14340 (expand_mult_highpart): Use an explicit encoding for the permutes.
14341 * optabs-query.c (can_mult_highpart_p): Likewise.
14342 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14343 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14344 (vectorizable_bswap): Likewise.
14345 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14346 explicit encoding for the power-of-2 permutes.
14347 (vect_permute_store_chain): Likewise.
14348 (vect_grouped_load_supported): Likewise.
14349 (vect_permute_load_chain): Likewise.
14351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14353 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14354 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14355 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14356 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14357 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14358 (vect_gen_perm_mask_any): Likewise.
14360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14362 * int-vector-builder.h: New file.
14363 * vec-perm-indices.h: Include int-vector-builder.h.
14364 (vec_perm_indices): Redefine as an int_vector_builder.
14365 (auto_vec_perm_indices): Delete.
14366 (vec_perm_builder): Redefine as a stand-alone class.
14367 (vec_perm_indices::vec_perm_indices): New function.
14368 (vec_perm_indices::clamp): Likewise.
14369 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14370 (vec_perm_indices::new_vector): New function.
14371 (vec_perm_indices::new_expanded_vector): Update for new
14372 vec_perm_indices class.
14373 (vec_perm_indices::rotate_inputs): New function.
14374 (vec_perm_indices::all_in_range_p): Operate directly on the
14375 encoded form, without computing elided elements.
14376 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14377 encoding. Update for new vec_perm_indices class.
14378 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14379 the given vec_perm_builder.
14380 (expand_vec_perm_var): Update vec_perm_builder constructor.
14381 (expand_mult_highpart): Use vec_perm_builder instead of
14382 auto_vec_perm_indices.
14383 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14384 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14385 or double series encoding as appropriate.
14386 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14387 vec_perm_indices instead of auto_vec_perm_indices.
14388 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14389 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14390 (vect_permute_store_chain): Likewise.
14391 (vect_grouped_load_supported): Likewise.
14392 (vect_permute_load_chain): Likewise.
14393 (vect_shift_permute_load_chain): Likewise.
14394 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14395 (vect_transform_slp_perm_load): Likewise.
14396 (vect_schedule_slp_instance): Likewise.
14397 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14398 (vectorizable_mask_load_store): Likewise.
14399 (vectorizable_bswap): Likewise.
14400 (vectorizable_store): Likewise.
14401 (vectorizable_load): Likewise.
14402 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14403 vec_perm_indices instead of auto_vec_perm_indices. Use
14404 tree_to_vec_perm_builder to read the vector from a tree.
14405 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14406 vec_perm_builder instead of a vec_perm_indices.
14407 (have_whole_vector_shift): Use vec_perm_builder and
14408 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14409 truncation to calc_vec_perm_mask_for_shift.
14410 (vect_create_epilog_for_reduction): Likewise.
14411 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14412 from auto_vec_perm_indices to vec_perm_indices.
14413 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14414 instead of changing individual elements.
14415 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14416 the vector in d.perm.
14417 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14418 from auto_vec_perm_indices to vec_perm_indices.
14419 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14420 instead of changing individual elements.
14421 (arm_vectorize_vec_perm_const): Use new_vector to install
14422 the vector in d.perm.
14423 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14424 Update vec_perm_builder constructor.
14425 (rs6000_expand_interleave): Likewise.
14426 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14427 (rs6000_expand_interleave): Likewise.
14429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14431 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14432 to qimode could truncate the indices.
14433 * optabs.c (expand_vec_perm_var): Likewise.
14435 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14437 * Makefile.in (OBJS): Add vec-perm-indices.o.
14438 * vec-perm-indices.h: New file.
14439 * vec-perm-indices.c: Likewise.
14440 * target.h (vec_perm_indices): Replace with a forward class
14442 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14443 * optabs.h: Include vec-perm-indices.h.
14444 (expand_vec_perm): Delete.
14445 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14446 (expand_vec_perm_const): Declare.
14447 * target.def (vec_perm_const_ok): Replace with...
14448 (vec_perm_const): ...this new hook.
14449 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14450 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14451 * doc/tm.texi: Regenerate.
14452 * optabs.def (vec_perm_const): Delete.
14453 * doc/md.texi (vec_perm_const): Likewise.
14454 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14455 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14456 expand_vec_perm for constant permutation vectors. Assert that
14457 the mode of variable permutation vectors is the integer equivalent
14458 of the mode that is being permuted.
14459 * optabs-query.h (selector_fits_mode_p): Declare.
14460 * optabs-query.c: Include vec-perm-indices.h.
14461 (selector_fits_mode_p): New function.
14462 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14463 is defined, instead of checking whether the vec_perm_const_optab
14464 exists. Use targetm.vectorize.vec_perm_const instead of
14465 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14466 fit in the vector mode before using a variable permute.
14467 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14468 vec_perm_indices instead of an rtx.
14469 (expand_vec_perm): Replace with...
14470 (expand_vec_perm_const): ...this new function. Take the selector
14471 as a vec_perm_indices rather than an rtx. Also take the mode of
14472 the selector. Update call to shift_amt_for_vec_perm_mask.
14473 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14474 Use vec_perm_indices::new_expanded_vector to expand the original
14475 selector into bytes. Check whether the indices fit in the vector
14476 mode before using a variable permute.
14477 (expand_vec_perm_var): Make global.
14478 (expand_mult_highpart): Use expand_vec_perm_const.
14479 * fold-const.c: Includes vec-perm-indices.h.
14480 * tree-ssa-forwprop.c: Likewise.
14481 * tree-vect-data-refs.c: Likewise.
14482 * tree-vect-generic.c: Likewise.
14483 * tree-vect-loop.c: Likewise.
14484 * tree-vect-slp.c: Likewise.
14485 * tree-vect-stmts.c: Likewise.
14486 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14488 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14489 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14490 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14491 (aarch64_vectorize_vec_perm_const): ...this new function.
14492 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14493 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14494 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14495 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14496 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14497 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14498 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14500 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14501 check for NEON modes.
14502 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14503 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14504 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14505 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14507 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14508 the old VEC_PERM_CONST conditions.
14509 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14510 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14511 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14512 (ia64_vectorize_vec_perm_const_ok): Merge into...
14513 (ia64_vectorize_vec_perm_const): ...this new function.
14514 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14515 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14516 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14517 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14518 * config/mips/mips.c (mips_expand_vec_perm_const)
14519 (mips_vectorize_vec_perm_const_ok): Merge into...
14520 (mips_vectorize_vec_perm_const): ...this new function.
14521 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14522 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14523 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14524 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14525 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14526 (rs6000_expand_vec_perm_const): Delete.
14527 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14529 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14530 (altivec_expand_vec_perm_const_le): Take each operand individually.
14531 Operate on constant selectors rather than rtxes.
14532 (altivec_expand_vec_perm_const): Likewise. Update call to
14533 altivec_expand_vec_perm_const_le.
14534 (rs6000_expand_vec_perm_const): Delete.
14535 (rs6000_vectorize_vec_perm_const_ok): Delete.
14536 (rs6000_vectorize_vec_perm_const): New function.
14537 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14538 an element count and rtx array.
14539 (rs6000_expand_extract_even): Update call accordingly.
14540 (rs6000_expand_interleave): Likewise.
14541 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14542 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14543 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14544 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14545 (rs6000_expand_vec_perm_const): Delete.
14546 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14547 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14548 (altivec_expand_vec_perm_const_le): Take each operand individually.
14549 Operate on constant selectors rather than rtxes.
14550 (altivec_expand_vec_perm_const): Likewise. Update call to
14551 altivec_expand_vec_perm_const_le.
14552 (rs6000_expand_vec_perm_const): Delete.
14553 (rs6000_vectorize_vec_perm_const_ok): Delete.
14554 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14555 reference to the SPE evmerge intructions.
14556 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14557 an element count and rtx array.
14558 (rs6000_expand_extract_even): Update call accordingly.
14559 (rs6000_expand_interleave): Likewise.
14560 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14561 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14563 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14567 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14568 vector mode and that that mode matches the mode of the data
14570 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14571 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14572 directly using expand_vec_perm_1 when forcing selectors into
14574 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14576 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14578 * optabs-query.h (can_vec_perm_p): Delete.
14579 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14580 * optabs-query.c (can_vec_perm_p): Split into...
14581 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14582 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14583 particular selector is valid.
14584 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14585 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14586 (vect_grouped_load_supported): Likewise.
14587 (vect_shift_permute_load_chain): Likewise.
14588 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14589 (vect_transform_slp_perm_load): Likewise.
14590 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14591 (vectorizable_bswap): Likewise.
14592 (vect_gen_perm_mask_checked): Likewise.
14593 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14594 implementations of variable permutation vectors into account
14595 when deciding which selector to use.
14596 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14597 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14598 with a false third argument.
14599 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14600 to test whether the constant selector is valid and can_vec_perm_var_p
14601 to test whether a variable selector is valid.
14603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14605 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14606 * optabs-query.c (can_vec_perm_p): Likewise.
14607 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14608 instead of vec_perm_indices.
14609 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14610 (vect_gen_perm_mask_checked): Likewise,
14611 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14612 (vect_gen_perm_mask_checked): Likewise,
14614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14616 * optabs-query.h (qimode_for_vec_perm): Declare.
14617 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14618 (qimode_for_vec_perm): ...this new function.
14619 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14621 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14623 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14624 does not have a conditional at the top.
14626 2018-01-02 Richard Biener <rguenther@suse.de>
14628 * ipa-inline.c (big_speedup_p): Fix expression.
14630 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14633 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14636 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14640 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14641 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14642 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14643 cond_taken_branch_cost 3->4.
14645 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14647 PR tree-optimization/83581
14648 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14649 TODO_cleanup_cfg if any changes have been made.
14651 PR middle-end/83608
14652 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14653 convert_modes if target mode has the right side, but different mode
14656 PR middle-end/83609
14657 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14658 last argument when extracting from CONCAT. If either from_real or
14659 from_imag is NULL, use expansion through memory. If result is not
14660 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14661 the parts directly to inner mode, if even that fails, use expansion
14664 PR middle-end/83623
14665 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14666 check for bswap in mode rather than HImode and use that in expand_unop
14669 Copyright (C) 2018 Free Software Foundation, Inc.
14671 Copying and distribution of this file, with or without modification,
14672 are permitted in any medium without royalty provided the copyright
14673 notice and this notice are preserved.