| blob | 1f48794ee74c031a68770a8b884f8293e984f65b |
1 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3 PR tree-optimization/84969
4 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5 builtin memset partitions if they set different rhs values.
7 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9 PR rtl-optimization/84989
10 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11 VEC_DUPLICATE with scalar result mode.
13 2018-03-21 Martin Liska <mliska@suse.cz>
15 PR ipa/84963
16 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17 not intended return statement.
19 2018-03-21 Martin Liska <mliska@suse.cz>
21 PR target/84988
22 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23 (chkp_find_bound_slots_1): Limit number of iterations.
25 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
27 PR target/84838
28 * Minor grammar fixes for x86 options.
30 2018-03-20 Jakub Jelinek <jakub@redhat.com>
32 PR debug/84875
33 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
34 holding REG_CFA_RESTORE notes, instead turn them into a USE.
36 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
38 PR target/83789
39 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
40 (altivec_lvx_<mode>_1op): Likewise.
41 (altivec_stvx_<mode>_2op): Likewise.
42 (altivec_stvx_<mode>_1op): Likewise.
43 (altivec_lvx_<VM2:mode>): New define_expand.
44 (altivec_stvx_<VM2:mode>): Likewise.
45 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
46 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
47 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
48 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
49 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
50 (rs6000_gen_lvx): Likewise.
51 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
52 (altivec_expand_stv_builtin): Likewise.
53 (altivec_expand_builtin): Likewise.
54 * config/rs6000/vector.md: Likewise.
56 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
58 PR target/82518
59 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
60 BYTES_BIG_ENDIAN.
62 2018-03-20 Richard Biener <rguenther@suse.de>
64 PR target/84986
65 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
66 sign-conversions as zero, fall back to standard scalar_stmt
67 cost for the rest.
69 2018-03-20 Martin Liska <mliska@suse.cz>
71 PR ipa/84825
72 * predict.c (rebuild_frequencies): Handle case when we have
73 PROFILE_ABSENT, but flag_guess_branch_prob is false.
75 2018-03-20 Jakub Jelinek <jakub@redhat.com>
77 PR target/84990
78 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
79 flag_section_anchors.
80 * varasm.c (use_blocks_for_decl_p): Remove hack for
81 dw2_force_const_mem.
83 PR target/84845
84 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
85 to ...
86 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
87 be created, use lowpart_subreg of operands[0] rather than operands[0]
88 itself.
89 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
90 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
91 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
92 and n constraint instead of aarch64_shift_imm_di and Usd.
93 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
94 (*aarch64_<optab>_reg_minus<mode>3): ... this.
96 2018-03-20 Sudakshina Das <sudi.das@arm.com>
98 PR target/82989
99 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
100 to favor GPR over NEON registers.
101 (<shift>di3_neon): Likewise.
103 2018-03-20 Tom de Vries <tom@codesourcery.com>
105 PR target/84952
106 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
107 (nvptx_process_pars): Emit bar.sync asap and alap.
109 2018-03-20 Tom de Vries <tom@codesourcery.com>
111 PR target/84954
112 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
113 seen_label if seen_label is already set.
115 2018-03-20 Jakub Jelinek <jakub@redhat.com>
117 PR target/84945
118 * config/i386/i386.c (fold_builtin_cpu): For features above 31
119 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
120 Use 1U instead of 1. Formatting fixes.
122 PR c/84953
123 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
124 instead of TREE_TYPE (s1) for the return value.
126 2018-03-19 Jakub Jelinek <jakub@redhat.com>
128 PR tree-optimization/84946
129 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
130 bitsize + bitsize in poly_uint64 rather than poly_int64.
132 PR sanitizer/78651
133 * dwarf2asm.c: Include fold-const.c.
134 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
135 of decl rather than decl itself.
137 PR rtl-optimization/84643
138 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
140 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
142 PR sanitizer/78651
143 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
144 calling assemble_variable.
146 2018-03-19 Sudakshina Das <sudi.das@arm.com>
148 PR target/81647
149 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
150 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
152 2018-03-19 Jim Wilson <jimw@sifive.com>
154 PR bootstrap/84856
155 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
156 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
157 (riscv_first_stack_step): Likewise.
158 (riscv_option_override): Use STACK_BOUNDARY instead of
159 MIN_STACK_BOUNDARY.
160 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
161 MIN_STACK_BOUNDARY.
162 (BIGGEST_ALIGNMENT): Set to 128.
163 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
164 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
165 STACK_BOUNDARY.
167 2018-03-19 Richard Biener <rguenther@suse.de>
169 PR tree-optimization/84933
170 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
171 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
173 2018-03-19 Richard Biener <rguenther@suse.de>
175 PR tree-optimization/84859
176 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
177 (cond_if_else_store_replacement): Perform sinking operation on
178 single-store BBs regardless of MAX_STORES_TO_SINK setting.
179 Generalize what a BB with a single eligible store is.
181 2018-03-19 Richard Biener <rguenther@suse.de>
183 PR tree-optimization/84929
184 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
185 chrec_is_positive against non-chrec arg.
187 2018-03-19 Tamar Christina <tamar.christina@arm.com>
189 PR target/84711
190 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
192 2018-03-18 Martin Liska <mliska@suse.cz>
194 PR rtl-optimization/84635
195 * regrename.c (build_def_use): Use matches_mode only when
196 matches >= 0.
198 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
200 PR tree-optimization/84913
201 * tree-vect-loop.c (vectorizable_reduction): Don't try to
202 vectorize chains of COND_EXPRs.
204 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
206 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
208 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
210 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
212 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
214 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
216 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
217 Kito Cheng <kito.cheng@gmail.com>
219 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
220 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
221 (nds32_adjust_reg_alloc_order): New function.
222 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
224 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
226 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
227 nds32_print_operand, nds32_print_operand_address): Use
228 HOST_WIDE_INT_PRINT_DEC instead.
230 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
234 2018-03-17 Jakub Jelinek <jakub@redhat.com>
236 PR target/84902
237 * config/i386/i386.c (initial_ix86_tune_features,
238 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
239 unsigned long long.
240 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
241 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
242 rather than 1u << ix86_tune. Formatting fix.
243 (ix86_option_override_internal): Change ix86_arch_mask from
244 unsigned int to unsigned HOST_WIDE_INT, initialize to
245 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
246 (ix86_function_specific_restore): Likewise.
248 2018-03-16 Jakub Jelinek <jakub@redhat.com>
250 PR target/84899
251 * postreload.c (reload_combine_recognize_pattern): Perform
252 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
253 truncate_int_for_mode the result for the destination's mode.
255 PR c/84909
256 * hsa-gen.c (mem_type_for_type): Fix comment typo.
257 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
258 Likewise.
259 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
260 Likewise.
262 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
264 PR target/84876
265 * lra-assigns.c (lra_split_hard_reg_for): Don't use
266 regno_allocno_class_array and sorted_pseudos.
267 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
268 insns where regno is used.
270 2018-03-16 Martin Liska <mliska@suse.cz>
272 PR ipa/84833
273 * multiple_target.c (create_dispatcher_calls): Redirect
274 reference in the symbol table.
276 2018-03-16 Martin Liska <mliska@suse.cz>
278 PR ipa/84722
279 * multiple_target.c (create_dispatcher_calls): Redirect also
280 an alias.
282 2018-03-16 Jakub Jelinek <jakub@redhat.com>
284 PR c++/79937
285 PR c++/82410
286 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
287 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
288 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
290 2018-03-16 Julia Koval <julia.koval@intel.com>
292 * doc/invoke.texi (Skylake Server): Add CLWB.
293 Cannonlake): Remove CLWB.
295 2018-03-16 Jakub Jelinek <jakub@redhat.com>
297 PR tree-optimization/84841
298 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
299 1 << 3.
300 (FLOAT_ONE_CONST_TYPE): Define.
301 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
302 (sort_by_operand_rank): Put entries with higher constant_type last
303 rather than first to match comments.
305 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
307 * config/nios2/nios2.md (movsi_internal): Fix thinko in
308 split predicate.
310 2018-03-15 Jakub Jelinek <jakub@redhat.com>
312 PR c++/79085
313 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
314 check and use address of target always.
316 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
318 PR target/84574
319 * config/i386/i386.c (indirect_thunk_needed): Update comments.
320 (indirect_thunk_bnd_needed): Likewise.
321 (indirect_thunks_used): Likewise.
322 (indirect_thunks_bnd_used): Likewise.
323 (indirect_return_needed): New.
324 (indirect_return_bnd_needed): Likewise.
325 (output_indirect_thunk_function): Add a bool argument for
326 function return.
327 (output_indirect_thunk_function): Don't generate alias for
328 function return thunk.
329 (ix86_code_end): Call output_indirect_thunk_function to generate
330 function return thunks.
331 (ix86_output_function_return): Set indirect_return_bnd_needed
332 and indirect_return_needed instead of indirect_thunk_bnd_needed
333 and indirect_thunk_needed.
335 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
337 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
338 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
339 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
341 2018-03-15 David Malcolm <dmalcolm@redhat.com>
342 Paul Hua <paul.hua.gm@gmail.com>
344 PR c/84852
345 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
347 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
349 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
350 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
351 resp. SFmode cases.
353 2018-03-15 Tamar Christina <tamar.christina@arm.com>
355 PR target/84711
356 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
357 instead of GET_MODE_SIZE when comparing Units.
359 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
361 PR target/68256
362 * varasm.c (hash_section): Return an unchangeble hash value
363 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
364 Return !aarch64_can_use_per_function_literal_pools_p ().
366 2018-03-15 Jakub Jelinek <jakub@redhat.com>
368 PR target/84860
369 * optabs.c (emit_conditional_move): Pass address of cmode's copy
370 rather than address of cmode as last argument to prepare_cmp_insn.
372 2018-03-15 Julia Koval <julia.koval@intel.com>
374 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
375 F_AVX512VNNI, F_AVX512BITALG): New.
377 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
379 PR target/83451
380 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
381 insn for floating-point loads and stores.
383 2018-03-14 Carl Love <cel@us.ibm.com>
385 * config/rs6000/rs6000-c.c: Add macro definitions for
386 ALTIVEC_BUILTIN_VEC_PERMXOR.
387 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
388 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
389 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
390 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
391 UNSPEC_VPERMXOR.
392 * config/doc/extend.texi: Add prototypes for vec_permxor.
394 2018-03-14 David Malcolm <dmalcolm@redhat.com>
396 PR c/84852
397 * diagnostic-show-locus.c (class layout_point): Convert m_line
398 from int to linenum_type.
399 (line_span::comparator): Use linenum "compare" function when
400 comparing line numbers.
401 (test_line_span): New function.
402 (layout_range::contains_point): Convert param "row" from int to
403 linenum_type.
404 (layout_range::intersects_line_p): Likewise.
405 (layout::will_show_line_p): Likewise.
406 (layout::print_source_line): Likewise.
407 (layout::should_print_annotation_line_p): Likewise.
408 (layout::print_annotation_line): Likewise.
409 (layout::print_leading_fixits): Likewise.
410 (layout::annotation_line_showed_range_p): Likewise.
411 (struct line_corrections): Likewise for field m_row.
412 (line_corrections::line_corrections): Likewise for param "row".
413 (layout::print_trailing_fixits): Likewise.
414 (layout::get_state_at_point): Likewise.
415 (layout::get_x_bound_for_row): Likewise.
416 (layout::print_line): Likewise.
417 (diagnostic_show_locus): Likewise for locals "last_line" and
418 "row".
419 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
420 * input.c (selftest::test_linenum_comparisons): New function.
421 (selftest::input_c_tests): Call it.
422 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
423 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
424 * selftest.h (ASSERT_GT): New macro.
425 (ASSERT_GT_AT): New macro.
426 (ASSERT_LT): New macro.
427 (ASSERT_LT_AT): New macro.
429 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
431 PR rtl-optimization/84780
432 * combine.c (distribute_links): Don't make a link based on pc_rtx.
434 2018-03-14 Martin Liska <mliska@suse.cz>
436 * tree.c (record_node_allocation_statistics): Use
437 get_stats_node_kind.
438 (get_stats_node_kind): New function extracted from
439 record_node_allocation_statistics.
440 (free_node): Use get_stats_node_kind.
442 2018-03-14 Richard Biener <rguenther@suse.de>
444 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
445 that the value-set of ANTIC_IN doesn't grow.
447 Revert
448 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
449 member.
450 (BB_VISITED_WITH_VISITED_SUCCS): New define.
451 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
453 2018-03-14 Julia Koval <julia.koval@intel.com>
455 * config.gcc (icelake-client, icelake-server): New.
456 (icelake): Remove.
457 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
458 (initial_ix86_arch_features): Ditto.
459 (PTA_SKYLAKE): Add SGX.
460 (PTA_ICELAKE): Remove.
461 (PTA_ICELAKE_CLIENT): New.
462 (PTA_ICELAKE_SERVER): New.
463 (ix86_option_override_internal): Split up icelake on icelake client and
464 icelake server.
465 (get_builtin_code_for_version): Ditto.
466 (fold_builtin_cpu): Ditto.
467 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
468 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
469 * config/i386/i386.h (processor_type): Ditto.
470 * doc/invoke.texi: Ditto.
472 2018-03-14 Jakub Jelinek <jakub@redhat.com>
474 PR sanitizer/83392
475 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
476 INTEGER_CST offset, add it together with bitpos / 8 and
477 sign extend based on POINTER_SIZE.
479 PR target/84844
480 Revert
481 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
483 PR target/78090
484 * config/i386/constraints.md (Yc): New register constraint.
485 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
486 Use Yc constraint for alternative 2 of operand 0. Remove
487 preferred_for_speed attribute.
489 2018-03-14 Richard Biener <rguenther@suse.de>
491 PR tree-optimization/84830
492 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
493 with the old one to avoid oscillations.
495 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
497 PR target/83712
498 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
499 pseudos.
500 (assign_by_spills): Return a flag of reload assignment failure.
501 Do not process the reload assignment failures. Do not spill other
502 reload pseudos if they has the same reg class. Update n if
503 necessary.
504 (lra_assign): Add a return arg. Set up from the result of
505 assign_by_spills call.
506 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
507 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
508 usage_insns if it is not NULL.
509 (spill_hard_reg_in_range): New function.
510 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
511 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
512 function prototypes.
513 (lra_assign): Change prototype.
514 * lra.c (lra): Add code to deal with fails by splitting hard reg
515 live ranges.
517 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
519 * config/riscv/riscv.opt (mrelax): New option.
520 * config/riscv/riscv.c (riscv_file_start): Emit ".option
521 "norelax" when riscv_mrelax is disabled.
522 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
524 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
526 PR target/84743
527 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
528 reassociation for int modes.
530 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
532 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
533 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
534 for big-endian.
535 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
536 * config/aarch64/aarch64-sve.md
537 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
538 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
539 (*extend<mode><Vwide>2): Rename to...
540 (aarch64_sve_extend<mode><Vwide>2): ...this.
541 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
542 renaming the old pattern to...
543 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
544 unsigned packs.
545 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
546 define_expand, renaming the old pattern to...
547 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
548 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
549 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
550 account when deciding which SVE instruction the optab should use.
551 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
553 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
555 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
556 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
557 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
558 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
559 (tlsdesc_small_<mode>): Turn a define_expand and use
560 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
561 (tlsdesc_small_advsimd_<mode>): ...this.
562 (tlsdesc_small_sve_<mode>): New pattern.
564 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
566 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
567 (UNSPEC_UMUL_HIGHPART): New constants.
568 (MUL_HIGHPART): New int iteraor.
569 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
570 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
571 define_expand.
572 (*<su>mul<mode>3_highpart): New define_insn.
574 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
576 PR lto/84805
577 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
578 incomplete types.
580 2018-03-13 Martin Liska <mliska@suse.cz>
582 PR ipa/84658.
583 * (sem_item_optimizer::sem_item_optimizer): Initialize new
584 vector.
585 (sem_item_optimizer::~sem_item_optimizer): Release it.
586 (sem_item_optimizer::merge_classes): Register variable aliases.
587 (sem_item_optimizer::fixup_pt_set): New function.
588 (sem_item_optimizer::fixup_points_to_sets): Likewise.
589 * ipa-icf.h: Declare new variables and functions.
591 2018-03-13 Jakub Jelinek <jakub@redhat.com>
593 PR middle-end/84834
594 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
595 integer_pow2p@2 and test integer_pow2p in condition.
596 (A < 0 ? C : 0): Similarly for @1.
598 PR middle-end/84831
599 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
600 characters starting at p contain '\0' character, don't look beyond
601 that.
603 PR target/84827
604 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
605 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
607 PR target/84828
608 * reg-stack.c (change_stack): Change update_end var from int to
609 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
610 also call set_block_for_insn on the newly added insns and rescan.
612 PR target/84786
613 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
614 on the last operand.
616 PR c++/84704
617 * tree.c (stabilize_reference_1): Return save_expr (e) for
618 STATEMENT_LIST even if it doesn't have side-effects.
620 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
622 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
624 2018-03-12 Renlin Li <renlin.li@arm.com>
626 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
627 aarch64_output_scalar_simd_mov_immediate.
629 2018-03-12 Martin Sebor <msebor@redhat.com>
631 PR tree-optimization/83456
632 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
633 for perfectly overlapping calls to memcpy.
634 (gimple_fold_builtin_memory_chk): Same.
635 (gimple_fold_builtin_strcpy): Handle no-warning.
636 (gimple_fold_builtin_stxcpy_chk): Same.
637 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
639 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
641 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
642 parameter. Use it for SFmode.
643 (rs6000_function_arg_advance_1): Adjust.
644 (rs6000_function_arg): Adjust.
645 (rs6000_gimplify_va_arg): Pass false for that new parameter.
647 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
649 PR rtl-optimization/84169
650 PR rtl-optimization/84780
651 * combine.c (can_combine_p): Check for a 2-insn combination whether
652 the destination register is used between the two insns, too.
654 2018-03-12 Richard Biener <rguenther@suse.de>
656 PR tree-optimization/84803
657 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
658 for refs DR analysis didn't process.
660 2018-03-12 Richard Biener <rguenther@suse.de>
662 PR tree-optimization/84777
663 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
664 force-vectorize loops ignore whether we are optimizing for size.
666 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
668 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
669 (TARGET_MD_ASM_ADJUST): Define.
671 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
672 Kito Cheng <kito.cheng@gmail.com>
673 Chung-Ju Wu <jasonwucj@gmail.com>
675 * config/nds32/nds32.c (nds32_compute_stack_frame,
676 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
677 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
678 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
679 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
680 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
681 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
682 * config/nds32/nds32.md (prologue, epilogue): Use macro
683 NDS32_V3PUSH_AVAILABLE_P to do checking.
685 2018-03-11 Jakub Jelinek <jakub@redhat.com>
687 PR debug/58150
688 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
689 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
690 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
691 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
692 addition of most attributes on !orig_type_die or the attribute not
693 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
695 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
696 Chung-Ju Wu <jasonwucj@gmail.com>
698 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
699 __NDS32_VH__ macro.
700 * config/nds32/nds32.opt (mvh): New option.
702 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
703 Chung-Ju Wu <jasonwucj@gmail.com>
705 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
706 function.
707 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
708 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
709 definition.
711 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
712 Chung-Ju Wu <jasonwucj@gmail.com>
714 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
715 function.
716 * config/nds32/nds32-multiple.md (strlensi): New pattern.
717 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
719 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
720 Kito Cheng <kito.cheng@gmail.com>
721 Chung-Ju Wu <jasonwucj@gmail.com>
723 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
724 UNSPEC_FFMISM and UNSPEC_FLMISM.
725 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
726 for ffb, ffmism and flmism.
727 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
728 (unspec_ffmism): Ditto.
729 (unspec_flmism): Ditto.
730 (nds32_expand_builtin_impl): Check if string extension is available.
731 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
732 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
734 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
736 Reverting patch:
737 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
739 PR target/83712
740 * lra-assigns.c (assign_by_spills): Return a flag of reload
741 assignment failure. Do not process the reload assignment
742 failures. Do not spill other reload pseudos if they has the same
743 reg class.
744 (lra_assign): Add a return arg. Set up from the result of
745 assign_by_spills call.
746 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
747 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
748 usage_insns if it is not NULL.
749 (spill_hard_reg_in_range): New function.
750 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
751 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
752 function prototypes.
753 (lra_assign): Change prototype.
754 * lra.c (lra): Add code to deal with fails by splitting hard reg
755 live ranges.
757 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
759 PR target/84807
760 * config/i386/i386.opt: Replace Enforcment with Enforcement.
762 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
764 PR debug/84620
765 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
766 (dw_val_node): Add val_symbolic_view.
767 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
768 (symview_upper_bound): New.
769 (new_line_info_table): Initialize symviews_since_reset.
770 (dwarf2out_source_line): Count symviews_since_reset and set
771 symview_upper_bound.
772 (dw_val_equal_p): Handle symview.
773 (add_AT_symview): New.
774 (print_dw_val): Handle symview.
775 (attr_checksum, attr_checksum_ordered): Likewise.
776 (same_dw_val_p, size_of_die): Likewise.
777 (value_format, output_die): Likewise.
778 (add_high_low_attributes): Use add_AT_symview for entry_view.
779 (dwarf2out_finish): Reset symview_upper_bound, clear
780 zero_view_p.
782 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
784 PR target/83969
785 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
786 Add strict argument and use it.
787 (rs6000_split_multireg_move): Update for new strict argument.
788 (mem_operand_gpr): Disallow all non-offsettable addresses.
789 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
791 2018-03-09 Jakub Jelinek <jakub@redhat.com>
793 PR target/84772
794 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
795 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
796 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
798 PR c++/84767
799 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
800 decl, use remap_type if we want to use the type.
802 2018-03-09 Martin Sebor <msebor@redhat.com>
804 PR tree-optimization/84526
805 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
806 Remove dead code.
807 (builtin_access::generic_overlap): Be prepared to handle non-array
808 base objects.
810 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
812 PR rtl-optimization/84682
813 * lra-constraints.c (process_address_1): Check is_address flag
814 for address constraints.
815 (process_alt_operands): Likewise.
816 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
817 preprocess_constraints.
818 * recog.h (preprocess_constraints): Add oploc parameter.
819 Adjust callers.
820 * recog.c (preprocess_constraints): Test address_operand for
821 CT_ADDRESS constraints.
823 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
825 PR target/83712
826 * lra-assigns.c (assign_by_spills): Return a flag of reload
827 assignment failure. Do not process the reload assignment
828 failures. Do not spill other reload pseudos if they has the same
829 reg class.
830 (lra_assign): Add a return arg. Set up from the result of
831 assign_by_spills call.
832 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
833 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
834 usage_insns if it is not NULL.
835 (spill_hard_reg_in_range): New function.
836 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
837 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
838 function prototypes.
839 (lra_assign): Change prototype.
840 * lra.c (lra): Add code to deal with fails by splitting hard reg
841 live ranges.
843 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
845 PR target/83193
846 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
847 Accept complain bool parameter. Only emit errors if it is true.
848 (arm_parse_cpu_option_name): Likewise.
849 (arm_target_thumb_only): Adjust callers of the above.
850 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
851 prototype to take a default true bool parameter.
852 (arm_parse_arch_option_name): Likewise.
854 2018-03-09 David Malcolm <dmalcolm@redhat.com>
855 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
857 PR jit/64089
858 PR jit/84288
859 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
860 * configure: Regenerate.
861 * configure.ac ("linker --version-script option"): New.
862 ("linker soname option"): New.
864 2018-03-09 Richard Biener <rguenther@suse.de>
866 PR tree-optimization/84775
867 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
868 immediate uses of predicate stmts and mark them modified.
870 Revert
871 PR tree-optimization/84178
872 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
873 to caller.
874 (version_loop_for_if_conversion): Delay update_ssa call.
875 (tree_if_conversion): Delay update_ssa until after predicate
876 insertion.
878 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
880 PR target/84763
881 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
882 when the function accesses prior frames.
884 2018-03-08 Jakub Jelinek <jakub@redhat.com>
886 PR debug/84456
887 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
888 gen_llsym, otherwise call maybe_gen_llsym.
890 PR inline-asm/84742
891 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
892 has ',' character inside of it.
894 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
896 PR target/84748
897 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
898 as clobbering CC_REGNUM.
900 2018-03-08 Richard Biener <rguenther@suse.de>
902 PR middle-end/84552
903 * tree-scalar-evolution.c: Include tree-into-ssa.h.
904 (follow_copies_to_constant): Do not follow SSA names registered
905 for update.
907 2018-03-08 Richard Biener <rguenther@suse.de>
909 PR tree-optimization/84178
910 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
911 to caller.
912 (version_loop_for_if_conversion): Delay update_ssa call.
913 (tree_if_conversion): Delay update_ssa until after predicate
914 insertion.
916 2018-03-08 David Malcolm <dmalcolm@redhat.com>
918 PR tree-optimization/84178
919 * tree-if-conv.c (release_bb_predicate): Remove the
920 the assertion that the stmts have NULL use_ops.
921 Discard the statements, asserting that they haven't
922 yet been added to a BB.
924 2018-03-08 Richard Biener <rguenther@suse.de>
926 PR tree-optimization/84746
927 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
928 (phi_translate): Pass in destination ANTIC_OUT set.
929 (phi_translate_1): Likewise. For a simplified result lookup
930 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
931 (phi_translate_set): Adjust.
932 (do_pre_regular_insertion): Likewise.
933 (do_pre_partial_partial_insertion): Likewise.
935 2018-03-08 Martin Liska <mliska@suse.cz>
937 PR gcov-profile/84735
938 * doc/gcov.texi: Document usage of profile files.
939 * gcov-io.h: Document changes in the format.
941 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
943 PR debug/84404
944 PR debug/84408
945 * dwarf2out.c (struct dw_line_info_table): Update comments for
946 view == -1.
947 (FORCE_RESET_NEXT_VIEW): New.
948 (FORCE_RESETTING_VIEW_P): New.
949 (RESETTING_VIEW_P): Check for -1 too.
950 (ZERO_VIEW_P): Likewise.
951 (new_line_info_table): Force-reset next view.
952 (dwarf2out_begin_function): Likewise.
953 (dwarf2out_source_line): Simplify zero_view_p initialization.
954 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
955 view directly. Omit view when omitting .loc at line 0.
957 2018-03-08 Jakub Jelinek <jakub@redhat.com>
959 PR tree-optimization/84740
960 * tree-switch-conversion.c (process_switch): Call build_constructors
961 only if info.phi_count is non-zero.
963 PR tree-optimization/84739
964 * tree-tailcall.c (find_tail_calls): Check call arguments against
965 DECL_ARGUMENTS (current_function_decl) rather than
966 DECL_ARGUMENTS (func) when checking for tail recursion.
968 2018-03-07 Jakub Jelinek <jakub@redhat.com>
970 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
971 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
972 Volker Reichelt's entry and add entries for people that perform
973 GCC fuzzy testing and report numerous bugs.
975 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
977 PR target/82411
978 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
979 readonly data in sdata, if that is disabled.
980 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
981 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
982 -mreadonly-in-sdata option.
984 2018-03-07 Martin Sebor <msebor@redhat.com>
986 PR tree-optimization/84468
987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
988 basic block when looking for nul assignment.
990 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
992 PR target/84277
993 * except.h (output_function_exception_table): Adjust prototype.
994 * except.c (output_function_exception_table): Remove FNNAME parameter
995 and add SECTION parameter. Ouput one part of the table at a time.
996 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
997 the first part of the exception table and emit unwind directives.
998 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
999 (i386_pe_seh_cold_init): Likewise.
1000 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1001 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1002 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1003 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1004 * config/i386/winnt.c: Include except.h.
1005 (struct seh_frame_state): Add reg_offset, after_prologue and
1006 in_cold_section fields.
1007 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1008 (i386_pe_seh_cold_init): New function.
1009 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1010 to seh->in_cold_section.
1011 (seh_emit_push): Record the offset of the push.
1012 (seh_emit_save): Record the offet of the save.
1013 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1014 Test seh->after_prologue to disregard the epilogue.
1015 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1016 (i386_pe_end_cold_function): New function.
1018 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1020 PR fortran/84565
1021 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1022 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1024 PR c++/84704
1025 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1026 on tmp_var.
1027 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1028 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1030 PR middle-end/84723
1031 * multiple_target.c: Include tree-inline.h and intl.h.
1032 (expand_target_clones): Diagnose and fail if node->definition and
1033 !tree_versionable_function_p (node->decl).
1035 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1037 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1038 sprint_ul.
1039 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1040 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1041 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1043 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1045 PR target/84710
1046 * combine.c (try_combine): Use reg_or_subregno instead of handling
1047 just paradoxical SUBREGs and REGs.
1049 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1051 * config/arc/arc.c (arc_finalize_pic): Remove function.
1052 (arc_must_save_register): We use single base PIC register, remove
1053 checks to save/restore the PIC register.
1054 (arc_expand_prologue): Likewise.
1055 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1056 Remove.
1057 (arc_verify_short): Likewise.
1058 (arc_attr_type): Likewise.
1059 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1060 (walk_stores): Likewise.
1061 (arc_address_cost): Make it static.
1062 (arc_verify_short): Likewise.
1063 (branch_dest): Likewise.
1064 (arc_attr_type): Likewise.
1065 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1066 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1067 (arc_final_prescan_insn): Remove inserting the nops due to
1068 hardware hazards. It is done in reorg step.
1069 (insn_length_variant_t): Remove.
1070 (insn_length_parameters_t): Likewise.
1071 (arc_insn_length_parameters): Likewise.
1072 (arc_get_insn_variants): Likewise.
1073 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1075 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1077 PR inline-asm/84683
1078 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1079 assertion failure.
1081 PR tree-optimization/84687
1082 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1083 on new_node->decl.
1084 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1086 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1088 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1089 Rename to ppc_speculation_barrier.
1090 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1091 __builtin_ppc_speculation_barrier.
1093 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1095 PR target/84700
1096 * combine.c (combine_simplify_rtx): Don't try to simplify if
1097 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1098 are equal to x.
1100 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1102 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1103 to 32 bytes when compiling for POWER9.
1105 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1107 PR target/84564
1108 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1109 regparm >= 3 with no arg reg available also for calls with
1110 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1112 PR target/84524
1113 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1114 orig,vex.
1115 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1117 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1119 PR target/84264
1120 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1122 2018-03-05 Richard Biener <rguenther@suse.de>
1124 PR tree-optimization/84486
1125 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1126 When inserting a __builtin_assume_aligned call set the LHS
1127 SSA name alignment info accordingly.
1129 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1131 PR tree-optimization/84114
1132 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1133 Avoid reassociation of FLOAT_MODE addition.
1135 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1137 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1138 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1139 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1140 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1141 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1142 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1143 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1144 and -mwbnoinvd.
1145 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1146 __builtin_ia32_wbinvd): New builtins.
1147 (SPECIAL_ARGS2): New.
1148 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1149 (SPECIAL_ARGS2): New.
1150 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1151 (ix86_valid_target_attribute_inner_p): Ditto.
1152 (ix86_init_mmx_sse_builtins): Add special_args2.
1153 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1154 TARGET_WBNOINVD_P): New.
1155 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1156 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1157 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1158 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1159 * config/i386/pconfigintrin.h: New file.
1160 * config/i386/wbnoinvdintrin.h: Ditto.
1161 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1162 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1164 2018-03-05 Richard Biener <rguenther@suse.de>
1166 PR tree-optimization/84670
1167 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1168 member.
1169 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1170 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1171 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1172 doesn't grow if all successors (recursively) were visited at least
1173 once.
1175 2018-03-05 Richard Biener <rguenther@suse.de>
1177 PR tree-optimization/84650
1178 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1179 if executed in the loop pipeline.
1181 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1183 * doc/configfiles.texi (Configuration Files): Move info about
1184 conditionalizing $target-protos.h to...
1185 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1186 differs from $target-protos.h.
1188 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1189 Chung-Ju Wu <jasonwucj@gmail.com>
1191 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1192 * config/nds32/nds32-multiple.md (setmemsi): Define.
1193 * config/nds32/nds32-memory-manipulation.c
1194 (nds32_gen_dup_4_byte_to_word_value): New.
1195 (emit_setmem_word_loop): New.
1196 (emit_setmem_byte_loop): New.
1197 (nds32_expand_setmem_loop): New.
1198 (nds32_expand_setmem_loop_v3m): New.
1199 (nds32_expand_setmem_unroll): New.
1200 (nds32_expand_setmem): New.
1202 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1203 Chung-Ju Wu <jasonwucj@gmail.com>
1205 * config/nds32/nds32-memory-manipulation.c
1206 (nds32_emit_load_store): New.
1207 (nds32_emit_post_inc_load_store): New.
1208 (nds32_emit_mem_move): New.
1209 (nds32_emit_mem_move_block): New.
1210 (nds32_expand_movmemsi_loop_unknown_size): New.
1211 (nds32_expand_movmemsi_loop_known_size): New.
1212 (nds32_expand_movmemsi_loop): New.
1213 (nds32_expand_movmemsi_unroll): New.
1214 (nds32_expand_movmemqi): Rename ...
1215 (nds32_expand_movmemsi): ... to this.
1216 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1217 (movmemsi): ... to this.
1218 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1219 (nds32_expand_movmemsi): ... to this.
1221 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1222 Monk Chiang <sh.chiang04@gmail.com>
1223 Chung-Ju Wu <jasonwucj@gmail.com>
1225 * config/nds32/nds32-protos.h
1226 (nds32_expand_load_multiple): New arguments.
1227 (nds32_expand_store_multiple): Ditto.
1228 (nds32_valid_multiple_load_store): Rename ...
1229 (nds32_valid_multiple_load_store_p): ... to this.
1230 * config/nds32/nds32-memory-manipulation.c
1231 (nds32_expand_load_multiple): Refine implementation.
1232 (nds32_expand_store_multiple): Ditto.
1233 * config/nds32/nds32-multiple.md
1234 (load_multiple): Update nds32_expand_load_multiple interface.
1235 (store_multiple): Update nds32_expand_store_multiple interface.
1236 * config/nds32/nds32-predicates.c
1237 (nds32_valid_multiple_load_store): Rename ...
1238 (nds32_valid_multiple_load_store_p): ... to this and refine
1239 implementation.
1240 * config/nds32/predicates.md
1241 (nds32_load_multiple_and_update_address_operation): New predicate.
1242 (nds32_store_multiple_and_update_address_operation): New predicate.
1244 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1245 Chung-Ju Wu <jasonwucj@gmail.com>
1247 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1248 (combo): New attribute.
1249 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1251 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1253 * config/nds32/nds32.opt: Change -mcmodel= default value.
1255 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1256 Monk Chiang <sh.chiang04@gmail.com>
1257 Chung-Ju Wu <jasonwucj@gmail.com>
1259 * config/nds32/constants.md (unspec_element): New enum.
1260 * config/nds32/constraints.md (Umw): New constraint.
1261 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1262 * config/nds32/nds32-intrinsic.md: Likewise.
1263 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1264 (nds32_valid_smw_lwm_base_p): New.
1265 (nds32_output_smw_single_word): New.
1266 (nds32_output_lmw_single_word): New.
1267 (nds32_expand_unaligned_load): New.
1268 (nds32_expand_unaligned_store): New.
1269 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1270 (nds32_output_smw_single_word): Declare.
1271 (nds32_output_lmw_single_word): Declare.
1272 (nds32_expand_unaligned_load): Declare.
1273 (nds32_expand_unaligned_store): Declare.
1274 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1275 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1276 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1277 NDS32_BUILTIN_UASTORE_DW.
1278 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1279 predicate.
1281 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1282 Kito Cheng <kito.cheng@gmail.com>
1283 Chung-Ju Wu <jasonwucj@gmail.com>
1285 * config/nds32/nds32-intrinsic.c
1286 (nds32_expand_builtin_null_ftype_reg): Delete.
1287 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1288 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1289 (nds32_read_argument): New.
1290 (nds32_legitimize_target): Ditto.
1291 (nds32_legitimize_argument): Ditto.
1292 (nds32_check_constant_argument): Ditto.
1293 (nds32_expand_unop_builtin): Ditto.
1294 (nds32_expand_unopimm_builtin): Ditto.
1295 (nds32_expand_binop_builtin): Ditto.
1296 (nds32_builtin_decl_impl): Ditto.
1297 (builtin_description): Ditto.
1298 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1299 (nds32_init_builtins_impl): Ditto.
1300 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1301 (nds32_builtin_decl): New.
1302 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1303 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1305 2018-03-02 Jeff Law <law@redhat.com>
1307 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1308 (redundant_insn, fill_simple_delay_slots): Likewise.
1309 (fill_slots_from_thread): Likewise.
1310 * resource.c (mark_referenced_resources): Likewise.
1311 (mark_set_resources, find_dead_or_set_registers): Likewise.
1313 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1315 * substring-locations.h (format_warning_va): Formatting fix for
1316 ATTRIBUTE_GCC_DIAG.
1317 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1318 argument.
1319 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1320 * substring-locations.c: Include intl.h.
1321 (format_warning_va): Turned into small wrapper around
1322 format_warning_n_va, renamed to ...
1323 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1324 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1325 use ngettext.
1326 (format_warning_at_substring_n): New function.
1327 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1328 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1329 format_warning_at_substring with just a shorter name instead of
1330 const function pointer.
1331 (fmtwarn_n): New function.
1332 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1333 appropriate, get rid of all the fmtstr temporaries, move conditionals
1334 with G_() wrapped string literals directly into fmtwarn arguments,
1335 cast dir.len to (int), formatting fixes.
1337 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1339 * doc/invoke.texi: Remove "Cilk Plus" references.
1341 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1342 Richard Biener <rguenther@suse.de>
1344 PR ipa/84628
1345 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1346 for error or warning attributes if CALL_FROM_THUNK_P is set.
1347 Formatting fixes.
1349 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1351 PR target/56540
1352 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1353 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1355 PR target/56540
1356 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1357 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1359 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1360 instead of -1U in last predictors element's probability member.
1362 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1364 PR ipa/83983
1365 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1366 arguments if they are comparable.
1368 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1370 PR tree-optimization/84634
1371 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1372 masks and masked_loop_p with a single loop_masks, making sure it's
1373 null for bb vectorization.
1375 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1377 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1378 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1379 loop->force_vectorize to check whether there is no alias.
1381 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1383 PR target/84614
1384 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1385 prototypes.
1386 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1387 comments.
1388 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1389 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1390 instead of a loop around prev_real_insn.
1391 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1392 prev_real_insn.
1394 PR inline-asm/84625
1395 * config/i386/i386.c (ix86_print_operand): Use conditional
1396 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1397 zero vector.
1399 2018-03-02 Richard Biener <rguenther@suse.de>
1401 PR tree-optimization/84427
1402 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1403 (bitmap_set_subtract_values): Rewrite to handle multiple
1404 exprs per value.
1405 (clean): Likewise.
1406 (prune_clobbered_mems): Likewise.
1407 (phi_translate): Take edge instead of pred/phiblock.
1408 (phi_translate_1): Likewise.
1409 (phi_translate_set): Likewise. Insert all translated
1410 exprs for a value into the set, keeping possibly multiple
1411 expressions per value.
1412 (compute_antic_aux): Adjust for phi_translate changes.
1413 When intersecting union the expressions and prune those
1414 not in the final value set, keeping possibly multiple
1415 expressions per value. Do not use value-insertion
1416 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1417 all expressions. Add verification that the value-sets
1418 only shrink during iteration.
1419 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1420 (do_pre_regular_insertion): Likewise.
1421 (do_pre_partial_partial_insertion): Likewise.
1423 2018-03-02 Richard Biener <rguenther@suse.de>
1425 PR target/82005
1426 * config/darwin.c (saved_debug_info_level): New static global.
1427 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1428 (darwin_asm_lto_end): Restore debug info generation settings.
1430 2018-03-01 Martin Liska <mliska@suse.cz>
1432 PR sanitizer/82484
1433 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1434 volatile arguments.
1436 2018-03-01 Richard Biener <rguenther@suse.de>
1438 PR debug/84645
1439 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1440 type annotation with LTO when debug was disabled at compile-time.
1442 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1444 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1445 XINT with INTVAL.
1446 (mips_final_postscan_insn): Likewise.
1448 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1450 PR rtl-optimization/84528
1451 * alias.c (init_alias_target): Add commentary.
1452 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1453 a unique base value if the frame pointer is not eliminated
1454 to the stack pointer.
1456 2018-03-01 Tom de Vries <tom@codesourcery.com>
1458 PR rtl-optimization/83327
1459 * lra-int.h (hard_regs_spilled_into): Declare.
1460 * lra.c (hard_regs_spilled_into): Define.
1461 (init_reg_info): Init hard_regs_spilled_into.
1462 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1463 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1464 (process_bb_lives): Handle hard_regs_spilled_into.
1465 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1466 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1468 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1470 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1471 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1472 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1473 * config/rs6000/aix72.h: New file.
1475 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1477 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1478 instead of warning_at with conditional singular and plural messages
1479 where possible.
1481 PR target/52991
1482 * stor-layout.c (update_alignment_for_field): For
1483 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1484 && !DECL_PACKED (field), do the alignment update, just use
1485 only desired_align instead of MAX (type_align, desired_align)
1486 as the alignment.
1487 (place_field): Don't do known_align < desired_align handling
1488 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1489 is non-NULL, instead do it after rli->prev_field handling and
1490 only if not within a bitfield word. For DECL_PACKED (field)
1491 use type_align of BITS_PER_UNIT.
1493 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1495 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1496 superfluous parentheses and trailing spaces.
1498 2018-02-28 Richard Biener <rguenther@suse.de>
1500 PR tree-optimization/84584
1501 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1502 SCoPs with fake exit edge.
1504 2018-02-28 Martin Liska <mliska@suse.cz>
1506 PR testsuite/84597
1507 * timevar.c (timer::print): Fix format to properly print 100%
1508 values.
1510 2018-02-28 Richard Biener <rguenther@suse.de>
1512 PR middle-end/84607
1513 * genmatch.c (capture_info::walk_match): Do not mark
1514 captured expressions without operands as expr_p given
1515 they act more like predicates and should be subject to
1516 "lost tail" side-effect preserving.
1518 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1520 PR rtl-optimization/81611
1521 * auto-inc-dec.c (attempt_change): Move dead note from
1522 mem_insn if it's the next use of regno
1523 (find_address): Take address use of reg holding
1524 non-incremented value. Add parm to limit search to the named
1525 reg only.
1526 (merge_in_block): Attempt to use a mem insn that is the next
1527 use of the original regno.
1529 2018-02-27 Martin Sebor <msebor@redhat.com>
1531 PR c++/83871
1532 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1533 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1535 2018-02-27 Martin Sebor <msebor@redhat.com>
1537 PR translation/84207
1538 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1539 n argument to unsigned HOST_WIDE_INT.
1540 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1541 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1542 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1543 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1545 2018-02-27 Richard Biener <rguenther@suse.de>
1547 PR tree-optimization/84512
1548 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1549 Do not use the estimate returned from record_stmt_cost for
1550 the scalar iteration cost but sum properly using add_stmt_cost.
1552 2018-02-27 Richard Biener <rguenther@suse.de>
1554 PR tree-optimization/84466
1555 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1556 Adjust last change to less strictly validate use operands.
1558 2018-02-27 Martin Liska <mliska@suse.cz>
1560 PR gcov-profile/84548
1561 * gcov.c (process_file): Allow partial overlap and consider it
1562 also as group functions.
1563 (output_lines): Properly calculate range of lines for a group.
1565 2018-02-27 Martin Liska <mliska@suse.cz>
1567 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1568 'ggc' suffixes. Change first column width.
1569 (timer::print): Fix formatting of the column.
1571 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1573 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1574 preserve inline entry blocks for the sake of debug inline
1575 entry point markers alone.
1576 (remove_unused_locals): Suggest in comments a better place to
1577 force the preservation of inline entry blocks that are
1578 otherwise unused, but do not preserve them.
1580 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1582 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1584 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1586 PR target/84039
1587 * config/i386/constraints.md (Bs): Replace
1588 ix86_indirect_branch_register with
1589 TARGET_INDIRECT_BRANCH_REGISTER.
1590 (Bw): Likewise.
1591 * config/i386/i386.md (indirect_jump): Likewise.
1592 (tablejump): Likewise.
1593 (*sibcall_memory): Likewise.
1594 (*sibcall_value_memory): Likewise.
1595 Peepholes of indirect call and jump via memory: Likewise.
1596 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1597 (*sibcall_value_GOT_32): Likewise.
1598 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1599 (GOT_memory_operand): Likewise.
1600 (call_insn_operand): Likewise.
1601 (sibcall_insn_operand): Likewise.
1602 (GOT32_symbol_operand): Likewise.
1603 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1605 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1607 PR rtl-optimization/83496
1608 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1609 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1610 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1611 redundant insn, if any.
1612 (relax_delay_slots): Likewise.
1613 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1615 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1617 PR tree-optimization/83965
1618 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1619 that grouped statements are part of a reduction chain. Return
1620 true if the statement is not marked as a reduction itself but
1621 is part of a group.
1622 (vect_recog_dot_prod_pattern): Don't check whether the statement
1623 is part of a group here.
1624 (vect_recog_sad_pattern): Likewise.
1625 (vect_recog_widen_sum_pattern): Likewise.
1627 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1629 PR debug/84545
1630 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1632 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1634 PR target/84530
1635 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1636 the bool argument.
1637 (ix86_output_indirect_function_return): New prototype.
1638 (ix86_split_simple_return_pop_internal): Likewise.
1639 * config/i386/i386.c (indirect_return_via_cx): New.
1640 (indirect_return_via_cx_bnd): Likewise.
1641 (indirect_thunk_name): Handle return va CX_REG.
1642 (output_indirect_thunk_function): Create alias for
1643 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1644 (ix86_output_indirect_jmp): Remove the bool argument.
1645 (ix86_output_indirect_function_return): New function.
1646 (ix86_split_simple_return_pop_internal): Likewise.
1647 * config/i386/i386.md (*indirect_jump): Don't pass false
1648 to ix86_output_indirect_jmp.
1649 (*tablejump_1): Likewise.
1650 (simple_return_pop_internal): Change it to define_insn_and_split.
1651 Call ix86_split_simple_return_pop_internal to split it for
1652 -mfunction-return=.
1653 (simple_return_indirect_internal): Call
1654 ix86_output_indirect_function_return instead of
1655 ix86_output_indirect_jmp.
1657 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1659 PR bootstrap/84405
1660 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1661 memset and value initialization afterwards.
1663 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1665 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1667 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1669 PR target/84521
1670 * common/config/aarch64/aarch64-common.c
1671 (aarch_option_optimization_table[]): Switch
1672 off fomit-frame-pointer
1674 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1675 Chung-Ju Wu <jasonwucj@gmail.com>
1677 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1678 volatile memory.
1679 (store_multiple): Ditto.
1681 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1683 * config.gcc: Add --with-cpu support for nds32 target.
1684 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1685 * config/nds32/nds32.opt: Add -mcpu= option.
1687 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1689 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1690 isel=yes): Warn for these deprecated options.
1692 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1694 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1695 ISA_2_5_MASKS_EMBEDDED.
1697 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1699 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1700 p->max as pointers rather than using iterative_hash_expr.
1702 2018-02-23 Carl Love <cel@us.ibm.com>
1704 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1705 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1706 BU_P8V_OVERLOAD_2.
1707 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1708 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1709 P8V_BUILTIN_VEC_VUNSIGNED2.
1711 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1713 PR target/81572
1714 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1715 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1716 LRA_UNKNOWN_ALT.
1717 * lra-constraints.c (curr_insn_transform): Set up
1718 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1719 LRA_UNKNOWN_ALT.
1720 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1721 * lra-eliminations.c (spill_pseudos): Ditto.
1722 (process_insn_for_elimination): Ditto.
1723 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1724 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1725 LRA_NON_CLOBBERED_ALT.
1727 2018-02-22 Martin Sebor <msebor@redhat.com>
1729 PR tree-optimization/84480
1730 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1731 to maybe_diag_stxncpy_trunc. Call it.
1732 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1733 from gimple_fold_builtin_strcpy. Print inlining stack.
1734 (handle_builtin_stxncpy): Print inlining stack.
1735 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1737 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1739 PR target/84176
1740 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1741 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1742 and -fcheck-pointer-bounds are used together.
1743 (indirect_thunk_prefix): New enum.
1744 (indirect_thunk_need_prefix): New function.
1745 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1746 "_nt" instead of "_bnd" for NOTRACK prefix.
1747 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1748 (output_indirect_thunk_function): Likewise.
1749 (): Likewise.
1750 (ix86_code_end): Update output_indirect_thunk_function calls.
1751 (ix86_output_indirect_branch_via_reg): Replace
1752 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1753 (ix86_output_indirect_branch_via_push): Likewise.
1754 (ix86_output_function_return): Likewise.
1755 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1756 incompatible with -fcf-protection=branch and
1757 -fcheck-pointer-bounds.
1759 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1761 PR target/83335
1762 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1763 Change gcc_assert call to output_operand_lossage.
1765 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1767 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1769 2018-02-22 DJ Delorie <dj@redhat.com>
1770 Sebastian Perta <sebastian.perta@renesas.com>
1771 Oleg Endo <olegendo@gcc.gnu.org>
1773 * config/rx/rx.c (rx_rtx_costs): New function.
1774 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1776 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1778 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1780 2018-02-22 Martin Liska <mliska@suse.cz>
1782 PR driver/83193
1783 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1784 Add "native" as a possible value.
1786 2018-02-22 Martin Liska <mliska@suse.cz>
1788 PR driver/83193
1789 * config/i386/i386.c (ix86_option_override_internal):
1790 Add "native" as a possible value for -march and -mtune.
1792 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1794 PR target/84502
1795 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1796 to all type variants.
1798 PR tree-optimization/84503
1799 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1800 width as info->bitpos + info->bitsize - start.
1801 (merged_store_group::merge_overlapping): Simplify width computation.
1802 (check_no_overlap): New function.
1803 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1804 start + width and last_order of the group, fail if check_no_overlap
1805 fails.
1806 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1807 to group if check_no_overlap fails.
1809 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1811 * config/rs6000/altivec.md: Delete contraint arguments to
1812 define_expand, define_split, and define_peephole2, and in
1813 define_insn_and_split if always unused.
1814 * config/rs6000/darwin.md: Ditto.
1815 * config/rs6000/dfp.md: Ditto.
1816 * config/rs6000/rs6000.md: Ditto.
1817 * config/rs6000/sync.md: Ditto.
1818 * config/rs6000/vector.md: Ditto.
1819 * config/rs6000/vsx.md: Ditto.
1821 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1823 * config/rs6000/altivec.md: Write output control strings as braced
1824 blocks instead of double-quoted strings.
1825 * config/rs6000/darwin.md: Ditto.
1826 * config/rs6000/rs6000.md: Ditto.
1827 * config/rs6000/vector.md: Ditto.
1828 * config/rs6000/vsx.md: Ditto.
1830 2018-02-21 Jason Merrill <jason@redhat.com>
1832 PR c++/84314 - ICE with templates and fastcall attribute.
1833 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1835 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1837 * ipa-cp.c (determine_versionability): Fix comment typos.
1839 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1841 PR c/84229
1842 * ipa-cp.c (determine_versionability): Do not version functions caling
1843 va_arg_pack.
1845 2018-02-21 Martin Liska <mliska@suse.cz>
1847 PR driver/83193
1848 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1849 Add "native" as a possible value.
1850 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1851 the macro when native cpu detection is available.
1853 2018-02-21 Martin Liska <mliska@suse.cz>
1855 PR driver/83193
1856 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1857 Add "native" as a possible value.
1858 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1859 when native cpu detection is available.
1861 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1862 Martin Sebor <msebor@redhat.com>
1864 PR tree-optimization/84478
1865 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1866 false.
1867 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1868 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1869 support which is conservatively correct, for 2 only stay conservative
1870 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1871 argument to the 2 argument get_range_strlen, adjust 6 arg
1872 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1873 false.
1874 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1875 (gimple_fold_builtin_strlen): Pass true as last argument to
1876 get_range_strlen.
1878 2018-02-20 Martin Sebor <msebor@redhat.com>
1880 PR middle-end/84095
1881 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1882 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1883 (builtin_memref::builtin_memref): Factor out parts into
1884 set_base_and_offset and call it.
1886 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1888 PR middle-end/84406
1889 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1890 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1891 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1892 search at the associated MODE_INT.
1894 2018-02-20 Jeff Law <law@redhat.com>
1896 PR middle-end/82123
1897 PR tree-optimization/81592
1898 PR middle-end/79257
1899 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1900 for range data rather than using global data.
1901 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1902 range data rather than using global data.
1903 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1904 pass it to children as needed.
1905 (struct directive::fmtresult): Similarly.
1906 (struct directive::set_width): Similarly.
1907 (struct directive::set_precision): Similarly.
1908 (format_integer, format_directive, parse_directive): Similarly.
1909 (format_none): Accept unnamed vr_values parameter.
1910 (format_percent, format_floating, format_character): Similarly.
1911 (format_string, format_plain): Similarly.
1912 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1913 the EVRP range analyzer for range data rather than using global data.
1914 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1915 gimple-ssa-evrp-analyze.h
1916 (class sprintf_dom_walker): Add after_dom_children member function.
1917 Add evrp_range_analyzer member.
1918 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1919 range analyzer as needed.
1920 (sprintf_dom_walker::after_dom_children): New member function.
1921 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1922 if not optimizing.
1923 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1924 (evrp_range_analyzer::pop_to_marker): Likewise.
1926 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1928 PR tree-optimization/84419
1929 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1930 with the required type if its current type is compatible but
1931 different.
1933 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1935 PR middle-end/82004
1936 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1937 after vectorization.
1939 2018-02-20 Martin Liska <mliska@suse.cz>
1941 PR driver/83193
1942 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1943 possible values if we don't have a hint.
1945 2018-02-20 Martin Liska <mliska@suse.cz>
1947 PR c/84310
1948 PR target/79747
1949 * final.c (shorten_branches): Build align_tab array with one
1950 more element.
1951 * opts.c (finish_options): Add alignment option limit check.
1952 (MAX_CODE_ALIGN): Likewise.
1953 (MAX_CODE_ALIGN_VALUE): Likewise.
1954 * doc/invoke.texi: Document maximum allowed option value for
1955 all -falign-* options.
1957 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1959 PR target/84146
1960 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1961 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1962 * var-tracking.c (emit_note_insn_var_location): Remove all references
1963 to NOTE_INSN_CALL_ARG_LOCATION.
1964 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1965 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1966 Use copy_rtx_if_shared.
1967 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1968 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1969 (dwarf2out_var_location): Remove handling of
1970 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1971 on call_insn.
1972 * final.c (final_scan_insn): Remove all references to
1973 NOTE_INSN_CALL_ARG_LOCATION.
1974 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1975 before dumping final insns.
1976 * except.c (emit_note_eh_region_end): Remove all references to
1977 NOTE_INSN_CALL_ARG_LOCATION.
1978 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1979 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1980 * config/arc/arc.c (hwloop_optimize): Likewise.
1981 * config/arm/arm.c (create_fix_barrier): Likewise.
1982 * config/s390/s390.c (s390_chunkify_start): Likewise.
1983 * config/sh/sh.c (find_barrier): Likewise.
1984 * config/i386/i386.c (rest_of_insert_endbranch,
1985 ix86_seh_fixup_eh_fallthru): Likewise.
1986 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1987 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1988 * config/frv/frv.c (frv_function_prologue): Likewise.
1989 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1990 reg note.
1991 (note_outside_basic_block_p): Remove all references to
1992 NOTE_INSN_CALL_ARG_LOCATION.
1993 * gengtype.c (adjust_field_rtx_def): Likewise.
1994 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1995 Likewise.
1996 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1997 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1999 PR c++/84444
2000 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2001 is ADDR_EXPR.
2003 PR tree-optimization/84452
2004 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2005 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2006 is NULL.
2008 2018-02-19 Martin Liska <mliska@suse.cz>
2010 PR sanitizer/82183
2011 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2013 2018-02-19 Martin Liska <mliska@suse.cz>
2014 Richard Sandiford <richard.sandiford@linaro.org>
2016 PR tree-optimization/82491
2017 * gimple-fold.c (get_base_constructor): Make earlier bail out
2018 to prevent ubsan.
2020 2018-02-19 Carl Love <cel@us.ibm.com>
2022 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2023 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2024 BU_P8V_OVERLOAD_1.
2025 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2026 P8V_BUILTIN_VEC_NEG.
2028 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2030 * config/rl78/rl78.md (movdf): New define expand.
2032 2018-02-19 Martin Liska <mliska@suse.cz>
2034 PR other/80589
2035 * doc/invoke.texi: Fix typo.
2036 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2038 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2040 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2041 handle rs6000_single_float and rs6000_double_float specially for
2042 e500 family CPUs.
2044 2018-02-16 Jeff Law <law@redhat.com>
2046 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2047 (pop_regs): Use it.
2049 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2051 PR ipa/84425
2052 * ipa-inline.c (inline_small_functions): Fix a typo.
2054 2018-02-16 Nathan Sidwell <nathan@acm.org>
2056 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2058 2018-02-16 Carl Love <cel@us.ibm.com>
2060 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2061 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2062 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2063 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2064 expansion to P8V_BUILTIN_VEC_FLOAT2.
2066 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2068 PR rtl-optimization/70023
2069 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2070 src_regno into account.
2072 2018-02-16 Carl Love <cel@us.ibm.com>
2074 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2075 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2076 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2077 * config/rs6000/rs6000.c: Remove case statements for
2078 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2079 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2080 and P9V_BUILTIN_VEC_VINSERT4B.
2081 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2082 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2083 * config/rs6000/vsx.md:
2084 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2085 vec_insert4b.
2087 2018-02-16 Carl Love <cel@us.ibm.com>
2089 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2090 vec_insert4b.
2091 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2092 definitions.
2093 * config/rs6000/rs6000-c.c: Add the definitions for
2094 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2095 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2096 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2097 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2098 definition for insert4b and define insn *insert3b_internal.
2099 * doc/extend.texi: Add documentation for vec_extract4b.
2101 2018-02-16 Nathan Sidwell <nathan@acm.org>
2103 * doc/extend.texi (Backwards Compatibility): Mention friend
2104 injection. Note for-scope is deprecated.
2105 * doc/invoke.texi (-ffriend-injection): Deprecate.
2107 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2109 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2110 that moved to I2, also allow destinations that are a paradoxical
2111 subreg (instead of a normal reg).
2113 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2115 PR target/83831
2116 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2117 to QImode.
2119 2018-02-16 Richard Biener <rguenther@suse.de>
2121 PR tree-optimization/84037
2122 PR tree-optimization/84016
2123 PR target/82862
2124 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2125 Adjust vec_construct for the fact we need additional higher latency
2126 128bit inserts for AVX256 and AVX512 vector builds.
2127 (ix86_add_stmt_cost): Scale vector construction cost for
2128 elementwise loads.
2130 2018-02-16 Richard Biener <rguenther@suse.de>
2132 PR tree-optimization/84417
2133 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2134 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2135 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2137 2018-02-16 Martin Liska <mliska@suse.cz>
2139 PR sanitizer/84307
2140 * internal-fn.def (ASAN_CHECK): Set proper flags.
2141 (ASAN_MARK): Likewise.
2143 2018-02-16 Julia Koval <julia.koval@intel.com>
2145 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2146 from PTA_CANNONLAKE.
2148 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2150 PR target/84272
2151 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2152 Use ++iter rather than iter++ for std::list iterators.
2153 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2154 defer deleting them until all nodes in the forest are processed. Do
2155 free even leaf nodes. Change to_process into auto_vec.
2157 PR bootstrap/84405
2158 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2159 * vec.h (vec_default_construct): Use memset instead of placement new
2160 if BROKEN_VALUE_INITIALIZATION is defined.
2161 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2162 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2163 is defined.
2165 PR rtl-optimization/83723
2166 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2167 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2168 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2169 recursive calls.
2170 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2171 callers.
2172 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2174 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2176 PR rtl-optimization/81443
2177 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2178 from inner REGs to paradoxical SUBREGs.
2180 2018-02-16 Richard Biener <rguenther@suse.de>
2182 PR tree-optimization/84399
2183 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2184 For operands we can analyze at their definition make sure we can
2185 analyze them at each use as well.
2187 2018-02-16 Richard Biener <rguenther@suse.de>
2189 PR tree-optimization/84190
2190 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2191 volatile accesses if the decl isn't volatile.
2193 2018-02-15 Jason Merrill <jason@redhat.com>
2195 PR c++/84314 - ICE with templates and fastcall attribute.
2196 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2197 TYPE_CANONICAL on an existing type.
2199 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2201 PR tree-optimization/84383
2202 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2203 dstoff nor call operand_equal_p if dstbase is NULL.
2205 PR tree-optimization/84334
2206 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2207 also a CONSTANT_CLASS_P, punt.
2209 2018-02-14 Jim Wilson <jimw@sifive.com>
2211 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2212 first SMALL_OPERAND check. New local min_second_step. Move assert
2213 to where locals are set. Add TARGET_RVC support.
2214 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2216 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2218 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2220 2018-02-14 Martin Sebor <msebor@redhat.com>
2222 PR tree-optimization/83698
2223 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2224 arrays constrain the offset range to their bounds.
2225 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2226 (builtin_access::overlap): Avoid setting the size of overlap if it's
2227 already been set.
2228 (maybe_diag_overlap): Also consider arrays when deciding what values
2229 of offsets to include in diagnostics.
2231 2018-02-14 Martin Sebor <msebor@redhat.com>
2233 PR c/84108
2234 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2235 that correspond to the kind of a declaration.
2237 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2239 PR target/83984
2240 * config/pa/pa.md: Load address of PIC label using the linkage table
2241 if the label is nonlocal.
2243 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2245 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2246 warning message if user requests -maltivec=be.
2247 * doc/invoke.texi: Document deprecation of -maltivec=be.
2249 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2251 PR target/84220
2252 * config/rs6000/rs6000-c.c: Update definitions for
2253 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2254 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2256 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2258 PR target/84239
2259 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2260 add _get_ssp intrinsics. Remove argument from
2261 __builtin_ia32_rdssp[d|q].
2262 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2263 * config/i386/i386-builtin.def: Remove argument from
2264 __builtin_ia32_rdssp[d|q].
2265 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2266 ix86_expand_special_args_builtin for _rdssp[d|q].
2267 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2268 Clear register before usage.
2269 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2270 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2272 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2274 PR tree-optimization/84357
2275 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2276 operand 1 of an ARRAY_REF too.
2278 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2280 PR target/83831
2281 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2282 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2283 declarations.
2284 (set_of_reg): New struct.
2285 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2286 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2287 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2288 functions.
2289 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2290 Split into bitclr, bitset, bitinvert patterns if appropriate.
2291 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2292 use rx_fuse_in_memory_bitop.
2293 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2294 to named insn, correct maximum insn length.
2296 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2298 PR target/79242
2299 * machmode.def: Define a complex mode for PARTIAL_INT.
2300 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2301 MODE_PARTIAL_INT.
2302 * doc/rtl.texi: Document CSPImode.
2303 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2304 handling.
2305 (msp430_hard_regno_nregs_with_padding): Likewise.
2307 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2309 PR target/84279
2310 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2312 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2314 PR rtl-optimization/84169
2315 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2316 we generated a parallel as new i3 and we split that to new i2 and i3
2317 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2318 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2319 those to i2, not i1. Partially rewrite this scan code.
2321 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2323 PR c/82210
2324 * stor-layout.c (place_field): For variable length fields, adjust
2325 offset_align afterwards not just based on the field's alignment,
2326 but also on the size.
2328 PR middle-end/84309
2329 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2330 of exps and logs in the use_exp2 case.
2332 2018-02-13 Jeff Law <law@redhat.com>
2334 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2335 entry for "vector".
2337 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2338 ARGS as unused.
2340 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2342 PR debug/84342
2343 PR debug/84319
2344 * common.opt (gas-loc-support, gas-locview-support): New.
2345 (ginline-points, ginternal-reset-location-views): New.
2346 * doc/invoke.texi: Document them. Use @itemx where intended.
2347 (gvariable-location-views): Adjust.
2348 * target.def (reset_location_view): New.
2349 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2350 (TARGET_RESET_LOCATION_VIEW): New.
2351 * doc/tm.texi: Rebuilt.
2352 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2353 (dwarf2out_default_as_locview_support): New.
2354 (output_asm_line_debug_info): Use option variables.
2355 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2356 (output_loc_list): Likewise.
2357 (add_high_low_attributes): Check option variables.
2358 Don't output entry view attribute in strict mode.
2359 (gen_inlined_subroutine_die): Check option variables.
2360 (dwarf2out_inline_entry): Likewise.
2361 (init_sections_and_labels): Likewise.
2362 (dwarf2out_early_finish): Likewise.
2363 (maybe_reset_location_view): New, from...
2364 (dwarf2out_var_location): ... here. Call it.
2365 * debug.h (dwarf2out_default_as_loc_support): Declare.
2366 (dwarf2out_default_as_locview_support): Declare.
2367 * hooks.c (hook_int_rtx_insn_0): New.
2368 * hooks.h (hook_int_rtx_insn_0): Declare.
2369 * toplev.c (process_options): Take -gas-loc-support and
2370 -gas-locview-support from dwarf2out. Enable
2371 -gvariable-location-views by default only with locview
2372 assembler support. Enable -ginternal-reset-location-views by
2373 default only if the target defines the corresponding hook.
2374 Enable -ginline-points by default if location views are
2375 enabled; force it disabled if statement frontiers are
2376 disabled.
2377 * tree-inline.c (expand_call_inline): Check option variables.
2378 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2380 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2382 PR tree-optimization/84321
2383 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2384 handling. Also check whether the anti-range contains any values
2385 that satisfy the mask; switch to a VR_RANGE if not.
2387 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2389 PR sanitizer/84340
2390 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2392 2018-02-13 Martin Jambor <mjambor@suse.cz>
2394 PR c++/83990
2395 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2396 of call statements, also set location of a load to a temporary.
2398 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2400 * config/rl78/rl78.c (add_vector_labels): New function.
2401 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2402 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2403 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2404 which checks that no arguments are passed.
2405 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2406 * doc/extend.texi: Documentation for the new attribute.
2408 2018-02-13 Andreas Schwab <schwab@suse.de>
2410 * config/riscv/linux.h (CPP_SPEC): Define.
2412 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2414 PR target/84335
2415 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2416 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2417 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2418 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2419 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2420 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2421 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2422 temporarily for AES and PCLMUL builtins.
2424 PR tree-optimization/84339
2425 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2426 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2427 Formatting fixes.
2429 PR middle-end/84309
2430 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2431 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2432 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2433 inline function.
2434 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2435 inline function.
2436 * omp-simd-clone.h: New file.
2437 * omp-simd-clone.c: Include omp-simd-clone.h.
2438 (expand_simd_clones): No longer static.
2439 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2440 cgraph.h and omp-simd-clone.h.
2441 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2442 (vect_recog_widen_shift_pattern): Formatting fix.
2443 (vect_pattern_recog_1): Don't check optab for calls.
2445 PR target/84336
2446 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2447 operands[2] into a REG before using gen_lowpart on it.
2449 2018-02-12 Jeff Law <law@redhat.com>
2451 PR target/83760
2452 * config/sh/sh.c (find_barrier): Consider a sibling call
2453 a barrier as well.
2455 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2456 successfully back substituting a reg.
2458 2018-02-12 Richard Biener <rguenther@suse.de>
2460 PR tree-optimization/84037
2461 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2462 parameter, move visited init to caller.
2463 (vect_slp_analyze_operations): Separate cost from validity
2464 check, initialize visited once for all instances.
2465 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2466 for all instances.
2467 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2468 out an assert.
2469 (vect_model_promotion_demotion_cost): Likewise.
2470 (vectorizable_bswap): Guard cost modeling with !slp_node
2471 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2472 SLP stmts.
2473 (vectorizable_call): Likewise.
2474 (vectorizable_conversion): Likewise.
2475 (vectorizable_assignment): Likewise.
2476 (vectorizable_shift): Likewise.
2477 (vectorizable_operation): Likewise.
2478 (vectorizable_store): Likewise.
2479 (vectorizable_load): Likewise.
2480 (vectorizable_condition): Likewise.
2481 (vectorizable_comparison): Likewise.
2483 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2485 PR sanitizer/84307
2486 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2487 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2488 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2490 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2492 PR middle-end/83665
2493 * params.def (inline-min-speedup): Increase from 8 to 15.
2494 (max-inline-insns-auto): Decrease from 40 to 30.
2495 * ipa-split.c (consider_split): Add some buffer for function to
2496 be considered inlining candidate.
2497 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2498 default values.
2500 2018-02-12 Richard Biener <rguenther@suse.de>
2502 PR tree-optimization/84037
2503 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2504 matched stmts if we cannot swap the non-matched ones.
2506 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2508 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2509 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2510 _mm_maskz_scalef_round_ss): New intrinsics.
2511 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2512 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2513 __builtin_ia32_scalefss_round): Remove.
2514 (__builtin_ia32_scalefsd_mask_round,
2515 __builtin_ia32_scalefss_mask_round): New intrinsics.
2516 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2517 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2518 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2519 "<round_constraint>")): Changed to ...
2520 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2521 "<round_scalar_constraint>")): ... this.
2522 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2523 %0, %1, %2<round_op3>}"): Changed to ...
2524 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2525 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2526 %2<round_scalar_mask_op3>}"): ... this.
2527 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2529 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2531 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2532 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2533 (_mm_maskz_sqrt_round_ss): New intrinsics.
2534 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2535 (__builtin_ia32_sqrtsd_mask_round)
2536 (__builtin_ia32_sqrtss_mask_round): New builtins.
2537 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2538 (__builtin_ia32_sqrtss_round): Remove.
2539 (__builtin_ia32_sqrtsd_mask_round)
2540 (__builtin_ia32_sqrtss_mask_round): New builtins.
2541 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2542 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2543 ((match_operand:VF_128 1 "vector_operand"
2544 "xBm,<round_constraint>")): Changed to ...
2545 ((match_operand:VF_128 1 "vector_operand"
2546 "xBm,<round_scalar_constraint>")): ... this.
2547 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2548 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2549 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2550 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2551 %<iptr>1<round_scalar_mask_op3>}): ... this.
2552 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2553 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2555 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2557 PR target/84266
2558 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2559 Cast vec_cmpeq result to correct type.
2560 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2561 Cast vec_cmpgt result to correct type.
2563 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2565 * final.c (final_scan_insn_1): Renamed from...
2566 (final_scan_insn): ... this. New wrapper, to recover
2567 seen from the outermost call in recursive ones.
2568 * config/sparc/sparc.c (output_return): Drop seen from call.
2569 (output_sibcall): Likewise.
2570 * config/visium/visium.c (output_branch): Likewise.
2572 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2574 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2575 function label.
2577 2018-02-10 Alan Modra <amodra@gmail.com>
2579 PR target/84300
2580 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2581 Specify LR as an input.
2583 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2585 PR sanitizer/83987
2586 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2587 remove_member_access_dummy_vars): New functions.
2588 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2589 lower_omp_1, execute_lower_omp): Use them.
2591 PR rtl-optimization/84308
2592 * shrink-wrap.c (spread_components): Release todo vector.
2594 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2596 PR rtl-optimization/57193
2597 * ira-color.c (struct allocno_color_data): Add member
2598 conflict_allocno_hard_prefs.
2599 (update_conflict_allocno_hard_prefs): New.
2600 (bucket_allocno_compare_func): Add a preference based on
2601 conflict_allocno_hard_prefs.
2602 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2603 (color_allocnos): Remove a dead code. Initiate
2604 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2606 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2608 PR target/84226
2609 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2610 constraint from =wa to wa. Avoid a subreg on the output operand,
2611 instead use a pseudo and subreg it in a move.
2612 (p9_xxbrd_<mode>): Changed to ...
2613 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2614 (p9_xxbrd_v2df): New expander.
2615 (p9_xxbrw_<mode>): Changed to ...
2616 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2617 (p9_xxbrw_v4sf): New expander.
2619 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2621 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2623 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2625 PR target/83926
2626 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2627 multiply in 32-bit mode.
2628 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2629 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2630 mode.
2632 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2634 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2635 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2636 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2637 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2639 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2641 PR lto/84213
2642 * dwarf2out.c (is_trivial_indirect_ref): New function.
2643 (dwarf2out_late_global_decl): Do not generate a location
2644 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2645 and that are not defined in the current unit.
2647 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2649 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2650 instead of a libcall for UNORDERED.
2652 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2654 PR target/82641
2655 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2656 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2658 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2660 PR target/PR84295
2661 * config/s390/s390.c (s390_set_current_function): Invoke
2662 s390_indirect_branch_settings also if fndecl didn't change.
2664 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2666 * config/rs6000/rs6000.md (blockage): Set length to zero.
2668 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2670 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2672 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2674 PR sanitizer/84285
2675 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2676 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2677 -static-lib*san.
2679 PR debug/84252
2680 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2681 PARALLEL incoming that failed vt_get_decl_and_offset check.
2683 PR middle-end/84237
2684 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2685 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2686 TREE_READONLY bit.
2687 (get_variable_section): For decls in named .bss* sections pass true as
2688 second argument to bss_initializer_p.
2690 2018-02-09 Marek Polacek <polacek@redhat.com>
2691 Jakub Jelinek <jakub@redhat.com>
2693 PR c++/83659
2694 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2695 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2696 Sync some changes from cxx_fold_indirect_ref.
2698 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2700 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2701 markers.
2702 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2703 (BLOCK_INLINE_ENTRY_LABEL): New.
2704 (dwarf2out_var_location): Disregard inline entry markers.
2705 (inline_entry_data): New struct.
2706 (inline_entry_data_hasher): New hashtable type.
2707 (inline_entry_data_hasher::hash): New.
2708 (inline_entry_data_hasher::equal): New.
2709 (inline_entry_data_table): New variable.
2710 (add_high_low_attributes): Add DW_AT_entry_pc and
2711 DW_AT_GNU_entry_view attributes if a pending entry is found
2712 in inline_entry_data_table. Add old entry_pc attribute only
2713 if debug nonbinding markers are disabled.
2714 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2715 markers are enabled.
2716 (block_within_block_p, dwarf2out_inline_entry): New.
2717 (dwarf2out_finish): Check that no entries remained in
2718 inline_entry_data_table.
2719 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2720 (final_scan_insn, notice_source_line): Likewise.
2721 (rest_of_clean_state): Skip inline entry markers.
2722 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2723 markers.
2724 * gimple.c (gimple_build_debug_inline_entry): New.
2725 * gimple.h (enum gimple_debug_subcode): Add
2726 GIMPLE_DEBUG_INLINE_ENTRY.
2727 (gimple_build_debug_inline_entry): Declare.
2728 (gimple_debug_inline_entry_p): New.
2729 (gimple_debug_nonbind_marker_p): Adjust.
2730 * insn-notes.def (INLINE_ENTRY): New.
2731 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2732 inline entry marker notes.
2733 (print_insn): Likewise.
2734 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2735 (INSN_DEBUG_MARKER_KIND): Likewise.
2736 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2737 * tree-inline.c (expand_call_inline): Build and insert
2738 debug_inline_entry stmt.
2739 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2740 inline entry blocks early, if nonbind markers are enabled.
2741 (dump_scope_block): Dump fragment info.
2742 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2743 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2744 (gimple_build_debug_inline_entry): New.
2745 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2746 Enable/disable inline entry points too.
2747 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2748 (DEBUG_INSN): Describe inline entry markers.
2750 * common.opt (gvariable-location-views): New.
2751 (gvariable-location-views=incompat5): New.
2752 * config.in: Rebuilt.
2753 * configure: Rebuilt.
2754 * configure.ac: Test assembler for view support.
2755 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2756 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2757 * dwarf2out.c (var_loc_view): New typedef.
2758 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2759 (dwarf2out_locviews_in_attribute): New.
2760 (dwarf2out_locviews_in_loclist): New.
2761 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2762 (enum dw_line_info_opcode): Add LI_adv_address.
2763 (struct dw_line_info_table): Add view.
2764 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2765 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2766 (zero_view_p): New variable.
2767 (ZERO_VIEW_P): New macro.
2768 (output_asm_line_debug_info): New.
2769 (struct var_loc_node): Add view.
2770 (add_AT_view_list, AT_loc_list): New.
2771 (add_var_loc_to_decl): Add view param. Test it against last.
2772 (new_loc_list): Add view params. Record them.
2773 (AT_loc_list_ptr): Handle loc and view lists.
2774 (view_list_to_loc_list_val_node): New.
2775 (print_dw_val): Handle dw_val_class_view_list.
2776 (size_of_die): Likewise.
2777 (value_format): Likewise.
2778 (loc_list_has_views): New.
2779 (gen_llsym): Set vl_symbol too.
2780 (maybe_gen_llsym, skip_loc_list_entry): New.
2781 (dwarf2out_maybe_output_loclist_view_pair): New.
2782 (output_loc_list): Output view list or entries too.
2783 (output_view_list_offset): New.
2784 (output_die): Handle dw_val_class_view_list.
2785 (output_dwarf_version): New.
2786 (output_compilation_unit_header): Use it.
2787 (output_skeleton_debug_sections): Likewise.
2788 (output_rnglists, output_line_info): Likewise.
2789 (output_pubnames, output_aranges): Update version comments.
2790 (output_one_line_info_table): Output view numbers in asm comments.
2791 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2792 Call maybe_gen_llsym.
2793 (loc_list_from_tree_1): Adjust.
2794 (add_AT_location_description): Create view list attribute if
2795 needed, check it's absent otherwise.
2796 (convert_cfa_to_fb_loc_list): Adjust.
2797 (maybe_emit_file): Call output_asm_line_debug_info for test.
2798 (dwarf2out_var_location): Reset views as needed. Precompute
2799 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2800 attribute. Set view.
2801 (new_line_info_table): Reset next view.
2802 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2803 (dwarf2out_source_line): Likewise. Output view resets and labels to
2804 the assembler, or select appropriate line info opcodes.
2805 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2806 (optimize_string_length): Catch it. Adjust.
2807 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2808 dw_val_class_view_list, and remove it if no longer needed.
2809 (hash_loc_list): Hash view numbers.
2810 (loc_list_hasher::equal): Compare them.
2811 (optimize_location_lists): Check whether a view list symbol is
2812 needed, and whether the locview attribute is present, and
2813 whether they match. Remove the locview attribute if no longer
2814 needed.
2815 (index_location_lists): Call skip_loc_list_entry for test.
2816 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2817 Use output_dwarf_version.
2818 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2819 (struct dw_val_node): Add val_view_list.
2820 * final.c (SEEN_NEXT_VIEW): New.
2821 (set_next_view_needed): New.
2822 (clear_next_view_needed): New.
2823 (maybe_output_next_view): New.
2824 (final_start_function): Rename to...
2825 (final_start_function_1): ... this. Take pointer to FIRST,
2826 add SEEN parameter. Emit param bindings in the initial view.
2827 (final_start_function): Reintroduce SEEN-less interface.
2828 (final): Rename to...
2829 (final_1): ... this. Take SEEN parameter. Output final pending
2830 next view at the end.
2831 (final): Reintroduce seen-less interface.
2832 (final_scan_insn): Output pending next view before switching
2833 sections or ending a block. Mark the next view as needed when
2834 outputting variable locations. Notify debug backend of section
2835 changes, and of location view changes.
2836 (rest_of_handle_final): Adjust.
2837 * toplev.c (process_options): Autodetect value for debug variable
2838 location views option. Warn on incompat5 without -gdwarf-5.
2839 * doc/invoke.texi (gvariable-location-views): New.
2840 (gvariable-location-views=incompat5): New.
2841 (gno-variable-location-views): New.
2843 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2845 PR tree-optimization/84136
2846 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2847 that the result of find_edge is non-NULL.
2849 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2851 PR target/83008
2852 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2853 storing integer register in SImode. Fix cost of 256 and 512
2854 byte aligned SSE register store.
2856 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2858 * config/i386/i386.c (ix86_multiplication_cost): Fix
2859 multiplication cost for TARGET_AVX512DQ.
2861 2018-02-08 Marek Polacek <polacek@redhat.com>
2863 PR tree-optimization/84238
2864 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2865 get_range_strlen.
2867 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2869 PR tree-optimization/84265
2870 * tree-vect-stmts.c (vectorizable_store): Don't treat
2871 VMAT_CONTIGUOUS accesses as grouped.
2872 (vectorizable_load): Likewise.
2874 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2876 PR tree-optimization/81635
2877 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2878 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2879 (test_round_for_mask): New functions.
2880 (wide_int_cc_tests): Call test_round_for_mask.
2881 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2882 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2883 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2884 range returned by get_range_info.
2886 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2888 PR ipa/81360
2889 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2890 * symtab.c: Include builtins.h
2891 (symtab_node::output_to_lto_symbol_table_p): Move here
2892 from lto-streamer-out.c:output_symbol_p.
2893 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2894 (output_symbol_p): Move all logic to symtab.c
2895 (produce_symtab): Update.
2897 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2899 * config/s390/s390-opts.h (enum indirect_branch): Define.
2900 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2901 (s390_indirect_branch_via_thunk)
2902 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2903 (enum s390_indirect_branch_type): Define.
2904 * config/s390/s390.c (struct s390_frame_layout, struct
2905 machine_function): Remove.
2906 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2907 (indirect_branch_table_label_no, indirect_branch_table_name):
2908 Define variables.
2909 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2910 (enum s390_indirect_branch_option): Define.
2911 (s390_return_addr_from_memory): New function.
2912 (s390_handle_string_attribute): New function.
2913 (s390_attribute_table): Add new attribute handler.
2914 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2915 (s390_indirect_branch_via_thunk): New function.
2916 (s390_indirect_branch_via_inline_thunk): New function.
2917 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2918 sibling call optimization for non z10 compiles.
2919 (s390_emit_call): Force indirect branch target to be a single
2920 register. Add r1 clobber for non-z10 compiles.
2921 (s390_emit_epilogue): Emit return jump via return_use expander.
2922 (s390_reorg): Handle JUMP_INSNs as execute targets.
2923 (s390_option_override_internal): Perform validity checks for the
2924 new command line options.
2925 (s390_indirect_branch_attrvalue): New function.
2926 (s390_indirect_branch_settings): New function.
2927 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2928 (s390_output_indirect_thunk_function): New function.
2929 (s390_code_end): Implement target hook.
2930 (s390_case_values_threshold): Implement target hook.
2931 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2932 macros.
2933 * config/s390/s390.h (struct s390_frame_layout)
2934 (struct machine_function): Move here from s390.c.
2935 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2936 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2937 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2938 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2939 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2940 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2941 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2942 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2943 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2944 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2945 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2946 (mnemonic attribute): Add values which aren't recognized
2947 automatically.
2948 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2949 pattern for branch conversion. Fix mnemonic attribute.
2950 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2951 indirect branch via thunk if requested.
2952 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2953 ("*indirect_jump"): Disable for branch conversion using out of
2954 line thunks.
2955 ("indirect_jump_via_thunk<mode>_z10")
2956 ("indirect_jump_via_thunk<mode>")
2957 ("indirect_jump_via_inlinethunk<mode>_z10")
2958 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2959 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2960 ("casesi_jump_via_inlinethunk<mode>_z10")
2961 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2962 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2963 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2964 ("*indirect2_jump"): Disable for branch conversion.
2965 ("casesi_jump"): Turn into expander and expand patterns for branch
2966 conversion.
2967 ("return_use"): New expander.
2968 ("*return"): Emit return via thunk and rename it to ...
2969 ("*return<mode>"): ... this one.
2970 * config/s390/s390.opt: Add new options and and enum for the
2971 option values.
2973 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2975 * lra-constraints.c (match_reload): Unconditionally use
2976 gen_lowpart_SUBREG, rather than selecting between that
2977 and equivalent gen_rtx_SUBREG code.
2979 2018-02-08 Richard Biener <rguenther@suse.de>
2981 PR tree-optimization/84233
2982 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2983 changed flag instead of boguously re-using phi_inserted.
2985 2018-02-08 Martin Jambor <mjambor@suse.cz>
2987 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2988 static local variables.
2990 2018-02-08 Richard Biener <rguenther@suse.de>
2992 PR tree-optimization/84278
2993 * tree-vect-stmts.c (vectorizable_store): When looking for
2994 smaller vector types to perform grouped strided loads/stores
2995 make sure the mode is supported by the target.
2996 (vectorizable_load): Likewise.
2998 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3000 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3001 Increase LDP/STP opportunities by adding adjacent callee-saves.
3003 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3005 PR rtl-optimization/84068
3006 PR rtl-optimization/83459
3007 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3009 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3011 PR tree-optimization/84224
3012 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3013 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3014 non-zero arguments.
3016 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3018 PR target/84113
3019 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3020 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3021 count, remove one USE.
3023 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3025 * doc/install.texi (Configuration): Document the
3026 --with-long-double-format={ibm,ieee} PowerPC configuration
3027 options.
3029 PR target/84154
3030 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3031 Convert from define_expand to be define_insn_and_split. Rework
3032 float/double/_Float128 conversions to QI/HI/SImode to work with
3033 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3034 conversions to QI/HImode types did a store and then a load to
3035 truncate the value. For conversions to VSX registers, don't split
3036 the insn, instead emit the code directly. Use the code iterator
3037 any_fix to combine signed and unsigned conversions.
3038 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3039 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3040 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3041 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3042 (fix_<mode>di2_hw): Likewise.
3043 (fixuns_<mode>di2_hw): Likewise.
3044 (fix_<mode>si2_hw): Likewise.
3045 (fixuns_<mode>si2_hw): Likewise.
3046 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3047 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3048 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3049 fix<uns>_trunc<SFDF:mode>si2_p8.
3050 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3051 used.
3052 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3053 (fix<uns>_<mode>_mem): Likewise.
3054 (fctiw<u>z_<mode>_mem): Likewise.
3055 (fix<uns>_<mode>_mem): Likewise.
3056 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3057 the register allocator from doing a direct move to the GPRs to do
3058 a store, and instead use the ISA 3.0 store byte/half-word from
3059 vector register instruction. For IEEE 128-bit floating point,
3060 also optimize stores of 32-bit ints.
3061 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3063 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3065 * genextract.c (push_pathstr_operand): New function to support
3066 [a-zA-Z].
3067 (walk_rtx): Call push_pathstr_operand.
3068 (print_path): Support [a-zA-Z].
3070 2018-02-07 Richard Biener <rguenther@suse.de>
3072 PR tree-optimization/84037
3073 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3074 (cse_and_gimplify_to_preheader): Declare.
3075 (vect_get_place_in_interleaving_chain): Likewise.
3076 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3077 ivexpr_map.
3078 (_loop_vec_info::~_loop_vec_info): Delete it.
3079 (cse_and_gimplify_to_preheader): New function.
3080 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3081 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3082 (vectorizable_load): Likewise. For grouped stores always base
3083 the IV on the first element.
3084 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3085 condition before gimplifying.
3087 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3089 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3090 *DIV_EXPR and *MOD_EXPR.
3092 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3094 PR target/84248
3095 * config/i386/i386.c (ix86_option_override_internal): Mask out
3096 the CF_SET bit when checking -fcf-protection.
3098 2018-02-07 Tom de Vries <tom@codesourcery.com>
3100 PR libgomp/84217
3101 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3102 enough.
3104 2018-02-07 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/84204
3107 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3108 this place.
3110 PR tree-optimization/84205
3111 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3112 special-case isl_ast_op_zdiv_r.
3114 PR tree-optimization/84223
3115 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3116 Only add conditions from within the region.
3117 (gather_bbs::after_dom_children): Adjust.
3119 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3121 PR target/84209
3122 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3123 * config/avr/avr.md: Only post-reload split REG-REG moves if
3124 either register is GENERAL_REG_P.
3126 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3128 PR tree-optimization/84235
3129 * tree-ssa-scopedtables.c
3130 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3131 if the subtraction is performed in floating point type where NaNs are
3132 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3133 build 1. Formatting fix.
3135 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3137 PR target/84146
3138 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3139 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3140 and skip it regardless of bb boundaries. Use CALL_P macro,
3141 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3142 unnecessarily, formatting fix.
3144 2018-02-06 Michael Collison <michael.collison@arm.com>
3146 * config/arm/thumb2.md:
3147 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3148 (*thumb_mov_notscc): Ditto.
3150 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3152 PR target/84154
3153 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3154 unsigned_fix, not "s".
3156 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3158 * configure.ac (gcc_fn_eh_frame_ro): New function.
3159 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3160 correct .eh_frame permissions.
3161 * configure: Regenerate.
3163 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3165 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3166 irrelevant options.
3168 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3170 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3171 Display warning message for -mno-speculate-indirect-jumps.
3173 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3175 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3176 Undocumented.
3177 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3179 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3181 PR tree-optimization/84225
3182 * tree-eh.c (find_trapping_overflow): Only call
3183 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3185 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3187 PR target/84145
3188 * config/i386/i386.c: Reimplement the check of possible options
3189 -mibt/-mshstk conbination. Change error messages.
3190 * doc/invoke.texi: Fix a typo: remove extra '='.
3192 2018-02-06 Marek Polacek <polacek@redhat.com>
3194 PR tree-optimization/84228
3195 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3197 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3199 PR target/82641
3200 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3201 emitted arch directives.
3202 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3203 __ARM_FEATURE_COPROC before changing architectures.
3205 2018-02-06 Richard Biener <rguenther@suse.de>
3207 * config/i386/i386.c (print_reg): Fix typo.
3208 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3210 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3212 * configure: Regenerate.
3214 2018-02-05 Martin Sebor <msebor@redhat.com>
3216 PR tree-optimization/83369
3217 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3218 inlining context.
3220 2018-02-05 Martin Liska <mliska@suse.cz>
3222 * doc/invoke.texi: Cherry-pick upstream r323995.
3224 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3226 * ira.c (ira_init_register_move_cost): Adjust comment.
3228 2018-02-05 Martin Liska <mliska@suse.cz>
3230 PR gcov-profile/84137
3231 * doc/gcov.texi: Fix typo in documentation.
3233 2018-02-05 Martin Liska <mliska@suse.cz>
3235 PR gcov-profile/83879
3236 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3237 using dlopen functionality.
3239 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3241 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3242 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3243 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3244 _mm_maskz_range_round_ss): New intrinsics.
3245 (__builtin_ia32_rangesd128_round)
3246 (__builtin_ia32_rangess128_round): Remove.
3247 (__builtin_ia32_rangesd128_mask_round,
3248 __builtin_ia32_rangess128_mask_round): New builtins.
3249 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3250 __builtin_ia32_rangess128_round): Remove.
3251 (__builtin_ia32_rangesd128_mask_round,
3252 __builtin_ia32_rangess128_mask_round): New builtins.
3253 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3254 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3255 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3256 "<round_saeonly_constraint>")): Changed to ...
3257 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3258 "<round_saeonly_scalar_constraint>")): ... this.
3259 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3260 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3261 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3262 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3263 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3265 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3267 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3268 options.
3269 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3270 Remove all values except native, 8540 and 8548.
3272 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3274 * config/i386/i386.c (ix86_output_function_return): Pass
3275 INVALID_REGNUM, instead of -1, as invalid register number to
3276 indirect_thunk_name and output_indirect_thunk.
3278 2018-02-02 Julia Koval <julia.koval@intel.com>
3280 * config.gcc: Add -march=icelake.
3281 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3283 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3284 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3285 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3286 (processor_target_table): Add icelake.
3287 (ix86_option_override_internal): Handle new PTAs.
3288 (get_builtin_code_for_version): Handle icelake.
3289 (M_INTEL_COREI7_ICELAKE): New.
3290 (fold_builtin_cpu): Handle icelake.
3291 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3292 * doc/invoke.texi: Add -march=icelake.
3294 2018-02-02 Julia Koval <julia.koval@intel.com>
3296 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3297 to wide_int_bitmask.
3298 * wide-int-bitmask.h: New.
3300 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3302 PR target/84066
3303 * config/i386/i386.md: Replace Pmode with word_mode in
3304 builtin_setjmp_setup and builtin_longjmp to support x32.
3306 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3308 PR target/56010
3309 PR target/83743
3310 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3311 #include "opts.h".
3312 (rs6000_supported_cpu_names): New static variable.
3313 (linux_cpu_translation_table): Likewise.
3314 (elf_platform) <cpu>: Define new static variable and use it.
3315 Translate kernel AT_PLATFORM name to canonical name if needed.
3316 Error if platform name is unknown.
3318 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3320 PR target/84089
3321 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3323 2018-02-01 Jeff Law <law@redhat.com>
3325 PR target/84128
3326 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3327 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3328 the scratch if RELEASE_VIA_POP is false.
3329 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3330 If we have to save a temporary register, decrement SIZE appropriately.
3331 Pass new arguments to release_scratch_register_on_entry.
3332 (ix86_adjust_stack_and_probe): Likewise.
3333 (ix86_emit_probe_stack_range): Pass new arguments to
3334 release_scratch_register_on_entry.
3336 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3338 PR rtl-optimization/84157
3339 * combine.c (change_zero_ext): Use REG_P predicate in
3340 front of HARD_REGISTER_P predicate.
3342 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3344 * config/avr/avr.c (avr_option_override): Move disabling of
3345 -fdelete-null-pointer-checks to...
3346 * common/config/avr/avr-common.c (avr_option_optimization_table):
3347 ...here.
3349 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3351 PR tree-optimization/81635
3352 * tree-data-ref.c (split_constant_offset_1): For types that
3353 wrap on overflow, try to use range info to prove that wrapping
3354 cannot occur.
3356 2018-02-01 Renlin Li <renlin.li@arm.com>
3358 PR target/83370
3359 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3360 TAILCALL_ADDR_REGS.
3361 (aarch64_register_move_cost): Likewise.
3362 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3363 TAILCALL_ADDR_REGS.
3364 (REG_CLASS_NAMES): Likewise.
3365 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3366 TAILCALL_ADDR_REGS. Remove IP registers.
3367 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3369 2018-02-01 Richard Biener <rguenther@suse.de>
3371 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3372 for specifying RPO order and allow NULL for that.
3373 * domwalk.c (dom_walker::dom_walker): Likewise.
3374 (dom_walker::walk): Handle NULL RPO order.
3375 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3376 in RPO order.
3377 (rewrite_update_dom_walker): Likewise.
3378 (mark_def_dom_walker): Likewise.
3380 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3382 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3383 (aarch64_maybe_expand_sve_subreg_move): Declare.
3384 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3385 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3386 predicate.
3387 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3388 that are semantically a reverse operation.
3389 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3390 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3391 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3392 functions.
3393 (aarch64_can_change_mode_class): For big-endian, forbid changes
3394 between two SVE modes if they have different element sizes.
3396 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3398 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3399 the TImode handling for big-endian targets.
3401 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3403 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3404 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3405 not just bytes.
3406 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3407 Remove BSWAP handing for big-endian targets and use the form of
3408 LD1RQ appropariate for the mode.
3410 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3412 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3413 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3414 duplicated element.
3416 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3418 PR tearget/83845
3419 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3420 check for operands that need to go through aarch64_sve_reload_be.
3422 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3424 PR tree-optimization/81661
3425 PR tree-optimization/84117
3426 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3427 * tree-eh.c: Include gimplify.h.
3428 (find_trapping_overflow, replace_trapping_overflow,
3429 rewrite_to_non_trapping_overflow): New functions.
3430 * tree-vect-loop.c: Include tree-eh.h.
3431 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3432 * tree-data-ref.c: Include tree-eh.h.
3433 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3435 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3437 PR rtl-optimization/84123
3438 * combine.c (change_zero_ext): Check if hard register satisfies
3439 can_change_dest_mode before calling gen_lowpart_SUBREG.
3441 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3443 PR target/82444
3444 * ira.c (ira_init_register_move_cost): Remove assert.
3446 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3448 PR rtl-optimization/84071
3449 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3450 * doc/tm.texi: Regenerate.
3452 2018-01-31 Richard Biener <rguenther@suse.de>
3454 PR tree-optimization/84132
3455 * tree-data-ref.c (analyze_miv_subscript): Properly
3456 check whether evolution_function_is_affine_multivariate_p
3457 before calling gcd_of_steps_may_divide_p.
3459 2018-01-31 Julia Koval <julia.koval@intel.com>
3461 PR target/83618
3462 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3463 * config/i386/i386.md (rdpid_rex64) New.
3464 (rdpid): Make 32bit only.
3466 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3468 PR lto/84105
3469 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3470 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3472 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3474 Revert
3475 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3477 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3479 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3481 PR rtl-optimization/84071
3482 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3483 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3485 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3487 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3488 (arc_attribute_table): Add 'aux' attribute.
3489 (arc_in_small_data_p): Consider aux like variables.
3490 (arc_is_aux_reg_p): New function.
3491 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3492 (arc_get_aux_arg): New function.
3493 (prepare_move_operands): Handle aux-register access.
3494 (arc_handle_aux_attribute): New function.
3495 * doc/extend.texi (ARC Variable attributes): Add subsection.
3497 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3499 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3500 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3501 (arc_attribute_table): Add 'uncached' attribute.
3502 (arc_print_operand): Print '.di' flag for uncached memory
3503 accesses.
3504 (arc_in_small_data_p): Do not consider for small data the uncached
3505 types.
3506 (arc_is_uncached_mem_p): New function.
3507 * config/arc/predicates.md (compact_store_memory_operand): Check
3508 for uncached memory accesses.
3509 (nonvol_nonimm_operand): Likewise.
3510 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3512 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3514 PR c/84100
3515 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3516 falign-loops=): Add Optimization flag.
3518 2018-01-30 Jeff Law <law@redhat.com>
3520 PR target/84064
3521 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3522 INT_REGISTERS_SAVED. Check it prior to calling
3523 get_scratch_register_on_entry.
3524 (ix86_adjust_stack_and_probe): Similarly.
3525 (ix86_emit_probe_stack_range): Similarly.
3526 (ix86_expand_prologue): Corresponding changes.
3528 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3530 PR target/40411
3531 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3532 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3534 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3536 PR target/84112
3537 * lra-constraints.c (curr_insn_transform): Process AND in the
3538 address.
3540 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3542 PR rtl-optimization/83986
3543 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3544 dependence against last_pending_memory_flush in addition to
3545 pending_jump_insns.
3547 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3549 PR tree-optimization/81611
3550 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3551 copies.
3553 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3555 PR target/83758
3556 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3557 a reg rtx.
3559 2018-01-30 Richard Biener <rguenther@suse.de>
3560 Jakub Jelinek <jakub@redhat.com>
3562 PR tree-optimization/84111
3563 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3564 inner loops added during recursion, as they don't have up-to-date
3565 SSA form.
3567 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3569 PR ipa/81360
3570 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3571 (can_inline_edge_by_limits_p): ... here.
3572 (can_early_inline_edge_p, check_callers,
3573 update_caller_keys, update_callee_keys, recursive_inlining,
3574 add_new_edges_to_heap, speculation_useful_p,
3575 inline_small_functions,
3576 inline_small_functions, flatten_function,
3577 inline_to_all_callers_1): Update.
3579 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3581 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3582 zeros correctly.
3584 2018-01-30 Richard Biener <rguenther@suse.de>
3586 PR tree-optimization/83008
3587 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3588 invariant and constant vector uses in stmts when they need
3589 more than one stmt.
3591 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3593 PR bootstrap/84017
3594 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3595 * configure: Regenerate.
3597 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3599 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3600 pattern.
3601 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3602 Use gen_rtx_REG rather than gen_lowpart.
3604 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3606 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3607 rather than 0 when creating partial subregs.
3609 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3611 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3612 of usage.
3614 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3616 PR target/81550
3617 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3618 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3619 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3620 flags. This restores the settings used before the 2017-07-24.
3621 Turning off pre increment/decrement/modify allows IVOPTS to
3622 optimize DF/SF loops where the index is an int.
3624 2018-01-29 Richard Biener <rguenther@suse.de>
3625 Kelvin Nilsen <kelvin@gcc.gnu.org>
3627 PR bootstrap/80867
3628 * tree-vect-stmts.c (vectorizable_call): Don't call
3629 targetm.vectorize_builtin_md_vectorized_function if callee is
3630 NULL.
3632 2018-01-22 Carl Love <cel@us.ibm.com>
3634 * doc/extend.tex: Fix typo in second arg in
3635 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3637 2018-01-29 Richard Biener <rguenther@suse.de>
3639 PR tree-optimization/84086
3640 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3641 (flush_ssaname_freelist): When SSA names were released reset
3642 the SCEV hash table.
3644 2018-01-29 Richard Biener <rguenther@suse.de>
3646 PR tree-optimization/84057
3647 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3648 removed paths when removing edges.
3650 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3652 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3653 -mfunction-return=@var{choice}.
3655 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3657 PR diagnostic/84034
3658 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3659 Handle CR like TAB.
3660 (layout::print_source_line): Likewise.
3661 (test_get_line_width_without_trailing_whitespace): Add test cases.
3663 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3665 PR middle-end/84040
3666 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3667 debug insns.
3669 2018-01-26 Jim Wilson <jimw@sifive.com>
3671 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3673 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3674 specified.
3676 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3678 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3679 and CMP + SUB-immediate -> SUBS.
3681 2018-01-26 Martin Sebor <msebor@redhat.com>
3683 PR tree-optimization/83896
3684 * tree-ssa-strlen.c (get_string_len): Rename...
3685 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3686 Avoid assuming length is constant.
3687 (handle_char_store): Use HOST_WIDE_INT for string length.
3689 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3691 PR target/81763
3692 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3693 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3695 2018-01-26 Richard Biener <rguenther@suse.de>
3697 PR rtl-optimization/84003
3698 * dse.c (record_store): Only record redundant stores when
3699 the earlier store aliases at least all accesses the later one does.
3701 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3703 PR rtl-optimization/83985
3704 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3705 REG_CFA_RESTORE insns.
3706 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3707 REG_CFA_RESTORE insns here.
3709 PR c/83989
3710 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3711 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3713 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3715 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3716 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3717 (arc_init): Likewise.
3718 (arc_override_options): Likewise.
3719 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3720 value.
3721 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3722 support.
3723 * config/arc/arc.h (TARGET_DBNZ): Define.
3724 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3725 properly set the tune attribute.
3726 (dbnz): Use TARGET_DBNZ guard.
3727 * config/arc/arc.opt (mtune): Add core3 option.
3729 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3731 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3732 recognize new pic like addresses.
3733 (arc_delegitimize_address): Clean up.
3735 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3737 * config/arc/arc-arches.def: Option mrf16 valid for all
3738 architectures.
3739 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3740 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3741 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3742 * config/arc/arc-tables.opt: Regenerate.
3743 * config/arc/arc.c (arc_conditional_register_usage): Handle
3744 reduced register file case.
3745 (arc_file_start): Set must have build attributes.
3746 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3747 mrf16 option value.
3748 * config/arc/arc.opt (mrf16): Add new option.
3749 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3750 * config/arc/genmultilib.awk: Handle new mrf16 option.
3751 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3752 * config/arc/t-multilib: Regenerate.
3753 * doc/invoke.texi (ARC Options): Document mrf16 option.
3755 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3757 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3758 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3759 (arc_attribute_table): Add 'secure_call' attribute.
3760 (arc_print_operand): Print secure call operand.
3761 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3762 secure.
3763 (arc_is_secure_call_p): New function. * config/arc/arc.md
3764 (call_i): Add support for sjli instruction.
3765 (call_value_i): Likewise.
3766 * config/arc/constraints.md (Csc): New constraint.
3768 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3769 John Eric Martin <John.Martin@emmicro-us.com>
3771 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3772 * config/arc/arc.c (_arc_jli_section): New struct.
3773 (arc_jli_section): New type.
3774 (rc_jli_sections): New static variable.
3775 (arc_handle_jli_attribute): New function.
3776 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3777 (arc_file_end): New function.
3778 (TARGET_ASM_FILE_END): Define.
3779 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3780 (arc_add_jli_section): New function.
3781 (jli_call_scan): Likewise.
3782 (arc_reorg): Call jli_call_scan.
3783 (arc_output_addsi): Remove 'S' from printing asm operand.
3784 (arc_is_jli_call_p): New function.
3785 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3786 operand.
3787 (movhi_insn): Likewise.
3788 (movsi_insn): Likewise.
3789 (movsi_set_cc_insn): Likewise.
3790 (loadqi_update): Likewise.
3791 (load_zeroextendqisi_update): Likewise.
3792 (load_signextendqisi_update): Likewise.
3793 (loadhi_update): Likewise.
3794 (load_zeroextendhisi_update): Likewise.
3795 (load_signextendhisi_update): Likewise.
3796 (loadsi_update): Likewise.
3797 (loadsf_update): Likewise.
3798 (movsicc_insn): Likewise.
3799 (bset_insn): Likewise.
3800 (bxor_insn): Likewise.
3801 (bclr_insn): Likewise.
3802 (bmsk_insn): Likewise.
3803 (bicsi3_insn): Likewise.
3804 (cmpsi_cc_c_insn): Likewise.
3805 (movsi_ne): Likewise.
3806 (movsi_cond_exec): Likewise.
3807 (clrsbsi2): Likewise.
3808 (norm_f): Likewise.
3809 (normw): Likewise.
3810 (swap): Likewise.
3811 (divaw): Likewise.
3812 (flag): Likewise.
3813 (sr): Likewise.
3814 (kflag): Likewise.
3815 (ffs): Likewise.
3816 (ffs_f): Likewise.
3817 (fls): Likewise.
3818 (call_i): Remove 'S' asm letter, add jli instruction.
3819 (call_value_i): Likewise.
3820 * config/arc/arc.op (mjli-always): New option.
3821 * config/arc/constraints.md (Cji): New constraint.
3822 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3823 operand.
3824 (subsf3_fpx): Likewise.
3825 (mulsf3_fpx): Likewise.
3826 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3827 asm operand.
3828 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3829 function attrbutes.
3830 * doc/invoke.texi (ARC): Document mjli-always option.
3832 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3834 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3835 avoid addition with 0 and use incw and decw where possible.
3837 2018-01-26 Richard Biener <rguenther@suse.de>
3839 PR tree-optimization/81082
3840 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3841 association if it requires casting to unsigned.
3842 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3843 from fold_plusminus_mult_expr to catch important cases late when
3844 range info is available.
3846 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3848 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3849 * configure.ac (hidden_linkonce): New test.
3850 * configure: Regenerate.
3851 * config.in: Regenerate.
3853 2018-01-26 Julia Koval <julia.koval@intel.com>
3855 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3856 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3857 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3858 _mm_mask_bitshuffle_epi64_mask): Fix type.
3859 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3860 USI_FTYPE_V4DI_V4DI_USI): Remove.
3861 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3862 __builtin_ia32_vpshufbitqmb256_mask,
3863 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3864 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3865 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3867 2018-01-26 Alan Modra <amodra@gmail.com>
3869 PR target/84033
3870 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3871 UNSPEC_VBPERMQ. Sort other unspecs.
3873 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3875 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3877 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3879 PR middle-end/83055
3880 * predict.c (drop_profile): Do not push/pop cfun; update also
3881 node->count.
3882 (handle_missing_profiles): Fix logic looking for zero profiles.
3884 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3886 PR middle-end/83977
3887 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3888 on functions with #pragma omp declare simd or functions with simd
3889 attribute.
3890 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3891 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3892 Remove trailing \n from warning_at calls.
3894 2018-01-25 Tom de Vries <tom@codesourcery.com>
3896 PR target/84028
3897 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3898 for neutered workers.
3900 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3902 PR target/68467
3903 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3904 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3906 2018-01-24 Jeff Law <law@redhat.com>
3908 PR target/83994
3909 * i386.c (get_probe_interval): Move to earlier point.
3910 (ix86_compute_frame_layout): If -fstack-clash-protection and
3911 the frame is larger than the probe interval, then use pushes
3912 to save registers rather than reg->mem moves.
3913 (ix86_expand_prologue): Remove conditional for int_registers_saved
3914 assertion.
3916 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3918 PR target/84014
3919 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3920 min/max for never referenced object.
3922 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3924 PR middle-end/83977
3925 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3926 here.
3927 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3928 attributes from DECL_ATTRIBUTES (decl) without affecting
3929 DECL_ATTRIBUTES (current_function_decl).
3930 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3931 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3933 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3935 PR tree-optimization/83979
3936 * fold-const.c (fold_comparison): Use constant_boolean_node
3937 instead of boolean_{true,false}_node.
3939 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3941 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3942 with zero counts.
3944 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3946 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3947 Simplify the clause that sets the length attribute.
3948 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3949 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3950 clause that sets the length attribute.
3951 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3953 2018-01-24 Tom de Vries <tom@codesourcery.com>
3955 PR target/83589
3956 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3957 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3958 Add strict parameter.
3959 (prevent_branch_around_nothing): Insert dummy insn between branch to
3960 label and label with no ptx insn inbetween.
3961 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3963 2018-01-24 Tom de Vries <tom@codesourcery.com>
3965 PR target/81352
3966 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3967 for neutered threads in warp.
3968 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3970 2018-01-24 Richard Biener <rguenther@suse.de>
3972 PR tree-optimization/83176
3973 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3974 operands.
3976 2018-01-24 Richard Biener <rguenther@suse.de>
3978 PR tree-optimization/82819
3979 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3980 code generating pluses that are no-ops in the target precision.
3982 2018-01-24 Richard Biener <rguenther@suse.de>
3984 PR middle-end/84000
3985 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3987 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3989 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3990 to merge probabilities.
3991 * predict.c (probably_never_executed): Also mark as cold functions
3992 with global 0 profile and guessed local profile.
3993 * profile-count.c (profile_probability::combine_with_count): New
3994 member function.
3995 * profile-count.h (profile_probability::operator*,
3996 profile_probability::operator*=, profile_probability::operator/,
3997 profile_probability::operator/=): Reduce precision to adjusted
3998 and set value to guessed on contradictory divisions.
3999 (profile_probability::combine_with_freq): Remove.
4000 (profile_probability::combine_wiht_count): Declare.
4001 (profile_count::force_nonzero):: Set to adjusted.
4002 (profile_count::probability_in):: Set quality to adjusted.
4003 * tree-ssa-tail-merge.c (replace_block_by): Use
4004 combine_with_count.
4006 2018-01-23 Andrew Waterman <andrew@sifive.com>
4007 Jim Wilson <jimw@sifive.com>
4009 * config/riscv/riscv.c (riscv_stack_boundary): New.
4010 (riscv_option_override): Set riscv_stack_boundary. Handle
4011 riscv_preferred_stack_boundary_arg.
4012 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4013 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4014 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4015 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4016 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4017 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4019 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4021 PR target/83905
4022 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4023 of struct ix86_frame.
4024 (ix86_expand_epilogue): Likewise. Add a local variable for
4025 the reg_save_offset field in struct ix86_frame.
4027 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4029 PR tree-optimization/82604
4030 * tree-loop-distribution.c (enum partition_kind): New enum item
4031 PKIND_PARTIAL_MEMSET.
4032 (partition_builtin_p): Support above new enum item.
4033 (generate_code_for_partition): Ditto.
4034 (compute_access_range): Differentiate cases that equality can be
4035 proven at all loops, the innermost loops or no loops.
4036 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4037 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4038 (finalize_partitions, distribute_loop): Don't fuse partition of
4039 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4040 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4041 parloop is enabled.
4043 2018-01-23 Martin Liska <mliska@suse.cz>
4045 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4046 order to ignore the predictor.
4047 (PRED_POLYMORPHIC_CALL): Likewise.
4048 (PRED_RECURSIVE_CALL): Likewise.
4050 2018-01-23 Martin Liska <mliska@suse.cz>
4052 * tree-profile.c (tree_profiling): Print function header to
4053 aware reader which function we are working on.
4054 * value-prof.c (gimple_find_values_to_profile): Do not print
4055 not interesting value histograms.
4057 2018-01-23 Martin Liska <mliska@suse.cz>
4059 * profile-count.h (enum profile_quality): Add
4060 profile_uninitialized as the first value. Do not number values
4061 as they are zero based.
4062 (profile_count::verify): Update sanity check.
4063 (profile_probability::verify): Likewise.
4065 2018-01-23 Nathan Sidwell <nathan@acm.org>
4067 * doc/invoke.texi (ffor-scope): Deprecate.
4069 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4071 PR tree-optimization/83510
4072 * domwalk.c (set_all_edges_as_executable): New function.
4073 (dom_walker::dom_walker): Convert bool param
4074 "skip_unreachable_blocks" to enum reachability. Move setup of
4075 edge flags to set_all_edges_as_executable and only do it when
4076 reachability is REACHABLE_BLOCKS.
4077 * domwalk.h (enum dom_walker::reachability): New enum.
4078 (dom_walker::dom_walker): Convert bool param
4079 "skip_unreachable_blocks" to enum reachability.
4080 (set_all_edges_as_executable): New decl.
4081 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4082 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4083 "reachability".
4084 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4085 but converting true to REACHABLE_BLOCKS.
4086 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4087 * tree-vrp.c
4088 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4089 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4090 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4091 REACHABLE_BLOCKS.
4092 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4093 if check_all_array_refs will be called.
4095 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4097 * tree.c (selftest::test_location_wrappers): Add more test
4098 coverage.
4100 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4102 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4103 (selftest::test_bit_in_range): Likewise.
4105 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4107 PR testsuite/83888
4108 * doc/sourcebuild.texi (vect_float): Say that the selector
4109 only describes the situation when -funsafe-math-optimizations is on.
4110 (vect_float_strict): Document.
4112 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4114 PR tree-optimization/83965
4115 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4116 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4117 instead of checking only for a reduction.
4118 (vect_recog_widen_sum_pattern): Likewise.
4120 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4122 * predict.c (probably_never_executed): Only use precise profile info.
4123 (compute_function_frequency): Skip after inlining hack since we now
4124 have quality checking.
4126 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4128 * profile-count.h (profile_probability::very_unlikely,
4129 profile_probability::unlikely, profile_probability::even): Set
4130 precision to guessed.
4132 2018-01-23 Richard Biener <rguenther@suse.de>
4134 PR tree-optimization/83963
4135 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4136 Properly terminate dominator walk when crossing the exit edge not
4137 when visiting its source block.
4139 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4141 PR c++/83918
4142 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4143 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4145 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4147 PR tree-optimization/83957
4148 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4149 semicolon after for body surrounded by braces.
4151 PR tree-optimization/83081
4152 * profile-count.h (profile_probability::split): New method.
4153 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4154 Use profile_probability::split.
4155 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4156 when splitting a single conditional jump into 2.
4158 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4160 PR tree-optimization/69452
4161 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4162 decl.
4164 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4166 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4167 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4168 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4172 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4173 * config/rl78/rl78.md (movdi): New define_expand.
4174 * config/rl78/rl78.c (rl78_split_movdi): New function.
4176 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4178 PR target/83862
4179 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4180 no longer used.
4181 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4182 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4183 128-bit to produce an UNSPEC move to get the double word with the
4184 signbit and then a shift directly to do signbit.
4185 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4186 implementation with a new version that just does either a direct
4187 move or a regular move. Move memory interface to separate insns.
4188 Move insns so they are next to the expander.
4189 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4190 with signbit move. Split big and little endian case.
4191 (signbit<mode>2_dm_mem_le): Likewise.
4192 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4193 (signbit<mode>2_dm2): Likewise.
4195 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4197 * config/rl78/rl78.md (anddi3): New define_expand.
4199 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4201 * config/rl78/rl78.md (umindi3): New define_expand.
4203 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4205 * config/rl78/rl78.md (smindi3): New define_expand.
4207 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4209 * config/rl78/rl78.md (smaxdi3): New define_expand.
4211 2018-01-22 Carl Love <cel@us.ibm.com>
4213 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4214 LVX_V1TI): Add macro expansion.
4215 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4216 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4217 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4218 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4219 Change check to determine if the instruction is a byte reversing
4220 entry. Fix typo in comment.
4221 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4222 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4223 Add def_builtin calls for new builtins.
4224 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4225 Add define_insn expansion.
4227 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4229 * config/rl78/rl78.md (umaxdi3): New define_expand.
4231 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4233 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4234 for non-QImode registers.
4236 2018-01-22 Richard Biener <rguenther@suse.de>
4238 PR tree-optimization/83963
4239 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4240 including the loop exit block.
4241 (scop_detection::merge_sese): Likewise.
4242 (scop_detection::add_scop): Do it here instead.
4244 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4246 * doc/sourcebuild.texi (arm_softfloat): Document.
4248 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4250 PR gcc/77734
4251 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4252 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4253 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4255 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4256 David Edelsohn <dje.gcc@gmail.com>
4258 PR target/83946
4259 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4260 Change "crset eq" to "crset 2".
4261 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4262 (*call_indirect_aix<mode>_nospec): Likewise.
4263 (*call_value_indirect_aix<mode>_nospec): Likewise.
4264 (*call_indirect_elfv2<mode>_nospec): Likewise.
4265 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4266 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4267 change assembly output from . to $.
4268 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4269 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4270 (*tablejump<mode>_internal1_nospec): Likewise.
4272 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4274 PR target/80870
4275 * config/sh/sh_optimize_sett_clrt.cc:
4276 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4278 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4280 PR tree-optimization/83940
4281 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4282 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4283 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4284 use it to pass back the definition type.
4285 (vect_check_store_rhs): Likewise rhs_dt_out.
4286 (vect_build_gather_load_calls): Add a mask_dt argument and use
4287 it instead of a call to vect_is_simple_use.
4288 (vectorizable_store): Update calls to vect_check_load_store_mask
4289 and vect_check_store_rhs. Use the dt returned by the latter instead
4290 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4291 instead of calls to vect_is_simple_use. Pass the scalar rather
4292 than the vector operand to vect_is_simple_use when handling
4293 second and subsequent copies of an rhs value.
4294 (vectorizable_load): Update calls to vect_check_load_store_mask
4295 and vect_build_gather_load_calls. Use the cached mask_dt and
4296 gs_info.offset_dt instead of calls to vect_is_simple_use.
4298 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4300 PR middle-end/83945
4301 * tree-emutls.c: Include gimplify.h.
4302 (lower_emutls_2): New function.
4303 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4304 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4305 it before further processing.
4307 PR target/83930
4308 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4309 UINTVAL (trueop1) instead of INTVAL (op1).
4311 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4313 PR debug/81570
4314 PR debug/83728
4315 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4316 INCOMING_FRAME_SP_OFFSET if not defined.
4317 (scan_trace): Add ENTRY argument. If true and
4318 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4319 emit a note to adjust the CFA offset.
4320 (create_cfi_notes): Adjust scan_trace callers.
4321 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4322 INCOMING_FRAME_SP_OFFSET in the CIE.
4323 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4324 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4325 Likewise.
4326 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4327 * doc/tm.texi: Regenerated.
4329 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4331 PR rtl-optimization/83147
4332 * lra-constraints.c (remove_inheritance_pseudos): Use
4333 lra_substitute_pseudo_within_insn.
4335 2018-01-19 Tom de Vries <tom@codesourcery.com>
4336 Cesar Philippidis <cesar@codesourcery.com>
4338 PR target/83920
4339 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4341 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4343 PR target/83790
4344 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4345 spaces for function labels.
4347 2018-01-19 Martin Liska <mliska@suse.cz>
4349 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4350 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4351 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4352 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4353 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4354 (PRED_CONST_RETURN): Change from 69 to 65.
4355 (PRED_NULL_RETURN): Change from 91 to 71.
4356 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4357 (PRED_LOOP_GUARD): Change from 66 to 73.
4359 2018-01-19 Martin Liska <mliska@suse.cz>
4361 * predict.c (predict_insn_def): Add new assert.
4362 (struct branch_predictor): Change type to signed integer.
4363 (test_prediction_value_range): Amend test to cover
4364 PROB_UNINITIALIZED.
4365 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4366 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4367 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4368 (PRED_LOOP_IV_COMPARE): Likewise.
4369 * predict.h (PROB_UNINITIALIZED): Define new constant.
4371 2018-01-19 Martin Liska <mliska@suse.cz>
4373 * predict.c (dump_prediction): Add new format for
4374 analyze_brprob.py script which is enabled with -details
4375 suboption.
4376 * profile-count.h (precise_p): New function.
4378 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4380 PR tree-optimization/83922
4381 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4382 there are no statements that need masking.
4383 (vect_active_double_reduction_p): New function.
4384 (vect_analyze_loop_operations): Use it when handling phis that
4385 are not in the loop header.
4387 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4389 PR tree-optimization/83914
4390 * tree-vect-loop.c (vectorizable_induction): Don't convert
4391 init_expr or apply the peeling adjustment for inductions
4392 that are nested within the vectorized loop.
4394 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4396 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4397 instead of NEG.
4399 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4401 PR sanitizer/81715
4402 PR testsuite/83882
4403 * function.h (gimplify_parameters): Add gimple_seq * argument.
4404 * function.c: Include gimple.h and options.h.
4405 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4406 for the added local temporaries if needed.
4407 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4408 if there are any parameter cleanups, wrap whole body into a
4409 try/finally with the cleanups.
4411 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4413 PR target/82964
4414 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4415 Use GET_MODE_CLASS for scalar floating point.
4417 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4419 PR ipa/82256
4420 patch by PaX Team
4421 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4422 Fix call of call_cgraph_insertion_hooks.
4424 2018-01-18 Martin Sebor <msebor@redhat.com>
4426 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4428 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4430 PR ipa/83619
4431 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4432 frequencies.
4434 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4436 PR other/70268
4437 * common.opt: (-ffile-prefix-map): New option.
4438 * opts.c (common_handle_option): Defer it.
4439 * opts-global.c (handle_common_deferred_options): Handle it.
4440 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4441 * file-prefix-map.h: New file.
4442 (remap_debug_filename, add_debug_prefix_map): ...here.
4443 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4444 * final.c (debug_prefix_map, add_debug_prefix_map
4445 remap_debug_filename): Move to...
4446 * file-prefix-map.c: New file.
4447 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4448 generalize, get rid of alloca(), use strrchr() instead of strchr().
4449 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4450 Implement in terms of add_prefix_map().
4451 (remap_macro_filename, remap_debug_filename): Implement in term of
4452 remap_filename().
4453 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4454 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4455 * dbxout.c: Include file-prefix-map.h.
4456 * varasm.c: Likewise.
4457 * vmsdbgout.c: Likewise.
4458 * xcoffout.c: Likewise.
4459 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4460 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4461 * doc/invoke.texi (-ffile-prefix-map): Document.
4462 (-fdebug-prefix-map): Update description.
4464 2018-01-18 Martin Liska <mliska@suse.cz>
4466 * config/i386/i386.c (indirect_thunk_name): Document that also
4467 lfence is emitted.
4468 (output_indirect_thunk): Document why both instructions
4469 (pause and lfence) are generated.
4471 2018-01-18 Richard Biener <rguenther@suse.de>
4473 PR tree-optimization/83887
4474 * graphite-scop-detection.c
4475 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4476 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4477 (scop_detection::merge_sese): Re-implement with a flood-fill
4478 algorithm that properly finds a SESE region if it exists.
4480 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4482 PR c/61240
4483 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4484 pointer_diff optimizations use view_convert instead of convert.
4486 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4488 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4489 Generate different code for -mno-speculate-indirect-jumps.
4490 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4491 (*call_indirect_aix<mode>): Disable for
4492 -mno-speculate-indirect-jumps.
4493 (*call_indirect_aix<mode>_nospec): New define_insn.
4494 (*call_value_indirect_aix<mode>): Disable for
4495 -mno-speculate-indirect-jumps.
4496 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4497 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4498 -mno-speculate-indirect-jumps.
4499 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4501 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4503 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4504 long double type, set the flags for noting the default long double
4505 type, even if we don't pass or return a long double type.
4507 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4509 PR ipa/83051
4510 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4511 failure.
4513 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4515 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4516 support for merge[hl].
4517 (fold_mergehl_helper): New helper function.
4518 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4519 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4520 (altivec_vmrglw_direct): Add xxmrglw insn.
4522 2018-01-17 Andrew Waterman <andrew@sifive.com>
4524 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4525 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4527 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4529 PR lto/83121
4530 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4531 call the lto_location_cache before reading the
4532 DECL_SOURCE_LOCATION of the types.
4534 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4535 Richard Sandiford <richard.sandiford@linaro.org>
4537 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4538 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4539 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4540 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4541 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4542 Add declaration.
4543 * config/aarch64/constraints.md (aarch64_movti_operand):
4544 Limit immediates.
4545 * config/aarch64/predicates.md (Uti): Add new constraint.
4547 2018-01-17 Carl Love <cel@us.ibm.com>
4549 * config/rs6000/vsx.md (define_expand xl_len_r,
4550 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4551 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4552 lxvll.
4553 (define_expand, define_insn): Move the shift left from the
4554 define_insn to the define_expand for lxvl and stxvl instructions.
4555 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4556 and XL_LEN_R definitions to PURE.
4558 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4560 * config/i386/i386.c (indirect_thunk_name): Declare regno
4561 as unsigned int. Compare regno with INVALID_REGNUM.
4562 (output_indirect_thunk): Ditto.
4563 (output_indirect_thunk_function): Ditto.
4564 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4565 in the call to output_indirect_thunk_function.
4567 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4569 PR middle-end/83884
4570 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4571 rather than the size of inner_type to determine the stack slot size
4572 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4574 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4576 PR target/83546
4577 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4578 to PTA_SILVERMONT.
4580 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4582 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4583 endian Linux systems to optionally enable multilibs for selecting
4584 the long double type if the user configured an explicit type.
4585 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4586 have no long double multilibs if not defined.
4587 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4588 warn if the user used -mabi={ieee,ibm}longdouble and we built
4589 multilibs for long double.
4590 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4591 appropriate multilib option.
4592 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4593 multilib options.
4594 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4595 for building long double multilibs.
4596 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4598 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4600 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4601 copies.
4603 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4604 64 bits.
4605 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4606 128 bits.
4608 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4609 variables.
4611 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4612 return value.
4614 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4616 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4617 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4619 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4621 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4622 different rtl trees depending on TARGET_64BIT.
4623 (rs6000_gen_lvx): Likewise.
4625 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4627 * config/visium/visium.md (nop): Tweak comment.
4628 (hazard_nop): Likewise.
4630 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4632 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4633 -mspeculate-indirect-jumps.
4634 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4635 for -mno-speculate-indirect-jumps.
4636 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4637 (*call_value_indirect_elfv2<mode>): Disable for
4638 -mno-speculate-indirect-jumps.
4639 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4640 (indirect_jump): Emit different RTL for
4641 -mno-speculate-indirect-jumps.
4642 (*indirect_jump<mode>): Disable for
4643 -mno-speculate-indirect-jumps.
4644 (*indirect_jump<mode>_nospec): New define_insn.
4645 (tablejump): Emit different RTL for
4646 -mno-speculate-indirect-jumps.
4647 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4648 (tablejumpsi_nospec): New define_expand.
4649 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4650 (tablejumpdi_nospec): New define_expand.
4651 (*tablejump<mode>_internal1): Disable for
4652 -mno-speculate-indirect-jumps.
4653 (*tablejump<mode>_internal1_nospec): New define_insn.
4654 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4655 option.
4657 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4659 * caller-save.c (insert_save): Drop unnecessary parameter. All
4660 callers updated.
4662 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4663 Richard Biener <rguenth@suse.de>
4665 PR libgomp/83590
4666 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4667 return early, inline manually is_gimple_sizepos. Make sure if we
4668 call gimplify_expr we don't end up with a gimple constant.
4669 * tree.c (variably_modified_type_p): Don't return true for
4670 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4671 * gimplify.h (is_gimple_sizepos): Remove.
4673 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4675 PR tree-optimization/83857
4676 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4677 vectorizable_live_operation for pure SLP statements.
4678 (vectorizable_live_operation): Handle PHIs.
4680 2018-01-16 Richard Biener <rguenther@suse.de>
4682 PR tree-optimization/83867
4683 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4684 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4686 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4688 PR c/83844
4689 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4690 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4691 If off is not INTEGER_CST, issue a may not be aligned warning
4692 rather than isn't aligned. Use isn%'t rather than isn't.
4693 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4694 into MULT_EXPR.
4695 <case MULT_EXPR>: Improve the case when bottom and one of the
4696 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4697 operand, in that case check if the other operand is multiple of
4698 bottom divided by the INTEGER_CST operand.
4700 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4702 PR target/83858
4703 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4704 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4705 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4706 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4707 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4708 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4709 (pa_function_arg_size): New function.
4711 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4713 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4714 in a separate statement.
4716 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4718 PR tree-optimization/83847
4719 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4720 group gathers and scatters.
4722 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4724 PR rtl-optimization/86620
4725 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4727 PR rtl-optimization/83213
4728 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4729 to last if both are JUMP_INSNs.
4731 PR tree-optimization/83843
4732 * gimple-ssa-store-merging.c
4733 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4734 store_immediate_info for bswap/nop orig_stores.
4736 2018-01-15 Andrew Waterman <andrew@sifive.com>
4738 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4739 !TARGET_MUL.
4740 <UDIV>: Increase cost if !TARGET_DIV.
4742 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4744 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4745 (define_attr "cr_logical_3op"): New.
4746 (cceq_ior_compare): Adjust.
4747 (cceq_ior_compare_complement): Adjust.
4748 (*cceq_rev_compare): Adjust.
4749 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4750 (is_cracked_insn): Adjust.
4751 (insn_must_be_first_in_group): Adjust.
4752 * config/rs6000/40x.md: Adjust.
4753 * config/rs6000/440.md: Adjust.
4754 * config/rs6000/476.md: Adjust.
4755 * config/rs6000/601.md: Adjust.
4756 * config/rs6000/603.md: Adjust.
4757 * config/rs6000/6xx.md: Adjust.
4758 * config/rs6000/7450.md: Adjust.
4759 * config/rs6000/7xx.md: Adjust.
4760 * config/rs6000/8540.md: Adjust.
4761 * config/rs6000/cell.md: Adjust.
4762 * config/rs6000/e300c2c3.md: Adjust.
4763 * config/rs6000/e500mc.md: Adjust.
4764 * config/rs6000/e500mc64.md: Adjust.
4765 * config/rs6000/e5500.md: Adjust.
4766 * config/rs6000/e6500.md: Adjust.
4767 * config/rs6000/mpc.md: Adjust.
4768 * config/rs6000/power4.md: Adjust.
4769 * config/rs6000/power5.md: Adjust.
4770 * config/rs6000/power6.md: Adjust.
4771 * config/rs6000/power7.md: Adjust.
4772 * config/rs6000/power8.md: Adjust.
4773 * config/rs6000/power9.md: Adjust.
4774 * config/rs6000/rs64.md: Adjust.
4775 * config/rs6000/titan.md: Adjust.
4777 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4779 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4780 ix86_indirect_branch_register logic.
4782 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4784 * config/i386/constraints.md (Bs): Update
4785 ix86_indirect_branch_register check. Don't check
4786 ix86_indirect_branch_register with GOT_memory_operand.
4787 (Bw): Likewise.
4788 * config/i386/predicates.md (GOT_memory_operand): Don't check
4789 ix86_indirect_branch_register here.
4790 (GOT32_symbol_operand): Likewise.
4792 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4794 * config/i386/predicates.md (constant_call_address_operand):
4795 Rewrite ix86_indirect_branch_register logic.
4796 (sibcall_insn_operand): Likewise.
4798 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4800 * config/i386/constraints.md (Bs): Replace
4801 ix86_indirect_branch_thunk_register with
4802 ix86_indirect_branch_register.
4803 (Bw): Likewise.
4804 * config/i386/i386.md (indirect_jump): Likewise.
4805 (tablejump): Likewise.
4806 (*sibcall_memory): Likewise.
4807 (*sibcall_value_memory): Likewise.
4808 Peepholes of indirect call and jump via memory: Likewise.
4809 * config/i386/i386.opt: Likewise.
4810 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4811 (GOT_memory_operand): Likewise.
4812 (call_insn_operand): Likewise.
4813 (sibcall_insn_operand): Likewise.
4814 (GOT32_symbol_operand): Likewise.
4816 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4818 PR middle-end/83837
4819 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4820 type rather than type addr's type points to.
4821 (expand_omp_atomic_mutex): Likewise.
4822 (expand_omp_atomic): Likewise.
4824 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4826 PR target/83839
4827 * config/i386/i386.c (output_indirect_thunk_function): Use
4828 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4829 for __x86_return_thunk.
4831 2018-01-15 Richard Biener <rguenther@suse.de>
4833 PR middle-end/83850
4834 * expmed.c (extract_bit_field_1): Fix typo.
4836 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4838 PR target/83687
4839 * config/arm/iterators.md (VF): New mode iterator.
4840 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4841 Remove integer-related logic from pattern.
4842 (neon_vabd<mode>_3): Likewise.
4844 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4846 PR middle-end/82694
4847 * common.opt (fstrict-overflow): No longer an alias.
4848 (fwrapv-pointer): New option.
4849 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4850 also for pointer types based on flag_wrapv_pointer.
4851 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4852 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4853 opts->x_flag_wrapv got set.
4854 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4855 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4856 POINTER_TYPE_OVERFLOW_UNDEFINED.
4857 * match.pd: Likewise in address comparison pattern.
4858 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4860 2018-01-15 Richard Biener <rguenther@suse.de>
4862 PR lto/83804
4863 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4864 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4865 Reset type names to their identifier if their TYPE_DECL doesn't
4866 have linkage (and thus is used for ODR and devirt).
4867 (save_debug_info_for_decl): Remove.
4868 (save_debug_info_for_type): Likewise.
4869 (add_tree_to_fld_list): Adjust.
4870 * tree-pretty-print.c (dump_generic_node): Make dumping of
4871 type names more robust.
4873 2018-01-15 Richard Biener <rguenther@suse.de>
4875 * BASE-VER: Bump to 8.0.1.
4877 2018-01-14 Martin Sebor <msebor@redhat.com>
4879 PR other/83508
4880 * builtins.c (check_access): Avoid warning when the no-warning bit
4881 is set.
4883 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4885 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4886 * ira-color (allocno_hard_regs_compare): Likewise.
4888 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4890 PR target/83013
4891 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4892 Use .pushsection/.popsection.
4894 2018-01-14 Martin Sebor <msebor@redhat.com>
4896 PR c++/81327
4897 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4899 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4901 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4902 entry from extra_headers.
4903 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4904 extra_headers, make the list bitwise identical to the i?86-*-* one.
4906 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4908 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4909 -mcmodel=large with -mindirect-branch=thunk,
4910 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4911 -mfunction-return=thunk-extern.
4912 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4913 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4914 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4916 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4918 * config/i386/i386.c (print_reg): Print the name of the full
4919 integer register without '%'.
4920 (ix86_print_operand): Handle 'V'.
4921 * doc/extend.texi: Document 'V' modifier.
4923 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4925 * config/i386/constraints.md (Bs): Disallow memory operand for
4926 -mindirect-branch-register.
4927 (Bw): Likewise.
4928 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4929 (GOT_memory_operand): Likewise.
4930 (call_insn_operand): Likewise.
4931 (sibcall_insn_operand): Likewise.
4932 (GOT32_symbol_operand): Likewise.
4933 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4934 for -mindirect-branch-register.
4935 (tablejump): Likewise.
4936 (*sibcall_memory): Likewise.
4937 (*sibcall_value_memory): Likewise.
4938 Disallow peepholes of indirect call and jump via memory for
4939 -mindirect-branch-register.
4940 (*call_pop): Replace m with Bw.
4941 (*call_value_pop): Likewise.
4942 (*sibcall_pop_memory): Replace m with Bs.
4943 * config/i386/i386.opt (mindirect-branch-register): New option.
4944 * doc/invoke.texi: Document -mindirect-branch-register option.
4946 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4948 * config/i386/i386-protos.h (ix86_output_function_return): New.
4949 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4950 set function_return_type.
4951 (indirect_thunk_name): Add ret_p to indicate thunk for function
4952 return.
4953 (output_indirect_thunk_function): Pass false to
4954 indirect_thunk_name.
4955 (ix86_output_indirect_branch_via_reg): Likewise.
4956 (ix86_output_indirect_branch_via_push): Likewise.
4957 (output_indirect_thunk_function): Create alias for function
4958 return thunk if regno < 0.
4959 (ix86_output_function_return): New function.
4960 (ix86_handle_fndecl_attribute): Handle function_return.
4961 (ix86_attribute_table): Add function_return.
4962 * config/i386/i386.h (machine_function): Add
4963 function_return_type.
4964 * config/i386/i386.md (simple_return_internal): Use
4965 ix86_output_function_return.
4966 (simple_return_internal_long): Likewise.
4967 * config/i386/i386.opt (mfunction-return=): New option.
4968 (indirect_branch): Mention -mfunction-return=.
4969 * doc/extend.texi: Document function_return function attribute.
4970 * doc/invoke.texi: Document -mfunction-return= option.
4972 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4974 * config/i386/i386-opts.h (indirect_branch): New.
4975 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4976 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4977 with local indirect jump when converting indirect call and jump.
4978 (ix86_set_indirect_branch_type): New.
4979 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4980 (indirectlabelno): New.
4981 (indirect_thunk_needed): Likewise.
4982 (indirect_thunk_bnd_needed): Likewise.
4983 (indirect_thunks_used): Likewise.
4984 (indirect_thunks_bnd_used): Likewise.
4985 (INDIRECT_LABEL): Likewise.
4986 (indirect_thunk_name): Likewise.
4987 (output_indirect_thunk): Likewise.
4988 (output_indirect_thunk_function): Likewise.
4989 (ix86_output_indirect_branch_via_reg): Likewise.
4990 (ix86_output_indirect_branch_via_push): Likewise.
4991 (ix86_output_indirect_branch): Likewise.
4992 (ix86_output_indirect_jmp): Likewise.
4993 (ix86_code_end): Call output_indirect_thunk_function if needed.
4994 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4995 needed.
4996 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4997 (ix86_attribute_table): Add indirect_branch.
4998 * config/i386/i386.h (machine_function): Add indirect_branch_type
4999 and has_local_indirect_jump.
5000 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5001 to true.
5002 (tablejump): Likewise.
5003 (*indirect_jump): Use ix86_output_indirect_jmp.
5004 (*tablejump_1): Likewise.
5005 (simple_return_indirect_internal): Likewise.
5006 * config/i386/i386.opt (mindirect-branch=): New option.
5007 (indirect_branch): New.
5008 (keep): Likewise.
5009 (thunk): Likewise.
5010 (thunk-inline): Likewise.
5011 (thunk-extern): Likewise.
5012 * doc/extend.texi: Document indirect_branch function attribute.
5013 * doc/invoke.texi: Document -mindirect-branch= option.
5015 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5017 PR ipa/83051
5018 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5020 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5022 * ipa-inline.c (want_inline_small_function_p): Return false if
5023 inlining has already failed with CIF_FINAL_ERROR.
5024 (update_caller_keys): Call want_inline_small_function_p before
5025 can_inline_edge_p.
5026 (update_callee_keys): Likewise.
5028 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5030 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5031 New function.
5032 (rs6000_quadword_masked_address_p): Likewise.
5033 (quad_aligned_load_p): Likewise.
5034 (quad_aligned_store_p): Likewise.
5035 (const_load_sequence_p): Add comment to describe the outer-most loop.
5036 (mimic_memory_attributes_and_flags): New function.
5037 (rs6000_gen_stvx): Likewise.
5038 (replace_swapped_aligned_store): Likewise.
5039 (rs6000_gen_lvx): Likewise.
5040 (replace_swapped_aligned_load): Likewise.
5041 (replace_swapped_load_constant): Capitalize argument name in
5042 comment describing this function.
5043 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5044 and stores that access quad-word aligned addresses and replace
5045 with stvx or lvx instructions when appropriate.
5046 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5047 New function prototype.
5048 (rs6000_quadword_masked_address_p): Likewise.
5049 (rs6000_gen_lvx): Likewise.
5050 (rs6000_gen_stvx): Likewise.
5051 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5052 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5053 when memory address is aligned.
5054 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5055 this split to select lvx instruction when memory address is aligned.
5056 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5057 instruction when memory address is aligned.
5058 (*vsx_le_perm_load_v16qi): Likewise.
5059 (four unnamed splitters): Modify to select the stvx instruction
5060 when memory is aligned.
5062 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5064 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5065 which appears in the queue multiple times.
5067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5068 Alan Hayward <alan.hayward@arm.com>
5069 David Sherwood <david.sherwood@arm.com>
5071 * tree-vectorizer.h (vec_lower_bound): New structure.
5072 (_loop_vec_info): Add check_nonzero and lower_bounds.
5073 (LOOP_VINFO_CHECK_NONZERO): New macro.
5074 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5075 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5076 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5077 fields. Make seg_len the distance travelled, not including the
5078 access size.
5079 (dr_direction_indicator): Declare.
5080 (dr_zero_step_indicator): Likewise.
5081 (dr_known_forward_stride_p): Likewise.
5082 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5083 tree-ssanames.h.
5084 (runtime_alias_check_p): Allow runtime alias checks with
5085 variable strides.
5086 (operator ==): Compare access_size and align.
5087 (prune_runtime_alias_test_list): Rework for new distinction between
5088 the access_size and seg_len.
5089 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5090 segment lengths.
5091 (get_segment_min_max): New function.
5092 (create_intersect_range_checks): Use it.
5093 (dr_step_indicator): New function.
5094 (dr_direction_indicator): Likewise.
5095 (dr_zero_step_indicator): Likewise.
5096 (dr_known_forward_stride_p): Likewise.
5097 * tree-loop-distribution.c (data_ref_segment_size): Return
5098 DR_STEP * (niters - 1).
5099 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5100 constructor.
5101 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5102 (vect_preserves_scalar_order_p): New function, split out from...
5103 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5104 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5105 (vect_vfa_access_size): New function.
5106 (vect_vfa_align): Likewise.
5107 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5108 (dump_lower_bound): New function.
5109 (vect_check_lower_bound): Likewise.
5110 (vect_small_gap_p): Likewise.
5111 (vectorizable_with_step_bound_p): Likewise.
5112 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5113 depencies if the vectorization factor is 1. Convert the checks
5114 for nonzero steps into checks on the bounds of DR_STEP. Try using
5115 a bunds check for variable steps if the minimum required step is
5116 relatively small. Update calls to the dr_with_seg_len
5117 constructor and to vect_compile_time_alias.
5118 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5119 function.
5120 (vect_loop_versioning): Call it.
5121 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5122 when retrying.
5123 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5126 Alan Hayward <alan.hayward@arm.com>
5127 David Sherwood <david.sherwood@arm.com>
5129 * doc/sourcebuild.texi (vect_scatter_store): Document.
5130 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5131 optabs.
5132 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5133 Document.
5134 * genopinit.c (main): Add supports_vec_scatter_store and
5135 supports_vec_scatter_store_cached to target_optabs.
5136 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5137 IFN_MASK_SCATTER_STORE.
5138 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5139 functions.
5140 * internal-fn.h (internal_store_fn_p): Declare.
5141 (internal_fn_stored_value_index): Likewise.
5142 * internal-fn.c (scatter_store_direct): New macro.
5143 (expand_scatter_store_optab_fn): New function.
5144 (direct_scatter_store_optab_supported_p): New macro.
5145 (internal_store_fn_p): New function.
5146 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5147 IFN_MASK_SCATTER_STORE.
5148 (internal_fn_mask_index): Likewise.
5149 (internal_fn_stored_value_index): New function.
5150 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5151 for scatter stores.
5152 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5153 * optabs-query.c (supports_vec_scatter_store_p): New function.
5154 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5155 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5156 true for scatter stores.
5157 (vect_gather_scatter_fn_p): Handle scatter stores too.
5158 (vect_check_gather_scatter): Consider using scatter stores if
5159 supports_vec_scatter_store_p.
5160 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5161 scatter stores too.
5162 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5163 internal_fn_stored_value_index.
5164 (check_load_store_masking): Handle scatter stores too.
5165 (vect_get_store_rhs): Make public.
5166 (vectorizable_call): Use internal_store_fn_p.
5167 (vectorizable_store): Handle scatter store internal functions.
5168 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5169 when deciding whether the end of the group has been reached.
5170 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5171 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5172 (mask_scatter_store<mode>): New insns.
5174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5175 Alan Hayward <alan.hayward@arm.com>
5176 David Sherwood <david.sherwood@arm.com>
5178 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5179 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5180 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5181 function.
5182 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5183 Use vect_truncate_gather_scatter_offset if we can't treat the
5184 operation as a normal gather load or scatter store.
5185 (get_group_load_store_type): Take the gather_scatter_info
5186 as argument. Try using a gather load or scatter store for
5187 single-element groups.
5188 (get_load_store_type): Update calls to get_group_load_store_type
5189 and vect_use_strided_gather_scatters_p.
5191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5192 Alan Hayward <alan.hayward@arm.com>
5193 David Sherwood <david.sherwood@arm.com>
5195 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5196 optional tree argument.
5197 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5198 null target hooks.
5199 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5200 but continue to use the current value as a fallback.
5201 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5202 to compare the updates.
5203 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5204 (get_load_store_type): Use it when handling a strided access.
5205 (vect_get_strided_load_store_ops): New function.
5206 (vect_get_data_ptr_increment): Likewise.
5207 (vectorizable_load): Handle strided gather loads. Always pass
5208 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5211 Alan Hayward <alan.hayward@arm.com>
5212 David Sherwood <david.sherwood@arm.com>
5214 * doc/md.texi (gather_load@var{m}): Document.
5215 (mask_gather_load@var{m}): Likewise.
5216 * genopinit.c (main): Add supports_vec_gather_load and
5217 supports_vec_gather_load_cached to target_optabs.
5218 * optabs-tree.c (init_tree_optimization_optabs): Use
5219 ggc_cleared_alloc to allocate target_optabs.
5220 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5221 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5222 functions.
5223 * internal-fn.h (internal_load_fn_p): Declare.
5224 (internal_gather_scatter_fn_p): Likewise.
5225 (internal_fn_mask_index): Likewise.
5226 (internal_gather_scatter_fn_supported_p): Likewise.
5227 * internal-fn.c (gather_load_direct): New macro.
5228 (expand_gather_load_optab_fn): New function.
5229 (direct_gather_load_optab_supported_p): New macro.
5230 (direct_internal_fn_optab): New function.
5231 (internal_load_fn_p): Likewise.
5232 (internal_gather_scatter_fn_p): Likewise.
5233 (internal_fn_mask_index): Likewise.
5234 (internal_gather_scatter_fn_supported_p): Likewise.
5235 * optabs-query.c (supports_at_least_one_mode_p): New function.
5236 (supports_vec_gather_load_p): Likewise.
5237 * optabs-query.h (supports_vec_gather_load_p): Declare.
5238 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5239 and memory_type field.
5240 (NUM_PATTERNS): Bump to 15.
5241 * tree-vect-data-refs.c: Include internal-fn.h.
5242 (vect_gather_scatter_fn_p): New function.
5243 (vect_describe_gather_scatter_call): Likewise.
5244 (vect_check_gather_scatter): Try using internal functions for
5245 gather loads. Recognize existing calls to a gather load function.
5246 (vect_analyze_data_refs): Consider using gather loads if
5247 supports_vec_gather_load_p.
5248 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5249 (vect_get_gather_scatter_offset_type): Likewise.
5250 (vect_convert_mask_for_vectype): Likewise.
5251 (vect_add_conversion_to_patterm): Likewise.
5252 (vect_try_gather_scatter_pattern): Likewise.
5253 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5254 (vect_vect_recog_func_ptrs): Add it.
5255 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5256 internal_fn_mask_index and internal_gather_scatter_fn_p.
5257 (check_load_store_masking): Take the gather_scatter_info as an
5258 argument and handle gather loads.
5259 (vect_get_gather_scatter_ops): New function.
5260 (vectorizable_call): Check internal_load_fn_p.
5261 (vectorizable_load): Likewise. Handle gather load internal
5262 functions.
5263 (vectorizable_store): Update call to check_load_store_masking.
5264 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5265 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5266 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5267 (aarch64_gather_scale_operand_d): New predicates.
5268 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5269 (mask_gather_load<mode>): New insns.
5271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5272 Alan Hayward <alan.hayward@arm.com>
5273 David Sherwood <david.sherwood@arm.com>
5275 * optabs.def (fold_left_plus_optab): New optab.
5276 * doc/md.texi (fold_left_plus_@var{m}): Document.
5277 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5278 * internal-fn.c (fold_left_direct): Define.
5279 (expand_fold_left_optab_fn): Likewise.
5280 (direct_fold_left_optab_supported_p): Likewise.
5281 * fold-const-call.c (fold_const_fold_left): New function.
5282 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5283 * tree-parloops.c (valid_reduction_p): New function.
5284 (gather_scalar_reductions): Use it.
5285 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5286 (vect_finish_replace_stmt): Declare.
5287 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5288 (needs_fold_left_reduction_p): New function, split out from...
5289 (vect_is_simple_reduction): ...here. Accept reductions that
5290 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5291 (vect_force_simple_reduction): Also store the reduction type in
5292 the assignment's STMT_VINFO_REDUC_TYPE.
5293 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5294 (merge_with_identity): New function.
5295 (vect_expand_fold_left): Likewise.
5296 (vectorize_fold_left_reduction): Likewise.
5297 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5298 scalar phi in place for it. Check for target support and reject
5299 cases that would reassociate the operation. Defer the transform
5300 phase to vectorize_fold_left_reduction.
5301 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5302 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5303 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5307 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5308 call to ifc_temp_var.
5310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5311 Alan Hayward <alan.hayward@arm.com>
5312 David Sherwood <david.sherwood@arm.com>
5314 * target.def (legitimize_address_displacement): Take the original
5315 offset as a poly_int.
5316 * targhooks.h (default_legitimize_address_displacement): Update
5317 accordingly.
5318 * targhooks.c (default_legitimize_address_displacement): Likewise.
5319 * doc/tm.texi: Regenerate.
5320 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5321 as an argument, moving assert of ad->disp == ad->disp_term to...
5322 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5323 Try calling targetm.legitimize_address_displacement before expanding
5324 the address rather than afterwards, and adjust for the new interface.
5325 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5326 Match the new hook interface. Handle SVE addresses.
5327 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5328 new hook interface.
5330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5332 * Makefile.in (OBJS): Add early-remat.o.
5333 * target.def (select_early_remat_modes): New hook.
5334 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5335 * doc/tm.texi: Regenerate.
5336 * targhooks.h (default_select_early_remat_modes): Declare.
5337 * targhooks.c (default_select_early_remat_modes): New function.
5338 * timevar.def (TV_EARLY_REMAT): New timevar.
5339 * passes.def (pass_early_remat): New pass.
5340 * tree-pass.h (make_pass_early_remat): Declare.
5341 * early-remat.c: New file.
5342 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5343 function.
5344 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5347 Alan Hayward <alan.hayward@arm.com>
5348 David Sherwood <david.sherwood@arm.com>
5350 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5351 vfm1 with a bound_epilog parameter.
5352 (vect_do_peeling): Update calls accordingly, and move the prologue
5353 call earlier in the function. Treat the base bound_epilog as 0 for
5354 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5355 this base when peeling for gaps.
5356 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5357 with fully-masked loops.
5358 (vect_estimate_min_profitable_iters): Handle the single peeled
5359 iteration in that case.
5361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5362 Alan Hayward <alan.hayward@arm.com>
5363 David Sherwood <david.sherwood@arm.com>
5365 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5366 single-element interleaving even if the size is not a power of 2.
5367 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5368 accesses for single-element interleaving if the group size is
5369 not a power of 2.
5371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5372 Alan Hayward <alan.hayward@arm.com>
5373 David Sherwood <david.sherwood@arm.com>
5375 * doc/md.texi (fold_extract_last_@var{m}): Document.
5376 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5377 * optabs.def (fold_extract_last_optab): New optab.
5378 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5379 * internal-fn.c (fold_extract_direct): New macro.
5380 (expand_fold_extract_optab_fn): Likewise.
5381 (direct_fold_extract_optab_supported_p): Likewise.
5382 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5383 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5384 EXTRACT_LAST_REDUCTION.
5385 (get_initial_def_for_reduction): Do not create an initial vector
5386 for EXTRACT_LAST_REDUCTION reductions.
5387 (vectorizable_reduction): Leave the scalar phi in place for
5388 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5389 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5390 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5391 transform phase to vectorizable_condition.
5392 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5393 split out from...
5394 (vect_finish_stmt_generation): ...here.
5395 (vect_finish_replace_stmt): New function.
5396 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5397 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5398 pattern.
5399 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5402 Alan Hayward <alan.hayward@arm.com>
5403 David Sherwood <david.sherwood@arm.com>
5405 * doc/md.texi (extract_last_@var{m}): Document.
5406 * optabs.def (extract_last_optab): New optab.
5407 * internal-fn.def (EXTRACT_LAST): New internal function.
5408 * internal-fn.c (cond_unary_direct): New macro.
5409 (expand_cond_unary_optab_fn): Likewise.
5410 (direct_cond_unary_optab_supported_p): Likewise.
5411 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5412 loops using EXTRACT_LAST.
5413 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5414 (extract_last_<mode>): ...this optab.
5415 (vec_extract<mode><Vel>): Update accordingly.
5417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5418 Alan Hayward <alan.hayward@arm.com>
5419 David Sherwood <david.sherwood@arm.com>
5421 * target.def (empty_mask_is_expensive): New hook.
5422 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5423 * doc/tm.texi: Regenerate.
5424 * targhooks.h (default_empty_mask_is_expensive): Declare.
5425 * targhooks.c (default_empty_mask_is_expensive): New function.
5426 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5427 if the target says that empty masks are expensive.
5428 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5429 New function.
5430 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5433 Alan Hayward <alan.hayward@arm.com>
5434 David Sherwood <david.sherwood@arm.com>
5436 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5437 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5438 (vect_use_loop_mask_for_alignment_p): New function.
5439 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5440 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5441 niters_skip argument. Make sure that the first niters_skip elements
5442 of the first iteration are inactive.
5443 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5444 Update call to vect_set_loop_masks_directly.
5445 (get_misalign_in_elems): New function, split out from...
5446 (vect_gen_prolog_loop_niters): ...here.
5447 (vect_update_init_of_dr): Take a code argument that specifies whether
5448 the adjustment should be added or subtracted.
5449 (vect_update_init_of_drs): Likewise.
5450 (vect_prepare_for_masked_peels): New function.
5451 (vect_do_peeling): Skip prologue peeling if we're using a mask
5452 instead. Update call to vect_update_inits_of_drs.
5453 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5454 mask_skip_niters.
5455 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5456 alignment. Do not include the number of peeled iterations in
5457 the minimum threshold in that case.
5458 (vectorizable_induction): Adjust the start value down by
5459 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5460 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5461 Take the number of skipped iterations into account when calculating
5462 the loop bounds.
5463 * tree-vect-stmts.c (vect_gen_while_not): New function.
5465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5466 Alan Hayward <alan.hayward@arm.com>
5467 David Sherwood <david.sherwood@arm.com>
5469 * doc/sourcebuild.texi (vect_fully_masked): Document.
5470 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5471 default value to 0.
5472 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5473 split out from...
5474 (vect_analyze_loop_2): ...here. Don't check the vectorization
5475 factor against the number of loop iterations if the loop is
5476 fully-masked.
5478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5479 Alan Hayward <alan.hayward@arm.com>
5480 David Sherwood <david.sherwood@arm.com>
5482 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5483 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5484 (dump_groups): Update accordingly.
5485 (iv_use::mem_type): New member variable.
5486 (address_p): New function.
5487 (record_use): Add a mem_type argument and initialize the new
5488 mem_type field.
5489 (record_group_use): Add a mem_type argument. Use address_p.
5490 Remove obsolete null checks of base_object. Update call to record_use.
5491 (find_interesting_uses_op): Update call to record_group_use.
5492 (find_interesting_uses_cond): Likewise.
5493 (find_interesting_uses_address): Likewise.
5494 (get_mem_type_for_internal_fn): New function.
5495 (find_address_like_use): Likewise.
5496 (find_interesting_uses_stmt): Try find_address_like_use before
5497 calling find_interesting_uses_op.
5498 (addr_offset_valid_p): Use the iv mem_type field as the type
5499 of the addressed memory.
5500 (add_autoinc_candidates): Likewise.
5501 (get_address_cost): Likewise.
5502 (split_small_address_groups_p): Use address_p.
5503 (split_address_groups): Likewise.
5504 (add_iv_candidate_for_use): Likewise.
5505 (autoinc_possible_for_pair): Likewise.
5506 (rewrite_groups): Likewise.
5507 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5508 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5509 (get_alias_ptr_type_for_ptr_address): New function.
5510 (rewrite_use_address): Rewrite address uses in calls that were
5511 identified by find_address_like_use.
5513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5514 Alan Hayward <alan.hayward@arm.com>
5515 David Sherwood <david.sherwood@arm.com>
5517 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5518 TARGET_MEM_REFs.
5519 * gimple-expr.h (is_gimple_addressable: Likewise.
5520 * gimple-expr.c (is_gimple_address): Likewise.
5521 * internal-fn.c (expand_call_mem_ref): New function.
5522 (expand_mask_load_optab_fn): Use it.
5523 (expand_mask_store_optab_fn): Likewise.
5525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5526 Alan Hayward <alan.hayward@arm.com>
5527 David Sherwood <david.sherwood@arm.com>
5529 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5530 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5531 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5532 (cond_umax@var{mode}): Document.
5533 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5534 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5535 (cond_umin_optab, cond_umax_optab): New optabs.
5536 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5537 (COND_IOR, COND_XOR): New internal functions.
5538 * internal-fn.h (get_conditional_internal_fn): Declare.
5539 * internal-fn.c (cond_binary_direct): New macro.
5540 (expand_cond_binary_optab_fn): Likewise.
5541 (direct_cond_binary_optab_supported_p): Likewise.
5542 (get_conditional_internal_fn): New function.
5543 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5544 Cope with reduction statements that are vectorized as calls rather
5545 than assignments.
5546 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5547 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5548 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5549 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5550 (UNSPEC_COND_EOR): New unspecs.
5551 (optab): Add mappings for them.
5552 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5553 (sve_int_op, sve_fp_op): New int attributes.
5555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5556 Alan Hayward <alan.hayward@arm.com>
5557 David Sherwood <david.sherwood@arm.com>
5559 * optabs.def (while_ult_optab): New optab.
5560 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5561 * internal-fn.def (WHILE_ULT): New internal function.
5562 * internal-fn.h (direct_internal_fn_supported_p): New override
5563 that takes two types as argument.
5564 * internal-fn.c (while_direct): New macro.
5565 (expand_while_optab_fn): New function.
5566 (convert_optab_supported_p): Likewise.
5567 (direct_while_optab_supported_p): New macro.
5568 * wide-int.h (wi::udiv_ceil): New function.
5569 * tree-vectorizer.h (rgroup_masks): New structure.
5570 (vec_loop_masks): New typedef.
5571 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5572 and fully_masked_p.
5573 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5574 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5575 (vect_max_vf): New function.
5576 (slpeel_make_loop_iterate_ntimes): Delete.
5577 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5578 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5579 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5580 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5581 internal-fn.h, stor-layout.h and optabs-query.h.
5582 (vect_set_loop_mask): New function.
5583 (add_preheader_seq): Likewise.
5584 (add_header_seq): Likewise.
5585 (interleave_supported_p): Likewise.
5586 (vect_maybe_permute_loop_masks): Likewise.
5587 (vect_set_loop_masks_directly): Likewise.
5588 (vect_set_loop_condition_masked): Likewise.
5589 (vect_set_loop_condition_unmasked): New function, split out from
5590 slpeel_make_loop_iterate_ntimes.
5591 (slpeel_make_loop_iterate_ntimes): Rename to..
5592 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5593 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5594 (vect_do_peeling): Update call accordingly.
5595 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5596 loops.
5597 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5598 mask_compare_type, can_fully_mask_p and fully_masked_p.
5599 (release_vec_loop_masks): New function.
5600 (_loop_vec_info): Use it to free the loop masks.
5601 (can_produce_all_loop_masks_p): New function.
5602 (vect_get_max_nscalars_per_iter): Likewise.
5603 (vect_verify_full_masking): Likewise.
5604 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5605 retries, and free the mask rgroups before retrying. Check loop-wide
5606 reasons for disallowing fully-masked loops. Make the final decision
5607 about whether use a fully-masked loop or not.
5608 (vect_estimate_min_profitable_iters): Do not assume that peeling
5609 for the number of iterations will be needed for fully-masked loops.
5610 (vectorizable_reduction): Disable fully-masked loops.
5611 (vectorizable_live_operation): Likewise.
5612 (vect_halve_mask_nunits): New function.
5613 (vect_double_mask_nunits): Likewise.
5614 (vect_record_loop_mask): Likewise.
5615 (vect_get_loop_mask): Likewise.
5616 (vect_transform_loop): Handle the case in which the final loop
5617 iteration might handle a partial vector. Call vect_set_loop_condition
5618 instead of slpeel_make_loop_iterate_ntimes.
5619 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5620 (check_load_store_masking): New function.
5621 (prepare_load_store_mask): Likewise.
5622 (vectorizable_store): Handle fully-masked loops.
5623 (vectorizable_load): Likewise.
5624 (supportable_widening_operation): Use vect_halve_mask_nunits for
5625 booleans.
5626 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5627 (vect_gen_while): New function.
5628 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5629 (aarch64_uqdec<mode>): New insn.
5631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5632 Alan Hayward <alan.hayward@arm.com>
5633 David Sherwood <david.sherwood@arm.com>
5635 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5636 (reduc_xor_scal_optab): New optabs.
5637 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5638 (reduc_xor_scal_@var{m}): Document.
5639 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5640 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5641 internal functions.
5642 * fold-const-call.c (fold_const_call): Handle them.
5643 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5644 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5645 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5646 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5647 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5648 (UNSPEC_XORV): New unspecs.
5649 (optab): Add entries for them.
5650 (BITWISEV): New int iterator.
5651 (bit_reduc_op): New int attributes.
5653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5654 Alan Hayward <alan.hayward@arm.com>
5655 David Sherwood <david.sherwood@arm.com>
5657 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5658 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5659 * optabs.def (vec_shl_insert_optab): New optab.
5660 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5661 (duplicate_and_interleave): Likewise.
5662 * tree-vect-loop.c: Include internal-fn.h.
5663 (neutral_op_for_slp_reduction): New function, split out from
5664 get_initial_defs_for_reduction.
5665 (get_initial_def_for_reduction): Handle option 2 for variable-length
5666 vectors by loading the neutral value into a vector and then shifting
5667 the initial value into element 0.
5668 (get_initial_defs_for_reduction): Replace the code argument with
5669 the neutral value calculated by neutral_op_for_slp_reduction.
5670 Use gimple_build_vector for constant-length vectors.
5671 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5672 but the first group_size elements have a neutral value.
5673 Use duplicate_and_interleave otherwise.
5674 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5675 Update call to get_initial_defs_for_reduction. Handle SLP
5676 reductions for variable-length vectors by creating one vector
5677 result for each scalar result, with the elements associated
5678 with other scalar results stubbed out with the neutral value.
5679 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5680 Require IFN_VEC_SHL_INSERT for double reductions on
5681 variable-length vectors, or SLP reductions that have
5682 a neutral value. Require can_duplicate_and_interleave_p
5683 support for variable-length unchained SLP reductions if there
5684 is no neutral value, such as for MIN/MAX reductions. Also require
5685 the number of vector elements to be a multiple of the number of
5686 SLP statements when doing variable-length unchained SLP reductions.
5687 Update call to vect_create_epilog_for_reduction.
5688 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5689 and remove initial values.
5690 (duplicate_and_interleave): Make public.
5691 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5692 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5695 Alan Hayward <alan.hayward@arm.com>
5696 David Sherwood <david.sherwood@arm.com>
5698 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5699 (can_duplicate_and_interleave_p): New function.
5700 (vect_get_and_check_slp_defs): Take the vector of statements
5701 rather than just the current one. Remove excess parentheses.
5702 Restriction rejectinon of vect_constant_def and vect_external_def
5703 for variable-length vectors to boolean types, or types for which
5704 can_duplicate_and_interleave_p is false.
5705 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5706 (duplicate_and_interleave): New function.
5707 (vect_get_constant_vectors): Use gimple_build_vector for
5708 constant-length vectors and suitable variable-length constant
5709 vectors. Use duplicate_and_interleave for other variable-length
5710 vectors. Don't defer the update when inserting new statements.
5712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5713 Alan Hayward <alan.hayward@arm.com>
5714 David Sherwood <david.sherwood@arm.com>
5716 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5717 min_profitable_iters doesn't go negative.
5719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5720 Alan Hayward <alan.hayward@arm.com>
5721 David Sherwood <david.sherwood@arm.com>
5723 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5724 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5725 * optabs.def (vec_mask_load_lanes_optab): New optab.
5726 (vec_mask_store_lanes_optab): Likewise.
5727 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5728 (MASK_STORE_LANES): Likewise.
5729 * internal-fn.c (mask_load_lanes_direct): New macro.
5730 (mask_store_lanes_direct): Likewise.
5731 (expand_mask_load_optab_fn): Handle masked operations.
5732 (expand_mask_load_lanes_optab_fn): New macro.
5733 (expand_mask_store_optab_fn): Handle masked operations.
5734 (expand_mask_store_lanes_optab_fn): New macro.
5735 (direct_mask_load_lanes_optab_supported_p): Likewise.
5736 (direct_mask_store_lanes_optab_supported_p): Likewise.
5737 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5738 parameter.
5739 (vect_load_lanes_supported): Likewise.
5740 * tree-vect-data-refs.c (strip_conversion): New function.
5741 (can_group_stmts_p): Likewise.
5742 (vect_analyze_data_ref_accesses): Use it instead of checking
5743 for a pair of assignments.
5744 (vect_store_lanes_supported): Take a masked_p parameter.
5745 (vect_load_lanes_supported): Likewise.
5746 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5747 vect_store_lanes_supported and vect_load_lanes_supported.
5748 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5749 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5750 parameter. Don't allow gaps for masked accesses.
5751 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5752 and vect_load_lanes_supported.
5753 (get_load_store_type): Take a masked_p parameter and update
5754 call to get_group_load_store_type.
5755 (vectorizable_store): Update call to get_load_store_type.
5756 Handle IFN_MASK_STORE_LANES.
5757 (vectorizable_load): Update call to get_load_store_type.
5758 Handle IFN_MASK_LOAD_LANES.
5760 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5761 Alan Hayward <alan.hayward@arm.com>
5762 David Sherwood <david.sherwood@arm.com>
5764 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5765 modes for SVE.
5766 * config/aarch64/aarch64-protos.h
5767 (aarch64_sve_struct_memory_operand_p): Declare.
5768 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5769 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5770 (VPRED, vpred): Handle SVE structure modes.
5771 * config/aarch64/constraints.md (Utx): New constraint.
5772 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5773 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5774 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5775 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5776 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5777 structure modes. Split into pieces after RA.
5778 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5779 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5780 New patterns.
5781 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5782 SVE structure modes.
5783 (aarch64_classify_address): Likewise.
5784 (sizetochar): Move earlier in file.
5785 (aarch64_print_operand): Handle SVE register lists.
5786 (aarch64_array_mode): New function.
5787 (aarch64_sve_struct_memory_operand_p): Likewise.
5788 (TARGET_ARRAY_MODE): Redefine.
5790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5791 Alan Hayward <alan.hayward@arm.com>
5792 David Sherwood <david.sherwood@arm.com>
5794 * target.def (array_mode): New target hook.
5795 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5796 * doc/tm.texi: Regenerate.
5797 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5798 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5799 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5800 targetm.array_mode.
5801 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5802 type sizes.
5804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5805 Alan Hayward <alan.hayward@arm.com>
5806 David Sherwood <david.sherwood@arm.com>
5808 * fold-const.c (fold_binary_loc): Check the argument types
5809 rather than the result type when testing for a vector operation.
5811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5813 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5814 * doc/tm.texi: Regenerate.
5816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5817 Alan Hayward <alan.hayward@arm.com>
5818 David Sherwood <david.sherwood@arm.com>
5820 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5821 (sve): Document new AArch64 extension.
5822 * doc/md.texi (w): Extend the description of the AArch64
5823 constraint to include SVE vectors.
5824 (Upl, Upa): Document new AArch64 predicate constraints.
5825 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5826 enum.
5827 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5828 (msve-vector-bits=): New option.
5829 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5830 SVE when these are disabled.
5831 (sve): New extension.
5832 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5833 modes. Adjust their number of units based on aarch64_sve_vg.
5834 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5835 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5836 aarch64_addr_query_type.
5837 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5838 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5839 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5840 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5841 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5842 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5843 (aarch64_simd_imm_zero_p): Delete.
5844 (aarch64_check_zero_based_sve_index_immediate): Declare.
5845 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5846 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5847 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5848 (aarch64_sve_float_mul_immediate_p): Likewise.
5849 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5850 rather than an rtx.
5851 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5852 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5853 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5854 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5855 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5856 (aarch64_regmode_natural_size): Likewise.
5857 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5858 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5859 left one place.
5860 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5861 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5862 for VG and the SVE predicate registers.
5863 (V_ALIASES): Add a "z"-prefixed alias.
5864 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5865 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5866 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5867 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5868 (REG_CLASS_NAMES): Add entries for them.
5869 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5870 and the predicate registers.
5871 (aarch64_sve_vg): Declare.
5872 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5873 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5874 (REGMODE_NATURAL_SIZE): Define.
5875 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5876 SVE macros.
5877 * config/aarch64/aarch64.c: Include cfgrtl.h.
5878 (simd_immediate_info): Add a constructor for series vectors,
5879 and an associated step field.
5880 (aarch64_sve_vg): New variable.
5881 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5882 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5883 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5884 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5885 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5886 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5887 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5888 (aarch64_get_mask_mode): New functions.
5889 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5890 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5891 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5892 predicate modes and predicate registers. Explicitly restrict
5893 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5894 to store a vector mode if it is recognized by
5895 aarch64_classify_vector_mode.
5896 (aarch64_regmode_natural_size): New function.
5897 (aarch64_hard_regno_caller_save_mode): Return the original mode
5898 for predicates.
5899 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5900 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5901 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5902 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5903 functions.
5904 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5905 does not overlap dest if the function is frame-related. Handle
5906 SVE constants.
5907 (aarch64_split_add_offset): New function.
5908 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5909 them aarch64_add_offset.
5910 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5911 and update call to aarch64_sub_sp.
5912 (aarch64_add_cfa_expression): New function.
5913 (aarch64_expand_prologue): Pass extra temporary registers to the
5914 functions above. Handle the case in which we need to emit new
5915 DW_CFA_expressions for registers that were originally saved
5916 relative to the stack pointer, but now have to be expressed
5917 relative to the frame pointer.
5918 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5919 functions above.
5920 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5921 IP0 and IP1 values for SVE frames.
5922 (aarch64_expand_vec_series): New function.
5923 (aarch64_expand_sve_widened_duplicate): Likewise.
5924 (aarch64_expand_sve_const_vector): Likewise.
5925 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5926 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5927 into the register, rather than emitting a SET directly.
5928 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5929 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5930 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5931 (offset_9bit_signed_scaled_p): New functions.
5932 (aarch64_replicate_bitmask_imm): New function.
5933 (aarch64_bitmask_imm): Use it.
5934 (aarch64_cannot_force_const_mem): Reject expressions involving
5935 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5936 (aarch64_classify_index): Handle SVE indices, by requiring
5937 a plain register index with a scale that matches the element size.
5938 (aarch64_classify_address): Handle SVE addresses. Assert that
5939 the mode of the address is VOIDmode or an integer mode.
5940 Update call to aarch64_classify_symbol.
5941 (aarch64_classify_symbolic_expression): Update call to
5942 aarch64_classify_symbol.
5943 (aarch64_const_vec_all_in_range_p): New function.
5944 (aarch64_print_vector_float_operand): Likewise.
5945 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5946 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5947 and the FP immediates 1.0 and 0.5.
5948 (aarch64_print_address_internal): Handle SVE addresses.
5949 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5950 (aarch64_regno_regclass): Handle predicate registers.
5951 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5952 data modes.
5953 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5954 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5955 (aarch64_convert_sve_vector_bits): New function.
5956 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5957 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5958 rather than an rtx.
5959 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5960 Handle SVE vector and predicate modes. Accept VL-based constants
5961 that need only one temporary register, and VL offsets that require
5962 no temporary registers.
5963 (aarch64_conditional_register_usage): Mark the predicate registers
5964 as fixed if SVE isn't available.
5965 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5966 Return true for SVE vector and predicate modes.
5967 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5968 rather than an unsigned int. Handle SVE modes.
5969 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5970 SVE modes.
5971 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5972 if SVE is enabled.
5973 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5974 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5975 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5976 (aarch64_sve_float_mul_immediate_p): New functions.
5977 (aarch64_sve_valid_immediate): New function.
5978 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5979 Explicitly reject structure modes. Check for INDEX constants.
5980 Handle PTRUE and PFALSE constants.
5981 (aarch64_check_zero_based_sve_index_immediate): New function.
5982 (aarch64_simd_imm_zero_p): Delete.
5983 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5984 vector modes. Accept constants in the range of CNT[BHWD].
5985 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5986 ask for an Advanced SIMD mode.
5987 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5988 (aarch64_simd_vector_alignment): Handle SVE predicates.
5989 (aarch64_vectorize_preferred_vector_alignment): New function.
5990 (aarch64_simd_vector_alignment_reachable): Use it instead of
5991 the vector size.
5992 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5993 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5994 functions.
5995 (MAX_VECT_LEN): Delete.
5996 (expand_vec_perm_d): Add a vec_flags field.
5997 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5998 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5999 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6000 for SVE modes.
6001 (aarch64_evpc_rev): Rename to...
6002 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6003 (aarch64_evpc_rev_global): New function.
6004 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6005 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6006 MAX_VECT_LEN.
6007 (aarch64_evpc_sve_tbl): New function.
6008 (aarch64_expand_vec_perm_const_1): Update after rename of
6009 aarch64_evpc_rev. Handle SVE permutes too, trying
6010 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6011 than aarch64_evpc_tbl.
6012 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6013 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6014 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6015 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6016 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6017 (aarch64_expand_sve_vcond): New functions.
6018 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6019 of aarch64_vector_mode_p.
6020 (aarch64_dwarf_poly_indeterminate_value): New function.
6021 (aarch64_compute_pressure_classes): Likewise.
6022 (aarch64_can_change_mode_class): Likewise.
6023 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6024 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6025 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6026 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6027 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6028 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6029 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6030 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6031 constraints.
6032 (Dn, Dl, Dr): Accept const as well as const_vector.
6033 (Dz): Likewise. Compare against CONST0_RTX.
6034 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6035 of "vector" where appropriate.
6036 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6037 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6038 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6039 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6040 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6041 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6042 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6043 (v_int_equiv): Extend to SVE modes.
6044 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6045 mode attributes.
6046 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6047 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6048 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6049 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6050 (SVE_COND_FP_CMP): New int iterators.
6051 (perm_hilo): Handle the new unpack unspecs.
6052 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6053 attributes.
6054 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6055 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6056 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6057 (aarch64_equality_operator, aarch64_constant_vector_operand)
6058 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6059 (aarch64_sve_nonimmediate_operand): Likewise.
6060 (aarch64_sve_general_operand): Likewise.
6061 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6062 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6063 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6064 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6065 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6066 (aarch64_sve_float_arith_immediate): Likewise.
6067 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6068 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6069 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6070 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6071 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6072 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6073 (aarch64_sve_float_arith_operand): Likewise.
6074 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6075 (aarch64_sve_float_mul_operand): Likewise.
6076 (aarch64_sve_vec_perm_operand): Likewise.
6077 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6078 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6079 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6080 as well as const_vector.
6081 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6082 in file. Use CONST0_RTX and CONSTM1_RTX.
6083 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6084 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6085 Use aarch64_simd_imm_zero.
6086 * config/aarch64/aarch64-sve.md: New file.
6087 * config/aarch64/aarch64.md: Include it.
6088 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6089 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6090 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6091 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6092 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6093 (sve): New attribute.
6094 (enabled): Disable instructions with the sve attribute unless
6095 TARGET_SVE.
6096 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6097 aarch64_expand_mov_immediate.
6098 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6099 CNT[BHSD] immediates.
6100 (movti): Split CONST_POLY_INT moves into two halves.
6101 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6102 Split additions that need a temporary here if the destination
6103 is the stack pointer.
6104 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6105 (*add<mode>3_poly_1): New instruction.
6106 (set_clobber_cc): New expander.
6108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6110 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6111 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6112 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6113 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6114 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6115 Change innermode from fixed_mode_size to machine_mode.
6116 (simplify_subreg): Update call accordingly. Handle a constant-sized
6117 subreg of a variable-length CONST_VECTOR.
6119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6120 Alan Hayward <alan.hayward@arm.com>
6121 David Sherwood <david.sherwood@arm.com>
6123 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6124 (add_offset_to_base): New function, split out from...
6125 (create_mem_ref): ...here. When handling a scale other than 1,
6126 check first whether the address is valid without the offset.
6127 Add it into the base if so, leaving the index and scale as-is.
6129 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6131 PR c++/83778
6132 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6133 fold_for_warn before checking if arg2 is INTEGER_CST.
6135 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6137 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6138 (store_multiple_operation): Delete.
6139 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6140 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6141 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6142 guarded by TARGET_STRING.
6143 (rs6000_output_load_multiple): Delete.
6144 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6145 OPTION_MASK_STRING / TARGET_STRING handling.
6146 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6147 (const rs6000_opt_masks) <"string">: Change mask to 0.
6148 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6149 (MASK_STRING): Delete.
6150 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6151 parts. Simplify.
6152 (load_multiple): Delete.
6153 (*ldmsi8): Delete.
6154 (*ldmsi7): Delete.
6155 (*ldmsi6): Delete.
6156 (*ldmsi5): Delete.
6157 (*ldmsi4): Delete.
6158 (*ldmsi3): Delete.
6159 (store_multiple): Delete.
6160 (*stmsi8): Delete.
6161 (*stmsi7): Delete.
6162 (*stmsi6): Delete.
6163 (*stmsi5): Delete.
6164 (*stmsi4): Delete.
6165 (*stmsi3): Delete.
6166 (movmemsi_8reg): Delete.
6167 (corresponding unnamed define_insn): Delete.
6168 (movmemsi_6reg): Delete.
6169 (corresponding unnamed define_insn): Delete.
6170 (movmemsi_4reg): Delete.
6171 (corresponding unnamed define_insn): Delete.
6172 (movmemsi_2reg): Delete.
6173 (corresponding unnamed define_insn): Delete.
6174 (movmemsi_1reg): Delete.
6175 (corresponding unnamed define_insn): Delete.
6176 * config/rs6000/rs6000.opt (mno-string): New.
6177 (mstring): Replace by deprecation warning stub.
6178 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6180 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6182 * regrename.c (regrename_do_replace): If replacing the same
6183 reg multiple times, try to reuse last created gen_raw_REG.
6185 PR debug/81155
6186 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6187 main to workaround a bug in GDB.
6189 2018-01-12 Tom de Vries <tom@codesourcery.com>
6191 PR target/83737
6192 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6194 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6196 PR rtl-optimization/80481
6197 * ira-color.c (get_cap_member): New function.
6198 (allocnos_conflict_by_live_ranges_p): Use it.
6199 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6200 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6202 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6204 PR target/83628
6205 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6206 (*saddl_se_1): Ditto.
6207 (*ssubsi_1): Ditto.
6208 (*ssubl_se_1): Ditto.
6210 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6212 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6213 rather than wi::to_widest for DR_INITs.
6214 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6215 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6216 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6217 INTEGER_CSTs.
6218 (vect_analyze_group_access_1): Note that here.
6220 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6222 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6223 polynomial type sizes.
6225 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6227 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6228 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6229 (gimple_add_tmp_var): Likewise.
6231 2018-01-12 Martin Liska <mliska@suse.cz>
6233 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6234 (gimple_alloc_sizes): Likewise.
6235 (dump_gimple_statistics): Use PRIu64 in printf format.
6236 * gimple.h: Change uint64_t to int.
6238 2018-01-12 Martin Liska <mliska@suse.cz>
6240 * tree-core.h: Use uint64_t instead of int.
6241 * tree.c (tree_node_counts): Likewise.
6242 (tree_node_sizes): Likewise.
6243 (dump_tree_statistics): Use PRIu64 in printf format.
6245 2018-01-12 Martin Liska <mliska@suse.cz>
6247 * Makefile.in: As qsort_chk is implemented in vec.c, add
6248 vec.o to linkage of gencfn-macros.
6249 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6250 passing the info to record_node_allocation_statistics.
6251 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6252 and pass the info.
6253 * ggc-common.c (struct ggc_usage): Add operator== and use
6254 it in operator< and compare function.
6255 * mem-stats.h (struct mem_usage): Likewise.
6256 * vec.c (struct vec_usage): Remove operator< and compare
6257 function. Can be simply inherited.
6259 2018-01-12 Martin Jambor <mjambor@suse.cz>
6261 PR target/81616
6262 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6263 * tree-ssa-math-opts.c: Include domwalk.h.
6264 (convert_mult_to_fma_1): New function.
6265 (fma_transformation_info): New type.
6266 (fma_deferring_state): Likewise.
6267 (cancel_fma_deferring): New function.
6268 (result_of_phi): Likewise.
6269 (last_fma_candidate_feeds_initial_phi): Likewise.
6270 (convert_mult_to_fma): Added deferring logic, split actual
6271 transformation to convert_mult_to_fma_1.
6272 (math_opts_dom_walker): New type.
6273 (math_opts_dom_walker::after_dom_children): New method, body moved
6274 here from pass_optimize_widening_mul::execute, added deferring logic
6275 bits.
6276 (pass_optimize_widening_mul::execute): Moved most of code to
6277 math_opts_dom_walker::after_dom_children.
6278 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6279 * config/i386/i386.c (ix86_option_override_internal): Added
6280 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6282 2018-01-12 Richard Biener <rguenther@suse.de>
6284 PR debug/83157
6285 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6286 inline instance vars.
6288 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6290 PR target/81819
6291 * config/rx/rx.c (rx_is_restricted_memory_address):
6292 Handle SUBREG case.
6294 2018-01-12 Richard Biener <rguenther@suse.de>
6296 PR tree-optimization/80846
6297 * target.def (split_reduction): New target hook.
6298 * targhooks.c (default_split_reduction): New function.
6299 * targhooks.h (default_split_reduction): Declare.
6300 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6301 target requests first reduce vectors by combining low and high
6302 parts.
6303 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6304 (get_vectype_for_scalar_type_and_size): Export.
6305 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6306 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6307 * doc/tm.texi: Regenerate.
6308 * config/i386/i386.c (ix86_split_reduction): Implement
6309 TARGET_VECTORIZE_SPLIT_REDUCTION.
6311 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6313 PR target/83368
6314 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6315 in PIC mode except for TARGET_VXWORKS_RTP.
6316 * config/sparc/sparc.c: Include cfgrtl.h.
6317 (TARGET_INIT_PIC_REG): Define.
6318 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6319 (sparc_pic_register_p): New predicate.
6320 (sparc_legitimate_address_p): Use it.
6321 (sparc_legitimize_pic_address): Likewise.
6322 (sparc_delegitimize_address): Likewise.
6323 (sparc_mode_dependent_address_p): Likewise.
6324 (gen_load_pcrel_sym): Remove 4th parameter.
6325 (load_got_register): Adjust call to above. Remove obsolete stuff.
6326 (sparc_expand_prologue): Do not call load_got_register here.
6327 (sparc_flat_expand_prologue): Likewise.
6328 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6329 (sparc_use_pseudo_pic_reg): New function.
6330 (sparc_init_pic_reg): Likewise.
6331 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6332 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6334 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6336 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6337 Add item for branch_cost.
6339 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6341 PR rtl-optimization/83565
6342 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6343 not extend the result to a larger mode for rotate operations.
6344 (num_sign_bit_copies1): Likewise.
6346 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6348 PR target/40411
6349 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6350 -symbolic.
6351 Use values-Xc.o for -pedantic.
6352 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6354 2018-01-12 Martin Liska <mliska@suse.cz>
6356 PR ipa/83054
6357 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6358 New function.
6359 (possible_polymorphic_call_targets): Use it.
6360 (ipa_devirt): Likewise.
6362 2018-01-12 Martin Liska <mliska@suse.cz>
6364 * profile-count.h (enum profile_quality): Use 0 as invalid
6365 enum value of profile_quality.
6367 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6369 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6370 -mext-string options.
6372 2018-01-12 Richard Biener <rguenther@suse.de>
6374 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6375 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6376 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6377 Likewise.
6378 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6380 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6382 * configure.ac (--with-long-double-format): Add support for the
6383 configuration option to change the default long double format on
6384 PowerPC systems.
6385 * config.gcc (powerpc*-linux*-*): Likewise.
6386 * configure: Regenerate.
6387 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6388 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6389 used without modification.
6391 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6393 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6394 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6395 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6396 MISC_BUILTIN_SPEC_BARRIER.
6397 (rs6000_init_builtins): Likewise.
6398 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6399 enum value.
6400 (speculation_barrier): New define_insn.
6401 * doc/extend.texi: Document __builtin_speculation_barrier.
6403 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6405 PR target/83203
6406 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6407 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6408 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6409 iterators.
6410 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6411 integral modes instead of "ss" and "sd".
6412 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6413 vectors with 32-bit and 64-bit elements.
6414 (vecdupssescalarmodesuffix): New mode attribute.
6415 (vec_dup<mode>): Use it.
6417 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6419 PR target/83330
6420 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6421 frame if argument is passed on stack.
6423 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6425 PR target/82682
6426 * ree.c (combine_reaching_defs): Optimize also
6427 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6428 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6430 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6432 PR middle-end/83189
6433 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6435 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6437 PR middle-end/83718
6438 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6439 after they are computed.
6441 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6443 PR tree-optimization/83695
6444 * gimple-loop-linterchange.cc
6445 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6446 reset cached scev information after interchange.
6447 (pass_linterchange::execute): Remove call to scev_reset_htab.
6449 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6451 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6452 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6453 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6454 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6455 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6456 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6457 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6458 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6459 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6460 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6461 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6462 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6463 (V_lane_reg): Likewise.
6464 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6465 New define_expand.
6466 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6467 (vfmal_lane_low<mode>_intrinsic,
6468 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6469 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6470 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6471 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6472 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6473 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6475 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6477 * config/arm/arm-cpus.in (fp16fml): New feature.
6478 (ALL_SIMD): Add fp16fml.
6479 (armv8.2-a): Add fp16fml as an option.
6480 (armv8.3-a): Likewise.
6481 (armv8.4-a): Add fp16fml as part of fp16.
6482 * config/arm/arm.h (TARGET_FP16FML): Define.
6483 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6484 when appropriate.
6485 * config/arm/arm-modes.def (V2HF): Define.
6486 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6487 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6488 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6489 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6490 vfmsl_low, vfmsl_high): New set of builtins.
6491 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6492 (vfml_op): New code attribute.
6493 (VFMLHALVES): New int iterator.
6494 (VFML, VFMLSEL): New mode attributes.
6495 (V_reg): Define mapping for V2HF.
6496 (V_hi, V_lo): New mode attributes.
6497 (VF_constraint): Likewise.
6498 (vfml_half, vfml_half_selector): New int attributes.
6499 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6500 define_expand.
6501 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6502 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6503 New define_insn.
6504 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6505 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6506 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6507 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6508 documentation.
6509 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6510 Document new effective target and option set.
6512 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6514 * config/arm/arm-cpus.in (armv8_4): New feature.
6515 (ARMv8_4a): New fgroup.
6516 (armv8.4-a): New arch.
6517 * config/arm/arm-tables.opt: Regenerate.
6518 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6519 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6520 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6521 Add matching rules for -march=armv8.4-a and extensions.
6522 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6524 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6526 PR target/81821
6527 * config/rx/rx.md (BW): New mode attribute.
6528 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6530 2018-01-11 Richard Biener <rguenther@suse.de>
6532 PR tree-optimization/83435
6533 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6534 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6535 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6537 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6538 Alan Hayward <alan.hayward@arm.com>
6539 David Sherwood <david.sherwood@arm.com>
6541 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6542 field.
6543 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6544 (aarch64_print_address_internal): Use it to check for a zero offset.
6546 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6547 Alan Hayward <alan.hayward@arm.com>
6548 David Sherwood <david.sherwood@arm.com>
6550 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6551 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6552 Return a poly_int64 rather than a HOST_WIDE_INT.
6553 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6554 rather than a HOST_WIDE_INT.
6555 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6556 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6557 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6558 final_offset from HOST_WIDE_INT to poly_int64.
6559 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6560 to_constant when getting the number of units in an Advanced SIMD
6561 mode.
6562 (aarch64_builtin_vectorized_function): Check for a constant number
6563 of units.
6564 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6565 GET_MODE_SIZE.
6566 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6567 attribute instead of GET_MODE_NUNITS.
6568 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6569 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6570 GET_MODE_SIZE for fixed-size registers.
6571 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6572 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6573 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6574 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6575 (aarch64_print_operand, aarch64_print_address_internal)
6576 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6577 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6578 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6579 Handle polynomial GET_MODE_SIZE.
6580 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6581 wider than SImode without modification.
6582 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6583 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6584 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6585 passing and returning SVE modes.
6586 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6587 rather than GEN_INT.
6588 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6589 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6590 (aarch64_allocate_and_probe_stack_space): Likewise.
6591 (aarch64_layout_frame): Cope with polynomial offsets.
6592 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6593 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6594 polynomial offsets.
6595 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6596 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6597 poly_int64 rather than a HOST_WIDE_INT.
6598 (aarch64_get_separate_components, aarch64_process_components)
6599 (aarch64_expand_prologue, aarch64_expand_epilogue)
6600 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6601 (aarch64_anchor_offset): New function, split out from...
6602 (aarch64_legitimize_address): ...here.
6603 (aarch64_builtin_vectorization_cost): Handle polynomial
6604 TYPE_VECTOR_SUBPARTS.
6605 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6606 GET_MODE_NUNITS.
6607 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6608 number of elements from the PARALLEL rather than the mode.
6609 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6610 rather than GET_MODE_BITSIZE.
6611 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6612 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6613 (aarch64_expand_vec_perm_const_1): Handle polynomial
6614 d->perm.length () and d->perm elements.
6615 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6616 Apply to_constant to d->perm elements.
6617 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6618 polynomial CONST_VECTOR_NUNITS.
6619 (aarch64_move_pointer): Take amount as a poly_int64 rather
6620 than an int.
6621 (aarch64_progress_pointer): Avoid temporary variable.
6622 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6623 the mode attribute instead of GET_MODE.
6625 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6626 Alan Hayward <alan.hayward@arm.com>
6627 David Sherwood <david.sherwood@arm.com>
6629 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6630 x exists before using it.
6631 (aarch64_add_constant_internal): Rename to...
6632 (aarch64_add_offset_1): ...this. Replace regnum with separate
6633 src and dest rtxes. Handle the case in which they're different,
6634 including when the offset is zero. Replace scratchreg with an rtx.
6635 Use 2 additions if there is no spare register into which we can
6636 move a 16-bit constant.
6637 (aarch64_add_constant): Delete.
6638 (aarch64_add_offset): Replace reg with separate src and dest
6639 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6640 Use aarch64_add_offset_1.
6641 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6642 an rtx rather than an int. Take the delta as a poly_int64
6643 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6644 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6645 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6646 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6647 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6648 and aarch64_add_sp.
6649 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6650 aarch64_add_constant.
6652 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6654 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6655 Use scalar_float_mode.
6657 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6659 * config/aarch64/aarch64-simd.md
6660 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6661 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6662 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6663 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6664 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6665 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6666 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6667 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6668 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6669 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6671 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6673 PR target/83514
6674 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6675 targ_options->x_arm_arch_string is non NULL.
6677 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6679 * config/aarch64/aarch64.h
6680 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6682 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6684 PR target/82096
6685 * expmed.c (emit_store_flag_force): Swap if const op0
6686 and change VOIDmode to mode of op0.
6688 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6690 PR rtl-optimization/83761
6691 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6692 than bytes to mode_for_size.
6694 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6696 PR middle-end/83189
6697 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6698 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6699 profile.
6701 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6703 PR middle-end/83575
6704 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6705 when in layout mode.
6706 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6707 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6708 partition fixup.
6710 2018-01-10 Michael Collison <michael.collison@arm.com>
6712 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6713 * config/aarch64/aarch64-option-extension.def: Add
6714 AARCH64_OPT_EXTENSION of 'fp16fml'.
6715 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6716 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6717 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6718 * config/aarch64/constraints.md (Ui7): New constraint.
6719 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6720 (VFMLA_SEL_W): Ditto.
6721 (f16quad): Ditto.
6722 (f16mac1): Ditto.
6723 (VFMLA16_LOW): New int iterator.
6724 (VFMLA16_HIGH): Ditto.
6725 (UNSPEC_FMLAL): New unspec.
6726 (UNSPEC_FMLSL): Ditto.
6727 (UNSPEC_FMLAL2): Ditto.
6728 (UNSPEC_FMLSL2): Ditto.
6729 (f16mac): New code attribute.
6730 * config/aarch64/aarch64-simd-builtins.def
6731 (aarch64_fmlal_lowv2sf): Ditto.
6732 (aarch64_fmlsl_lowv2sf): Ditto.
6733 (aarch64_fmlalq_lowv4sf): Ditto.
6734 (aarch64_fmlslq_lowv4sf): Ditto.
6735 (aarch64_fmlal_highv2sf): Ditto.
6736 (aarch64_fmlsl_highv2sf): Ditto.
6737 (aarch64_fmlalq_highv4sf): Ditto.
6738 (aarch64_fmlslq_highv4sf): Ditto.
6739 (aarch64_fmlal_lane_lowv2sf): Ditto.
6740 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6741 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6742 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6743 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6744 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6745 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6746 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6747 (aarch64_fmlal_lane_highv2sf): Ditto.
6748 (aarch64_fmlsl_lane_highv2sf): Ditto.
6749 (aarch64_fmlal_laneq_highv2sf): Ditto.
6750 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6751 (aarch64_fmlalq_lane_highv4sf): Ditto.
6752 (aarch64_fmlsl_lane_highv4sf): Ditto.
6753 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6754 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6755 * config/aarch64/aarch64-simd.md:
6756 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6757 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6758 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6759 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6760 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6761 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6762 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6763 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6764 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6765 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6766 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6767 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6768 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6769 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6770 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6771 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6772 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6773 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6774 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6775 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6776 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6777 (vfmlsl_low_u32): Ditto.
6778 (vfmlalq_low_u32): Ditto.
6779 (vfmlslq_low_u32): Ditto.
6780 (vfmlal_high_u32): Ditto.
6781 (vfmlsl_high_u32): Ditto.
6782 (vfmlalq_high_u32): Ditto.
6783 (vfmlslq_high_u32): Ditto.
6784 (vfmlal_lane_low_u32): Ditto.
6785 (vfmlsl_lane_low_u32): Ditto.
6786 (vfmlal_laneq_low_u32): Ditto.
6787 (vfmlsl_laneq_low_u32): Ditto.
6788 (vfmlalq_lane_low_u32): Ditto.
6789 (vfmlslq_lane_low_u32): Ditto.
6790 (vfmlalq_laneq_low_u32): Ditto.
6791 (vfmlslq_laneq_low_u32): Ditto.
6792 (vfmlal_lane_high_u32): Ditto.
6793 (vfmlsl_lane_high_u32): Ditto.
6794 (vfmlal_laneq_high_u32): Ditto.
6795 (vfmlsl_laneq_high_u32): Ditto.
6796 (vfmlalq_lane_high_u32): Ditto.
6797 (vfmlslq_lane_high_u32): Ditto.
6798 (vfmlalq_laneq_high_u32): Ditto.
6799 (vfmlslq_laneq_high_u32): Ditto.
6800 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6801 (AARCH64_FL_FOR_ARCH8_4): New.
6802 (AARCH64_ISA_F16FML): New ISA flag.
6803 (TARGET_F16FML): New feature flag for fp16fml.
6804 (doc/invoke.texi): Document new fp16fml option.
6806 2018-01-10 Michael Collison <michael.collison@arm.com>
6808 * config/aarch64/aarch64-builtins.c:
6809 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6810 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6811 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6812 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6813 (AARCH64_ISA_SHA3): New ISA flag.
6814 (TARGET_SHA3): New feature flag for sha3.
6815 * config/aarch64/iterators.md (sha512_op): New int attribute.
6816 (CRYPTO_SHA512): New int iterator.
6817 (UNSPEC_SHA512H): New unspec.
6818 (UNSPEC_SHA512H2): Ditto.
6819 (UNSPEC_SHA512SU0): Ditto.
6820 (UNSPEC_SHA512SU1): Ditto.
6821 * config/aarch64/aarch64-simd-builtins.def
6822 (aarch64_crypto_sha512hqv2di): New builtin.
6823 (aarch64_crypto_sha512h2qv2di): Ditto.
6824 (aarch64_crypto_sha512su0qv2di): Ditto.
6825 (aarch64_crypto_sha512su1qv2di): Ditto.
6826 (aarch64_eor3qv8hi): Ditto.
6827 (aarch64_rax1qv2di): Ditto.
6828 (aarch64_xarqv2di): Ditto.
6829 (aarch64_bcaxqv8hi): Ditto.
6830 * config/aarch64/aarch64-simd.md:
6831 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6832 (aarch64_crypto_sha512su0qv2di): Ditto.
6833 (aarch64_crypto_sha512su1qv2di): Ditto.
6834 (aarch64_eor3qv8hi): Ditto.
6835 (aarch64_rax1qv2di): Ditto.
6836 (aarch64_xarqv2di): Ditto.
6837 (aarch64_bcaxqv8hi): Ditto.
6838 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6839 (vsha512h2q_u64): Ditto.
6840 (vsha512su0q_u64): Ditto.
6841 (vsha512su1q_u64): Ditto.
6842 (veor3q_u16): Ditto.
6843 (vrax1q_u64): Ditto.
6844 (vxarq_u64): Ditto.
6845 (vbcaxq_u16): Ditto.
6846 * config/arm/types.md (crypto_sha512): New type attribute.
6847 (crypto_sha3): Ditto.
6848 (doc/invoke.texi): Document new sha3 option.
6850 2018-01-10 Michael Collison <michael.collison@arm.com>
6852 * config/aarch64/aarch64-builtins.c:
6853 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6854 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6855 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6856 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6857 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6858 (AARCH64_ISA_SM4): New ISA flag.
6859 (TARGET_SM4): New feature flag for sm4.
6860 * config/aarch64/aarch64-simd-builtins.def
6861 (aarch64_sm3ss1qv4si): Ditto.
6862 (aarch64_sm3tt1aq4si): Ditto.
6863 (aarch64_sm3tt1bq4si): Ditto.
6864 (aarch64_sm3tt2aq4si): Ditto.
6865 (aarch64_sm3tt2bq4si): Ditto.
6866 (aarch64_sm3partw1qv4si): Ditto.
6867 (aarch64_sm3partw2qv4si): Ditto.
6868 (aarch64_sm4eqv4si): Ditto.
6869 (aarch64_sm4ekeyqv4si): Ditto.
6870 * config/aarch64/aarch64-simd.md:
6871 (aarch64_sm3ss1qv4si): Ditto.
6872 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6873 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6874 (aarch64_sm4eqv4si): Ditto.
6875 (aarch64_sm4ekeyqv4si): Ditto.
6876 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6877 (sm3part_op): Ditto.
6878 (CRYPTO_SM3TT): Ditto.
6879 (CRYPTO_SM3PART): Ditto.
6880 (UNSPEC_SM3SS1): New unspec.
6881 (UNSPEC_SM3TT1A): Ditto.
6882 (UNSPEC_SM3TT1B): Ditto.
6883 (UNSPEC_SM3TT2A): Ditto.
6884 (UNSPEC_SM3TT2B): Ditto.
6885 (UNSPEC_SM3PARTW1): Ditto.
6886 (UNSPEC_SM3PARTW2): Ditto.
6887 (UNSPEC_SM4E): Ditto.
6888 (UNSPEC_SM4EKEY): Ditto.
6889 * config/aarch64/constraints.md (Ui2): New constraint.
6890 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6891 * config/arm/types.md (crypto_sm3): New type attribute.
6892 (crypto_sm4): Ditto.
6893 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6894 (vsm3tt1aq_u32): Ditto.
6895 (vsm3tt1bq_u32): Ditto.
6896 (vsm3tt2aq_u32): Ditto.
6897 (vsm3tt2bq_u32): Ditto.
6898 (vsm3partw1q_u32): Ditto.
6899 (vsm3partw2q_u32): Ditto.
6900 (vsm4eq_u32): Ditto.
6901 (vsm4ekeyq_u32): Ditto.
6902 (doc/invoke.texi): Document new sm4 option.
6904 2018-01-10 Michael Collison <michael.collison@arm.com>
6906 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6907 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6908 (AARCH64_FL_FOR_ARCH8_4): New.
6909 (AARCH64_FL_V8_4): New flag.
6910 (doc/invoke.texi): Document new armv8.4-a option.
6912 2018-01-10 Michael Collison <michael.collison@arm.com>
6914 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6915 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6916 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6917 * config/aarch64/aarch64-option-extension.def: Add
6918 AARCH64_OPT_EXTENSION of 'sha2'.
6919 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6920 (crypto): Disable sha2 and aes if crypto disabled.
6921 (crypto): Enable aes and sha2 if enabled.
6922 (simd): Disable sha2 and aes if simd disabled.
6923 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6924 New flags.
6925 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6926 (TARGET_SHA2): New feature flag for sha2.
6927 (TARGET_AES): New feature flag for aes.
6928 * config/aarch64/aarch64-simd.md:
6929 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6930 conditional on TARGET_AES.
6931 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6932 (aarch64_crypto_sha1hsi): Make pattern conditional
6933 on TARGET_SHA2.
6934 (aarch64_crypto_sha1hv4si): Ditto.
6935 (aarch64_be_crypto_sha1hv4si): Ditto.
6936 (aarch64_crypto_sha1su1v4si): Ditto.
6937 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6938 (aarch64_crypto_sha1su0v4si): Ditto.
6939 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6940 (aarch64_crypto_sha256su0v4si): Ditto.
6941 (aarch64_crypto_sha256su1v4si): Ditto.
6942 (doc/invoke.texi): Document new aes and sha2 options.
6944 2018-01-10 Martin Sebor <msebor@redhat.com>
6946 PR tree-optimization/83781
6947 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6948 as string arrays.
6950 2018-01-11 Martin Sebor <msebor@gmail.com>
6951 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6953 PR tree-optimization/83501
6954 PR tree-optimization/81703
6956 * tree-ssa-strlen.c (get_string_cst): Rename...
6957 (get_string_len): ...to this. Handle global constants.
6958 (handle_char_store): Adjust.
6960 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6961 Jim Wilson <jimw@sifive.com>
6963 * config/riscv/riscv-protos.h (riscv_output_return): New.
6964 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6965 (riscv_attribute_table, riscv_output_return),
6966 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6967 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6968 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6969 function.
6970 (riscv_expand_prologue): Add early return for naked function.
6971 (riscv_expand_epilogue): Likewise.
6972 (riscv_function_ok_for_sibcall): Return false for naked function.
6973 (riscv_set_current_function): New.
6974 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6975 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6976 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6977 * doc/extend.texi (RISC-V Function Attributes): New.
6979 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6981 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6982 check for 128-bit long double before checking TCmode.
6983 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6984 128-bit long doubles before checking TFmode or TCmode.
6985 (FLOAT128_IBM_P): Likewise.
6987 2018-01-10 Martin Sebor <msebor@redhat.com>
6989 PR tree-optimization/83671
6990 * builtins.c (c_strlen): Unconditionally return zero for the empty
6991 string.
6992 Use -Warray-bounds for warnings.
6993 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6994 for non-constant array indices with COMPONENT_REF, arrays of
6995 arrays, and pointers to arrays.
6996 (gimple_fold_builtin_strlen): Determine and set length range for
6997 non-constant character arrays.
6999 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7001 PR middle-end/81897
7002 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7003 empty blocks.
7005 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7007 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7009 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7011 PR target/83399
7012 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7013 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7014 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7015 indexed_or_indirect_operand predicate.
7016 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7017 (*vsx_le_perm_load_v8hi): Likewise.
7018 (*vsx_le_perm_load_v16qi): Likewise.
7019 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7020 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7021 (*vsx_le_perm_store_v8hi): Likewise.
7022 (*vsx_le_perm_store_v16qi): Likewise.
7023 (eight unnamed splitters): Likewise.
7025 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7027 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7028 * config/rs6000/emmintrin.h: Likewise.
7029 * config/rs6000/mmintrin.h: Likewise.
7030 * config/rs6000/xmmintrin.h: Likewise.
7032 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7034 PR c++/43486
7035 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7036 "public_flag".
7037 * tree.c (tree_nop_conversion): Return true for location wrapper
7038 nodes.
7039 (maybe_wrap_with_location): New function.
7040 (selftest::check_strip_nops): New function.
7041 (selftest::test_location_wrappers): New function.
7042 (selftest::tree_c_tests): Call it.
7043 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7044 (maybe_wrap_with_location): New decl.
7045 (EXPR_LOCATION_WRAPPER_P): New macro.
7046 (location_wrapper_p): New inline function.
7047 (tree_strip_any_location_wrapper): New inline function.
7049 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7051 PR target/83735
7052 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7053 stack_realign_offset for the largest alignment of stack slot
7054 actually used.
7055 (ix86_find_max_used_stack_alignment): New function.
7056 (ix86_finalize_stack_frame_flags): Use it. Set
7057 max_used_stack_alignment if we don't realign stack.
7058 * config/i386/i386.h (machine_function): Add
7059 max_used_stack_alignment.
7061 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7063 * config/arm/arm.opt (-mbranch-cost): New option.
7064 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7065 account.
7067 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7069 PR target/83629
7070 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7071 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7073 2018-01-10 Richard Biener <rguenther@suse.de>
7075 PR debug/83765
7076 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7077 early out so it also covers the case where we have a non-NULL
7078 origin.
7080 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7082 PR tree-optimization/83753
7083 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7084 for non-strided grouped accesses if the number of elements is 1.
7086 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7088 PR target/81616
7089 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7090 * i386.h (TARGET_USE_GATHER): Define.
7091 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7093 2018-01-10 Martin Liska <mliska@suse.cz>
7095 PR bootstrap/82831
7096 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7097 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7098 partitioning.
7099 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7100 CLEANUP_NO_PARTITIONING is not set.
7102 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7104 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7105 for vectors, as a partial revert of r254296.
7106 * rtl.h (const_vec_p): Delete.
7107 (const_vec_duplicate_p): Don't test for vector CONSTs.
7108 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7109 * expmed.c (make_tree): Likewise.
7111 Revert:
7112 * common.md (E, F): Use CONSTANT_P instead of checking for
7113 CONST_VECTOR.
7114 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7115 checking for CONST_VECTOR.
7117 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7119 PR middle-end/83575
7120 * predict.c (force_edge_cold): Handle in more sane way edges
7121 with no prediction.
7123 2018-01-09 Carl Love <cel@us.ibm.com>
7125 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7126 V4SI, V4SF types.
7127 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7128 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7129 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7130 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7131 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7132 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7133 * config/rs6000/rs6000-protos.h: Add extern defition for
7134 rs6000_generate_float2_double_code.
7135 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7136 function.
7137 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7138 (float2_v2df): Add define_expand.
7140 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7142 PR target/83628
7143 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7144 op_mode in the force_to_mode call.
7146 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7148 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7149 instead of checking each element individually.
7150 (aarch64_evpc_uzp): Likewise.
7151 (aarch64_evpc_zip): Likewise.
7152 (aarch64_evpc_ext): Likewise.
7153 (aarch64_evpc_rev): Likewise.
7154 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7155 instead of checking each element individually. Return true without
7156 generating rtl if
7157 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7158 whether all selected elements come from the same input, instead of
7159 checking each element individually. Remove calls to gen_rtx_REG,
7160 start_sequence and end_sequence and instead assert that no rtl is
7161 generated.
7163 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7165 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7166 order of HIGH and CONST checks.
7168 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7170 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7171 if the destination isn't an SSA_NAME.
7173 2018-01-09 Richard Biener <rguenther@suse.de>
7175 PR tree-optimization/83668
7176 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7177 move prologue...
7178 (canonicalize_loop_form): ... here, renamed from ...
7179 (canonicalize_loop_closed_ssa_form): ... this and amended to
7180 swap successor edges for loop exit blocks to make us use
7181 the RPO order we need for initial schedule generation.
7183 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7185 PR tree-optimization/64811
7186 * match.pd: When optimizing comparisons with Inf, avoid
7187 introducing or losing exceptions from comparisons with NaN.
7189 2018-01-09 Martin Liska <mliska@suse.cz>
7191 PR sanitizer/82517
7192 * asan.c (shadow_mem_size): Add gcc_assert.
7194 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7196 Don't save registers in main().
7198 PR target/83738
7199 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7200 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7201 * config/avr/avr.c (avr_set_current_function): Don't error if
7202 naked, OS_task or OS_main are specified at the same time.
7203 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7204 OS_main.
7205 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7206 attribute.
7207 * common/config/avr/avr-common.c (avr_option_optimization_table):
7208 Switch on -mmain-is-OS_task for optimizing compilations.
7210 2018-01-09 Richard Biener <rguenther@suse.de>
7212 PR tree-optimization/83572
7213 * graphite.c: Include cfganal.h.
7214 (graphite_transform_loops): Connect infinite loops to exit
7215 and remove fake edges at the end.
7217 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7219 * ipa-inline.c (edge_badness): Revert accidental checkin.
7221 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7223 PR ipa/80763
7224 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7225 symbols; not inline clones.
7227 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7229 PR target/83507
7230 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7231 hard registers. Formatting fixes.
7233 PR preprocessor/83722
7234 * gcc.c (try_generate_repro): Pass
7235 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7236 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7237 do_report_bug.
7239 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7240 Kito Cheng <kito.cheng@gmail.com>
7242 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7243 (riscv_leaf_function_p): Delete.
7244 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7246 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7248 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7249 function.
7250 (do_ifelse): New function.
7251 (do_isel): New function.
7252 (do_sub3): New function.
7253 (do_add3): New function.
7254 (do_load_mask_compare): New function.
7255 (do_overlap_load_compare): New function.
7256 (expand_compare_loop): New function.
7257 (expand_block_compare): Call expand_compare_loop() when appropriate.
7258 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7259 option description.
7260 (-mblock-compare-inline-loop-limit): New option.
7262 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7264 PR target/83677
7265 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7266 Reverse order of second and third operands in first alternative.
7267 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7268 of first and second elements in UNSPEC_VPERMR vector.
7269 (altivec_expand_vec_perm_le): Likewise.
7271 2018-01-08 Jeff Law <law@redhat.com>
7273 PR rtl-optimizatin/81308
7274 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7275 (process_switch): If group_case_labels makes a change, then set
7276 cfg_altered.
7277 (pass_convert_switch::execute): If a switch is converted, then
7278 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7280 PR rtl-optimization/81308
7281 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7282 splitting insns.
7284 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7286 PR target/83663 - Revert r255946
7287 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7288 generation for cases where splatting a value is not useful.
7289 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7290 across a vec_duplicate and a paradoxical subreg forming a vector
7291 mode to a vec_concat.
7293 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7295 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7296 -march=armv8.3-a variants.
7297 * config/arm/t-multilib: Likewise.
7298 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7300 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7302 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7303 to generate rtl.
7304 (cceq_ior_compare_complement): Give it a name so I can use it, and
7305 change boolean_or_operator predicate to boolean_operator so it can
7306 be used to generate a crand.
7307 (eqne): New code iterator.
7308 (bd/bd_neg): New code_attrs.
7309 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7310 a single define_insn.
7311 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7312 decrement (bdnzt/bdnzf/bdzt/bdzf).
7313 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7314 with the new names of the branch decrement patterns, and added the
7315 names of the branch decrement conditional patterns.
7317 2018-01-08 Richard Biener <rguenther@suse.de>
7319 PR tree-optimization/83563
7320 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7321 cache.
7323 2018-01-08 Richard Biener <rguenther@suse.de>
7325 PR middle-end/83713
7326 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7328 2018-01-08 Richard Biener <rguenther@suse.de>
7330 PR tree-optimization/83685
7331 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7332 references to abnormals.
7334 2018-01-08 Richard Biener <rguenther@suse.de>
7336 PR lto/83719
7337 * dwarf2out.c (output_indirect_strings): Handle empty
7338 skeleton_debug_str_hash.
7339 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7341 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7343 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7344 (emit_store_direct): Likewise.
7345 (arc_trampoline_adjust_address): Likewise.
7346 (arc_asm_trampoline_template): New function.
7347 (arc_initialize_trampoline): Use asm_trampoline_template.
7348 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7349 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7350 * config/arc/arc.md (flush_icache): Delete pattern.
7352 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7354 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7355 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7356 munaligned-access.
7358 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7360 PR target/83681
7361 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7362 by not USED_FOR_TARGET.
7363 (make_pass_resolve_sw_modes): Likewise.
7365 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7367 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7368 USED_FOR_TARGET.
7370 2018-01-08 Richard Biener <rguenther@suse.de>
7372 PR middle-end/83580
7373 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7375 2018-01-08 Richard Biener <rguenther@suse.de>
7377 PR middle-end/83517
7378 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7380 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7382 PR middle-end/81897
7383 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7384 basic blocks with a small number of successors.
7385 (convert_control_dep_chain_into_preds): Improve handling of
7386 forwarder blocks.
7387 (dump_predicates): Split apart into...
7388 (dump_pred_chain): ...here...
7389 (dump_pred_info): ...and here.
7390 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7391 (can_chain_union_be_invalidated_p): Improve check for invalidation
7392 of paths.
7393 (uninit_uses_cannot_happen): Avoid unnecessary if
7394 convert_control_dep_chain_into_preds yielded nothing.
7396 2018-01-06 Martin Sebor <msebor@redhat.com>
7398 PR tree-optimization/83640
7399 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7400 subtracting negative offset from size.
7401 (builtin_access::overlap): Adjust offset bounds of the access to fall
7402 within the size of the object if possible.
7404 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7406 PR rtl-optimization/83699
7407 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7408 extract_bit_field_as_subreg to cases in which the extracted
7409 value is also a vector.
7411 * lra-constraints.c (process_alt_operands): Test for the equivalence
7412 substitutions when detecting a possible reload cycle.
7414 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7416 PR debug/83480
7417 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7418 by default if flag_selective_schedling{,2}. Formatting fixes.
7420 PR rtl-optimization/83682
7421 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7422 if it has non-VECTOR_MODE element mode.
7423 (vec_duplicate_p): Likewise.
7425 PR middle-end/83694
7426 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7427 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7429 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7431 PR target/83604
7432 * config/i386/i386-builtin.def
7433 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7434 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7435 Require also OPTION_MASK_ISA_AVX512F in addition to
7436 OPTION_MASK_ISA_GFNI.
7437 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7438 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7439 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7440 to OPTION_MASK_ISA_GFNI.
7441 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7442 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7443 OPTION_MASK_ISA_AVX512BW.
7444 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7445 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7446 addition to OPTION_MASK_ISA_GFNI.
7447 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7448 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7449 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7450 to OPTION_MASK_ISA_GFNI.
7451 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7452 a requirement for all ISAs rather than any of them with a few
7453 exceptions.
7454 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7455 processing.
7456 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7457 bitmasks to be enabled with 3 exceptions, instead of requiring any
7458 enabled ISA with lots of exceptions.
7459 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7460 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7461 Change avx512bw in isa attribute to avx512f.
7462 * config/i386/sgxintrin.h: Add license boilerplate.
7463 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7464 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7465 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7466 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7467 defined.
7468 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7469 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7470 temporarily sse2 rather than sse if not enabled already.
7472 PR target/83604
7473 * config/i386/sse.md (VI248_VLBW): Rename to ...
7474 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7475 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7476 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7477 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7478 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7479 mode iterator instead of VI248_VLBW.
7481 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7483 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7484 (record_modified): Skip clobbers; add debug output.
7485 (param_change_prob): Use sreal frequencies.
7487 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7489 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7490 punt for user-aligned variables.
7492 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7494 * tree-chrec.c (chrec_contains_symbols): Return true for
7495 POLY_INT_CST.
7497 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7499 PR target/82439
7500 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7501 of (x|y) == x for BICS pattern.
7503 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7505 PR tree-optimization/83605
7506 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7507 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7508 can throw.
7510 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7512 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7513 * config/epiphany/rtems.h: New file.
7515 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7516 Uros Bizjak <ubizjak@gmail.com>
7518 PR target/83554
7519 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7520 QIreg_operand instead of register_operand predicate.
7521 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7522 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7523 comments instead of -fmitigate[-_]rop.
7525 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7527 PR bootstrap/81926
7528 * cgraphunit.c (symbol_table::compile): Switch to text_section
7529 before calling assembly_start debug hook.
7530 * run-rtl-passes.c (run_rtl_passes): Likewise.
7531 Include output.h.
7533 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7535 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7536 range_int_cst_p rather than !symbolic_range_p before calling
7537 extract_range_from_multiplicative_op_1.
7539 2018-01-04 Jeff Law <law@redhat.com>
7541 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7542 redundant test in assertion.
7544 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7546 * doc/rtl.texi: Document machine_mode wrapper classes.
7548 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7550 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7551 using tree_to_uhwi.
7553 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7555 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7556 the VEC_PERM_EXPR fold to fail.
7558 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7560 PR debug/83585
7561 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7562 to switched_sections.
7564 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7566 PR target/83680
7567 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7568 test for d.testing.
7570 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7572 PR target/83387
7573 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7574 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7576 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7578 PR debug/83666
7579 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7580 is BLKmode and bitpos not zero or mode change is needed.
7582 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7584 PR target/83675
7585 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7586 TARGET_VIS2.
7588 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7590 PR target/83628
7591 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7592 instead of MULT rtx. Update all corresponding splitters.
7593 (*saddl_se): Ditto.
7594 (*ssub<modesuffix>): Ditto.
7595 (*ssubl_se): Ditto.
7596 (*cmp_sadd_di): Update split patterns.
7597 (*cmp_sadd_si): Ditto.
7598 (*cmp_sadd_sidi): Ditto.
7599 (*cmp_ssub_di): Ditto.
7600 (*cmp_ssub_si): Ditto.
7601 (*cmp_ssub_sidi): Ditto.
7602 * config/alpha/predicates.md (const23_operand): New predicate.
7603 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7604 Look for ASHIFT, not MULT inner operand.
7605 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7607 2018-01-04 Martin Liska <mliska@suse.cz>
7609 PR gcov-profile/83669
7610 * gcov.c (output_intermediate_file): Add version to intermediate
7611 gcov file.
7612 * doc/gcov.texi: Document new field 'version' in intermediate
7613 file format. Fix location of '-k' option of gcov command.
7615 2018-01-04 Martin Liska <mliska@suse.cz>
7617 PR ipa/82352
7618 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7620 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7622 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7624 2018-01-03 Martin Sebor <msebor@redhat.com>
7626 PR tree-optimization/83655
7627 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7628 checking calls with invalid arguments.
7630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7632 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7633 (vectorizable_mask_load_store): Delete.
7634 (vectorizable_call): Return false for masked loads and stores.
7635 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7636 instead of gimple_assign_rhs1.
7637 (vectorizable_load): Handle IFN_MASK_LOAD.
7638 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7642 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7643 split out from..,
7644 (vectorizable_mask_load_store): ...here.
7645 (vectorizable_load): ...and here.
7647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7649 * tree-vect-stmts.c (vect_build_all_ones_mask)
7650 (vect_build_zero_merge_argument): New functions, split out from...
7651 (vectorizable_load): ...here.
7653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7655 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7656 split out from...
7657 (vectorizable_mask_load_store): ...here.
7658 (vectorizable_store): ...and here.
7660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7662 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7663 split out from...
7664 (vectorizable_mask_load_store): ...here.
7666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7668 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7669 (vect_model_store_cost): Take a vec_load_store_type instead of a
7670 vect_def_type.
7671 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7672 (vect_model_store_cost): Take a vec_load_store_type instead of a
7673 vect_def_type.
7674 (vectorizable_mask_load_store): Update accordingly.
7675 (vectorizable_store): Likewise.
7676 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7680 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7681 IFN_MASK_LOAD calls here rather than...
7682 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7685 Alan Hayward <alan.hayward@arm.com>
7686 David Sherwood <david.sherwood@arm.com>
7688 * expmed.c (extract_bit_field_1): For vector extracts,
7689 fall back to extract_bit_field_as_subreg if vec_extract
7690 isn't available.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7697 they are variable or constant sized.
7698 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7699 slots for constant-sized data.
7701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7702 Alan Hayward <alan.hayward@arm.com>
7703 David Sherwood <david.sherwood@arm.com>
7705 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7706 handling COND_EXPRs with boolean comparisons, try to find a better
7707 basis for the mask type than the boolean itself.
7709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7711 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7712 is calculated and how it can be overridden.
7713 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7714 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7715 if defined.
7716 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7717 if nonzero.
7719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7720 Alan Hayward <alan.hayward@arm.com>
7721 David Sherwood <david.sherwood@arm.com>
7723 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7724 Remove the mode argument.
7725 (aarch64_simd_valid_immediate): Remove the mode and inverse
7726 arguments.
7727 * config/aarch64/iterators.md (bitsize): New iterator.
7728 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7729 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7730 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7731 aarch64_simd_valid_immediate.
7732 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7733 (aarch64_reg_or_bic_imm): Likewise.
7734 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7735 with an insn_type enum and msl with a modifier_type enum.
7736 Replace element_width with a scalar_mode. Change the shift
7737 to unsigned int. Add constructors for scalar_float_mode and
7738 scalar_int_mode elements.
7739 (aarch64_vect_float_const_representable_p): Delete.
7740 (aarch64_can_const_movi_rtx_p)
7741 (aarch64_simd_scalar_immediate_valid_for_move)
7742 (aarch64_simd_make_constant): Update call to
7743 aarch64_simd_valid_immediate.
7744 (aarch64_advsimd_valid_immediate_hs): New function.
7745 (aarch64_advsimd_valid_immediate): Likewise.
7746 (aarch64_simd_valid_immediate): Remove mode and inverse
7747 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7748 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7749 and aarch64_float_const_representable_p on the result.
7750 (aarch64_output_simd_mov_immediate): Remove mode argument.
7751 Update call to aarch64_simd_valid_immediate and use of
7752 simd_immediate_info.
7753 (aarch64_output_scalar_simd_mov_immediate): Update call
7754 accordingly.
7756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7757 Alan Hayward <alan.hayward@arm.com>
7758 David Sherwood <david.sherwood@arm.com>
7760 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7761 (mode_nunits): Likewise CONST_MODE_NUNITS.
7762 * machmode.def (ADJUST_NUNITS): Document.
7763 * genmodes.c (mode_data::need_nunits_adj): New field.
7764 (blank_mode): Update accordingly.
7765 (adj_nunits): New variable.
7766 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7767 parameter.
7768 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7769 listed in adj_nunits.
7770 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7771 listed in adj_nunits. Don't emit case statements for such modes.
7772 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7773 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7774 nothing if adj_nunits is nonnull.
7775 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7776 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7777 (emit_mode_fbit): Update use of print_maybe_const_decl.
7778 (emit_move_size): Likewise. Treat the array as non-const
7779 if adj_nunits.
7780 (emit_mode_adjustments): Handle adj_nunits.
7782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7785 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7786 (VECTOR_MODES): Use it.
7787 (make_vector_modes): Take the prefix as an argument.
7789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7790 Alan Hayward <alan.hayward@arm.com>
7791 David Sherwood <david.sherwood@arm.com>
7793 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7794 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7795 for MODE_VECTOR_BOOL.
7796 * machmode.def (VECTOR_BOOL_MODE): Document.
7797 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7798 (make_vector_bool_mode): New function.
7799 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7800 MODE_VECTOR_BOOL.
7801 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7802 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7803 Likewise.
7804 * stor-layout.c (int_mode_for_mode): Likewise.
7805 * tree.c (build_vector_type_for_mode): Likewise.
7806 * varasm.c (output_constant_pool_2): Likewise.
7807 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7808 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7809 for MODE_VECTOR_BOOL.
7810 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7811 of mode class checks.
7812 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7813 instead of a list of mode class checks.
7814 (expand_vector_scalar_condition): Likewise.
7815 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7818 Alan Hayward <alan.hayward@arm.com>
7819 David Sherwood <david.sherwood@arm.com>
7821 * machmode.h (mode_size): Change from unsigned short to
7822 poly_uint16_pod.
7823 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7824 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7825 or if measurement_type is not polynomial.
7826 (fixed_size_mode::includes_p): Check for constant-sized modes.
7827 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7828 return a poly_uint16 rather than an unsigned short.
7829 (emit_mode_size): Change the type of mode_size from unsigned short
7830 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7831 (emit_mode_adjustments): Cope with polynomial vector sizes.
7832 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7833 for GET_MODE_SIZE.
7834 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7835 for GET_MODE_SIZE.
7836 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7837 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7838 * caller-save.c (setup_save_areas): Likewise.
7839 (replace_reg_with_saved_mem): Likewise.
7840 * calls.c (emit_library_call_value_1): Likewise.
7841 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7842 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7843 (gen_lowpart_for_combine): Likewise.
7844 * convert.c (convert_to_integer_1): Likewise.
7845 * cse.c (equiv_constant, cse_insn): Likewise.
7846 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7847 (cselib_subst_to_values): Likewise.
7848 * dce.c (word_dce_process_block): Likewise.
7849 * df-problems.c (df_word_lr_mark_ref): Likewise.
7850 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7851 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7852 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7853 (rtl_for_decl_location): Likewise.
7854 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7855 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7856 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7857 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7858 (expand_expr_real_1): Likewise.
7859 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7860 (pad_below): Likewise.
7861 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7862 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7863 * ira.c (get_subreg_tracking_sizes): Likewise.
7864 * ira-build.c (ira_create_allocno_objects): Likewise.
7865 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7866 (ira_sort_regnos_for_alter_reg): Likewise.
7867 * ira-costs.c (record_operand_costs): Likewise.
7868 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7869 (resolve_simple_move): Likewise.
7870 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7871 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7872 (lra_constraints): Likewise.
7873 (CONST_POOL_OK_P): Reject variable-sized modes.
7874 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7875 (add_pseudo_to_slot, lra_spill): Likewise.
7876 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7877 * optabs-query.c (get_best_extraction_insn): Likewise.
7878 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7879 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7880 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7881 * recog.c (offsettable_address_addr_space_p): Likewise.
7882 * regcprop.c (maybe_mode_change): Likewise.
7883 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7884 * regrename.c (build_def_use): Likewise.
7885 * regstat.c (dump_reg_info): Likewise.
7886 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7887 (find_reloads, find_reloads_subreg_address): Likewise.
7888 * reload1.c (eliminate_regs_1): Likewise.
7889 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7890 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7891 (simplify_binary_operation_1, simplify_subreg): Likewise.
7892 * targhooks.c (default_function_arg_padding): Likewise.
7893 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7894 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7895 (verify_gimple_assign_ternary): Likewise.
7896 * tree-inline.c (estimate_move_cost): Likewise.
7897 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7898 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7899 (get_address_cost_ainc): Likewise.
7900 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7901 (vect_supportable_dr_alignment): Likewise.
7902 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7903 (vectorizable_reduction): Likewise.
7904 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7905 (vectorizable_operation, vectorizable_load): Likewise.
7906 * tree.c (build_same_sized_truth_vector_type): Likewise.
7907 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7908 * var-tracking.c (emit_note_insn_var_location): Likewise.
7909 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7910 (ADDR_VEC_ALIGN): Likewise.
7912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7913 Alan Hayward <alan.hayward@arm.com>
7914 David Sherwood <david.sherwood@arm.com>
7916 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7917 unsigned short.
7918 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7919 or if measurement_type is polynomial.
7920 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7921 * combine.c (make_extraction): Likewise.
7922 * dse.c (find_shift_sequence): Likewise.
7923 * dwarf2out.c (mem_loc_descriptor): Likewise.
7924 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7925 (extract_bit_field, extract_low_bits): Likewise.
7926 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7927 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7928 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7929 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7930 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7931 * reload.c (find_reloads): Likewise.
7932 * reload1.c (alter_reg): Likewise.
7933 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7934 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7935 * tree-if-conv.c (predicate_mem_writes): Likewise.
7936 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7937 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7938 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7939 * valtrack.c (dead_debug_insert_temp): Likewise.
7940 * varasm.c (mergeable_constant_section): Likewise.
7941 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7944 Alan Hayward <alan.hayward@arm.com>
7945 David Sherwood <david.sherwood@arm.com>
7947 * expr.c (expand_assignment): Cope with polynomial mode sizes
7948 when assigning to a CONCAT.
7950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7951 Alan Hayward <alan.hayward@arm.com>
7952 David Sherwood <david.sherwood@arm.com>
7954 * machmode.h (mode_precision): Change from unsigned short to
7955 poly_uint16_pod.
7956 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7957 short.
7958 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7959 or if measurement_type is not polynomial.
7960 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7961 in which the mode is already known to be a scalar_int_mode.
7962 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7963 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7964 initializer.
7965 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7966 for GET_MODE_PRECISION.
7967 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7968 for GET_MODE_PRECISION.
7969 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7970 as polynomial.
7971 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7972 (expand_field_assignment, make_extraction): Likewise.
7973 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7974 (get_last_value): Likewise.
7975 * convert.c (convert_to_integer_1): Likewise.
7976 * cse.c (cse_insn): Likewise.
7977 * expr.c (expand_expr_real_1): Likewise.
7978 * lra-constraints.c (simplify_operand_subreg): Likewise.
7979 * optabs-query.c (can_atomic_load_p): Likewise.
7980 * optabs.c (expand_atomic_load): Likewise.
7981 (expand_atomic_store): Likewise.
7982 * ree.c (combine_reaching_defs): Likewise.
7983 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7984 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7985 * tree.h (type_has_mode_precision_p): Likewise.
7986 * ubsan.c (instrument_si_overflow): Likewise.
7988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7989 Alan Hayward <alan.hayward@arm.com>
7990 David Sherwood <david.sherwood@arm.com>
7992 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7993 polynomial numbers of units.
7994 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7995 (valid_vector_subparts_p): New function.
7996 (build_vector_type): Remove temporary shim and take the number
7997 of units as a poly_uint64 rather than an int.
7998 (build_opaque_vector_type): Take the number of units as a
7999 poly_uint64 rather than an int.
8000 * tree.c (build_vector_from_ctor): Handle polynomial
8001 TYPE_VECTOR_SUBPARTS.
8002 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8003 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8004 (build_vector_from_val): If the number of units is variable,
8005 use build_vec_duplicate_cst for constant operands and
8006 VEC_DUPLICATE_EXPR otherwise.
8007 (make_vector_type): Remove temporary is_constant ().
8008 (build_vector_type, build_opaque_vector_type): Take the number of
8009 units as a poly_uint64 rather than an int.
8010 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8011 VECTOR_CST_NELTS.
8012 * cfgexpand.c (expand_debug_expr): Likewise.
8013 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8014 (store_constructor, expand_expr_real_1): Likewise.
8015 (const_scalar_mask_from_tree): Likewise.
8016 * fold-const-call.c (fold_const_reduction): Likewise.
8017 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8018 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8019 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8020 (fold_relational_const): Likewise.
8021 (native_interpret_vector): Likewise. Change the size from an
8022 int to an unsigned int.
8023 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8024 TYPE_VECTOR_SUBPARTS.
8025 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8026 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8027 duplicating a non-constant operand into a variable-length vector.
8028 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8029 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8030 * ipa-icf.c (sem_variable::equals): Likewise.
8031 * match.pd: Likewise.
8032 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8033 * print-tree.c (print_node): Likewise.
8034 * stor-layout.c (layout_type): Likewise.
8035 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8036 * tree-cfg.c (verify_gimple_comparison): Likewise.
8037 (verify_gimple_assign_binary): Likewise.
8038 (verify_gimple_assign_ternary): Likewise.
8039 (verify_gimple_assign_single): Likewise.
8040 * tree-pretty-print.c (dump_generic_node): Likewise.
8041 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8042 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8043 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8044 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8045 (vect_shift_permute_load_chain): Likewise.
8046 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8047 (expand_vector_condition, optimize_vector_constructor): Likewise.
8048 (lower_vec_perm, get_compute_type): Likewise.
8049 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8050 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8051 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8052 (vect_recog_mask_conversion_pattern): Likewise.
8053 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8054 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8055 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8056 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8057 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8058 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8059 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8060 (supportable_widening_operation): Likewise.
8061 (supportable_narrowing_operation): Likewise.
8062 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8063 Likewise.
8064 * varasm.c (output_constant): Likewise.
8066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8067 Alan Hayward <alan.hayward@arm.com>
8068 David Sherwood <david.sherwood@arm.com>
8070 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8071 so that both the length == 3 and length != 3 cases set up their
8072 own permute vectors. Add comments explaining why we know the
8073 number of elements is constant.
8074 (vect_permute_load_chain): Likewise.
8076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8077 Alan Hayward <alan.hayward@arm.com>
8078 David Sherwood <david.sherwood@arm.com>
8080 * machmode.h (mode_nunits): Change from unsigned char to
8081 poly_uint16_pod.
8082 (ONLY_FIXED_SIZE_MODES): New macro.
8083 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8084 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8085 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8086 New typedefs.
8087 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8088 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8089 or if measurement_type is not polynomial.
8090 * genmodes.c (ZERO_COEFFS): New macro.
8091 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8092 poly_uint16.
8093 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8094 Use ZERO_COEFFS when emitting initializers.
8095 * data-streamer.h (bp_pack_poly_value): New function.
8096 (bp_unpack_poly_value): Likewise.
8097 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8098 for GET_MODE_NUNITS.
8099 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8100 for GET_MODE_NUNITS.
8101 * tree.c (make_vector_type): Remove temporary shim and make
8102 the real function take the number of units as a poly_uint64
8103 rather than an int.
8104 (build_vector_type_for_mode): Handle polynomial nunits.
8105 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8106 * emit-rtl.c (const_vec_series_p_1): Likewise.
8107 (gen_rtx_CONST_VECTOR): Likewise.
8108 * fold-const.c (test_vec_duplicate_folding): Likewise.
8109 * genrecog.c (validate_pattern): Likewise.
8110 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8111 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8112 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8113 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8114 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8115 * rtlanal.c (subreg_get_info): Likewise.
8116 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8117 (vect_grouped_load_supported): Likewise.
8118 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8119 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8120 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8121 (simplify_const_unary_operation, simplify_binary_operation_1)
8122 (simplify_const_binary_operation, simplify_ternary_operation)
8123 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8124 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8125 instead of CONST_VECTOR_NUNITS.
8126 * varasm.c (output_constant_pool_2): Likewise.
8127 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8128 explicit-encoded elements in the XVEC for variable-length vectors.
8130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8132 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8135 Alan Hayward <alan.hayward@arm.com>
8136 David Sherwood <david.sherwood@arm.com>
8138 * coretypes.h (fixed_size_mode): Declare.
8139 (fixed_size_mode_pod): New typedef.
8140 * builtins.h (target_builtins::x_apply_args_mode)
8141 (target_builtins::x_apply_result_mode): Change type to
8142 fixed_size_mode_pod.
8143 * builtins.c (apply_args_size, apply_result_size, result_vector)
8144 (expand_builtin_apply_args_1, expand_builtin_apply)
8145 (expand_builtin_return): Update accordingly.
8147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8149 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8150 * cselib.c (cselib_hash_rtx): Likewise.
8151 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8152 CONST_VECTOR encoding.
8154 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8155 Jeff Law <law@redhat.com>
8157 PR target/83641
8158 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8159 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8160 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8161 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8163 PR target/83641
8164 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8165 explicitly probe *sp in a noreturn function if there were any callee
8166 register saves or frame pointer is needed.
8168 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8170 PR debug/83621
8171 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8172 BLKmode for ternary, binary or unary expressions.
8174 PR debug/83645
8175 * var-tracking.c (delete_vta_debug_insn): New inline function.
8176 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8177 insns from get_insns () to NULL instead of each bb separately.
8178 Use delete_vta_debug_insn. No longer static.
8179 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8180 delete_vta_debug_insns callers.
8181 * rtl.h (delete_vta_debug_insns): Declare.
8182 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8183 instead of variable_tracking_main.
8185 2018-01-03 Martin Sebor <msebor@redhat.com>
8187 PR tree-optimization/83603
8188 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8189 arguments past the endof the argument list in functions declared
8190 without a prototype.
8191 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8192 Avoid checking when arguments are null.
8194 2018-01-03 Martin Sebor <msebor@redhat.com>
8196 PR c/83559
8197 * doc/extend.texi (attribute const): Fix a typo.
8198 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8199 issuing -Wsuggest-attribute for void functions.
8201 2018-01-03 Martin Sebor <msebor@redhat.com>
8203 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8204 offset_int::from instead of wide_int::to_shwi.
8205 (maybe_diag_overlap): Remove assertion.
8206 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8207 * gimple-ssa-sprintf.c (format_directive): Same.
8208 (parse_directive): Same.
8209 (sprintf_dom_walker::compute_format_length): Same.
8210 (try_substitute_return_value): Same.
8212 2018-01-03 Jeff Law <law@redhat.com>
8214 PR middle-end/83654
8215 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8216 non-constant residual for zero at runtime and avoid probing in
8217 that case. Reorganize code for trailing problem to mirror handling
8218 of the residual.
8220 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8222 PR tree-optimization/83501
8223 * tree-ssa-strlen.c (get_string_cst): New.
8224 (handle_char_store): Call get_string_cst.
8226 2018-01-03 Martin Liska <mliska@suse.cz>
8228 PR tree-optimization/83593
8229 * tree-ssa-strlen.c: Include tree-cfg.h.
8230 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8231 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8232 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8233 to false.
8234 (strlen_dom_walker::before_dom_children): Call
8235 gimple_purge_dead_eh_edges. Dump tranformation with details
8236 dump flags.
8237 (strlen_dom_walker::before_dom_children): Update call by adding
8238 new argument cleanup_eh.
8239 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8241 2018-01-03 Martin Liska <mliska@suse.cz>
8243 PR ipa/83549
8244 * cif-code.def (VARIADIC_THUNK): New enum value.
8245 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8246 thunks.
8248 2018-01-03 Jan Beulich <jbeulich@suse.com>
8250 * sse.md (mov<mode>_internal): Tighten condition for when to use
8251 vmovdqu<ssescalarsize> for TI and OI modes.
8253 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8255 Update copyright years.
8257 2018-01-03 Martin Liska <mliska@suse.cz>
8259 PR ipa/83594
8260 * ipa-visibility.c (function_and_variable_visibility): Skip
8261 functions with noipa attribure.
8263 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8265 * gcc.c (process_command): Update copyright notice dates.
8266 * gcov-dump.c (print_version): Ditto.
8267 * gcov.c (print_version): Ditto.
8268 * gcov-tool.c (print_version): Ditto.
8269 * gengtype.c (create_file): Ditto.
8270 * doc/cpp.texi: Bump @copying's copyright year.
8271 * doc/cppinternals.texi: Ditto.
8272 * doc/gcc.texi: Ditto.
8273 * doc/gccint.texi: Ditto.
8274 * doc/gcov.texi: Ditto.
8275 * doc/install.texi: Ditto.
8276 * doc/invoke.texi: Ditto.
8278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8280 * vector-builder.h (vector_builder::m_full_nelts): Change from
8281 unsigned int to poly_uint64.
8282 (vector_builder::full_nelts): Update prototype accordingly.
8283 (vector_builder::new_vector): Likewise.
8284 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8285 (vector_builder::operator ==): Likewise.
8286 (vector_builder::finalize): Likewise.
8287 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8288 Take the number of elements as a poly_uint64 rather than an
8289 unsigned int.
8290 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8291 from unsigned int to poly_uint64.
8292 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8293 (vec_perm_indices::new_vector): Likewise.
8294 (vec_perm_indices::length): Likewise.
8295 (vec_perm_indices::nelts_per_input): Likewise.
8296 (vec_perm_indices::input_nelts): Likewise.
8297 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8298 number of elements per input as a poly_uint64 rather than an
8299 unsigned int. Use the original encoding for variable-length
8300 vectors, rather than clamping each individual element.
8301 For the second and subsequent elements in each pattern,
8302 clamp the step and base before clamping their sum.
8303 (vec_perm_indices::series_p): Handle polynomial element counts.
8304 (vec_perm_indices::all_in_range_p): Likewise.
8305 (vec_perm_indices_to_tree): Likewise.
8306 (vec_perm_indices_to_rtx): Likewise.
8307 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8308 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8309 (tree_vector_builder::new_binary_operation): Handle polynomial
8310 element counts. Return false if we need to know the number
8311 of elements at compile time.
8312 * fold-const.c (fold_vec_perm): Punt if the number of elements
8313 isn't known at compile time.
8315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8317 * vec-perm-indices.h (vec_perm_builder): Change element type
8318 from HOST_WIDE_INT to poly_int64.
8319 (vec_perm_indices::element_type): Update accordingly.
8320 (vec_perm_indices::clamp): Handle polynomial element_types.
8321 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8322 (vec_perm_indices::all_in_range_p): Likewise.
8323 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8324 than shwi trees.
8325 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8326 polynomial vec_perm_indices element types.
8327 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8328 * fold-const.c (fold_vec_perm): Likewise.
8329 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8330 * tree-vect-generic.c (lower_vec_perm): Likewise.
8331 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8332 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8333 element type to HOST_WIDE_INT.
8335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8336 Alan Hayward <alan.hayward@arm.com>
8337 David Sherwood <david.sherwood@arm.com>
8339 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8340 rather than an int. Use plus_constant.
8341 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8342 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8345 Alan Hayward <alan.hayward@arm.com>
8346 David Sherwood <david.sherwood@arm.com>
8348 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8349 a HOST_WIDE_INT to a poly_int64.
8351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8352 Alan Hayward <alan.hayward@arm.com>
8353 David Sherwood <david.sherwood@arm.com>
8355 * calls.c (load_register_parameters): Cope with polynomial
8356 mode sizes. Require a constant size for BLKmode parameters
8357 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8358 forces a parameter to be padded at the lsb end in order to
8359 fill a complete number of words, require the parameter size
8360 to be ordered wrt UNITS_PER_WORD.
8362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8363 Alan Hayward <alan.hayward@arm.com>
8364 David Sherwood <david.sherwood@arm.com>
8366 * reload1.c (spill_stack_slot_width): Change element type
8367 from unsigned int to poly_uint64_pod.
8368 (alter_reg): Treat mode sizes as polynomial.
8370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8371 Alan Hayward <alan.hayward@arm.com>
8372 David Sherwood <david.sherwood@arm.com>
8374 * reload.c (complex_word_subreg_p): New function.
8375 (reload_inner_reg_of_subreg, push_reload): Use it.
8377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8378 Alan Hayward <alan.hayward@arm.com>
8379 David Sherwood <david.sherwood@arm.com>
8381 * lra-constraints.c (process_alt_operands): Reject matched
8382 operands whose sizes aren't ordered.
8383 (match_reload): Refer to this check here.
8385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8386 Alan Hayward <alan.hayward@arm.com>
8387 David Sherwood <david.sherwood@arm.com>
8389 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8390 that the mode size is in the set {1, 2, 4, 8, 16}.
8392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8393 Alan Hayward <alan.hayward@arm.com>
8394 David Sherwood <david.sherwood@arm.com>
8396 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8397 Use plus_constant instead of gen_rtx_PLUS.
8399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8400 Alan Hayward <alan.hayward@arm.com>
8401 David Sherwood <david.sherwood@arm.com>
8403 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8404 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8405 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8406 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8407 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8408 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8409 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8410 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8411 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8412 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8413 a poly_int64.
8414 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8415 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8416 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8417 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8418 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8419 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8420 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8421 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8422 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8423 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8424 function.
8425 * expr.c (emit_move_resolve_push): Treat the input and result
8426 of PUSH_ROUNDING as a poly_int64.
8427 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8428 (emit_push_insn): Likewise.
8429 * lra-eliminations.c (mark_not_eliminable): Likewise.
8430 * recog.c (push_operand): Likewise.
8431 * reload1.c (elimination_effects): Likewise.
8432 * rtlanal.c (nonzero_bits1): Likewise.
8433 * calls.c (store_one_arg): Likewise. Require the padding to be
8434 known at compile time.
8436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8437 Alan Hayward <alan.hayward@arm.com>
8438 David Sherwood <david.sherwood@arm.com>
8440 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8441 Use plus_constant instead of gen_rtx_PLUS.
8443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8444 Alan Hayward <alan.hayward@arm.com>
8445 David Sherwood <david.sherwood@arm.com>
8447 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8448 rather than an int.
8450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8451 Alan Hayward <alan.hayward@arm.com>
8452 David Sherwood <david.sherwood@arm.com>
8454 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8455 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8456 via stack temporaries. Treat the mode size as polynomial too.
8458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8459 Alan Hayward <alan.hayward@arm.com>
8460 David Sherwood <david.sherwood@arm.com>
8462 * expr.c (expand_expr_real_2): When handling conversions involving
8463 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8464 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8465 as a poly_uint64 too.
8467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8468 Alan Hayward <alan.hayward@arm.com>
8469 David Sherwood <david.sherwood@arm.com>
8471 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8474 Alan Hayward <alan.hayward@arm.com>
8475 David Sherwood <david.sherwood@arm.com>
8477 * combine.c (can_change_dest_mode): Handle polynomial
8478 REGMODE_NATURAL_SIZE.
8479 * expmed.c (store_bit_field_1): Likewise.
8480 * expr.c (store_constructor): Likewise.
8481 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8482 and polynomial REGMODE_NATURAL_SIZE.
8483 (gen_lowpart_common): Likewise.
8484 * reginfo.c (record_subregs_of_mode): Likewise.
8485 * rtlanal.c (read_modify_subreg_p): Likewise.
8487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8488 Alan Hayward <alan.hayward@arm.com>
8489 David Sherwood <david.sherwood@arm.com>
8491 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8492 numbers of elements.
8494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8495 Alan Hayward <alan.hayward@arm.com>
8496 David Sherwood <david.sherwood@arm.com>
8498 * match.pd: Cope with polynomial numbers of vector elements.
8500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8501 Alan Hayward <alan.hayward@arm.com>
8502 David Sherwood <david.sherwood@arm.com>
8504 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8505 in a POINTER_PLUS_EXPR.
8507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8508 Alan Hayward <alan.hayward@arm.com>
8509 David Sherwood <david.sherwood@arm.com>
8511 * omp-simd-clone.c (simd_clone_subparts): New function.
8512 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8513 (ipa_simd_modify_function_body): Likewise.
8515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8516 Alan Hayward <alan.hayward@arm.com>
8517 David Sherwood <david.sherwood@arm.com>
8519 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8520 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8521 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8522 (expand_vector_condition, vector_element): Likewise.
8523 (subparts_gt): New function.
8524 (get_compute_type): Use subparts_gt.
8525 (count_type_subparts): Delete.
8526 (expand_vector_operations_1): Use subparts_gt instead of
8527 count_type_subparts.
8529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8530 Alan Hayward <alan.hayward@arm.com>
8531 David Sherwood <david.sherwood@arm.com>
8533 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8534 (vect_compile_time_alias): ...this new function. Do the calculation
8535 on poly_ints rather than trees.
8536 (vect_prune_runtime_alias_test_list): Update call accordingly.
8538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8539 Alan Hayward <alan.hayward@arm.com>
8540 David Sherwood <david.sherwood@arm.com>
8542 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8543 numbers of units.
8544 (vect_schedule_slp_instance): Likewise.
8546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8547 Alan Hayward <alan.hayward@arm.com>
8548 David Sherwood <david.sherwood@arm.com>
8550 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8551 constant and extern definitions for variable-length vectors.
8552 (vect_get_constant_vectors): Note that the number of units
8553 is known to be constant.
8555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8556 Alan Hayward <alan.hayward@arm.com>
8557 David Sherwood <david.sherwood@arm.com>
8559 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8560 of units as polynomial. Choose between WIDE and NARROW based
8561 on multiple_p.
8563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8564 Alan Hayward <alan.hayward@arm.com>
8565 David Sherwood <david.sherwood@arm.com>
8567 * tree-vect-stmts.c (simd_clone_subparts): New function.
8568 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8571 Alan Hayward <alan.hayward@arm.com>
8572 David Sherwood <david.sherwood@arm.com>
8574 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8575 vectors as polynomial. Use build_index_vector for
8576 IFN_GOMP_SIMD_LANE.
8578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8579 Alan Hayward <alan.hayward@arm.com>
8580 David Sherwood <david.sherwood@arm.com>
8582 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8583 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8584 for variable-length vectors.
8585 (vectorizable_mask_load_store): Treat the number of units as
8586 polynomial, asserting that it is constant if the condition has
8587 already been enforced.
8588 (vectorizable_store, vectorizable_load): Likewise.
8590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8591 Alan Hayward <alan.hayward@arm.com>
8592 David Sherwood <david.sherwood@arm.com>
8594 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8595 of units as polynomial. Punt if we can't tell at compile time
8596 which vector contains the final result.
8598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8599 Alan Hayward <alan.hayward@arm.com>
8600 David Sherwood <david.sherwood@arm.com>
8602 * tree-vect-loop.c (vectorizable_induction): Treat the number
8603 of units as polynomial. Punt on SLP inductions. Use an integer
8604 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8605 cast of such a series for variable-length floating-point
8606 reductions.
8608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8609 Alan Hayward <alan.hayward@arm.com>
8610 David Sherwood <david.sherwood@arm.com>
8612 * tree.h (build_index_vector): Declare.
8613 * tree.c (build_index_vector): New function.
8614 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8615 of units as polynomial, forcibly converting it to a constant if
8616 vectorizable_reduction has already enforced the condition.
8617 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8618 to create a {1,2,3,...} vector.
8619 (vectorizable_reduction): Treat the number of units as polynomial.
8620 Choose vectype_in based on the largest scalar element size rather
8621 than the smallest number of units. Enforce the restrictions
8622 relied on above.
8624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8625 Alan Hayward <alan.hayward@arm.com>
8626 David Sherwood <david.sherwood@arm.com>
8628 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8629 number of units as polynomial.
8631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8632 Alan Hayward <alan.hayward@arm.com>
8633 David Sherwood <david.sherwood@arm.com>
8635 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8636 * target.def (autovectorize_vector_sizes): Return the vector sizes
8637 by pointer, using vector_sizes rather than a bitmask.
8638 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8639 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8640 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8641 Likewise.
8642 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8643 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8644 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8645 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8646 * omp-general.c (omp_max_vf): Likewise.
8647 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8648 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8649 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8650 * tree-vect-slp.c (vect_slp_bb): Likewise.
8651 * doc/tm.texi: Regenerate.
8652 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8653 to a poly_uint64.
8654 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8655 the vector size as a poly_uint64 rather than an unsigned int.
8656 (current_vector_size): Change from an unsigned int to a poly_uint64.
8657 (get_vectype_for_scalar_type): Update accordingly.
8658 * tree.h (build_truth_vector_type): Take the size and number of
8659 units as a poly_uint64 rather than an unsigned int.
8660 (build_vector_type): Add a temporary overload that takes
8661 the number of units as a poly_uint64 rather than an unsigned int.
8662 * tree.c (make_vector_type): Likewise.
8663 (build_truth_vector_type): Take the number of units as a poly_uint64
8664 rather than an unsigned int.
8666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8667 Alan Hayward <alan.hayward@arm.com>
8668 David Sherwood <david.sherwood@arm.com>
8670 * target.def (get_mask_mode): Take the number of units and length
8671 as poly_uint64s rather than unsigned ints.
8672 * targhooks.h (default_get_mask_mode): Update accordingly.
8673 * targhooks.c (default_get_mask_mode): Likewise.
8674 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8675 * doc/tm.texi: Regenerate.
8677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8678 Alan Hayward <alan.hayward@arm.com>
8679 David Sherwood <david.sherwood@arm.com>
8681 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8682 * omp-general.c (omp_max_vf): Likewise.
8683 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8684 (expand_omp_simd): Handle polynomial safelen.
8685 * omp-low.c (omplow_simd_context): Add a default constructor.
8686 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8687 (lower_rec_simd_input_clauses): Update accordingly.
8688 (lower_rec_input_clauses): Likewise.
8690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8691 Alan Hayward <alan.hayward@arm.com>
8692 David Sherwood <david.sherwood@arm.com>
8694 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8695 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8696 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8697 (vect_analyze_slp_cost): Likewise.
8698 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8699 (vect_model_load_cost): Likewise.
8701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8702 Alan Hayward <alan.hayward@arm.com>
8703 David Sherwood <david.sherwood@arm.com>
8705 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8706 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8707 from an unsigned int * to a poly_uint64_pod *.
8708 (calculate_unrolling_factor): New function.
8709 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8712 Alan Hayward <alan.hayward@arm.com>
8713 David Sherwood <david.sherwood@arm.com>
8715 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8716 from an unsigned int to a poly_uint64.
8717 (_loop_vec_info::slp_unrolling_factor): Likewise.
8718 (_loop_vec_info::vectorization_factor): Change from an int
8719 to a poly_uint64.
8720 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8721 (vect_get_num_vectors): New function.
8722 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8723 (vect_get_num_copies): Use vect_get_num_vectors.
8724 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8725 to an unsigned int *.
8726 (vect_analyze_data_refs): Change min_vf from an int * to a
8727 poly_uint64 *.
8728 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8729 than an unsigned HOST_WIDE_INT.
8730 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8731 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8732 to an unsigned int *.
8733 (vect_analyze_data_ref_dependences): Likewise.
8734 (vect_compute_data_ref_alignment): Handle polynomial vf.
8735 (vect_enhance_data_refs_alignment): Likewise.
8736 (vect_prune_runtime_alias_test_list): Likewise.
8737 (vect_shift_permute_load_chain): Likewise.
8738 (vect_supportable_dr_alignment): Likewise.
8739 (dependence_distance_ge_vf): Take the vectorization factor as a
8740 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8741 (vect_analyze_data_refs): Change min_vf from an int * to a
8742 poly_uint64 *.
8743 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8744 vfm1 as a poly_uint64 rather than an int. Make the same change
8745 for the returned bound_scalar.
8746 (vect_gen_vector_loop_niters): Handle polynomial vf.
8747 (vect_do_peeling): Likewise. Update call to
8748 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8749 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8750 be constant.
8751 * tree-vect-loop.c (vect_determine_vectorization_factor)
8752 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8753 (vect_get_known_peeling_cost): Likewise.
8754 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8755 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8756 (vect_transform_loop): Likewise. Use the lowest possible VF when
8757 updating the upper bounds of the loop.
8758 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8759 rather than an int.
8760 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8761 polynomial unroll factors.
8762 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8763 (vect_make_slp_decision): Likewise.
8764 (vect_supported_load_permutation_p): Likewise, and polynomial
8765 vf too.
8766 (vect_analyze_slp_cost): Handle polynomial vf.
8767 (vect_slp_analyze_node_operations): Likewise.
8768 (vect_slp_analyze_bb_1): Likewise.
8769 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8770 than an unsigned HOST_WIDE_INT.
8771 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8772 (vectorizable_load): Handle polynomial vf.
8773 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8774 a poly_uint64.
8775 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8778 Alan Hayward <alan.hayward@arm.com>
8779 David Sherwood <david.sherwood@arm.com>
8781 * match.pd: Handle bit operations involving three constants
8782 and try to fold one pair.
8784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8786 * tree-vect-loop-manip.c: Include gimple-fold.h.
8787 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8788 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8789 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8790 Add a path that uses a step of VF instead of 1, but disable it
8791 for now.
8792 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8793 and niters_no_overflow parameters. Update calls to
8794 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8795 Create a new SSA name if the latter choses to use a ste other
8796 than zero, and return it via niters_vector_mult_vf_var.
8797 * tree-vect-loop.c (vect_transform_loop): Update calls to
8798 vect_do_peeling, vect_gen_vector_loop_niters and
8799 slpeel_make_loop_iterate_ntimes.
8800 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8801 (vect_gen_vector_loop_niters): Update declarations after above changes.
8803 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8805 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8806 128-bit round to integer instructions.
8807 (ceil<mode>2): Likewise.
8808 (btrunc<mode>2): Likewise.
8809 (round<mode>2): Likewise.
8811 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8813 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8814 unaligned VSX load/store on P8/P9.
8815 (expand_block_clear): Allow the use of unaligned VSX
8816 load/store on P8/P9.
8818 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8820 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8821 New function.
8822 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8823 swap associated with both a load and a store.
8825 2018-01-02 Andrew Waterman <andrew@sifive.com>
8827 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8828 * config/riscv/riscv.md (clear_cache): Use it.
8830 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8832 * web.c: Remove out-of-date comment.
8834 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8836 * expr.c (fixup_args_size_notes): Check that any existing
8837 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8838 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8839 (emit_single_push_insn): ...here.
8841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8843 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8844 (const_vector_encoded_nelts): New function.
8845 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8846 (const_vector_int_elt, const_vector_elt): Declare.
8847 * emit-rtl.c (const_vector_int_elt_1): New function.
8848 (const_vector_elt): Likewise.
8849 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8850 of CONST_VECTOR_ELT.
8852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8854 * expr.c: Include rtx-vector-builder.h.
8855 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8856 directly on the tree encoding.
8857 (const_vector_from_tree): Likewise.
8858 * optabs.c: Include rtx-vector-builder.h.
8859 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8860 sequence of "u" values.
8861 * vec-perm-indices.c: Include rtx-vector-builder.h.
8862 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8863 directly on the vec_perm_indices encoding.
8865 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8867 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8868 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8869 * rtx-vector-builder.h: New file.
8870 * rtx-vector-builder.c: Likewise.
8871 * rtl.h (rtx_def::u2): Add a const_vector field.
8872 (CONST_VECTOR_NPATTERNS): New macro.
8873 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8874 (CONST_VECTOR_DUPLICATE_P): Likewise.
8875 (CONST_VECTOR_STEPPED_P): Likewise.
8876 (CONST_VECTOR_ENCODED_ELT): Likewise.
8877 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8878 (unwrap_const_vec_duplicate): Likewise.
8879 (const_vec_series_p): Check for a non-duplicated vector encoding.
8880 Say that the function only returns true for integer vectors.
8881 * emit-rtl.c: Include rtx-vector-builder.h.
8882 (gen_const_vec_duplicate_1): Delete.
8883 (gen_const_vector): Call gen_const_vec_duplicate instead of
8884 gen_const_vec_duplicate_1.
8885 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8886 (gen_const_vec_duplicate): Use rtx_vector_builder.
8887 (gen_const_vec_series): Likewise.
8888 (gen_rtx_CONST_VECTOR): Likewise.
8889 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8890 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8891 Build a new vector rather than modifying a CONST_VECTOR in-place.
8892 (handle_special_swappables): Update call accordingly.
8893 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8894 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8895 Build a new vector rather than modifying a CONST_VECTOR in-place.
8896 (handle_special_swappables): Update call accordingly.
8898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8900 * simplify-rtx.c (simplify_const_binary_operation): Use
8901 CONST_VECTOR_ELT instead of XVECEXP.
8903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8905 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8906 the selector elements to be different from the data elements
8907 if the selector is a VECTOR_CST.
8908 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8909 ssizetype for the selector.
8911 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8913 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8914 before testing each element individually.
8915 * tree-vect-generic.c (lower_vec_perm): Likewise.
8917 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8919 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8920 * selftest-run-tests.c (selftest::run_tests): Call it.
8921 * vector-builder.h (vector_builder::operator ==): New function.
8922 (vector_builder::operator !=): Likewise.
8923 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8924 (vec_perm_indices::all_from_input_p): New function.
8925 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8926 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8927 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8928 instead of reading the VECTOR_CST directly. Detect whether both
8929 vector inputs are the same before constructing the vec_perm_indices,
8930 and update the number of inputs argument accordingly. Use the
8931 utility functions added above. Only construct sel2 if we need to.
8933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8935 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8936 the broadcast of the low byte.
8937 (expand_mult_highpart): Use an explicit encoding for the permutes.
8938 * optabs-query.c (can_mult_highpart_p): Likewise.
8939 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8940 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8941 (vectorizable_bswap): Likewise.
8942 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8943 explicit encoding for the power-of-2 permutes.
8944 (vect_permute_store_chain): Likewise.
8945 (vect_grouped_load_supported): Likewise.
8946 (vect_permute_load_chain): Likewise.
8948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8950 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8951 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8952 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8953 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8954 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8955 (vect_gen_perm_mask_any): Likewise.
8957 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8959 * int-vector-builder.h: New file.
8960 * vec-perm-indices.h: Include int-vector-builder.h.
8961 (vec_perm_indices): Redefine as an int_vector_builder.
8962 (auto_vec_perm_indices): Delete.
8963 (vec_perm_builder): Redefine as a stand-alone class.
8964 (vec_perm_indices::vec_perm_indices): New function.
8965 (vec_perm_indices::clamp): Likewise.
8966 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8967 (vec_perm_indices::new_vector): New function.
8968 (vec_perm_indices::new_expanded_vector): Update for new
8969 vec_perm_indices class.
8970 (vec_perm_indices::rotate_inputs): New function.
8971 (vec_perm_indices::all_in_range_p): Operate directly on the
8972 encoded form, without computing elided elements.
8973 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8974 encoding. Update for new vec_perm_indices class.
8975 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8976 the given vec_perm_builder.
8977 (expand_vec_perm_var): Update vec_perm_builder constructor.
8978 (expand_mult_highpart): Use vec_perm_builder instead of
8979 auto_vec_perm_indices.
8980 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8981 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8982 or double series encoding as appropriate.
8983 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8984 vec_perm_indices instead of auto_vec_perm_indices.
8985 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8986 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8987 (vect_permute_store_chain): Likewise.
8988 (vect_grouped_load_supported): Likewise.
8989 (vect_permute_load_chain): Likewise.
8990 (vect_shift_permute_load_chain): Likewise.
8991 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8992 (vect_transform_slp_perm_load): Likewise.
8993 (vect_schedule_slp_instance): Likewise.
8994 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8995 (vectorizable_mask_load_store): Likewise.
8996 (vectorizable_bswap): Likewise.
8997 (vectorizable_store): Likewise.
8998 (vectorizable_load): Likewise.
8999 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9000 vec_perm_indices instead of auto_vec_perm_indices. Use
9001 tree_to_vec_perm_builder to read the vector from a tree.
9002 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9003 vec_perm_builder instead of a vec_perm_indices.
9004 (have_whole_vector_shift): Use vec_perm_builder and
9005 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9006 truncation to calc_vec_perm_mask_for_shift.
9007 (vect_create_epilog_for_reduction): Likewise.
9008 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9009 from auto_vec_perm_indices to vec_perm_indices.
9010 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9011 instead of changing individual elements.
9012 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9013 the vector in d.perm.
9014 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9015 from auto_vec_perm_indices to vec_perm_indices.
9016 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9017 instead of changing individual elements.
9018 (arm_vectorize_vec_perm_const): Use new_vector to install
9019 the vector in d.perm.
9020 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9021 Update vec_perm_builder constructor.
9022 (rs6000_expand_interleave): Likewise.
9023 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9024 (rs6000_expand_interleave): Likewise.
9026 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9028 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9029 to qimode could truncate the indices.
9030 * optabs.c (expand_vec_perm_var): Likewise.
9032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9034 * Makefile.in (OBJS): Add vec-perm-indices.o.
9035 * vec-perm-indices.h: New file.
9036 * vec-perm-indices.c: Likewise.
9037 * target.h (vec_perm_indices): Replace with a forward class
9038 declaration.
9039 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9040 * optabs.h: Include vec-perm-indices.h.
9041 (expand_vec_perm): Delete.
9042 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9043 (expand_vec_perm_const): Declare.
9044 * target.def (vec_perm_const_ok): Replace with...
9045 (vec_perm_const): ...this new hook.
9046 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9047 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9048 * doc/tm.texi: Regenerate.
9049 * optabs.def (vec_perm_const): Delete.
9050 * doc/md.texi (vec_perm_const): Likewise.
9051 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9052 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9053 expand_vec_perm for constant permutation vectors. Assert that
9054 the mode of variable permutation vectors is the integer equivalent
9055 of the mode that is being permuted.
9056 * optabs-query.h (selector_fits_mode_p): Declare.
9057 * optabs-query.c: Include vec-perm-indices.h.
9058 (selector_fits_mode_p): New function.
9059 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9060 is defined, instead of checking whether the vec_perm_const_optab
9061 exists. Use targetm.vectorize.vec_perm_const instead of
9062 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9063 fit in the vector mode before using a variable permute.
9064 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9065 vec_perm_indices instead of an rtx.
9066 (expand_vec_perm): Replace with...
9067 (expand_vec_perm_const): ...this new function. Take the selector
9068 as a vec_perm_indices rather than an rtx. Also take the mode of
9069 the selector. Update call to shift_amt_for_vec_perm_mask.
9070 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9071 Use vec_perm_indices::new_expanded_vector to expand the original
9072 selector into bytes. Check whether the indices fit in the vector
9073 mode before using a variable permute.
9074 (expand_vec_perm_var): Make global.
9075 (expand_mult_highpart): Use expand_vec_perm_const.
9076 * fold-const.c: Includes vec-perm-indices.h.
9077 * tree-ssa-forwprop.c: Likewise.
9078 * tree-vect-data-refs.c: Likewise.
9079 * tree-vect-generic.c: Likewise.
9080 * tree-vect-loop.c: Likewise.
9081 * tree-vect-slp.c: Likewise.
9082 * tree-vect-stmts.c: Likewise.
9083 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9084 Delete.
9085 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9086 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9087 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9088 (aarch64_vectorize_vec_perm_const): ...this new function.
9089 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9090 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9091 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9092 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9093 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9094 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9095 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9096 into...
9097 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9098 check for NEON modes.
9099 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9100 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9101 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9102 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9103 into...
9104 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9105 the old VEC_PERM_CONST conditions.
9106 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9107 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9108 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9109 (ia64_vectorize_vec_perm_const_ok): Merge into...
9110 (ia64_vectorize_vec_perm_const): ...this new function.
9111 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9112 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9113 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9114 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9115 * config/mips/mips.c (mips_expand_vec_perm_const)
9116 (mips_vectorize_vec_perm_const_ok): Merge into...
9117 (mips_vectorize_vec_perm_const): ...this new function.
9118 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9119 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9120 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9121 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9122 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9123 (rs6000_expand_vec_perm_const): Delete.
9124 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9125 Delete.
9126 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9127 (altivec_expand_vec_perm_const_le): Take each operand individually.
9128 Operate on constant selectors rather than rtxes.
9129 (altivec_expand_vec_perm_const): Likewise. Update call to
9130 altivec_expand_vec_perm_const_le.
9131 (rs6000_expand_vec_perm_const): Delete.
9132 (rs6000_vectorize_vec_perm_const_ok): Delete.
9133 (rs6000_vectorize_vec_perm_const): New function.
9134 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9135 an element count and rtx array.
9136 (rs6000_expand_extract_even): Update call accordingly.
9137 (rs6000_expand_interleave): Likewise.
9138 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9139 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9140 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9141 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9142 (rs6000_expand_vec_perm_const): Delete.
9143 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9144 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9145 (altivec_expand_vec_perm_const_le): Take each operand individually.
9146 Operate on constant selectors rather than rtxes.
9147 (altivec_expand_vec_perm_const): Likewise. Update call to
9148 altivec_expand_vec_perm_const_le.
9149 (rs6000_expand_vec_perm_const): Delete.
9150 (rs6000_vectorize_vec_perm_const_ok): Delete.
9151 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9152 reference to the SPE evmerge intructions.
9153 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9154 an element count and rtx array.
9155 (rs6000_expand_extract_even): Update call accordingly.
9156 (rs6000_expand_interleave): Likewise.
9157 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9158 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9159 new function.
9160 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9164 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9165 vector mode and that that mode matches the mode of the data
9166 being permuted.
9167 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9168 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9169 directly using expand_vec_perm_1 when forcing selectors into
9170 registers.
9171 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9175 * optabs-query.h (can_vec_perm_p): Delete.
9176 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9177 * optabs-query.c (can_vec_perm_p): Split into...
9178 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9179 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9180 particular selector is valid.
9181 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9182 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9183 (vect_grouped_load_supported): Likewise.
9184 (vect_shift_permute_load_chain): Likewise.
9185 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9186 (vect_transform_slp_perm_load): Likewise.
9187 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9188 (vectorizable_bswap): Likewise.
9189 (vect_gen_perm_mask_checked): Likewise.
9190 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9191 implementations of variable permutation vectors into account
9192 when deciding which selector to use.
9193 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9194 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9195 with a false third argument.
9196 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9197 to test whether the constant selector is valid and can_vec_perm_var_p
9198 to test whether a variable selector is valid.
9200 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9202 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9203 * optabs-query.c (can_vec_perm_p): Likewise.
9204 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9205 instead of vec_perm_indices.
9206 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9207 (vect_gen_perm_mask_checked): Likewise,
9208 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9209 (vect_gen_perm_mask_checked): Likewise,
9211 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9213 * optabs-query.h (qimode_for_vec_perm): Declare.
9214 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9215 (qimode_for_vec_perm): ...this new function.
9216 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9218 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9220 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9221 does not have a conditional at the top.
9223 2018-01-02 Richard Biener <rguenther@suse.de>
9225 * ipa-inline.c (big_speedup_p): Fix expression.
9227 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9229 PR target/81616
9230 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9231 for generic 4->6.
9233 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9235 PR target/81616
9236 Generic tuning.
9237 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9238 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9239 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9240 cond_taken_branch_cost 3->4.
9242 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9244 PR tree-optimization/83581
9245 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9246 TODO_cleanup_cfg if any changes have been made.
9248 PR middle-end/83608
9249 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9250 convert_modes if target mode has the right side, but different mode
9251 class.
9253 PR middle-end/83609
9254 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9255 last argument when extracting from CONCAT. If either from_real or
9256 from_imag is NULL, use expansion through memory. If result is not
9257 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9258 the parts directly to inner mode, if even that fails, use expansion
9259 through memory.
9261 PR middle-end/83623
9262 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9263 check for bswap in mode rather than HImode and use that in expand_unop
9264 too.
9265 \f
9266 Copyright (C) 2018 Free Software Foundation, Inc.
9268 Copying and distribution of this file, with or without modification,
9269 are permitted in any medium without royalty provided the copyright
9270 notice and this notice are preserved.