| blob | 18d727e8ee3503bc31828f61aad0dc018399e117 |
1 2018-03-20 Richard Biener <rguenther@suse.de>
3 PR target/84986
4 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5 sign-conversions as zero, fall back to standard scalar_stmt
6 cost for the rest.
8 2018-03-20 Martin Liska <mliska@suse.cz>
10 PR ipa/84825
11 * predict.c (rebuild_frequencies): Handle case when we have
12 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16 PR target/84990
17 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18 flag_section_anchors.
19 * varasm.c (use_blocks_for_decl_p): Remove hack for
20 dw2_force_const_mem.
22 PR target/84845
23 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
24 to ...
25 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
26 be created, use lowpart_subreg of operands[0] rather than operands[0]
27 itself.
28 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
29 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
30 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
31 and n constraint instead of aarch64_shift_imm_di and Usd.
32 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
33 (*aarch64_<optab>_reg_minus<mode>3): ... this.
35 2018-03-20 Sudakshina Das <sudi.das@arm.com>
37 PR target/82989
38 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
39 to favor GPR over NEON registers.
40 (<shift>di3_neon): Likewise.
42 2018-03-20 Tom de Vries <tom@codesourcery.com>
44 PR target/84952
45 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
46 (nvptx_process_pars): Emit bar.sync asap and alap.
48 2018-03-20 Tom de Vries <tom@codesourcery.com>
50 PR target/84954
51 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
52 seen_label if seen_label is already set.
54 2018-03-20 Jakub Jelinek <jakub@redhat.com>
56 PR target/84945
57 * config/i386/i386.c (fold_builtin_cpu): For features above 31
58 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
59 Use 1U instead of 1. Formatting fixes.
61 PR c/84953
62 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
63 instead of TREE_TYPE (s1) for the return value.
65 2018-03-19 Jakub Jelinek <jakub@redhat.com>
67 PR tree-optimization/84946
68 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
69 bitsize + bitsize in poly_uint64 rather than poly_int64.
71 PR sanitizer/78651
72 * dwarf2asm.c: Include fold-const.c.
73 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
74 of decl rather than decl itself.
76 PR rtl-optimization/84643
77 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
79 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
81 PR sanitizer/78651
82 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
83 calling assemble_variable.
85 2018-03-19 Sudakshina Das <sudi.das@arm.com>
87 PR target/81647
88 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
89 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
91 2018-03-19 Jim Wilson <jimw@sifive.com>
93 PR bootstrap/84856
94 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
95 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
96 (riscv_first_stack_step): Likewise.
97 (riscv_option_override): Use STACK_BOUNDARY instead of
98 MIN_STACK_BOUNDARY.
99 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
100 MIN_STACK_BOUNDARY.
101 (BIGGEST_ALIGNMENT): Set to 128.
102 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
103 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
104 STACK_BOUNDARY.
106 2018-03-19 Richard Biener <rguenther@suse.de>
108 PR tree-optimization/84933
109 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
110 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
112 2018-03-19 Richard Biener <rguenther@suse.de>
114 PR tree-optimization/84859
115 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
116 (cond_if_else_store_replacement): Perform sinking operation on
117 single-store BBs regardless of MAX_STORES_TO_SINK setting.
118 Generalize what a BB with a single eligible store is.
120 2018-03-19 Richard Biener <rguenther@suse.de>
122 PR tree-optimization/84929
123 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
124 chrec_is_positive against non-chrec arg.
126 2018-03-19 Tamar Christina <tamar.christina@arm.com>
128 PR target/84711
129 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
131 2018-03-18 Martin Liska <mliska@suse.cz>
133 PR rtl-optimization/84635
134 * regrename.c (build_def_use): Use matches_mode only when
135 matches >= 0.
137 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
139 PR tree-optimization/84913
140 * tree-vect-loop.c (vectorizable_reduction): Don't try to
141 vectorize chains of COND_EXPRs.
143 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
145 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
147 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
149 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
151 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
153 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
155 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
156 Kito Cheng <kito.cheng@gmail.com>
158 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
159 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
160 (nds32_adjust_reg_alloc_order): New function.
161 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
163 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
165 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
166 nds32_print_operand, nds32_print_operand_address): Use
167 HOST_WIDE_INT_PRINT_DEC instead.
169 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
171 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
173 2018-03-17 Jakub Jelinek <jakub@redhat.com>
175 PR target/84902
176 * config/i386/i386.c (initial_ix86_tune_features,
177 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
178 unsigned long long.
179 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
180 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
181 rather than 1u << ix86_tune. Formatting fix.
182 (ix86_option_override_internal): Change ix86_arch_mask from
183 unsigned int to unsigned HOST_WIDE_INT, initialize to
184 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
185 (ix86_function_specific_restore): Likewise.
187 2018-03-16 Jakub Jelinek <jakub@redhat.com>
189 PR target/84899
190 * postreload.c (reload_combine_recognize_pattern): Perform
191 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
192 truncate_int_for_mode the result for the destination's mode.
194 PR c/84909
195 * hsa-gen.c (mem_type_for_type): Fix comment typo.
196 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
197 Likewise.
198 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
199 Likewise.
201 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
203 PR target/84876
204 * lra-assigns.c (lra_split_hard_reg_for): Don't use
205 regno_allocno_class_array and sorted_pseudos.
206 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
207 insns where regno is used.
209 2018-03-16 Martin Liska <mliska@suse.cz>
211 PR ipa/84833
212 * multiple_target.c (create_dispatcher_calls): Redirect
213 reference in the symbol table.
215 2018-03-16 Martin Liska <mliska@suse.cz>
217 PR ipa/84722
218 * multiple_target.c (create_dispatcher_calls): Redirect also
219 an alias.
221 2018-03-16 Jakub Jelinek <jakub@redhat.com>
223 PR c++/79937
224 PR c++/82410
225 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
226 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
227 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
229 2018-03-16 Julia Koval <julia.koval@intel.com>
231 * doc/invoke.texi (Skylake Server): Add CLWB.
232 Cannonlake): Remove CLWB.
234 2018-03-16 Jakub Jelinek <jakub@redhat.com>
236 PR tree-optimization/84841
237 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
238 1 << 3.
239 (FLOAT_ONE_CONST_TYPE): Define.
240 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
241 (sort_by_operand_rank): Put entries with higher constant_type last
242 rather than first to match comments.
244 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
246 * config/nios2/nios2.md (movsi_internal): Fix thinko in
247 split predicate.
249 2018-03-15 Jakub Jelinek <jakub@redhat.com>
251 PR c++/79085
252 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
253 check and use address of target always.
255 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
257 PR target/84574
258 * config/i386/i386.c (indirect_thunk_needed): Update comments.
259 (indirect_thunk_bnd_needed): Likewise.
260 (indirect_thunks_used): Likewise.
261 (indirect_thunks_bnd_used): Likewise.
262 (indirect_return_needed): New.
263 (indirect_return_bnd_needed): Likewise.
264 (output_indirect_thunk_function): Add a bool argument for
265 function return.
266 (output_indirect_thunk_function): Don't generate alias for
267 function return thunk.
268 (ix86_code_end): Call output_indirect_thunk_function to generate
269 function return thunks.
270 (ix86_output_function_return): Set indirect_return_bnd_needed
271 and indirect_return_needed instead of indirect_thunk_bnd_needed
272 and indirect_thunk_needed.
274 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
276 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
277 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
278 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
280 2018-03-15 David Malcolm <dmalcolm@redhat.com>
281 Paul Hua <paul.hua.gm@gmail.com>
283 PR c/84852
284 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
286 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
288 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
289 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
290 resp. SFmode cases.
292 2018-03-15 Tamar Christina <tamar.christina@arm.com>
294 PR target/84711
295 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
296 instead of GET_MODE_SIZE when comparing Units.
298 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
300 PR target/68256
301 * varasm.c (hash_section): Return an unchangeble hash value
302 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
303 Return !aarch64_can_use_per_function_literal_pools_p ().
305 2018-03-15 Jakub Jelinek <jakub@redhat.com>
307 PR target/84860
308 * optabs.c (emit_conditional_move): Pass address of cmode's copy
309 rather than address of cmode as last argument to prepare_cmp_insn.
311 2018-03-15 Julia Koval <julia.koval@intel.com>
313 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
314 F_AVX512VNNI, F_AVX512BITALG): New.
316 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
318 PR target/83451
319 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
320 insn for floating-point loads and stores.
322 2018-03-14 Carl Love <cel@us.ibm.com>
324 * config/rs6000/rs6000-c.c: Add macro definitions for
325 ALTIVEC_BUILTIN_VEC_PERMXOR.
326 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
327 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
328 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
329 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
330 UNSPEC_VPERMXOR.
331 * config/doc/extend.texi: Add prototypes for vec_permxor.
333 2018-03-14 David Malcolm <dmalcolm@redhat.com>
335 PR c/84852
336 * diagnostic-show-locus.c (class layout_point): Convert m_line
337 from int to linenum_type.
338 (line_span::comparator): Use linenum "compare" function when
339 comparing line numbers.
340 (test_line_span): New function.
341 (layout_range::contains_point): Convert param "row" from int to
342 linenum_type.
343 (layout_range::intersects_line_p): Likewise.
344 (layout::will_show_line_p): Likewise.
345 (layout::print_source_line): Likewise.
346 (layout::should_print_annotation_line_p): Likewise.
347 (layout::print_annotation_line): Likewise.
348 (layout::print_leading_fixits): Likewise.
349 (layout::annotation_line_showed_range_p): Likewise.
350 (struct line_corrections): Likewise for field m_row.
351 (line_corrections::line_corrections): Likewise for param "row".
352 (layout::print_trailing_fixits): Likewise.
353 (layout::get_state_at_point): Likewise.
354 (layout::get_x_bound_for_row): Likewise.
355 (layout::print_line): Likewise.
356 (diagnostic_show_locus): Likewise for locals "last_line" and
357 "row".
358 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
359 * input.c (selftest::test_linenum_comparisons): New function.
360 (selftest::input_c_tests): Call it.
361 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
362 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
363 * selftest.h (ASSERT_GT): New macro.
364 (ASSERT_GT_AT): New macro.
365 (ASSERT_LT): New macro.
366 (ASSERT_LT_AT): New macro.
368 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
370 PR rtl-optimization/84780
371 * combine.c (distribute_links): Don't make a link based on pc_rtx.
373 2018-03-14 Martin Liska <mliska@suse.cz>
375 * tree.c (record_node_allocation_statistics): Use
376 get_stats_node_kind.
377 (get_stats_node_kind): New function extracted from
378 record_node_allocation_statistics.
379 (free_node): Use get_stats_node_kind.
381 2018-03-14 Richard Biener <rguenther@suse.de>
383 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
384 that the value-set of ANTIC_IN doesn't grow.
386 Revert
387 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
388 member.
389 (BB_VISITED_WITH_VISITED_SUCCS): New define.
390 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
392 2018-03-14 Julia Koval <julia.koval@intel.com>
394 * config.gcc (icelake-client, icelake-server): New.
395 (icelake): Remove.
396 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
397 (initial_ix86_arch_features): Ditto.
398 (PTA_SKYLAKE): Add SGX.
399 (PTA_ICELAKE): Remove.
400 (PTA_ICELAKE_CLIENT): New.
401 (PTA_ICELAKE_SERVER): New.
402 (ix86_option_override_internal): Split up icelake on icelake client and
403 icelake server.
404 (get_builtin_code_for_version): Ditto.
405 (fold_builtin_cpu): Ditto.
406 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
407 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
408 * config/i386/i386.h (processor_type): Ditto.
409 * doc/invoke.texi: Ditto.
411 2018-03-14 Jakub Jelinek <jakub@redhat.com>
413 PR sanitizer/83392
414 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
415 INTEGER_CST offset, add it together with bitpos / 8 and
416 sign extend based on POINTER_SIZE.
418 PR target/84844
419 Revert
420 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
422 PR target/78090
423 * config/i386/constraints.md (Yc): New register constraint.
424 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
425 Use Yc constraint for alternative 2 of operand 0. Remove
426 preferred_for_speed attribute.
428 2018-03-14 Richard Biener <rguenther@suse.de>
430 PR tree-optimization/84830
431 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
432 with the old one to avoid oscillations.
434 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
436 PR target/83712
437 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
438 pseudos.
439 (assign_by_spills): Return a flag of reload assignment failure.
440 Do not process the reload assignment failures. Do not spill other
441 reload pseudos if they has the same reg class. Update n if
442 necessary.
443 (lra_assign): Add a return arg. Set up from the result of
444 assign_by_spills call.
445 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
446 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
447 usage_insns if it is not NULL.
448 (spill_hard_reg_in_range): New function.
449 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
450 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
451 function prototypes.
452 (lra_assign): Change prototype.
453 * lra.c (lra): Add code to deal with fails by splitting hard reg
454 live ranges.
456 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
458 * config/riscv/riscv.opt (mrelax): New option.
459 * config/riscv/riscv.c (riscv_file_start): Emit ".option
460 "norelax" when riscv_mrelax is disabled.
461 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
463 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
465 PR target/84743
466 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
467 reassociation for int modes.
469 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
471 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
472 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
473 for big-endian.
474 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
475 * config/aarch64/aarch64-sve.md
476 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
477 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
478 (*extend<mode><Vwide>2): Rename to...
479 (aarch64_sve_extend<mode><Vwide>2): ...this.
480 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
481 renaming the old pattern to...
482 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
483 unsigned packs.
484 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
485 define_expand, renaming the old pattern to...
486 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
487 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
488 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
489 account when deciding which SVE instruction the optab should use.
490 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
492 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
494 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
495 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
496 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
497 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
498 (tlsdesc_small_<mode>): Turn a define_expand and use
499 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
500 (tlsdesc_small_advsimd_<mode>): ...this.
501 (tlsdesc_small_sve_<mode>): New pattern.
503 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
505 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
506 (UNSPEC_UMUL_HIGHPART): New constants.
507 (MUL_HIGHPART): New int iteraor.
508 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
509 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
510 define_expand.
511 (*<su>mul<mode>3_highpart): New define_insn.
513 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
515 PR lto/84805
516 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
517 incomplete types.
519 2018-03-13 Martin Liska <mliska@suse.cz>
521 PR ipa/84658.
522 * (sem_item_optimizer::sem_item_optimizer): Initialize new
523 vector.
524 (sem_item_optimizer::~sem_item_optimizer): Release it.
525 (sem_item_optimizer::merge_classes): Register variable aliases.
526 (sem_item_optimizer::fixup_pt_set): New function.
527 (sem_item_optimizer::fixup_points_to_sets): Likewise.
528 * ipa-icf.h: Declare new variables and functions.
530 2018-03-13 Jakub Jelinek <jakub@redhat.com>
532 PR middle-end/84834
533 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
534 integer_pow2p@2 and test integer_pow2p in condition.
535 (A < 0 ? C : 0): Similarly for @1.
537 PR middle-end/84831
538 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
539 characters starting at p contain '\0' character, don't look beyond
540 that.
542 PR target/84827
543 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
544 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
546 PR target/84828
547 * reg-stack.c (change_stack): Change update_end var from int to
548 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
549 also call set_block_for_insn on the newly added insns and rescan.
551 PR target/84786
552 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
553 on the last operand.
555 PR c++/84704
556 * tree.c (stabilize_reference_1): Return save_expr (e) for
557 STATEMENT_LIST even if it doesn't have side-effects.
559 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
561 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
563 2018-03-12 Renlin Li <renlin.li@arm.com>
565 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
566 aarch64_output_scalar_simd_mov_immediate.
568 2018-03-12 Martin Sebor <msebor@redhat.com>
570 PR tree-optimization/83456
571 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
572 for perfectly overlapping calls to memcpy.
573 (gimple_fold_builtin_memory_chk): Same.
574 (gimple_fold_builtin_strcpy): Handle no-warning.
575 (gimple_fold_builtin_stxcpy_chk): Same.
576 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
578 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
580 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
581 parameter. Use it for SFmode.
582 (rs6000_function_arg_advance_1): Adjust.
583 (rs6000_function_arg): Adjust.
584 (rs6000_gimplify_va_arg): Pass false for that new parameter.
586 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
588 PR rtl-optimization/84169
589 PR rtl-optimization/84780
590 * combine.c (can_combine_p): Check for a 2-insn combination whether
591 the destination register is used between the two insns, too.
593 2018-03-12 Richard Biener <rguenther@suse.de>
595 PR tree-optimization/84803
596 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
597 for refs DR analysis didn't process.
599 2018-03-12 Richard Biener <rguenther@suse.de>
601 PR tree-optimization/84777
602 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
603 force-vectorize loops ignore whether we are optimizing for size.
605 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
607 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
608 (TARGET_MD_ASM_ADJUST): Define.
610 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
611 Kito Cheng <kito.cheng@gmail.com>
612 Chung-Ju Wu <jasonwucj@gmail.com>
614 * config/nds32/nds32.c (nds32_compute_stack_frame,
615 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
616 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
617 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
618 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
619 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
620 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
621 * config/nds32/nds32.md (prologue, epilogue): Use macro
622 NDS32_V3PUSH_AVAILABLE_P to do checking.
624 2018-03-11 Jakub Jelinek <jakub@redhat.com>
626 PR debug/58150
627 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
628 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
629 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
630 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
631 addition of most attributes on !orig_type_die or the attribute not
632 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
634 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
635 Chung-Ju Wu <jasonwucj@gmail.com>
637 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
638 __NDS32_VH__ macro.
639 * config/nds32/nds32.opt (mvh): New option.
641 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
642 Chung-Ju Wu <jasonwucj@gmail.com>
644 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
645 function.
646 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
647 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
648 definition.
650 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
651 Chung-Ju Wu <jasonwucj@gmail.com>
653 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
654 function.
655 * config/nds32/nds32-multiple.md (strlensi): New pattern.
656 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
658 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
659 Kito Cheng <kito.cheng@gmail.com>
660 Chung-Ju Wu <jasonwucj@gmail.com>
662 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
663 UNSPEC_FFMISM and UNSPEC_FLMISM.
664 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
665 for ffb, ffmism and flmism.
666 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
667 (unspec_ffmism): Ditto.
668 (unspec_flmism): Ditto.
669 (nds32_expand_builtin_impl): Check if string extension is available.
670 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
671 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
673 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
675 Reverting patch:
676 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
678 PR target/83712
679 * lra-assigns.c (assign_by_spills): Return a flag of reload
680 assignment failure. Do not process the reload assignment
681 failures. Do not spill other reload pseudos if they has the same
682 reg class.
683 (lra_assign): Add a return arg. Set up from the result of
684 assign_by_spills call.
685 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
686 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
687 usage_insns if it is not NULL.
688 (spill_hard_reg_in_range): New function.
689 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
690 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
691 function prototypes.
692 (lra_assign): Change prototype.
693 * lra.c (lra): Add code to deal with fails by splitting hard reg
694 live ranges.
696 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
698 PR target/84807
699 * config/i386/i386.opt: Replace Enforcment with Enforcement.
701 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
703 PR debug/84620
704 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
705 (dw_val_node): Add val_symbolic_view.
706 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
707 (symview_upper_bound): New.
708 (new_line_info_table): Initialize symviews_since_reset.
709 (dwarf2out_source_line): Count symviews_since_reset and set
710 symview_upper_bound.
711 (dw_val_equal_p): Handle symview.
712 (add_AT_symview): New.
713 (print_dw_val): Handle symview.
714 (attr_checksum, attr_checksum_ordered): Likewise.
715 (same_dw_val_p, size_of_die): Likewise.
716 (value_format, output_die): Likewise.
717 (add_high_low_attributes): Use add_AT_symview for entry_view.
718 (dwarf2out_finish): Reset symview_upper_bound, clear
719 zero_view_p.
721 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
723 PR target/83969
724 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
725 Add strict argument and use it.
726 (rs6000_split_multireg_move): Update for new strict argument.
727 (mem_operand_gpr): Disallow all non-offsettable addresses.
728 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
730 2018-03-09 Jakub Jelinek <jakub@redhat.com>
732 PR target/84772
733 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
734 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
735 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
737 PR c++/84767
738 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
739 decl, use remap_type if we want to use the type.
741 2018-03-09 Martin Sebor <msebor@redhat.com>
743 PR tree-optimization/84526
744 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
745 Remove dead code.
746 (builtin_access::generic_overlap): Be prepared to handle non-array
747 base objects.
749 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
751 PR rtl-optimization/84682
752 * lra-constraints.c (process_address_1): Check is_address flag
753 for address constraints.
754 (process_alt_operands): Likewise.
755 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
756 preprocess_constraints.
757 * recog.h (preprocess_constraints): Add oploc parameter.
758 Adjust callers.
759 * recog.c (preprocess_constraints): Test address_operand for
760 CT_ADDRESS constraints.
762 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
764 PR target/83712
765 * lra-assigns.c (assign_by_spills): Return a flag of reload
766 assignment failure. Do not process the reload assignment
767 failures. Do not spill other reload pseudos if they has the same
768 reg class.
769 (lra_assign): Add a return arg. Set up from the result of
770 assign_by_spills call.
771 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
772 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
773 usage_insns if it is not NULL.
774 (spill_hard_reg_in_range): New function.
775 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
776 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
777 function prototypes.
778 (lra_assign): Change prototype.
779 * lra.c (lra): Add code to deal with fails by splitting hard reg
780 live ranges.
782 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
784 PR target/83193
785 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
786 Accept complain bool parameter. Only emit errors if it is true.
787 (arm_parse_cpu_option_name): Likewise.
788 (arm_target_thumb_only): Adjust callers of the above.
789 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
790 prototype to take a default true bool parameter.
791 (arm_parse_arch_option_name): Likewise.
793 2018-03-09 David Malcolm <dmalcolm@redhat.com>
794 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
796 PR jit/64089
797 PR jit/84288
798 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
799 * configure: Regenerate.
800 * configure.ac ("linker --version-script option"): New.
801 ("linker soname option"): New.
803 2018-03-09 Richard Biener <rguenther@suse.de>
805 PR tree-optimization/84775
806 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
807 immediate uses of predicate stmts and mark them modified.
809 Revert
810 PR tree-optimization/84178
811 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
812 to caller.
813 (version_loop_for_if_conversion): Delay update_ssa call.
814 (tree_if_conversion): Delay update_ssa until after predicate
815 insertion.
817 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
819 PR target/84763
820 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
821 when the function accesses prior frames.
823 2018-03-08 Jakub Jelinek <jakub@redhat.com>
825 PR debug/84456
826 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
827 gen_llsym, otherwise call maybe_gen_llsym.
829 PR inline-asm/84742
830 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
831 has ',' character inside of it.
833 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
835 PR target/84748
836 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
837 as clobbering CC_REGNUM.
839 2018-03-08 Richard Biener <rguenther@suse.de>
841 PR middle-end/84552
842 * tree-scalar-evolution.c: Include tree-into-ssa.h.
843 (follow_copies_to_constant): Do not follow SSA names registered
844 for update.
846 2018-03-08 Richard Biener <rguenther@suse.de>
848 PR tree-optimization/84178
849 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
850 to caller.
851 (version_loop_for_if_conversion): Delay update_ssa call.
852 (tree_if_conversion): Delay update_ssa until after predicate
853 insertion.
855 2018-03-08 David Malcolm <dmalcolm@redhat.com>
857 PR tree-optimization/84178
858 * tree-if-conv.c (release_bb_predicate): Remove the
859 the assertion that the stmts have NULL use_ops.
860 Discard the statements, asserting that they haven't
861 yet been added to a BB.
863 2018-03-08 Richard Biener <rguenther@suse.de>
865 PR tree-optimization/84746
866 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
867 (phi_translate): Pass in destination ANTIC_OUT set.
868 (phi_translate_1): Likewise. For a simplified result lookup
869 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
870 (phi_translate_set): Adjust.
871 (do_pre_regular_insertion): Likewise.
872 (do_pre_partial_partial_insertion): Likewise.
874 2018-03-08 Martin Liska <mliska@suse.cz>
876 PR gcov-profile/84735
877 * doc/gcov.texi: Document usage of profile files.
878 * gcov-io.h: Document changes in the format.
880 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
882 PR debug/84404
883 PR debug/84408
884 * dwarf2out.c (struct dw_line_info_table): Update comments for
885 view == -1.
886 (FORCE_RESET_NEXT_VIEW): New.
887 (FORCE_RESETTING_VIEW_P): New.
888 (RESETTING_VIEW_P): Check for -1 too.
889 (ZERO_VIEW_P): Likewise.
890 (new_line_info_table): Force-reset next view.
891 (dwarf2out_begin_function): Likewise.
892 (dwarf2out_source_line): Simplify zero_view_p initialization.
893 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
894 view directly. Omit view when omitting .loc at line 0.
896 2018-03-08 Jakub Jelinek <jakub@redhat.com>
898 PR tree-optimization/84740
899 * tree-switch-conversion.c (process_switch): Call build_constructors
900 only if info.phi_count is non-zero.
902 PR tree-optimization/84739
903 * tree-tailcall.c (find_tail_calls): Check call arguments against
904 DECL_ARGUMENTS (current_function_decl) rather than
905 DECL_ARGUMENTS (func) when checking for tail recursion.
907 2018-03-07 Jakub Jelinek <jakub@redhat.com>
909 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
910 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
911 Volker Reichelt's entry and add entries for people that perform
912 GCC fuzzy testing and report numerous bugs.
914 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
916 PR target/82411
917 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
918 readonly data in sdata, if that is disabled.
919 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
920 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
921 -mreadonly-in-sdata option.
923 2018-03-07 Martin Sebor <msebor@redhat.com>
925 PR tree-optimization/84468
926 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
927 basic block when looking for nul assignment.
929 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
931 PR target/84277
932 * except.h (output_function_exception_table): Adjust prototype.
933 * except.c (output_function_exception_table): Remove FNNAME parameter
934 and add SECTION parameter. Ouput one part of the table at a time.
935 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
936 the first part of the exception table and emit unwind directives.
937 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
938 (i386_pe_seh_cold_init): Likewise.
939 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
940 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
941 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
942 (ix86_output_call_insn): Emit a nop in one more case for SEH.
943 * config/i386/winnt.c: Include except.h.
944 (struct seh_frame_state): Add reg_offset, after_prologue and
945 in_cold_section fields.
946 (i386_pe_seh_end_prologue): Set seh->after_prologue.
947 (i386_pe_seh_cold_init): New function.
948 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
949 to seh->in_cold_section.
950 (seh_emit_push): Record the offset of the push.
951 (seh_emit_save): Record the offet of the save.
952 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
953 Test seh->after_prologue to disregard the epilogue.
954 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
955 (i386_pe_end_cold_function): New function.
957 2018-03-07 Jakub Jelinek <jakub@redhat.com>
959 PR fortran/84565
960 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
961 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
963 PR c++/84704
964 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
965 on tmp_var.
966 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
967 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
969 PR middle-end/84723
970 * multiple_target.c: Include tree-inline.h and intl.h.
971 (expand_target_clones): Diagnose and fail if node->definition and
972 !tree_versionable_function_p (node->decl).
974 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
976 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
977 sprint_ul.
978 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
979 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
980 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
982 2018-03-06 Jakub Jelinek <jakub@redhat.com>
984 PR target/84710
985 * combine.c (try_combine): Use reg_or_subregno instead of handling
986 just paradoxical SUBREGs and REGs.
988 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
990 * config/arc/arc.c (arc_finalize_pic): Remove function.
991 (arc_must_save_register): We use single base PIC register, remove
992 checks to save/restore the PIC register.
993 (arc_expand_prologue): Likewise.
994 * config/arc/arc-protos.h (arc_set_default_type_attributes):
995 Remove.
996 (arc_verify_short): Likewise.
997 (arc_attr_type): Likewise.
998 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
999 (walk_stores): Likewise.
1000 (arc_address_cost): Make it static.
1001 (arc_verify_short): Likewise.
1002 (branch_dest): Likewise.
1003 (arc_attr_type): Likewise.
1004 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1005 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1006 (arc_final_prescan_insn): Remove inserting the nops due to
1007 hardware hazards. It is done in reorg step.
1008 (insn_length_variant_t): Remove.
1009 (insn_length_parameters_t): Likewise.
1010 (arc_insn_length_parameters): Likewise.
1011 (arc_get_insn_variants): Likewise.
1012 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1014 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1016 PR inline-asm/84683
1017 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1018 assertion failure.
1020 PR tree-optimization/84687
1021 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1022 on new_node->decl.
1023 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1025 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1027 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1028 Rename to ppc_speculation_barrier.
1029 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1030 __builtin_ppc_speculation_barrier.
1032 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1034 PR target/84700
1035 * combine.c (combine_simplify_rtx): Don't try to simplify if
1036 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1037 are equal to x.
1039 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1041 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1042 to 32 bytes when compiling for POWER9.
1044 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1046 PR target/84564
1047 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1048 regparm >= 3 with no arg reg available also for calls with
1049 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1051 PR target/84524
1052 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1053 orig,vex.
1054 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1056 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1058 PR target/84264
1059 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1061 2018-03-05 Richard Biener <rguenther@suse.de>
1063 PR tree-optimization/84486
1064 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1065 When inserting a __builtin_assume_aligned call set the LHS
1066 SSA name alignment info accordingly.
1068 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1070 PR tree-optimization/84114
1071 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1072 Avoid reassociation of FLOAT_MODE addition.
1074 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1076 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1077 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1078 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1079 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1080 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1081 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1082 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1083 and -mwbnoinvd.
1084 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1085 __builtin_ia32_wbinvd): New builtins.
1086 (SPECIAL_ARGS2): New.
1087 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1088 (SPECIAL_ARGS2): New.
1089 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1090 (ix86_valid_target_attribute_inner_p): Ditto.
1091 (ix86_init_mmx_sse_builtins): Add special_args2.
1092 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1093 TARGET_WBNOINVD_P): New.
1094 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1095 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1096 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1097 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1098 * config/i386/pconfigintrin.h: New file.
1099 * config/i386/wbnoinvdintrin.h: Ditto.
1100 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1101 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1103 2018-03-05 Richard Biener <rguenther@suse.de>
1105 PR tree-optimization/84670
1106 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1107 member.
1108 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1109 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1110 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1111 doesn't grow if all successors (recursively) were visited at least
1112 once.
1114 2018-03-05 Richard Biener <rguenther@suse.de>
1116 PR tree-optimization/84650
1117 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1118 if executed in the loop pipeline.
1120 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1122 * doc/configfiles.texi (Configuration Files): Move info about
1123 conditionalizing $target-protos.h to...
1124 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1125 differs from $target-protos.h.
1127 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1128 Chung-Ju Wu <jasonwucj@gmail.com>
1130 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1131 * config/nds32/nds32-multiple.md (setmemsi): Define.
1132 * config/nds32/nds32-memory-manipulation.c
1133 (nds32_gen_dup_4_byte_to_word_value): New.
1134 (emit_setmem_word_loop): New.
1135 (emit_setmem_byte_loop): New.
1136 (nds32_expand_setmem_loop): New.
1137 (nds32_expand_setmem_loop_v3m): New.
1138 (nds32_expand_setmem_unroll): New.
1139 (nds32_expand_setmem): New.
1141 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1142 Chung-Ju Wu <jasonwucj@gmail.com>
1144 * config/nds32/nds32-memory-manipulation.c
1145 (nds32_emit_load_store): New.
1146 (nds32_emit_post_inc_load_store): New.
1147 (nds32_emit_mem_move): New.
1148 (nds32_emit_mem_move_block): New.
1149 (nds32_expand_movmemsi_loop_unknown_size): New.
1150 (nds32_expand_movmemsi_loop_known_size): New.
1151 (nds32_expand_movmemsi_loop): New.
1152 (nds32_expand_movmemsi_unroll): New.
1153 (nds32_expand_movmemqi): Rename ...
1154 (nds32_expand_movmemsi): ... to this.
1155 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1156 (movmemsi): ... to this.
1157 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1158 (nds32_expand_movmemsi): ... to this.
1160 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1161 Monk Chiang <sh.chiang04@gmail.com>
1162 Chung-Ju Wu <jasonwucj@gmail.com>
1164 * config/nds32/nds32-protos.h
1165 (nds32_expand_load_multiple): New arguments.
1166 (nds32_expand_store_multiple): Ditto.
1167 (nds32_valid_multiple_load_store): Rename ...
1168 (nds32_valid_multiple_load_store_p): ... to this.
1169 * config/nds32/nds32-memory-manipulation.c
1170 (nds32_expand_load_multiple): Refine implementation.
1171 (nds32_expand_store_multiple): Ditto.
1172 * config/nds32/nds32-multiple.md
1173 (load_multiple): Update nds32_expand_load_multiple interface.
1174 (store_multiple): Update nds32_expand_store_multiple interface.
1175 * config/nds32/nds32-predicates.c
1176 (nds32_valid_multiple_load_store): Rename ...
1177 (nds32_valid_multiple_load_store_p): ... to this and refine
1178 implementation.
1179 * config/nds32/predicates.md
1180 (nds32_load_multiple_and_update_address_operation): New predicate.
1181 (nds32_store_multiple_and_update_address_operation): New predicate.
1183 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1184 Chung-Ju Wu <jasonwucj@gmail.com>
1186 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1187 (combo): New attribute.
1188 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1190 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1192 * config/nds32/nds32.opt: Change -mcmodel= default value.
1194 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1195 Monk Chiang <sh.chiang04@gmail.com>
1196 Chung-Ju Wu <jasonwucj@gmail.com>
1198 * config/nds32/constants.md (unspec_element): New enum.
1199 * config/nds32/constraints.md (Umw): New constraint.
1200 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1201 * config/nds32/nds32-intrinsic.md: Likewise.
1202 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1203 (nds32_valid_smw_lwm_base_p): New.
1204 (nds32_output_smw_single_word): New.
1205 (nds32_output_lmw_single_word): New.
1206 (nds32_expand_unaligned_load): New.
1207 (nds32_expand_unaligned_store): New.
1208 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1209 (nds32_output_smw_single_word): Declare.
1210 (nds32_output_lmw_single_word): Declare.
1211 (nds32_expand_unaligned_load): Declare.
1212 (nds32_expand_unaligned_store): Declare.
1213 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1214 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1215 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1216 NDS32_BUILTIN_UASTORE_DW.
1217 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1218 predicate.
1220 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1221 Kito Cheng <kito.cheng@gmail.com>
1222 Chung-Ju Wu <jasonwucj@gmail.com>
1224 * config/nds32/nds32-intrinsic.c
1225 (nds32_expand_builtin_null_ftype_reg): Delete.
1226 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1227 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1228 (nds32_read_argument): New.
1229 (nds32_legitimize_target): Ditto.
1230 (nds32_legitimize_argument): Ditto.
1231 (nds32_check_constant_argument): Ditto.
1232 (nds32_expand_unop_builtin): Ditto.
1233 (nds32_expand_unopimm_builtin): Ditto.
1234 (nds32_expand_binop_builtin): Ditto.
1235 (nds32_builtin_decl_impl): Ditto.
1236 (builtin_description): Ditto.
1237 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1238 (nds32_init_builtins_impl): Ditto.
1239 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1240 (nds32_builtin_decl): New.
1241 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1242 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1244 2018-03-02 Jeff Law <law@redhat.com>
1246 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1247 (redundant_insn, fill_simple_delay_slots): Likewise.
1248 (fill_slots_from_thread): Likewise.
1249 * resource.c (mark_referenced_resources): Likewise.
1250 (mark_set_resources, find_dead_or_set_registers): Likewise.
1252 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1254 * substring-locations.h (format_warning_va): Formatting fix for
1255 ATTRIBUTE_GCC_DIAG.
1256 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1257 argument.
1258 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1259 * substring-locations.c: Include intl.h.
1260 (format_warning_va): Turned into small wrapper around
1261 format_warning_n_va, renamed to ...
1262 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1263 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1264 use ngettext.
1265 (format_warning_at_substring_n): New function.
1266 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1267 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1268 format_warning_at_substring with just a shorter name instead of
1269 const function pointer.
1270 (fmtwarn_n): New function.
1271 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1272 appropriate, get rid of all the fmtstr temporaries, move conditionals
1273 with G_() wrapped string literals directly into fmtwarn arguments,
1274 cast dir.len to (int), formatting fixes.
1276 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1278 * doc/invoke.texi: Remove "Cilk Plus" references.
1280 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1281 Richard Biener <rguenther@suse.de>
1283 PR ipa/84628
1284 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1285 for error or warning attributes if CALL_FROM_THUNK_P is set.
1286 Formatting fixes.
1288 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1290 PR target/56540
1291 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1292 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1294 PR target/56540
1295 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1296 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1298 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1299 instead of -1U in last predictors element's probability member.
1301 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1303 PR ipa/83983
1304 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1305 arguments if they are comparable.
1307 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1309 PR tree-optimization/84634
1310 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1311 masks and masked_loop_p with a single loop_masks, making sure it's
1312 null for bb vectorization.
1314 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1316 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1317 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1318 loop->force_vectorize to check whether there is no alias.
1320 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1322 PR target/84614
1323 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1324 prototypes.
1325 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1326 comments.
1327 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1328 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1329 instead of a loop around prev_real_insn.
1330 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1331 prev_real_insn.
1333 PR inline-asm/84625
1334 * config/i386/i386.c (ix86_print_operand): Use conditional
1335 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1336 zero vector.
1338 2018-03-02 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/84427
1341 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1342 (bitmap_set_subtract_values): Rewrite to handle multiple
1343 exprs per value.
1344 (clean): Likewise.
1345 (prune_clobbered_mems): Likewise.
1346 (phi_translate): Take edge instead of pred/phiblock.
1347 (phi_translate_1): Likewise.
1348 (phi_translate_set): Likewise. Insert all translated
1349 exprs for a value into the set, keeping possibly multiple
1350 expressions per value.
1351 (compute_antic_aux): Adjust for phi_translate changes.
1352 When intersecting union the expressions and prune those
1353 not in the final value set, keeping possibly multiple
1354 expressions per value. Do not use value-insertion
1355 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1356 all expressions. Add verification that the value-sets
1357 only shrink during iteration.
1358 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1359 (do_pre_regular_insertion): Likewise.
1360 (do_pre_partial_partial_insertion): Likewise.
1362 2018-03-02 Richard Biener <rguenther@suse.de>
1364 PR target/82005
1365 * config/darwin.c (saved_debug_info_level): New static global.
1366 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1367 (darwin_asm_lto_end): Restore debug info generation settings.
1369 2018-03-01 Martin Liska <mliska@suse.cz>
1371 PR sanitizer/82484
1372 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1373 volatile arguments.
1375 2018-03-01 Richard Biener <rguenther@suse.de>
1377 PR debug/84645
1378 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1379 type annotation with LTO when debug was disabled at compile-time.
1381 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1383 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1384 XINT with INTVAL.
1385 (mips_final_postscan_insn): Likewise.
1387 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1389 PR rtl-optimization/84528
1390 * alias.c (init_alias_target): Add commentary.
1391 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1392 a unique base value if the frame pointer is not eliminated
1393 to the stack pointer.
1395 2018-03-01 Tom de Vries <tom@codesourcery.com>
1397 PR rtl-optimization/83327
1398 * lra-int.h (hard_regs_spilled_into): Declare.
1399 * lra.c (hard_regs_spilled_into): Define.
1400 (init_reg_info): Init hard_regs_spilled_into.
1401 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1402 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1403 (process_bb_lives): Handle hard_regs_spilled_into.
1404 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1405 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1407 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1409 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1410 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1411 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1412 * config/rs6000/aix72.h: New file.
1414 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1416 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1417 instead of warning_at with conditional singular and plural messages
1418 where possible.
1420 PR target/52991
1421 * stor-layout.c (update_alignment_for_field): For
1422 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1423 && !DECL_PACKED (field), do the alignment update, just use
1424 only desired_align instead of MAX (type_align, desired_align)
1425 as the alignment.
1426 (place_field): Don't do known_align < desired_align handling
1427 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1428 is non-NULL, instead do it after rli->prev_field handling and
1429 only if not within a bitfield word. For DECL_PACKED (field)
1430 use type_align of BITS_PER_UNIT.
1432 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1434 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1435 superfluous parentheses and trailing spaces.
1437 2018-02-28 Richard Biener <rguenther@suse.de>
1439 PR tree-optimization/84584
1440 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1441 SCoPs with fake exit edge.
1443 2018-02-28 Martin Liska <mliska@suse.cz>
1445 PR testsuite/84597
1446 * timevar.c (timer::print): Fix format to properly print 100%
1447 values.
1449 2018-02-28 Richard Biener <rguenther@suse.de>
1451 PR middle-end/84607
1452 * genmatch.c (capture_info::walk_match): Do not mark
1453 captured expressions without operands as expr_p given
1454 they act more like predicates and should be subject to
1455 "lost tail" side-effect preserving.
1457 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1459 PR rtl-optimization/81611
1460 * auto-inc-dec.c (attempt_change): Move dead note from
1461 mem_insn if it's the next use of regno
1462 (find_address): Take address use of reg holding
1463 non-incremented value. Add parm to limit search to the named
1464 reg only.
1465 (merge_in_block): Attempt to use a mem insn that is the next
1466 use of the original regno.
1468 2018-02-27 Martin Sebor <msebor@redhat.com>
1470 PR c++/83871
1471 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1472 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1474 2018-02-27 Martin Sebor <msebor@redhat.com>
1476 PR translation/84207
1477 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1478 n argument to unsigned HOST_WIDE_INT.
1479 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1480 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1481 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1482 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1484 2018-02-27 Richard Biener <rguenther@suse.de>
1486 PR tree-optimization/84512
1487 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1488 Do not use the estimate returned from record_stmt_cost for
1489 the scalar iteration cost but sum properly using add_stmt_cost.
1491 2018-02-27 Richard Biener <rguenther@suse.de>
1493 PR tree-optimization/84466
1494 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1495 Adjust last change to less strictly validate use operands.
1497 2018-02-27 Martin Liska <mliska@suse.cz>
1499 PR gcov-profile/84548
1500 * gcov.c (process_file): Allow partial overlap and consider it
1501 also as group functions.
1502 (output_lines): Properly calculate range of lines for a group.
1504 2018-02-27 Martin Liska <mliska@suse.cz>
1506 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1507 'ggc' suffixes. Change first column width.
1508 (timer::print): Fix formatting of the column.
1510 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1512 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1513 preserve inline entry blocks for the sake of debug inline
1514 entry point markers alone.
1515 (remove_unused_locals): Suggest in comments a better place to
1516 force the preservation of inline entry blocks that are
1517 otherwise unused, but do not preserve them.
1519 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1521 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1523 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1525 PR target/84039
1526 * config/i386/constraints.md (Bs): Replace
1527 ix86_indirect_branch_register with
1528 TARGET_INDIRECT_BRANCH_REGISTER.
1529 (Bw): Likewise.
1530 * config/i386/i386.md (indirect_jump): Likewise.
1531 (tablejump): Likewise.
1532 (*sibcall_memory): Likewise.
1533 (*sibcall_value_memory): Likewise.
1534 Peepholes of indirect call and jump via memory: Likewise.
1535 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1536 (*sibcall_value_GOT_32): Likewise.
1537 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1538 (GOT_memory_operand): Likewise.
1539 (call_insn_operand): Likewise.
1540 (sibcall_insn_operand): Likewise.
1541 (GOT32_symbol_operand): Likewise.
1542 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1544 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1546 PR rtl-optimization/83496
1547 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1548 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1549 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1550 redundant insn, if any.
1551 (relax_delay_slots): Likewise.
1552 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1554 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1556 PR tree-optimization/83965
1557 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1558 that grouped statements are part of a reduction chain. Return
1559 true if the statement is not marked as a reduction itself but
1560 is part of a group.
1561 (vect_recog_dot_prod_pattern): Don't check whether the statement
1562 is part of a group here.
1563 (vect_recog_sad_pattern): Likewise.
1564 (vect_recog_widen_sum_pattern): Likewise.
1566 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1568 PR debug/84545
1569 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1571 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1573 PR target/84530
1574 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1575 the bool argument.
1576 (ix86_output_indirect_function_return): New prototype.
1577 (ix86_split_simple_return_pop_internal): Likewise.
1578 * config/i386/i386.c (indirect_return_via_cx): New.
1579 (indirect_return_via_cx_bnd): Likewise.
1580 (indirect_thunk_name): Handle return va CX_REG.
1581 (output_indirect_thunk_function): Create alias for
1582 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1583 (ix86_output_indirect_jmp): Remove the bool argument.
1584 (ix86_output_indirect_function_return): New function.
1585 (ix86_split_simple_return_pop_internal): Likewise.
1586 * config/i386/i386.md (*indirect_jump): Don't pass false
1587 to ix86_output_indirect_jmp.
1588 (*tablejump_1): Likewise.
1589 (simple_return_pop_internal): Change it to define_insn_and_split.
1590 Call ix86_split_simple_return_pop_internal to split it for
1591 -mfunction-return=.
1592 (simple_return_indirect_internal): Call
1593 ix86_output_indirect_function_return instead of
1594 ix86_output_indirect_jmp.
1596 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1598 PR bootstrap/84405
1599 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1600 memset and value initialization afterwards.
1602 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1604 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1606 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1608 PR target/84521
1609 * common/config/aarch64/aarch64-common.c
1610 (aarch_option_optimization_table[]): Switch
1611 off fomit-frame-pointer
1613 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1614 Chung-Ju Wu <jasonwucj@gmail.com>
1616 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1617 volatile memory.
1618 (store_multiple): Ditto.
1620 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1622 * config.gcc: Add --with-cpu support for nds32 target.
1623 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1624 * config/nds32/nds32.opt: Add -mcpu= option.
1626 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1628 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1629 isel=yes): Warn for these deprecated options.
1631 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1633 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1634 ISA_2_5_MASKS_EMBEDDED.
1636 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1638 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1639 p->max as pointers rather than using iterative_hash_expr.
1641 2018-02-23 Carl Love <cel@us.ibm.com>
1643 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1644 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1645 BU_P8V_OVERLOAD_2.
1646 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1647 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1648 P8V_BUILTIN_VEC_VUNSIGNED2.
1650 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1652 PR target/81572
1653 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1654 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1655 LRA_UNKNOWN_ALT.
1656 * lra-constraints.c (curr_insn_transform): Set up
1657 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1658 LRA_UNKNOWN_ALT.
1659 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1660 * lra-eliminations.c (spill_pseudos): Ditto.
1661 (process_insn_for_elimination): Ditto.
1662 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1663 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1664 LRA_NON_CLOBBERED_ALT.
1666 2018-02-22 Martin Sebor <msebor@redhat.com>
1668 PR tree-optimization/84480
1669 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1670 to maybe_diag_stxncpy_trunc. Call it.
1671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1672 from gimple_fold_builtin_strcpy. Print inlining stack.
1673 (handle_builtin_stxncpy): Print inlining stack.
1674 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1676 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1678 PR target/84176
1679 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1680 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1681 and -fcheck-pointer-bounds are used together.
1682 (indirect_thunk_prefix): New enum.
1683 (indirect_thunk_need_prefix): New function.
1684 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1685 "_nt" instead of "_bnd" for NOTRACK prefix.
1686 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1687 (output_indirect_thunk_function): Likewise.
1688 (): Likewise.
1689 (ix86_code_end): Update output_indirect_thunk_function calls.
1690 (ix86_output_indirect_branch_via_reg): Replace
1691 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1692 (ix86_output_indirect_branch_via_push): Likewise.
1693 (ix86_output_function_return): Likewise.
1694 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1695 incompatible with -fcf-protection=branch and
1696 -fcheck-pointer-bounds.
1698 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1700 PR target/83335
1701 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1702 Change gcc_assert call to output_operand_lossage.
1704 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1706 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1708 2018-02-22 DJ Delorie <dj@redhat.com>
1709 Sebastian Perta <sebastian.perta@renesas.com>
1710 Oleg Endo <olegendo@gcc.gnu.org>
1712 * config/rx/rx.c (rx_rtx_costs): New function.
1713 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1715 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1717 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1719 2018-02-22 Martin Liska <mliska@suse.cz>
1721 PR driver/83193
1722 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1723 Add "native" as a possible value.
1725 2018-02-22 Martin Liska <mliska@suse.cz>
1727 PR driver/83193
1728 * config/i386/i386.c (ix86_option_override_internal):
1729 Add "native" as a possible value for -march and -mtune.
1731 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1733 PR target/84502
1734 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1735 to all type variants.
1737 PR tree-optimization/84503
1738 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1739 width as info->bitpos + info->bitsize - start.
1740 (merged_store_group::merge_overlapping): Simplify width computation.
1741 (check_no_overlap): New function.
1742 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1743 start + width and last_order of the group, fail if check_no_overlap
1744 fails.
1745 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1746 to group if check_no_overlap fails.
1748 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1750 * config/rs6000/altivec.md: Delete contraint arguments to
1751 define_expand, define_split, and define_peephole2, and in
1752 define_insn_and_split if always unused.
1753 * config/rs6000/darwin.md: Ditto.
1754 * config/rs6000/dfp.md: Ditto.
1755 * config/rs6000/rs6000.md: Ditto.
1756 * config/rs6000/sync.md: Ditto.
1757 * config/rs6000/vector.md: Ditto.
1758 * config/rs6000/vsx.md: Ditto.
1760 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1762 * config/rs6000/altivec.md: Write output control strings as braced
1763 blocks instead of double-quoted strings.
1764 * config/rs6000/darwin.md: Ditto.
1765 * config/rs6000/rs6000.md: Ditto.
1766 * config/rs6000/vector.md: Ditto.
1767 * config/rs6000/vsx.md: Ditto.
1769 2018-02-21 Jason Merrill <jason@redhat.com>
1771 PR c++/84314 - ICE with templates and fastcall attribute.
1772 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1774 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1776 * ipa-cp.c (determine_versionability): Fix comment typos.
1778 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1780 PR c/84229
1781 * ipa-cp.c (determine_versionability): Do not version functions caling
1782 va_arg_pack.
1784 2018-02-21 Martin Liska <mliska@suse.cz>
1786 PR driver/83193
1787 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1788 Add "native" as a possible value.
1789 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1790 the macro when native cpu detection is available.
1792 2018-02-21 Martin Liska <mliska@suse.cz>
1794 PR driver/83193
1795 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1796 Add "native" as a possible value.
1797 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1798 when native cpu detection is available.
1800 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1801 Martin Sebor <msebor@redhat.com>
1803 PR tree-optimization/84478
1804 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1805 false.
1806 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1807 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1808 support which is conservatively correct, for 2 only stay conservative
1809 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1810 argument to the 2 argument get_range_strlen, adjust 6 arg
1811 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1812 false.
1813 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1814 (gimple_fold_builtin_strlen): Pass true as last argument to
1815 get_range_strlen.
1817 2018-02-20 Martin Sebor <msebor@redhat.com>
1819 PR middle-end/84095
1820 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1821 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1822 (builtin_memref::builtin_memref): Factor out parts into
1823 set_base_and_offset and call it.
1825 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1827 PR middle-end/84406
1828 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1829 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1830 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1831 search at the associated MODE_INT.
1833 2018-02-20 Jeff Law <law@redhat.com>
1835 PR middle-end/82123
1836 PR tree-optimization/81592
1837 PR middle-end/79257
1838 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1839 for range data rather than using global data.
1840 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1841 range data rather than using global data.
1842 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1843 pass it to children as needed.
1844 (struct directive::fmtresult): Similarly.
1845 (struct directive::set_width): Similarly.
1846 (struct directive::set_precision): Similarly.
1847 (format_integer, format_directive, parse_directive): Similarly.
1848 (format_none): Accept unnamed vr_values parameter.
1849 (format_percent, format_floating, format_character): Similarly.
1850 (format_string, format_plain): Similarly.
1851 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1852 the EVRP range analyzer for range data rather than using global data.
1853 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1854 gimple-ssa-evrp-analyze.h
1855 (class sprintf_dom_walker): Add after_dom_children member function.
1856 Add evrp_range_analyzer member.
1857 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1858 range analyzer as needed.
1859 (sprintf_dom_walker::after_dom_children): New member function.
1860 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1861 if not optimizing.
1862 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1863 (evrp_range_analyzer::pop_to_marker): Likewise.
1865 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1867 PR tree-optimization/84419
1868 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1869 with the required type if its current type is compatible but
1870 different.
1872 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1874 PR middle-end/82004
1875 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1876 after vectorization.
1878 2018-02-20 Martin Liska <mliska@suse.cz>
1880 PR driver/83193
1881 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1882 possible values if we don't have a hint.
1884 2018-02-20 Martin Liska <mliska@suse.cz>
1886 PR c/84310
1887 PR target/79747
1888 * final.c (shorten_branches): Build align_tab array with one
1889 more element.
1890 * opts.c (finish_options): Add alignment option limit check.
1891 (MAX_CODE_ALIGN): Likewise.
1892 (MAX_CODE_ALIGN_VALUE): Likewise.
1893 * doc/invoke.texi: Document maximum allowed option value for
1894 all -falign-* options.
1896 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1898 PR target/84146
1899 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1900 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1901 * var-tracking.c (emit_note_insn_var_location): Remove all references
1902 to NOTE_INSN_CALL_ARG_LOCATION.
1903 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1904 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1905 Use copy_rtx_if_shared.
1906 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1907 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1908 (dwarf2out_var_location): Remove handling of
1909 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1910 on call_insn.
1911 * final.c (final_scan_insn): Remove all references to
1912 NOTE_INSN_CALL_ARG_LOCATION.
1913 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1914 before dumping final insns.
1915 * except.c (emit_note_eh_region_end): Remove all references to
1916 NOTE_INSN_CALL_ARG_LOCATION.
1917 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1918 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1919 * config/arc/arc.c (hwloop_optimize): Likewise.
1920 * config/arm/arm.c (create_fix_barrier): Likewise.
1921 * config/s390/s390.c (s390_chunkify_start): Likewise.
1922 * config/sh/sh.c (find_barrier): Likewise.
1923 * config/i386/i386.c (rest_of_insert_endbranch,
1924 ix86_seh_fixup_eh_fallthru): Likewise.
1925 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1926 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1927 * config/frv/frv.c (frv_function_prologue): Likewise.
1928 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1929 reg note.
1930 (note_outside_basic_block_p): Remove all references to
1931 NOTE_INSN_CALL_ARG_LOCATION.
1932 * gengtype.c (adjust_field_rtx_def): Likewise.
1933 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1934 Likewise.
1935 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1936 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1938 PR c++/84444
1939 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1940 is ADDR_EXPR.
1942 PR tree-optimization/84452
1943 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1944 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1945 is NULL.
1947 2018-02-19 Martin Liska <mliska@suse.cz>
1949 PR sanitizer/82183
1950 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1952 2018-02-19 Martin Liska <mliska@suse.cz>
1953 Richard Sandiford <richard.sandiford@linaro.org>
1955 PR tree-optimization/82491
1956 * gimple-fold.c (get_base_constructor): Make earlier bail out
1957 to prevent ubsan.
1959 2018-02-19 Carl Love <cel@us.ibm.com>
1961 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1962 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1963 BU_P8V_OVERLOAD_1.
1964 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1965 P8V_BUILTIN_VEC_NEG.
1967 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1969 * config/rl78/rl78.md (movdf): New define expand.
1971 2018-02-19 Martin Liska <mliska@suse.cz>
1973 PR other/80589
1974 * doc/invoke.texi: Fix typo.
1975 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1977 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1979 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1980 handle rs6000_single_float and rs6000_double_float specially for
1981 e500 family CPUs.
1983 2018-02-16 Jeff Law <law@redhat.com>
1985 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1986 (pop_regs): Use it.
1988 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1990 PR ipa/84425
1991 * ipa-inline.c (inline_small_functions): Fix a typo.
1993 2018-02-16 Nathan Sidwell <nathan@acm.org>
1995 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1997 2018-02-16 Carl Love <cel@us.ibm.com>
1999 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2000 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2001 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2002 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2003 expansion to P8V_BUILTIN_VEC_FLOAT2.
2005 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2007 PR rtl-optimization/70023
2008 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2009 src_regno into account.
2011 2018-02-16 Carl Love <cel@us.ibm.com>
2013 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2014 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2015 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2016 * config/rs6000/rs6000.c: Remove case statements for
2017 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2018 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2019 and P9V_BUILTIN_VEC_VINSERT4B.
2020 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2021 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2022 * config/rs6000/vsx.md:
2023 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2024 vec_insert4b.
2026 2018-02-16 Carl Love <cel@us.ibm.com>
2028 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2029 vec_insert4b.
2030 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2031 definitions.
2032 * config/rs6000/rs6000-c.c: Add the definitions for
2033 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2034 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2035 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2036 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2037 definition for insert4b and define insn *insert3b_internal.
2038 * doc/extend.texi: Add documentation for vec_extract4b.
2040 2018-02-16 Nathan Sidwell <nathan@acm.org>
2042 * doc/extend.texi (Backwards Compatibility): Mention friend
2043 injection. Note for-scope is deprecated.
2044 * doc/invoke.texi (-ffriend-injection): Deprecate.
2046 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2048 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2049 that moved to I2, also allow destinations that are a paradoxical
2050 subreg (instead of a normal reg).
2052 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2054 PR target/83831
2055 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2056 to QImode.
2058 2018-02-16 Richard Biener <rguenther@suse.de>
2060 PR tree-optimization/84037
2061 PR tree-optimization/84016
2062 PR target/82862
2063 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2064 Adjust vec_construct for the fact we need additional higher latency
2065 128bit inserts for AVX256 and AVX512 vector builds.
2066 (ix86_add_stmt_cost): Scale vector construction cost for
2067 elementwise loads.
2069 2018-02-16 Richard Biener <rguenther@suse.de>
2071 PR tree-optimization/84417
2072 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2073 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2074 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2076 2018-02-16 Martin Liska <mliska@suse.cz>
2078 PR sanitizer/84307
2079 * internal-fn.def (ASAN_CHECK): Set proper flags.
2080 (ASAN_MARK): Likewise.
2082 2018-02-16 Julia Koval <julia.koval@intel.com>
2084 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2085 from PTA_CANNONLAKE.
2087 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2089 PR target/84272
2090 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2091 Use ++iter rather than iter++ for std::list iterators.
2092 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2093 defer deleting them until all nodes in the forest are processed. Do
2094 free even leaf nodes. Change to_process into auto_vec.
2096 PR bootstrap/84405
2097 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2098 * vec.h (vec_default_construct): Use memset instead of placement new
2099 if BROKEN_VALUE_INITIALIZATION is defined.
2100 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2101 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2102 is defined.
2104 PR rtl-optimization/83723
2105 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2106 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2107 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2108 recursive calls.
2109 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2110 callers.
2111 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2113 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2115 PR rtl-optimization/81443
2116 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2117 from inner REGs to paradoxical SUBREGs.
2119 2018-02-16 Richard Biener <rguenther@suse.de>
2121 PR tree-optimization/84399
2122 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2123 For operands we can analyze at their definition make sure we can
2124 analyze them at each use as well.
2126 2018-02-16 Richard Biener <rguenther@suse.de>
2128 PR tree-optimization/84190
2129 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2130 volatile accesses if the decl isn't volatile.
2132 2018-02-15 Jason Merrill <jason@redhat.com>
2134 PR c++/84314 - ICE with templates and fastcall attribute.
2135 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2136 TYPE_CANONICAL on an existing type.
2138 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2140 PR tree-optimization/84383
2141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2142 dstoff nor call operand_equal_p if dstbase is NULL.
2144 PR tree-optimization/84334
2145 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2146 also a CONSTANT_CLASS_P, punt.
2148 2018-02-14 Jim Wilson <jimw@sifive.com>
2150 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2151 first SMALL_OPERAND check. New local min_second_step. Move assert
2152 to where locals are set. Add TARGET_RVC support.
2153 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2155 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2157 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2159 2018-02-14 Martin Sebor <msebor@redhat.com>
2161 PR tree-optimization/83698
2162 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2163 arrays constrain the offset range to their bounds.
2164 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2165 (builtin_access::overlap): Avoid setting the size of overlap if it's
2166 already been set.
2167 (maybe_diag_overlap): Also consider arrays when deciding what values
2168 of offsets to include in diagnostics.
2170 2018-02-14 Martin Sebor <msebor@redhat.com>
2172 PR c/84108
2173 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2174 that correspond to the kind of a declaration.
2176 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2178 PR target/83984
2179 * config/pa/pa.md: Load address of PIC label using the linkage table
2180 if the label is nonlocal.
2182 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2184 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2185 warning message if user requests -maltivec=be.
2186 * doc/invoke.texi: Document deprecation of -maltivec=be.
2188 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2190 PR target/84220
2191 * config/rs6000/rs6000-c.c: Update definitions for
2192 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2193 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2195 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2197 PR target/84239
2198 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2199 add _get_ssp intrinsics. Remove argument from
2200 __builtin_ia32_rdssp[d|q].
2201 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2202 * config/i386/i386-builtin.def: Remove argument from
2203 __builtin_ia32_rdssp[d|q].
2204 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2205 ix86_expand_special_args_builtin for _rdssp[d|q].
2206 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2207 Clear register before usage.
2208 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2209 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2211 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2213 PR tree-optimization/84357
2214 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2215 operand 1 of an ARRAY_REF too.
2217 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2219 PR target/83831
2220 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2221 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2222 declarations.
2223 (set_of_reg): New struct.
2224 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2225 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2226 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2227 functions.
2228 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2229 Split into bitclr, bitset, bitinvert patterns if appropriate.
2230 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2231 use rx_fuse_in_memory_bitop.
2232 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2233 to named insn, correct maximum insn length.
2235 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2237 PR target/79242
2238 * machmode.def: Define a complex mode for PARTIAL_INT.
2239 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2240 MODE_PARTIAL_INT.
2241 * doc/rtl.texi: Document CSPImode.
2242 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2243 handling.
2244 (msp430_hard_regno_nregs_with_padding): Likewise.
2246 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2248 PR target/84279
2249 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2251 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2253 PR rtl-optimization/84169
2254 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2255 we generated a parallel as new i3 and we split that to new i2 and i3
2256 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2257 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2258 those to i2, not i1. Partially rewrite this scan code.
2260 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2262 PR c/82210
2263 * stor-layout.c (place_field): For variable length fields, adjust
2264 offset_align afterwards not just based on the field's alignment,
2265 but also on the size.
2267 PR middle-end/84309
2268 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2269 of exps and logs in the use_exp2 case.
2271 2018-02-13 Jeff Law <law@redhat.com>
2273 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2274 entry for "vector".
2276 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2277 ARGS as unused.
2279 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2281 PR debug/84342
2282 PR debug/84319
2283 * common.opt (gas-loc-support, gas-locview-support): New.
2284 (ginline-points, ginternal-reset-location-views): New.
2285 * doc/invoke.texi: Document them. Use @itemx where intended.
2286 (gvariable-location-views): Adjust.
2287 * target.def (reset_location_view): New.
2288 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2289 (TARGET_RESET_LOCATION_VIEW): New.
2290 * doc/tm.texi: Rebuilt.
2291 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2292 (dwarf2out_default_as_locview_support): New.
2293 (output_asm_line_debug_info): Use option variables.
2294 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2295 (output_loc_list): Likewise.
2296 (add_high_low_attributes): Check option variables.
2297 Don't output entry view attribute in strict mode.
2298 (gen_inlined_subroutine_die): Check option variables.
2299 (dwarf2out_inline_entry): Likewise.
2300 (init_sections_and_labels): Likewise.
2301 (dwarf2out_early_finish): Likewise.
2302 (maybe_reset_location_view): New, from...
2303 (dwarf2out_var_location): ... here. Call it.
2304 * debug.h (dwarf2out_default_as_loc_support): Declare.
2305 (dwarf2out_default_as_locview_support): Declare.
2306 * hooks.c (hook_int_rtx_insn_0): New.
2307 * hooks.h (hook_int_rtx_insn_0): Declare.
2308 * toplev.c (process_options): Take -gas-loc-support and
2309 -gas-locview-support from dwarf2out. Enable
2310 -gvariable-location-views by default only with locview
2311 assembler support. Enable -ginternal-reset-location-views by
2312 default only if the target defines the corresponding hook.
2313 Enable -ginline-points by default if location views are
2314 enabled; force it disabled if statement frontiers are
2315 disabled.
2316 * tree-inline.c (expand_call_inline): Check option variables.
2317 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2319 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2321 PR tree-optimization/84321
2322 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2323 handling. Also check whether the anti-range contains any values
2324 that satisfy the mask; switch to a VR_RANGE if not.
2326 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2328 PR sanitizer/84340
2329 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2331 2018-02-13 Martin Jambor <mjambor@suse.cz>
2333 PR c++/83990
2334 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2335 of call statements, also set location of a load to a temporary.
2337 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2339 * config/rl78/rl78.c (add_vector_labels): New function.
2340 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2341 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2342 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2343 which checks that no arguments are passed.
2344 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2345 * doc/extend.texi: Documentation for the new attribute.
2347 2018-02-13 Andreas Schwab <schwab@suse.de>
2349 * config/riscv/linux.h (CPP_SPEC): Define.
2351 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2353 PR target/84335
2354 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2355 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2356 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2357 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2358 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2359 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2360 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2361 temporarily for AES and PCLMUL builtins.
2363 PR tree-optimization/84339
2364 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2365 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2366 Formatting fixes.
2368 PR middle-end/84309
2369 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2370 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2371 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2372 inline function.
2373 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2374 inline function.
2375 * omp-simd-clone.h: New file.
2376 * omp-simd-clone.c: Include omp-simd-clone.h.
2377 (expand_simd_clones): No longer static.
2378 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2379 cgraph.h and omp-simd-clone.h.
2380 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2381 (vect_recog_widen_shift_pattern): Formatting fix.
2382 (vect_pattern_recog_1): Don't check optab for calls.
2384 PR target/84336
2385 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2386 operands[2] into a REG before using gen_lowpart on it.
2388 2018-02-12 Jeff Law <law@redhat.com>
2390 PR target/83760
2391 * config/sh/sh.c (find_barrier): Consider a sibling call
2392 a barrier as well.
2394 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2395 successfully back substituting a reg.
2397 2018-02-12 Richard Biener <rguenther@suse.de>
2399 PR tree-optimization/84037
2400 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2401 parameter, move visited init to caller.
2402 (vect_slp_analyze_operations): Separate cost from validity
2403 check, initialize visited once for all instances.
2404 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2405 for all instances.
2406 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2407 out an assert.
2408 (vect_model_promotion_demotion_cost): Likewise.
2409 (vectorizable_bswap): Guard cost modeling with !slp_node
2410 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2411 SLP stmts.
2412 (vectorizable_call): Likewise.
2413 (vectorizable_conversion): Likewise.
2414 (vectorizable_assignment): Likewise.
2415 (vectorizable_shift): Likewise.
2416 (vectorizable_operation): Likewise.
2417 (vectorizable_store): Likewise.
2418 (vectorizable_load): Likewise.
2419 (vectorizable_condition): Likewise.
2420 (vectorizable_comparison): Likewise.
2422 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2424 PR sanitizer/84307
2425 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2426 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2427 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2429 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2431 PR middle-end/83665
2432 * params.def (inline-min-speedup): Increase from 8 to 15.
2433 (max-inline-insns-auto): Decrease from 40 to 30.
2434 * ipa-split.c (consider_split): Add some buffer for function to
2435 be considered inlining candidate.
2436 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2437 default values.
2439 2018-02-12 Richard Biener <rguenther@suse.de>
2441 PR tree-optimization/84037
2442 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2443 matched stmts if we cannot swap the non-matched ones.
2445 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2447 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2448 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2449 _mm_maskz_scalef_round_ss): New intrinsics.
2450 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2451 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2452 __builtin_ia32_scalefss_round): Remove.
2453 (__builtin_ia32_scalefsd_mask_round,
2454 __builtin_ia32_scalefss_mask_round): New intrinsics.
2455 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2456 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2457 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2458 "<round_constraint>")): Changed to ...
2459 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2460 "<round_scalar_constraint>")): ... this.
2461 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2462 %0, %1, %2<round_op3>}"): Changed to ...
2463 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2464 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2465 %2<round_scalar_mask_op3>}"): ... this.
2466 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2468 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2470 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2471 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2472 (_mm_maskz_sqrt_round_ss): New intrinsics.
2473 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2474 (__builtin_ia32_sqrtsd_mask_round)
2475 (__builtin_ia32_sqrtss_mask_round): New builtins.
2476 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2477 (__builtin_ia32_sqrtss_round): Remove.
2478 (__builtin_ia32_sqrtsd_mask_round)
2479 (__builtin_ia32_sqrtss_mask_round): New builtins.
2480 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2481 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2482 ((match_operand:VF_128 1 "vector_operand"
2483 "xBm,<round_constraint>")): Changed to ...
2484 ((match_operand:VF_128 1 "vector_operand"
2485 "xBm,<round_scalar_constraint>")): ... this.
2486 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2487 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2488 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2489 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2490 %<iptr>1<round_scalar_mask_op3>}): ... this.
2491 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2492 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2494 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2496 PR target/84266
2497 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2498 Cast vec_cmpeq result to correct type.
2499 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2500 Cast vec_cmpgt result to correct type.
2502 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2504 * final.c (final_scan_insn_1): Renamed from...
2505 (final_scan_insn): ... this. New wrapper, to recover
2506 seen from the outermost call in recursive ones.
2507 * config/sparc/sparc.c (output_return): Drop seen from call.
2508 (output_sibcall): Likewise.
2509 * config/visium/visium.c (output_branch): Likewise.
2511 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2513 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2514 function label.
2516 2018-02-10 Alan Modra <amodra@gmail.com>
2518 PR target/84300
2519 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2520 Specify LR as an input.
2522 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2524 PR sanitizer/83987
2525 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2526 remove_member_access_dummy_vars): New functions.
2527 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2528 lower_omp_1, execute_lower_omp): Use them.
2530 PR rtl-optimization/84308
2531 * shrink-wrap.c (spread_components): Release todo vector.
2533 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2535 PR rtl-optimization/57193
2536 * ira-color.c (struct allocno_color_data): Add member
2537 conflict_allocno_hard_prefs.
2538 (update_conflict_allocno_hard_prefs): New.
2539 (bucket_allocno_compare_func): Add a preference based on
2540 conflict_allocno_hard_prefs.
2541 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2542 (color_allocnos): Remove a dead code. Initiate
2543 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2545 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2547 PR target/84226
2548 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2549 constraint from =wa to wa. Avoid a subreg on the output operand,
2550 instead use a pseudo and subreg it in a move.
2551 (p9_xxbrd_<mode>): Changed to ...
2552 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2553 (p9_xxbrd_v2df): New expander.
2554 (p9_xxbrw_<mode>): Changed to ...
2555 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2556 (p9_xxbrw_v4sf): New expander.
2558 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2560 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2562 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2564 PR target/83926
2565 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2566 multiply in 32-bit mode.
2567 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2568 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2569 mode.
2571 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2573 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2574 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2575 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2576 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2578 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2580 PR lto/84213
2581 * dwarf2out.c (is_trivial_indirect_ref): New function.
2582 (dwarf2out_late_global_decl): Do not generate a location
2583 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2584 and that are not defined in the current unit.
2586 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2588 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2589 instead of a libcall for UNORDERED.
2591 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2593 PR target/82641
2594 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2595 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2597 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2599 PR target/PR84295
2600 * config/s390/s390.c (s390_set_current_function): Invoke
2601 s390_indirect_branch_settings also if fndecl didn't change.
2603 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2605 * config/rs6000/rs6000.md (blockage): Set length to zero.
2607 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2609 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2611 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2613 PR sanitizer/84285
2614 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2615 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2616 -static-lib*san.
2618 PR debug/84252
2619 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2620 PARALLEL incoming that failed vt_get_decl_and_offset check.
2622 PR middle-end/84237
2623 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2624 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2625 TREE_READONLY bit.
2626 (get_variable_section): For decls in named .bss* sections pass true as
2627 second argument to bss_initializer_p.
2629 2018-02-09 Marek Polacek <polacek@redhat.com>
2630 Jakub Jelinek <jakub@redhat.com>
2632 PR c++/83659
2633 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2634 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2635 Sync some changes from cxx_fold_indirect_ref.
2637 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2639 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2640 markers.
2641 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2642 (BLOCK_INLINE_ENTRY_LABEL): New.
2643 (dwarf2out_var_location): Disregard inline entry markers.
2644 (inline_entry_data): New struct.
2645 (inline_entry_data_hasher): New hashtable type.
2646 (inline_entry_data_hasher::hash): New.
2647 (inline_entry_data_hasher::equal): New.
2648 (inline_entry_data_table): New variable.
2649 (add_high_low_attributes): Add DW_AT_entry_pc and
2650 DW_AT_GNU_entry_view attributes if a pending entry is found
2651 in inline_entry_data_table. Add old entry_pc attribute only
2652 if debug nonbinding markers are disabled.
2653 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2654 markers are enabled.
2655 (block_within_block_p, dwarf2out_inline_entry): New.
2656 (dwarf2out_finish): Check that no entries remained in
2657 inline_entry_data_table.
2658 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2659 (final_scan_insn, notice_source_line): Likewise.
2660 (rest_of_clean_state): Skip inline entry markers.
2661 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2662 markers.
2663 * gimple.c (gimple_build_debug_inline_entry): New.
2664 * gimple.h (enum gimple_debug_subcode): Add
2665 GIMPLE_DEBUG_INLINE_ENTRY.
2666 (gimple_build_debug_inline_entry): Declare.
2667 (gimple_debug_inline_entry_p): New.
2668 (gimple_debug_nonbind_marker_p): Adjust.
2669 * insn-notes.def (INLINE_ENTRY): New.
2670 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2671 inline entry marker notes.
2672 (print_insn): Likewise.
2673 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2674 (INSN_DEBUG_MARKER_KIND): Likewise.
2675 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2676 * tree-inline.c (expand_call_inline): Build and insert
2677 debug_inline_entry stmt.
2678 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2679 inline entry blocks early, if nonbind markers are enabled.
2680 (dump_scope_block): Dump fragment info.
2681 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2682 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2683 (gimple_build_debug_inline_entry): New.
2684 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2685 Enable/disable inline entry points too.
2686 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2687 (DEBUG_INSN): Describe inline entry markers.
2689 * common.opt (gvariable-location-views): New.
2690 (gvariable-location-views=incompat5): New.
2691 * config.in: Rebuilt.
2692 * configure: Rebuilt.
2693 * configure.ac: Test assembler for view support.
2694 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2695 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2696 * dwarf2out.c (var_loc_view): New typedef.
2697 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2698 (dwarf2out_locviews_in_attribute): New.
2699 (dwarf2out_locviews_in_loclist): New.
2700 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2701 (enum dw_line_info_opcode): Add LI_adv_address.
2702 (struct dw_line_info_table): Add view.
2703 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2704 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2705 (zero_view_p): New variable.
2706 (ZERO_VIEW_P): New macro.
2707 (output_asm_line_debug_info): New.
2708 (struct var_loc_node): Add view.
2709 (add_AT_view_list, AT_loc_list): New.
2710 (add_var_loc_to_decl): Add view param. Test it against last.
2711 (new_loc_list): Add view params. Record them.
2712 (AT_loc_list_ptr): Handle loc and view lists.
2713 (view_list_to_loc_list_val_node): New.
2714 (print_dw_val): Handle dw_val_class_view_list.
2715 (size_of_die): Likewise.
2716 (value_format): Likewise.
2717 (loc_list_has_views): New.
2718 (gen_llsym): Set vl_symbol too.
2719 (maybe_gen_llsym, skip_loc_list_entry): New.
2720 (dwarf2out_maybe_output_loclist_view_pair): New.
2721 (output_loc_list): Output view list or entries too.
2722 (output_view_list_offset): New.
2723 (output_die): Handle dw_val_class_view_list.
2724 (output_dwarf_version): New.
2725 (output_compilation_unit_header): Use it.
2726 (output_skeleton_debug_sections): Likewise.
2727 (output_rnglists, output_line_info): Likewise.
2728 (output_pubnames, output_aranges): Update version comments.
2729 (output_one_line_info_table): Output view numbers in asm comments.
2730 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2731 Call maybe_gen_llsym.
2732 (loc_list_from_tree_1): Adjust.
2733 (add_AT_location_description): Create view list attribute if
2734 needed, check it's absent otherwise.
2735 (convert_cfa_to_fb_loc_list): Adjust.
2736 (maybe_emit_file): Call output_asm_line_debug_info for test.
2737 (dwarf2out_var_location): Reset views as needed. Precompute
2738 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2739 attribute. Set view.
2740 (new_line_info_table): Reset next view.
2741 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2742 (dwarf2out_source_line): Likewise. Output view resets and labels to
2743 the assembler, or select appropriate line info opcodes.
2744 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2745 (optimize_string_length): Catch it. Adjust.
2746 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2747 dw_val_class_view_list, and remove it if no longer needed.
2748 (hash_loc_list): Hash view numbers.
2749 (loc_list_hasher::equal): Compare them.
2750 (optimize_location_lists): Check whether a view list symbol is
2751 needed, and whether the locview attribute is present, and
2752 whether they match. Remove the locview attribute if no longer
2753 needed.
2754 (index_location_lists): Call skip_loc_list_entry for test.
2755 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2756 Use output_dwarf_version.
2757 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2758 (struct dw_val_node): Add val_view_list.
2759 * final.c (SEEN_NEXT_VIEW): New.
2760 (set_next_view_needed): New.
2761 (clear_next_view_needed): New.
2762 (maybe_output_next_view): New.
2763 (final_start_function): Rename to...
2764 (final_start_function_1): ... this. Take pointer to FIRST,
2765 add SEEN parameter. Emit param bindings in the initial view.
2766 (final_start_function): Reintroduce SEEN-less interface.
2767 (final): Rename to...
2768 (final_1): ... this. Take SEEN parameter. Output final pending
2769 next view at the end.
2770 (final): Reintroduce seen-less interface.
2771 (final_scan_insn): Output pending next view before switching
2772 sections or ending a block. Mark the next view as needed when
2773 outputting variable locations. Notify debug backend of section
2774 changes, and of location view changes.
2775 (rest_of_handle_final): Adjust.
2776 * toplev.c (process_options): Autodetect value for debug variable
2777 location views option. Warn on incompat5 without -gdwarf-5.
2778 * doc/invoke.texi (gvariable-location-views): New.
2779 (gvariable-location-views=incompat5): New.
2780 (gno-variable-location-views): New.
2782 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2784 PR tree-optimization/84136
2785 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2786 that the result of find_edge is non-NULL.
2788 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2790 PR target/83008
2791 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2792 storing integer register in SImode. Fix cost of 256 and 512
2793 byte aligned SSE register store.
2795 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2797 * config/i386/i386.c (ix86_multiplication_cost): Fix
2798 multiplication cost for TARGET_AVX512DQ.
2800 2018-02-08 Marek Polacek <polacek@redhat.com>
2802 PR tree-optimization/84238
2803 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2804 get_range_strlen.
2806 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2808 PR tree-optimization/84265
2809 * tree-vect-stmts.c (vectorizable_store): Don't treat
2810 VMAT_CONTIGUOUS accesses as grouped.
2811 (vectorizable_load): Likewise.
2813 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2815 PR tree-optimization/81635
2816 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2817 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2818 (test_round_for_mask): New functions.
2819 (wide_int_cc_tests): Call test_round_for_mask.
2820 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2821 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2822 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2823 range returned by get_range_info.
2825 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2827 PR ipa/81360
2828 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2829 * symtab.c: Include builtins.h
2830 (symtab_node::output_to_lto_symbol_table_p): Move here
2831 from lto-streamer-out.c:output_symbol_p.
2832 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2833 (output_symbol_p): Move all logic to symtab.c
2834 (produce_symtab): Update.
2836 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2838 * config/s390/s390-opts.h (enum indirect_branch): Define.
2839 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2840 (s390_indirect_branch_via_thunk)
2841 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2842 (enum s390_indirect_branch_type): Define.
2843 * config/s390/s390.c (struct s390_frame_layout, struct
2844 machine_function): Remove.
2845 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2846 (indirect_branch_table_label_no, indirect_branch_table_name):
2847 Define variables.
2848 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2849 (enum s390_indirect_branch_option): Define.
2850 (s390_return_addr_from_memory): New function.
2851 (s390_handle_string_attribute): New function.
2852 (s390_attribute_table): Add new attribute handler.
2853 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2854 (s390_indirect_branch_via_thunk): New function.
2855 (s390_indirect_branch_via_inline_thunk): New function.
2856 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2857 sibling call optimization for non z10 compiles.
2858 (s390_emit_call): Force indirect branch target to be a single
2859 register. Add r1 clobber for non-z10 compiles.
2860 (s390_emit_epilogue): Emit return jump via return_use expander.
2861 (s390_reorg): Handle JUMP_INSNs as execute targets.
2862 (s390_option_override_internal): Perform validity checks for the
2863 new command line options.
2864 (s390_indirect_branch_attrvalue): New function.
2865 (s390_indirect_branch_settings): New function.
2866 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2867 (s390_output_indirect_thunk_function): New function.
2868 (s390_code_end): Implement target hook.
2869 (s390_case_values_threshold): Implement target hook.
2870 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2871 macros.
2872 * config/s390/s390.h (struct s390_frame_layout)
2873 (struct machine_function): Move here from s390.c.
2874 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2875 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2876 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2877 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2878 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2879 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2880 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2881 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2882 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2883 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2884 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2885 (mnemonic attribute): Add values which aren't recognized
2886 automatically.
2887 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2888 pattern for branch conversion. Fix mnemonic attribute.
2889 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2890 indirect branch via thunk if requested.
2891 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2892 ("*indirect_jump"): Disable for branch conversion using out of
2893 line thunks.
2894 ("indirect_jump_via_thunk<mode>_z10")
2895 ("indirect_jump_via_thunk<mode>")
2896 ("indirect_jump_via_inlinethunk<mode>_z10")
2897 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2898 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2899 ("casesi_jump_via_inlinethunk<mode>_z10")
2900 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2901 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2902 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2903 ("*indirect2_jump"): Disable for branch conversion.
2904 ("casesi_jump"): Turn into expander and expand patterns for branch
2905 conversion.
2906 ("return_use"): New expander.
2907 ("*return"): Emit return via thunk and rename it to ...
2908 ("*return<mode>"): ... this one.
2909 * config/s390/s390.opt: Add new options and and enum for the
2910 option values.
2912 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2914 * lra-constraints.c (match_reload): Unconditionally use
2915 gen_lowpart_SUBREG, rather than selecting between that
2916 and equivalent gen_rtx_SUBREG code.
2918 2018-02-08 Richard Biener <rguenther@suse.de>
2920 PR tree-optimization/84233
2921 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2922 changed flag instead of boguously re-using phi_inserted.
2924 2018-02-08 Martin Jambor <mjambor@suse.cz>
2926 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2927 static local variables.
2929 2018-02-08 Richard Biener <rguenther@suse.de>
2931 PR tree-optimization/84278
2932 * tree-vect-stmts.c (vectorizable_store): When looking for
2933 smaller vector types to perform grouped strided loads/stores
2934 make sure the mode is supported by the target.
2935 (vectorizable_load): Likewise.
2937 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2939 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2940 Increase LDP/STP opportunities by adding adjacent callee-saves.
2942 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2944 PR rtl-optimization/84068
2945 PR rtl-optimization/83459
2946 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2948 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2950 PR tree-optimization/84224
2951 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2952 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2953 non-zero arguments.
2955 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2957 PR target/84113
2958 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2959 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2960 count, remove one USE.
2962 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2964 * doc/install.texi (Configuration): Document the
2965 --with-long-double-format={ibm,ieee} PowerPC configuration
2966 options.
2968 PR target/84154
2969 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2970 Convert from define_expand to be define_insn_and_split. Rework
2971 float/double/_Float128 conversions to QI/HI/SImode to work with
2972 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2973 conversions to QI/HImode types did a store and then a load to
2974 truncate the value. For conversions to VSX registers, don't split
2975 the insn, instead emit the code directly. Use the code iterator
2976 any_fix to combine signed and unsigned conversions.
2977 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2978 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2979 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2980 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2981 (fix_<mode>di2_hw): Likewise.
2982 (fixuns_<mode>di2_hw): Likewise.
2983 (fix_<mode>si2_hw): Likewise.
2984 (fixuns_<mode>si2_hw): Likewise.
2985 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2986 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2987 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2988 fix<uns>_trunc<SFDF:mode>si2_p8.
2989 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2990 used.
2991 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2992 (fix<uns>_<mode>_mem): Likewise.
2993 (fctiw<u>z_<mode>_mem): Likewise.
2994 (fix<uns>_<mode>_mem): Likewise.
2995 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2996 the register allocator from doing a direct move to the GPRs to do
2997 a store, and instead use the ISA 3.0 store byte/half-word from
2998 vector register instruction. For IEEE 128-bit floating point,
2999 also optimize stores of 32-bit ints.
3000 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3002 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3004 * genextract.c (push_pathstr_operand): New function to support
3005 [a-zA-Z].
3006 (walk_rtx): Call push_pathstr_operand.
3007 (print_path): Support [a-zA-Z].
3009 2018-02-07 Richard Biener <rguenther@suse.de>
3011 PR tree-optimization/84037
3012 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3013 (cse_and_gimplify_to_preheader): Declare.
3014 (vect_get_place_in_interleaving_chain): Likewise.
3015 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3016 ivexpr_map.
3017 (_loop_vec_info::~_loop_vec_info): Delete it.
3018 (cse_and_gimplify_to_preheader): New function.
3019 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3020 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3021 (vectorizable_load): Likewise. For grouped stores always base
3022 the IV on the first element.
3023 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3024 condition before gimplifying.
3026 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3028 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3029 *DIV_EXPR and *MOD_EXPR.
3031 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3033 PR target/84248
3034 * config/i386/i386.c (ix86_option_override_internal): Mask out
3035 the CF_SET bit when checking -fcf-protection.
3037 2018-02-07 Tom de Vries <tom@codesourcery.com>
3039 PR libgomp/84217
3040 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3041 enough.
3043 2018-02-07 Richard Biener <rguenther@suse.de>
3045 PR tree-optimization/84204
3046 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3047 this place.
3049 PR tree-optimization/84205
3050 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3051 special-case isl_ast_op_zdiv_r.
3053 PR tree-optimization/84223
3054 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3055 Only add conditions from within the region.
3056 (gather_bbs::after_dom_children): Adjust.
3058 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3060 PR target/84209
3061 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3062 * config/avr/avr.md: Only post-reload split REG-REG moves if
3063 either register is GENERAL_REG_P.
3065 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3067 PR tree-optimization/84235
3068 * tree-ssa-scopedtables.c
3069 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3070 if the subtraction is performed in floating point type where NaNs are
3071 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3072 build 1. Formatting fix.
3074 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3076 PR target/84146
3077 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3078 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3079 and skip it regardless of bb boundaries. Use CALL_P macro,
3080 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3081 unnecessarily, formatting fix.
3083 2018-02-06 Michael Collison <michael.collison@arm.com>
3085 * config/arm/thumb2.md:
3086 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3087 (*thumb_mov_notscc): Ditto.
3089 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3091 PR target/84154
3092 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3093 unsigned_fix, not "s".
3095 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3097 * configure.ac (gcc_fn_eh_frame_ro): New function.
3098 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3099 correct .eh_frame permissions.
3100 * configure: Regenerate.
3102 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3104 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3105 irrelevant options.
3107 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3109 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3110 Display warning message for -mno-speculate-indirect-jumps.
3112 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3114 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3115 Undocumented.
3116 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3118 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3120 PR tree-optimization/84225
3121 * tree-eh.c (find_trapping_overflow): Only call
3122 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3124 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3126 PR target/84145
3127 * config/i386/i386.c: Reimplement the check of possible options
3128 -mibt/-mshstk conbination. Change error messages.
3129 * doc/invoke.texi: Fix a typo: remove extra '='.
3131 2018-02-06 Marek Polacek <polacek@redhat.com>
3133 PR tree-optimization/84228
3134 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3136 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3138 PR target/82641
3139 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3140 emitted arch directives.
3141 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3142 __ARM_FEATURE_COPROC before changing architectures.
3144 2018-02-06 Richard Biener <rguenther@suse.de>
3146 * config/i386/i386.c (print_reg): Fix typo.
3147 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3149 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3151 * configure: Regenerate.
3153 2018-02-05 Martin Sebor <msebor@redhat.com>
3155 PR tree-optimization/83369
3156 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3157 inlining context.
3159 2018-02-05 Martin Liska <mliska@suse.cz>
3161 * doc/invoke.texi: Cherry-pick upstream r323995.
3163 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3165 * ira.c (ira_init_register_move_cost): Adjust comment.
3167 2018-02-05 Martin Liska <mliska@suse.cz>
3169 PR gcov-profile/84137
3170 * doc/gcov.texi: Fix typo in documentation.
3172 2018-02-05 Martin Liska <mliska@suse.cz>
3174 PR gcov-profile/83879
3175 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3176 using dlopen functionality.
3178 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3180 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3181 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3182 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3183 _mm_maskz_range_round_ss): New intrinsics.
3184 (__builtin_ia32_rangesd128_round)
3185 (__builtin_ia32_rangess128_round): Remove.
3186 (__builtin_ia32_rangesd128_mask_round,
3187 __builtin_ia32_rangess128_mask_round): New builtins.
3188 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3189 __builtin_ia32_rangess128_round): Remove.
3190 (__builtin_ia32_rangesd128_mask_round,
3191 __builtin_ia32_rangess128_mask_round): New builtins.
3192 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3193 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3194 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3195 "<round_saeonly_constraint>")): Changed to ...
3196 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3197 "<round_saeonly_scalar_constraint>")): ... this.
3198 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3199 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3200 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3201 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3202 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3204 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3206 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3207 options.
3208 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3209 Remove all values except native, 8540 and 8548.
3211 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3213 * config/i386/i386.c (ix86_output_function_return): Pass
3214 INVALID_REGNUM, instead of -1, as invalid register number to
3215 indirect_thunk_name and output_indirect_thunk.
3217 2018-02-02 Julia Koval <julia.koval@intel.com>
3219 * config.gcc: Add -march=icelake.
3220 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3221 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3222 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3223 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3224 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3225 (processor_target_table): Add icelake.
3226 (ix86_option_override_internal): Handle new PTAs.
3227 (get_builtin_code_for_version): Handle icelake.
3228 (M_INTEL_COREI7_ICELAKE): New.
3229 (fold_builtin_cpu): Handle icelake.
3230 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3231 * doc/invoke.texi: Add -march=icelake.
3233 2018-02-02 Julia Koval <julia.koval@intel.com>
3235 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3236 to wide_int_bitmask.
3237 * wide-int-bitmask.h: New.
3239 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3241 PR target/84066
3242 * config/i386/i386.md: Replace Pmode with word_mode in
3243 builtin_setjmp_setup and builtin_longjmp to support x32.
3245 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3247 PR target/56010
3248 PR target/83743
3249 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3250 #include "opts.h".
3251 (rs6000_supported_cpu_names): New static variable.
3252 (linux_cpu_translation_table): Likewise.
3253 (elf_platform) <cpu>: Define new static variable and use it.
3254 Translate kernel AT_PLATFORM name to canonical name if needed.
3255 Error if platform name is unknown.
3257 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3259 PR target/84089
3260 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3262 2018-02-01 Jeff Law <law@redhat.com>
3264 PR target/84128
3265 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3266 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3267 the scratch if RELEASE_VIA_POP is false.
3268 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3269 If we have to save a temporary register, decrement SIZE appropriately.
3270 Pass new arguments to release_scratch_register_on_entry.
3271 (ix86_adjust_stack_and_probe): Likewise.
3272 (ix86_emit_probe_stack_range): Pass new arguments to
3273 release_scratch_register_on_entry.
3275 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3277 PR rtl-optimization/84157
3278 * combine.c (change_zero_ext): Use REG_P predicate in
3279 front of HARD_REGISTER_P predicate.
3281 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3283 * config/avr/avr.c (avr_option_override): Move disabling of
3284 -fdelete-null-pointer-checks to...
3285 * common/config/avr/avr-common.c (avr_option_optimization_table):
3286 ...here.
3288 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3290 PR tree-optimization/81635
3291 * tree-data-ref.c (split_constant_offset_1): For types that
3292 wrap on overflow, try to use range info to prove that wrapping
3293 cannot occur.
3295 2018-02-01 Renlin Li <renlin.li@arm.com>
3297 PR target/83370
3298 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3299 TAILCALL_ADDR_REGS.
3300 (aarch64_register_move_cost): Likewise.
3301 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3302 TAILCALL_ADDR_REGS.
3303 (REG_CLASS_NAMES): Likewise.
3304 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3305 TAILCALL_ADDR_REGS. Remove IP registers.
3306 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3308 2018-02-01 Richard Biener <rguenther@suse.de>
3310 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3311 for specifying RPO order and allow NULL for that.
3312 * domwalk.c (dom_walker::dom_walker): Likewise.
3313 (dom_walker::walk): Handle NULL RPO order.
3314 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3315 in RPO order.
3316 (rewrite_update_dom_walker): Likewise.
3317 (mark_def_dom_walker): Likewise.
3319 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3321 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3322 (aarch64_maybe_expand_sve_subreg_move): Declare.
3323 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3324 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3325 predicate.
3326 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3327 that are semantically a reverse operation.
3328 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3329 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3330 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3331 functions.
3332 (aarch64_can_change_mode_class): For big-endian, forbid changes
3333 between two SVE modes if they have different element sizes.
3335 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3337 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3338 the TImode handling for big-endian targets.
3340 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3342 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3343 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3344 not just bytes.
3345 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3346 Remove BSWAP handing for big-endian targets and use the form of
3347 LD1RQ appropariate for the mode.
3349 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3351 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3352 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3353 duplicated element.
3355 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3357 PR tearget/83845
3358 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3359 check for operands that need to go through aarch64_sve_reload_be.
3361 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3363 PR tree-optimization/81661
3364 PR tree-optimization/84117
3365 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3366 * tree-eh.c: Include gimplify.h.
3367 (find_trapping_overflow, replace_trapping_overflow,
3368 rewrite_to_non_trapping_overflow): New functions.
3369 * tree-vect-loop.c: Include tree-eh.h.
3370 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3371 * tree-data-ref.c: Include tree-eh.h.
3372 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3374 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3376 PR rtl-optimization/84123
3377 * combine.c (change_zero_ext): Check if hard register satisfies
3378 can_change_dest_mode before calling gen_lowpart_SUBREG.
3380 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3382 PR target/82444
3383 * ira.c (ira_init_register_move_cost): Remove assert.
3385 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3387 PR rtl-optimization/84071
3388 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3389 * doc/tm.texi: Regenerate.
3391 2018-01-31 Richard Biener <rguenther@suse.de>
3393 PR tree-optimization/84132
3394 * tree-data-ref.c (analyze_miv_subscript): Properly
3395 check whether evolution_function_is_affine_multivariate_p
3396 before calling gcd_of_steps_may_divide_p.
3398 2018-01-31 Julia Koval <julia.koval@intel.com>
3400 PR target/83618
3401 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3402 * config/i386/i386.md (rdpid_rex64) New.
3403 (rdpid): Make 32bit only.
3405 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3407 PR lto/84105
3408 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3409 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3411 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3413 Revert
3414 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3416 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3418 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3420 PR rtl-optimization/84071
3421 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3422 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3424 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3426 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3427 (arc_attribute_table): Add 'aux' attribute.
3428 (arc_in_small_data_p): Consider aux like variables.
3429 (arc_is_aux_reg_p): New function.
3430 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3431 (arc_get_aux_arg): New function.
3432 (prepare_move_operands): Handle aux-register access.
3433 (arc_handle_aux_attribute): New function.
3434 * doc/extend.texi (ARC Variable attributes): Add subsection.
3436 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3438 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3439 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3440 (arc_attribute_table): Add 'uncached' attribute.
3441 (arc_print_operand): Print '.di' flag for uncached memory
3442 accesses.
3443 (arc_in_small_data_p): Do not consider for small data the uncached
3444 types.
3445 (arc_is_uncached_mem_p): New function.
3446 * config/arc/predicates.md (compact_store_memory_operand): Check
3447 for uncached memory accesses.
3448 (nonvol_nonimm_operand): Likewise.
3449 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3451 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3453 PR c/84100
3454 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3455 falign-loops=): Add Optimization flag.
3457 2018-01-30 Jeff Law <law@redhat.com>
3459 PR target/84064
3460 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3461 INT_REGISTERS_SAVED. Check it prior to calling
3462 get_scratch_register_on_entry.
3463 (ix86_adjust_stack_and_probe): Similarly.
3464 (ix86_emit_probe_stack_range): Similarly.
3465 (ix86_expand_prologue): Corresponding changes.
3467 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3469 PR target/40411
3470 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3471 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3473 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3475 PR target/84112
3476 * lra-constraints.c (curr_insn_transform): Process AND in the
3477 address.
3479 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3481 PR rtl-optimization/83986
3482 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3483 dependence against last_pending_memory_flush in addition to
3484 pending_jump_insns.
3486 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3488 PR tree-optimization/81611
3489 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3490 copies.
3492 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3494 PR target/83758
3495 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3496 a reg rtx.
3498 2018-01-30 Richard Biener <rguenther@suse.de>
3499 Jakub Jelinek <jakub@redhat.com>
3501 PR tree-optimization/84111
3502 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3503 inner loops added during recursion, as they don't have up-to-date
3504 SSA form.
3506 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3508 PR ipa/81360
3509 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3510 (can_inline_edge_by_limits_p): ... here.
3511 (can_early_inline_edge_p, check_callers,
3512 update_caller_keys, update_callee_keys, recursive_inlining,
3513 add_new_edges_to_heap, speculation_useful_p,
3514 inline_small_functions,
3515 inline_small_functions, flatten_function,
3516 inline_to_all_callers_1): Update.
3518 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3520 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3521 zeros correctly.
3523 2018-01-30 Richard Biener <rguenther@suse.de>
3525 PR tree-optimization/83008
3526 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3527 invariant and constant vector uses in stmts when they need
3528 more than one stmt.
3530 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3532 PR bootstrap/84017
3533 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3534 * configure: Regenerate.
3536 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3538 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3539 pattern.
3540 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3541 Use gen_rtx_REG rather than gen_lowpart.
3543 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3545 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3546 rather than 0 when creating partial subregs.
3548 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3550 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3551 of usage.
3553 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3555 PR target/81550
3556 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3557 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3558 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3559 flags. This restores the settings used before the 2017-07-24.
3560 Turning off pre increment/decrement/modify allows IVOPTS to
3561 optimize DF/SF loops where the index is an int.
3563 2018-01-29 Richard Biener <rguenther@suse.de>
3564 Kelvin Nilsen <kelvin@gcc.gnu.org>
3566 PR bootstrap/80867
3567 * tree-vect-stmts.c (vectorizable_call): Don't call
3568 targetm.vectorize_builtin_md_vectorized_function if callee is
3569 NULL.
3571 2018-01-22 Carl Love <cel@us.ibm.com>
3573 * doc/extend.tex: Fix typo in second arg in
3574 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3576 2018-01-29 Richard Biener <rguenther@suse.de>
3578 PR tree-optimization/84086
3579 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3580 (flush_ssaname_freelist): When SSA names were released reset
3581 the SCEV hash table.
3583 2018-01-29 Richard Biener <rguenther@suse.de>
3585 PR tree-optimization/84057
3586 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3587 removed paths when removing edges.
3589 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3591 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3592 -mfunction-return=@var{choice}.
3594 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3596 PR diagnostic/84034
3597 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3598 Handle CR like TAB.
3599 (layout::print_source_line): Likewise.
3600 (test_get_line_width_without_trailing_whitespace): Add test cases.
3602 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3604 PR middle-end/84040
3605 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3606 debug insns.
3608 2018-01-26 Jim Wilson <jimw@sifive.com>
3610 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3612 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3613 specified.
3615 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3617 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3618 and CMP + SUB-immediate -> SUBS.
3620 2018-01-26 Martin Sebor <msebor@redhat.com>
3622 PR tree-optimization/83896
3623 * tree-ssa-strlen.c (get_string_len): Rename...
3624 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3625 Avoid assuming length is constant.
3626 (handle_char_store): Use HOST_WIDE_INT for string length.
3628 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3630 PR target/81763
3631 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3632 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3634 2018-01-26 Richard Biener <rguenther@suse.de>
3636 PR rtl-optimization/84003
3637 * dse.c (record_store): Only record redundant stores when
3638 the earlier store aliases at least all accesses the later one does.
3640 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3642 PR rtl-optimization/83985
3643 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3644 REG_CFA_RESTORE insns.
3645 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3646 REG_CFA_RESTORE insns here.
3648 PR c/83989
3649 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3650 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3652 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3654 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3655 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3656 (arc_init): Likewise.
3657 (arc_override_options): Likewise.
3658 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3659 value.
3660 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3661 support.
3662 * config/arc/arc.h (TARGET_DBNZ): Define.
3663 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3664 properly set the tune attribute.
3665 (dbnz): Use TARGET_DBNZ guard.
3666 * config/arc/arc.opt (mtune): Add core3 option.
3668 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3670 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3671 recognize new pic like addresses.
3672 (arc_delegitimize_address): Clean up.
3674 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3676 * config/arc/arc-arches.def: Option mrf16 valid for all
3677 architectures.
3678 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3679 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3680 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3681 * config/arc/arc-tables.opt: Regenerate.
3682 * config/arc/arc.c (arc_conditional_register_usage): Handle
3683 reduced register file case.
3684 (arc_file_start): Set must have build attributes.
3685 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3686 mrf16 option value.
3687 * config/arc/arc.opt (mrf16): Add new option.
3688 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3689 * config/arc/genmultilib.awk: Handle new mrf16 option.
3690 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3691 * config/arc/t-multilib: Regenerate.
3692 * doc/invoke.texi (ARC Options): Document mrf16 option.
3694 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3696 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3697 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3698 (arc_attribute_table): Add 'secure_call' attribute.
3699 (arc_print_operand): Print secure call operand.
3700 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3701 secure.
3702 (arc_is_secure_call_p): New function. * config/arc/arc.md
3703 (call_i): Add support for sjli instruction.
3704 (call_value_i): Likewise.
3705 * config/arc/constraints.md (Csc): New constraint.
3707 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3708 John Eric Martin <John.Martin@emmicro-us.com>
3710 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3711 * config/arc/arc.c (_arc_jli_section): New struct.
3712 (arc_jli_section): New type.
3713 (rc_jli_sections): New static variable.
3714 (arc_handle_jli_attribute): New function.
3715 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3716 (arc_file_end): New function.
3717 (TARGET_ASM_FILE_END): Define.
3718 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3719 (arc_add_jli_section): New function.
3720 (jli_call_scan): Likewise.
3721 (arc_reorg): Call jli_call_scan.
3722 (arc_output_addsi): Remove 'S' from printing asm operand.
3723 (arc_is_jli_call_p): New function.
3724 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3725 operand.
3726 (movhi_insn): Likewise.
3727 (movsi_insn): Likewise.
3728 (movsi_set_cc_insn): Likewise.
3729 (loadqi_update): Likewise.
3730 (load_zeroextendqisi_update): Likewise.
3731 (load_signextendqisi_update): Likewise.
3732 (loadhi_update): Likewise.
3733 (load_zeroextendhisi_update): Likewise.
3734 (load_signextendhisi_update): Likewise.
3735 (loadsi_update): Likewise.
3736 (loadsf_update): Likewise.
3737 (movsicc_insn): Likewise.
3738 (bset_insn): Likewise.
3739 (bxor_insn): Likewise.
3740 (bclr_insn): Likewise.
3741 (bmsk_insn): Likewise.
3742 (bicsi3_insn): Likewise.
3743 (cmpsi_cc_c_insn): Likewise.
3744 (movsi_ne): Likewise.
3745 (movsi_cond_exec): Likewise.
3746 (clrsbsi2): Likewise.
3747 (norm_f): Likewise.
3748 (normw): Likewise.
3749 (swap): Likewise.
3750 (divaw): Likewise.
3751 (flag): Likewise.
3752 (sr): Likewise.
3753 (kflag): Likewise.
3754 (ffs): Likewise.
3755 (ffs_f): Likewise.
3756 (fls): Likewise.
3757 (call_i): Remove 'S' asm letter, add jli instruction.
3758 (call_value_i): Likewise.
3759 * config/arc/arc.op (mjli-always): New option.
3760 * config/arc/constraints.md (Cji): New constraint.
3761 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3762 operand.
3763 (subsf3_fpx): Likewise.
3764 (mulsf3_fpx): Likewise.
3765 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3766 asm operand.
3767 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3768 function attrbutes.
3769 * doc/invoke.texi (ARC): Document mjli-always option.
3771 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3773 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3774 avoid addition with 0 and use incw and decw where possible.
3776 2018-01-26 Richard Biener <rguenther@suse.de>
3778 PR tree-optimization/81082
3779 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3780 association if it requires casting to unsigned.
3781 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3782 from fold_plusminus_mult_expr to catch important cases late when
3783 range info is available.
3785 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3787 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3788 * configure.ac (hidden_linkonce): New test.
3789 * configure: Regenerate.
3790 * config.in: Regenerate.
3792 2018-01-26 Julia Koval <julia.koval@intel.com>
3794 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3795 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3796 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3797 _mm_mask_bitshuffle_epi64_mask): Fix type.
3798 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3799 USI_FTYPE_V4DI_V4DI_USI): Remove.
3800 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3801 __builtin_ia32_vpshufbitqmb256_mask,
3802 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3803 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3804 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3806 2018-01-26 Alan Modra <amodra@gmail.com>
3808 PR target/84033
3809 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3810 UNSPEC_VBPERMQ. Sort other unspecs.
3812 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3814 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3816 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3818 PR middle-end/83055
3819 * predict.c (drop_profile): Do not push/pop cfun; update also
3820 node->count.
3821 (handle_missing_profiles): Fix logic looking for zero profiles.
3823 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3825 PR middle-end/83977
3826 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3827 on functions with #pragma omp declare simd or functions with simd
3828 attribute.
3829 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3830 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3831 Remove trailing \n from warning_at calls.
3833 2018-01-25 Tom de Vries <tom@codesourcery.com>
3835 PR target/84028
3836 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3837 for neutered workers.
3839 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3841 PR target/68467
3842 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3843 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3845 2018-01-24 Jeff Law <law@redhat.com>
3847 PR target/83994
3848 * i386.c (get_probe_interval): Move to earlier point.
3849 (ix86_compute_frame_layout): If -fstack-clash-protection and
3850 the frame is larger than the probe interval, then use pushes
3851 to save registers rather than reg->mem moves.
3852 (ix86_expand_prologue): Remove conditional for int_registers_saved
3853 assertion.
3855 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3857 PR target/84014
3858 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3859 min/max for never referenced object.
3861 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3863 PR middle-end/83977
3864 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3865 here.
3866 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3867 attributes from DECL_ATTRIBUTES (decl) without affecting
3868 DECL_ATTRIBUTES (current_function_decl).
3869 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3870 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3872 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3874 PR tree-optimization/83979
3875 * fold-const.c (fold_comparison): Use constant_boolean_node
3876 instead of boolean_{true,false}_node.
3878 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3880 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3881 with zero counts.
3883 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3885 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3886 Simplify the clause that sets the length attribute.
3887 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3888 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3889 clause that sets the length attribute.
3890 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3892 2018-01-24 Tom de Vries <tom@codesourcery.com>
3894 PR target/83589
3895 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3896 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3897 Add strict parameter.
3898 (prevent_branch_around_nothing): Insert dummy insn between branch to
3899 label and label with no ptx insn inbetween.
3900 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3902 2018-01-24 Tom de Vries <tom@codesourcery.com>
3904 PR target/81352
3905 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3906 for neutered threads in warp.
3907 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3909 2018-01-24 Richard Biener <rguenther@suse.de>
3911 PR tree-optimization/83176
3912 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3913 operands.
3915 2018-01-24 Richard Biener <rguenther@suse.de>
3917 PR tree-optimization/82819
3918 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3919 code generating pluses that are no-ops in the target precision.
3921 2018-01-24 Richard Biener <rguenther@suse.de>
3923 PR middle-end/84000
3924 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3926 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3928 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3929 to merge probabilities.
3930 * predict.c (probably_never_executed): Also mark as cold functions
3931 with global 0 profile and guessed local profile.
3932 * profile-count.c (profile_probability::combine_with_count): New
3933 member function.
3934 * profile-count.h (profile_probability::operator*,
3935 profile_probability::operator*=, profile_probability::operator/,
3936 profile_probability::operator/=): Reduce precision to adjusted
3937 and set value to guessed on contradictory divisions.
3938 (profile_probability::combine_with_freq): Remove.
3939 (profile_probability::combine_wiht_count): Declare.
3940 (profile_count::force_nonzero):: Set to adjusted.
3941 (profile_count::probability_in):: Set quality to adjusted.
3942 * tree-ssa-tail-merge.c (replace_block_by): Use
3943 combine_with_count.
3945 2018-01-23 Andrew Waterman <andrew@sifive.com>
3946 Jim Wilson <jimw@sifive.com>
3948 * config/riscv/riscv.c (riscv_stack_boundary): New.
3949 (riscv_option_override): Set riscv_stack_boundary. Handle
3950 riscv_preferred_stack_boundary_arg.
3951 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3952 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3953 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3954 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3955 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3956 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3958 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3960 PR target/83905
3961 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3962 of struct ix86_frame.
3963 (ix86_expand_epilogue): Likewise. Add a local variable for
3964 the reg_save_offset field in struct ix86_frame.
3966 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3968 PR tree-optimization/82604
3969 * tree-loop-distribution.c (enum partition_kind): New enum item
3970 PKIND_PARTIAL_MEMSET.
3971 (partition_builtin_p): Support above new enum item.
3972 (generate_code_for_partition): Ditto.
3973 (compute_access_range): Differentiate cases that equality can be
3974 proven at all loops, the innermost loops or no loops.
3975 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3976 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3977 (finalize_partitions, distribute_loop): Don't fuse partition of
3978 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3979 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3980 parloop is enabled.
3982 2018-01-23 Martin Liska <mliska@suse.cz>
3984 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3985 order to ignore the predictor.
3986 (PRED_POLYMORPHIC_CALL): Likewise.
3987 (PRED_RECURSIVE_CALL): Likewise.
3989 2018-01-23 Martin Liska <mliska@suse.cz>
3991 * tree-profile.c (tree_profiling): Print function header to
3992 aware reader which function we are working on.
3993 * value-prof.c (gimple_find_values_to_profile): Do not print
3994 not interesting value histograms.
3996 2018-01-23 Martin Liska <mliska@suse.cz>
3998 * profile-count.h (enum profile_quality): Add
3999 profile_uninitialized as the first value. Do not number values
4000 as they are zero based.
4001 (profile_count::verify): Update sanity check.
4002 (profile_probability::verify): Likewise.
4004 2018-01-23 Nathan Sidwell <nathan@acm.org>
4006 * doc/invoke.texi (ffor-scope): Deprecate.
4008 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4010 PR tree-optimization/83510
4011 * domwalk.c (set_all_edges_as_executable): New function.
4012 (dom_walker::dom_walker): Convert bool param
4013 "skip_unreachable_blocks" to enum reachability. Move setup of
4014 edge flags to set_all_edges_as_executable and only do it when
4015 reachability is REACHABLE_BLOCKS.
4016 * domwalk.h (enum dom_walker::reachability): New enum.
4017 (dom_walker::dom_walker): Convert bool param
4018 "skip_unreachable_blocks" to enum reachability.
4019 (set_all_edges_as_executable): New decl.
4020 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4021 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4022 "reachability".
4023 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4024 but converting true to REACHABLE_BLOCKS.
4025 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4026 * tree-vrp.c
4027 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4028 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4029 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4030 REACHABLE_BLOCKS.
4031 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4032 if check_all_array_refs will be called.
4034 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4036 * tree.c (selftest::test_location_wrappers): Add more test
4037 coverage.
4039 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4041 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4042 (selftest::test_bit_in_range): Likewise.
4044 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4046 PR testsuite/83888
4047 * doc/sourcebuild.texi (vect_float): Say that the selector
4048 only describes the situation when -funsafe-math-optimizations is on.
4049 (vect_float_strict): Document.
4051 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4053 PR tree-optimization/83965
4054 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4055 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4056 instead of checking only for a reduction.
4057 (vect_recog_widen_sum_pattern): Likewise.
4059 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4061 * predict.c (probably_never_executed): Only use precise profile info.
4062 (compute_function_frequency): Skip after inlining hack since we now
4063 have quality checking.
4065 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4067 * profile-count.h (profile_probability::very_unlikely,
4068 profile_probability::unlikely, profile_probability::even): Set
4069 precision to guessed.
4071 2018-01-23 Richard Biener <rguenther@suse.de>
4073 PR tree-optimization/83963
4074 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4075 Properly terminate dominator walk when crossing the exit edge not
4076 when visiting its source block.
4078 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4080 PR c++/83918
4081 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4082 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4084 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4086 PR tree-optimization/83957
4087 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4088 semicolon after for body surrounded by braces.
4090 PR tree-optimization/83081
4091 * profile-count.h (profile_probability::split): New method.
4092 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4093 Use profile_probability::split.
4094 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4095 when splitting a single conditional jump into 2.
4097 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4099 PR tree-optimization/69452
4100 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4101 decl.
4103 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4105 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4106 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4107 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4109 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4111 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4112 * config/rl78/rl78.md (movdi): New define_expand.
4113 * config/rl78/rl78.c (rl78_split_movdi): New function.
4115 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4117 PR target/83862
4118 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4119 no longer used.
4120 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4121 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4122 128-bit to produce an UNSPEC move to get the double word with the
4123 signbit and then a shift directly to do signbit.
4124 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4125 implementation with a new version that just does either a direct
4126 move or a regular move. Move memory interface to separate insns.
4127 Move insns so they are next to the expander.
4128 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4129 with signbit move. Split big and little endian case.
4130 (signbit<mode>2_dm_mem_le): Likewise.
4131 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4132 (signbit<mode>2_dm2): Likewise.
4134 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4136 * config/rl78/rl78.md (anddi3): New define_expand.
4138 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4140 * config/rl78/rl78.md (umindi3): New define_expand.
4142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4144 * config/rl78/rl78.md (smindi3): New define_expand.
4146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4148 * config/rl78/rl78.md (smaxdi3): New define_expand.
4150 2018-01-22 Carl Love <cel@us.ibm.com>
4152 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4153 LVX_V1TI): Add macro expansion.
4154 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4155 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4156 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4157 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4158 Change check to determine if the instruction is a byte reversing
4159 entry. Fix typo in comment.
4160 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4161 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4162 Add def_builtin calls for new builtins.
4163 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4164 Add define_insn expansion.
4166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4168 * config/rl78/rl78.md (umaxdi3): New define_expand.
4170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4172 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4173 for non-QImode registers.
4175 2018-01-22 Richard Biener <rguenther@suse.de>
4177 PR tree-optimization/83963
4178 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4179 including the loop exit block.
4180 (scop_detection::merge_sese): Likewise.
4181 (scop_detection::add_scop): Do it here instead.
4183 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4185 * doc/sourcebuild.texi (arm_softfloat): Document.
4187 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4189 PR gcc/77734
4190 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4191 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4192 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4194 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4195 David Edelsohn <dje.gcc@gmail.com>
4197 PR target/83946
4198 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4199 Change "crset eq" to "crset 2".
4200 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4201 (*call_indirect_aix<mode>_nospec): Likewise.
4202 (*call_value_indirect_aix<mode>_nospec): Likewise.
4203 (*call_indirect_elfv2<mode>_nospec): Likewise.
4204 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4205 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4206 change assembly output from . to $.
4207 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4208 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4209 (*tablejump<mode>_internal1_nospec): Likewise.
4211 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4213 PR target/80870
4214 * config/sh/sh_optimize_sett_clrt.cc:
4215 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4217 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4219 PR tree-optimization/83940
4220 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4221 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4222 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4223 use it to pass back the definition type.
4224 (vect_check_store_rhs): Likewise rhs_dt_out.
4225 (vect_build_gather_load_calls): Add a mask_dt argument and use
4226 it instead of a call to vect_is_simple_use.
4227 (vectorizable_store): Update calls to vect_check_load_store_mask
4228 and vect_check_store_rhs. Use the dt returned by the latter instead
4229 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4230 instead of calls to vect_is_simple_use. Pass the scalar rather
4231 than the vector operand to vect_is_simple_use when handling
4232 second and subsequent copies of an rhs value.
4233 (vectorizable_load): Update calls to vect_check_load_store_mask
4234 and vect_build_gather_load_calls. Use the cached mask_dt and
4235 gs_info.offset_dt instead of calls to vect_is_simple_use.
4237 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4239 PR middle-end/83945
4240 * tree-emutls.c: Include gimplify.h.
4241 (lower_emutls_2): New function.
4242 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4243 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4244 it before further processing.
4246 PR target/83930
4247 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4248 UINTVAL (trueop1) instead of INTVAL (op1).
4250 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4252 PR debug/81570
4253 PR debug/83728
4254 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4255 INCOMING_FRAME_SP_OFFSET if not defined.
4256 (scan_trace): Add ENTRY argument. If true and
4257 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4258 emit a note to adjust the CFA offset.
4259 (create_cfi_notes): Adjust scan_trace callers.
4260 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4261 INCOMING_FRAME_SP_OFFSET in the CIE.
4262 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4263 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4264 Likewise.
4265 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4266 * doc/tm.texi: Regenerated.
4268 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4270 PR rtl-optimization/83147
4271 * lra-constraints.c (remove_inheritance_pseudos): Use
4272 lra_substitute_pseudo_within_insn.
4274 2018-01-19 Tom de Vries <tom@codesourcery.com>
4275 Cesar Philippidis <cesar@codesourcery.com>
4277 PR target/83920
4278 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4280 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4282 PR target/83790
4283 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4284 spaces for function labels.
4286 2018-01-19 Martin Liska <mliska@suse.cz>
4288 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4289 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4290 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4291 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4292 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4293 (PRED_CONST_RETURN): Change from 69 to 65.
4294 (PRED_NULL_RETURN): Change from 91 to 71.
4295 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4296 (PRED_LOOP_GUARD): Change from 66 to 73.
4298 2018-01-19 Martin Liska <mliska@suse.cz>
4300 * predict.c (predict_insn_def): Add new assert.
4301 (struct branch_predictor): Change type to signed integer.
4302 (test_prediction_value_range): Amend test to cover
4303 PROB_UNINITIALIZED.
4304 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4305 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4306 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4307 (PRED_LOOP_IV_COMPARE): Likewise.
4308 * predict.h (PROB_UNINITIALIZED): Define new constant.
4310 2018-01-19 Martin Liska <mliska@suse.cz>
4312 * predict.c (dump_prediction): Add new format for
4313 analyze_brprob.py script which is enabled with -details
4314 suboption.
4315 * profile-count.h (precise_p): New function.
4317 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4319 PR tree-optimization/83922
4320 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4321 there are no statements that need masking.
4322 (vect_active_double_reduction_p): New function.
4323 (vect_analyze_loop_operations): Use it when handling phis that
4324 are not in the loop header.
4326 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4328 PR tree-optimization/83914
4329 * tree-vect-loop.c (vectorizable_induction): Don't convert
4330 init_expr or apply the peeling adjustment for inductions
4331 that are nested within the vectorized loop.
4333 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4335 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4336 instead of NEG.
4338 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4340 PR sanitizer/81715
4341 PR testsuite/83882
4342 * function.h (gimplify_parameters): Add gimple_seq * argument.
4343 * function.c: Include gimple.h and options.h.
4344 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4345 for the added local temporaries if needed.
4346 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4347 if there are any parameter cleanups, wrap whole body into a
4348 try/finally with the cleanups.
4350 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4352 PR target/82964
4353 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4354 Use GET_MODE_CLASS for scalar floating point.
4356 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4358 PR ipa/82256
4359 patch by PaX Team
4360 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4361 Fix call of call_cgraph_insertion_hooks.
4363 2018-01-18 Martin Sebor <msebor@redhat.com>
4365 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4367 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4369 PR ipa/83619
4370 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4371 frequencies.
4373 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4375 PR other/70268
4376 * common.opt: (-ffile-prefix-map): New option.
4377 * opts.c (common_handle_option): Defer it.
4378 * opts-global.c (handle_common_deferred_options): Handle it.
4379 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4380 * file-prefix-map.h: New file.
4381 (remap_debug_filename, add_debug_prefix_map): ...here.
4382 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4383 * final.c (debug_prefix_map, add_debug_prefix_map
4384 remap_debug_filename): Move to...
4385 * file-prefix-map.c: New file.
4386 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4387 generalize, get rid of alloca(), use strrchr() instead of strchr().
4388 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4389 Implement in terms of add_prefix_map().
4390 (remap_macro_filename, remap_debug_filename): Implement in term of
4391 remap_filename().
4392 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4393 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4394 * dbxout.c: Include file-prefix-map.h.
4395 * varasm.c: Likewise.
4396 * vmsdbgout.c: Likewise.
4397 * xcoffout.c: Likewise.
4398 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4399 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4400 * doc/invoke.texi (-ffile-prefix-map): Document.
4401 (-fdebug-prefix-map): Update description.
4403 2018-01-18 Martin Liska <mliska@suse.cz>
4405 * config/i386/i386.c (indirect_thunk_name): Document that also
4406 lfence is emitted.
4407 (output_indirect_thunk): Document why both instructions
4408 (pause and lfence) are generated.
4410 2018-01-18 Richard Biener <rguenther@suse.de>
4412 PR tree-optimization/83887
4413 * graphite-scop-detection.c
4414 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4415 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4416 (scop_detection::merge_sese): Re-implement with a flood-fill
4417 algorithm that properly finds a SESE region if it exists.
4419 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4421 PR c/61240
4422 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4423 pointer_diff optimizations use view_convert instead of convert.
4425 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4427 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4428 Generate different code for -mno-speculate-indirect-jumps.
4429 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4430 (*call_indirect_aix<mode>): Disable for
4431 -mno-speculate-indirect-jumps.
4432 (*call_indirect_aix<mode>_nospec): New define_insn.
4433 (*call_value_indirect_aix<mode>): Disable for
4434 -mno-speculate-indirect-jumps.
4435 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4436 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4437 -mno-speculate-indirect-jumps.
4438 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4440 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4442 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4443 long double type, set the flags for noting the default long double
4444 type, even if we don't pass or return a long double type.
4446 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4448 PR ipa/83051
4449 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4450 failure.
4452 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4454 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4455 support for merge[hl].
4456 (fold_mergehl_helper): New helper function.
4457 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4458 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4459 (altivec_vmrglw_direct): Add xxmrglw insn.
4461 2018-01-17 Andrew Waterman <andrew@sifive.com>
4463 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4464 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4466 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4468 PR lto/83121
4469 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4470 call the lto_location_cache before reading the
4471 DECL_SOURCE_LOCATION of the types.
4473 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4474 Richard Sandiford <richard.sandiford@linaro.org>
4476 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4477 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4478 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4479 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4480 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4481 Add declaration.
4482 * config/aarch64/constraints.md (aarch64_movti_operand):
4483 Limit immediates.
4484 * config/aarch64/predicates.md (Uti): Add new constraint.
4486 2018-01-17 Carl Love <cel@us.ibm.com>
4488 * config/rs6000/vsx.md (define_expand xl_len_r,
4489 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4490 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4491 lxvll.
4492 (define_expand, define_insn): Move the shift left from the
4493 define_insn to the define_expand for lxvl and stxvl instructions.
4494 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4495 and XL_LEN_R definitions to PURE.
4497 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4499 * config/i386/i386.c (indirect_thunk_name): Declare regno
4500 as unsigned int. Compare regno with INVALID_REGNUM.
4501 (output_indirect_thunk): Ditto.
4502 (output_indirect_thunk_function): Ditto.
4503 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4504 in the call to output_indirect_thunk_function.
4506 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4508 PR middle-end/83884
4509 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4510 rather than the size of inner_type to determine the stack slot size
4511 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4513 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4515 PR target/83546
4516 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4517 to PTA_SILVERMONT.
4519 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4521 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4522 endian Linux systems to optionally enable multilibs for selecting
4523 the long double type if the user configured an explicit type.
4524 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4525 have no long double multilibs if not defined.
4526 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4527 warn if the user used -mabi={ieee,ibm}longdouble and we built
4528 multilibs for long double.
4529 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4530 appropriate multilib option.
4531 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4532 multilib options.
4533 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4534 for building long double multilibs.
4535 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4537 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4539 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4540 copies.
4542 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4543 64 bits.
4544 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4545 128 bits.
4547 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4548 variables.
4550 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4551 return value.
4553 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4556 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4558 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4560 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4561 different rtl trees depending on TARGET_64BIT.
4562 (rs6000_gen_lvx): Likewise.
4564 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4566 * config/visium/visium.md (nop): Tweak comment.
4567 (hazard_nop): Likewise.
4569 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4571 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4572 -mspeculate-indirect-jumps.
4573 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4574 for -mno-speculate-indirect-jumps.
4575 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4576 (*call_value_indirect_elfv2<mode>): Disable for
4577 -mno-speculate-indirect-jumps.
4578 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4579 (indirect_jump): Emit different RTL for
4580 -mno-speculate-indirect-jumps.
4581 (*indirect_jump<mode>): Disable for
4582 -mno-speculate-indirect-jumps.
4583 (*indirect_jump<mode>_nospec): New define_insn.
4584 (tablejump): Emit different RTL for
4585 -mno-speculate-indirect-jumps.
4586 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4587 (tablejumpsi_nospec): New define_expand.
4588 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4589 (tablejumpdi_nospec): New define_expand.
4590 (*tablejump<mode>_internal1): Disable for
4591 -mno-speculate-indirect-jumps.
4592 (*tablejump<mode>_internal1_nospec): New define_insn.
4593 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4594 option.
4596 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4598 * caller-save.c (insert_save): Drop unnecessary parameter. All
4599 callers updated.
4601 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4602 Richard Biener <rguenth@suse.de>
4604 PR libgomp/83590
4605 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4606 return early, inline manually is_gimple_sizepos. Make sure if we
4607 call gimplify_expr we don't end up with a gimple constant.
4608 * tree.c (variably_modified_type_p): Don't return true for
4609 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4610 * gimplify.h (is_gimple_sizepos): Remove.
4612 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4614 PR tree-optimization/83857
4615 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4616 vectorizable_live_operation for pure SLP statements.
4617 (vectorizable_live_operation): Handle PHIs.
4619 2018-01-16 Richard Biener <rguenther@suse.de>
4621 PR tree-optimization/83867
4622 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4623 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4625 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4627 PR c/83844
4628 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4629 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4630 If off is not INTEGER_CST, issue a may not be aligned warning
4631 rather than isn't aligned. Use isn%'t rather than isn't.
4632 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4633 into MULT_EXPR.
4634 <case MULT_EXPR>: Improve the case when bottom and one of the
4635 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4636 operand, in that case check if the other operand is multiple of
4637 bottom divided by the INTEGER_CST operand.
4639 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4641 PR target/83858
4642 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4643 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4644 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4645 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4646 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4647 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4648 (pa_function_arg_size): New function.
4650 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4652 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4653 in a separate statement.
4655 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4657 PR tree-optimization/83847
4658 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4659 group gathers and scatters.
4661 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4663 PR rtl-optimization/86620
4664 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4666 PR rtl-optimization/83213
4667 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4668 to last if both are JUMP_INSNs.
4670 PR tree-optimization/83843
4671 * gimple-ssa-store-merging.c
4672 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4673 store_immediate_info for bswap/nop orig_stores.
4675 2018-01-15 Andrew Waterman <andrew@sifive.com>
4677 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4678 !TARGET_MUL.
4679 <UDIV>: Increase cost if !TARGET_DIV.
4681 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4683 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4684 (define_attr "cr_logical_3op"): New.
4685 (cceq_ior_compare): Adjust.
4686 (cceq_ior_compare_complement): Adjust.
4687 (*cceq_rev_compare): Adjust.
4688 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4689 (is_cracked_insn): Adjust.
4690 (insn_must_be_first_in_group): Adjust.
4691 * config/rs6000/40x.md: Adjust.
4692 * config/rs6000/440.md: Adjust.
4693 * config/rs6000/476.md: Adjust.
4694 * config/rs6000/601.md: Adjust.
4695 * config/rs6000/603.md: Adjust.
4696 * config/rs6000/6xx.md: Adjust.
4697 * config/rs6000/7450.md: Adjust.
4698 * config/rs6000/7xx.md: Adjust.
4699 * config/rs6000/8540.md: Adjust.
4700 * config/rs6000/cell.md: Adjust.
4701 * config/rs6000/e300c2c3.md: Adjust.
4702 * config/rs6000/e500mc.md: Adjust.
4703 * config/rs6000/e500mc64.md: Adjust.
4704 * config/rs6000/e5500.md: Adjust.
4705 * config/rs6000/e6500.md: Adjust.
4706 * config/rs6000/mpc.md: Adjust.
4707 * config/rs6000/power4.md: Adjust.
4708 * config/rs6000/power5.md: Adjust.
4709 * config/rs6000/power6.md: Adjust.
4710 * config/rs6000/power7.md: Adjust.
4711 * config/rs6000/power8.md: Adjust.
4712 * config/rs6000/power9.md: Adjust.
4713 * config/rs6000/rs64.md: Adjust.
4714 * config/rs6000/titan.md: Adjust.
4716 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4718 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4719 ix86_indirect_branch_register logic.
4721 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4723 * config/i386/constraints.md (Bs): Update
4724 ix86_indirect_branch_register check. Don't check
4725 ix86_indirect_branch_register with GOT_memory_operand.
4726 (Bw): Likewise.
4727 * config/i386/predicates.md (GOT_memory_operand): Don't check
4728 ix86_indirect_branch_register here.
4729 (GOT32_symbol_operand): Likewise.
4731 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4733 * config/i386/predicates.md (constant_call_address_operand):
4734 Rewrite ix86_indirect_branch_register logic.
4735 (sibcall_insn_operand): Likewise.
4737 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4739 * config/i386/constraints.md (Bs): Replace
4740 ix86_indirect_branch_thunk_register with
4741 ix86_indirect_branch_register.
4742 (Bw): Likewise.
4743 * config/i386/i386.md (indirect_jump): Likewise.
4744 (tablejump): Likewise.
4745 (*sibcall_memory): Likewise.
4746 (*sibcall_value_memory): Likewise.
4747 Peepholes of indirect call and jump via memory: Likewise.
4748 * config/i386/i386.opt: Likewise.
4749 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4750 (GOT_memory_operand): Likewise.
4751 (call_insn_operand): Likewise.
4752 (sibcall_insn_operand): Likewise.
4753 (GOT32_symbol_operand): Likewise.
4755 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4757 PR middle-end/83837
4758 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4759 type rather than type addr's type points to.
4760 (expand_omp_atomic_mutex): Likewise.
4761 (expand_omp_atomic): Likewise.
4763 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4765 PR target/83839
4766 * config/i386/i386.c (output_indirect_thunk_function): Use
4767 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4768 for __x86_return_thunk.
4770 2018-01-15 Richard Biener <rguenther@suse.de>
4772 PR middle-end/83850
4773 * expmed.c (extract_bit_field_1): Fix typo.
4775 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4777 PR target/83687
4778 * config/arm/iterators.md (VF): New mode iterator.
4779 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4780 Remove integer-related logic from pattern.
4781 (neon_vabd<mode>_3): Likewise.
4783 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4785 PR middle-end/82694
4786 * common.opt (fstrict-overflow): No longer an alias.
4787 (fwrapv-pointer): New option.
4788 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4789 also for pointer types based on flag_wrapv_pointer.
4790 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4791 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4792 opts->x_flag_wrapv got set.
4793 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4794 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4795 POINTER_TYPE_OVERFLOW_UNDEFINED.
4796 * match.pd: Likewise in address comparison pattern.
4797 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4799 2018-01-15 Richard Biener <rguenther@suse.de>
4801 PR lto/83804
4802 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4803 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4804 Reset type names to their identifier if their TYPE_DECL doesn't
4805 have linkage (and thus is used for ODR and devirt).
4806 (save_debug_info_for_decl): Remove.
4807 (save_debug_info_for_type): Likewise.
4808 (add_tree_to_fld_list): Adjust.
4809 * tree-pretty-print.c (dump_generic_node): Make dumping of
4810 type names more robust.
4812 2018-01-15 Richard Biener <rguenther@suse.de>
4814 * BASE-VER: Bump to 8.0.1.
4816 2018-01-14 Martin Sebor <msebor@redhat.com>
4818 PR other/83508
4819 * builtins.c (check_access): Avoid warning when the no-warning bit
4820 is set.
4822 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4824 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4825 * ira-color (allocno_hard_regs_compare): Likewise.
4827 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4829 PR target/83013
4830 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4831 Use .pushsection/.popsection.
4833 2018-01-14 Martin Sebor <msebor@redhat.com>
4835 PR c++/81327
4836 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4838 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4840 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4841 entry from extra_headers.
4842 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4843 extra_headers, make the list bitwise identical to the i?86-*-* one.
4845 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4847 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4848 -mcmodel=large with -mindirect-branch=thunk,
4849 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4850 -mfunction-return=thunk-extern.
4851 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4852 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4853 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4855 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4857 * config/i386/i386.c (print_reg): Print the name of the full
4858 integer register without '%'.
4859 (ix86_print_operand): Handle 'V'.
4860 * doc/extend.texi: Document 'V' modifier.
4862 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4864 * config/i386/constraints.md (Bs): Disallow memory operand for
4865 -mindirect-branch-register.
4866 (Bw): Likewise.
4867 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4868 (GOT_memory_operand): Likewise.
4869 (call_insn_operand): Likewise.
4870 (sibcall_insn_operand): Likewise.
4871 (GOT32_symbol_operand): Likewise.
4872 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4873 for -mindirect-branch-register.
4874 (tablejump): Likewise.
4875 (*sibcall_memory): Likewise.
4876 (*sibcall_value_memory): Likewise.
4877 Disallow peepholes of indirect call and jump via memory for
4878 -mindirect-branch-register.
4879 (*call_pop): Replace m with Bw.
4880 (*call_value_pop): Likewise.
4881 (*sibcall_pop_memory): Replace m with Bs.
4882 * config/i386/i386.opt (mindirect-branch-register): New option.
4883 * doc/invoke.texi: Document -mindirect-branch-register option.
4885 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4887 * config/i386/i386-protos.h (ix86_output_function_return): New.
4888 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4889 set function_return_type.
4890 (indirect_thunk_name): Add ret_p to indicate thunk for function
4891 return.
4892 (output_indirect_thunk_function): Pass false to
4893 indirect_thunk_name.
4894 (ix86_output_indirect_branch_via_reg): Likewise.
4895 (ix86_output_indirect_branch_via_push): Likewise.
4896 (output_indirect_thunk_function): Create alias for function
4897 return thunk if regno < 0.
4898 (ix86_output_function_return): New function.
4899 (ix86_handle_fndecl_attribute): Handle function_return.
4900 (ix86_attribute_table): Add function_return.
4901 * config/i386/i386.h (machine_function): Add
4902 function_return_type.
4903 * config/i386/i386.md (simple_return_internal): Use
4904 ix86_output_function_return.
4905 (simple_return_internal_long): Likewise.
4906 * config/i386/i386.opt (mfunction-return=): New option.
4907 (indirect_branch): Mention -mfunction-return=.
4908 * doc/extend.texi: Document function_return function attribute.
4909 * doc/invoke.texi: Document -mfunction-return= option.
4911 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4913 * config/i386/i386-opts.h (indirect_branch): New.
4914 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4915 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4916 with local indirect jump when converting indirect call and jump.
4917 (ix86_set_indirect_branch_type): New.
4918 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4919 (indirectlabelno): New.
4920 (indirect_thunk_needed): Likewise.
4921 (indirect_thunk_bnd_needed): Likewise.
4922 (indirect_thunks_used): Likewise.
4923 (indirect_thunks_bnd_used): Likewise.
4924 (INDIRECT_LABEL): Likewise.
4925 (indirect_thunk_name): Likewise.
4926 (output_indirect_thunk): Likewise.
4927 (output_indirect_thunk_function): Likewise.
4928 (ix86_output_indirect_branch_via_reg): Likewise.
4929 (ix86_output_indirect_branch_via_push): Likewise.
4930 (ix86_output_indirect_branch): Likewise.
4931 (ix86_output_indirect_jmp): Likewise.
4932 (ix86_code_end): Call output_indirect_thunk_function if needed.
4933 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4934 needed.
4935 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4936 (ix86_attribute_table): Add indirect_branch.
4937 * config/i386/i386.h (machine_function): Add indirect_branch_type
4938 and has_local_indirect_jump.
4939 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4940 to true.
4941 (tablejump): Likewise.
4942 (*indirect_jump): Use ix86_output_indirect_jmp.
4943 (*tablejump_1): Likewise.
4944 (simple_return_indirect_internal): Likewise.
4945 * config/i386/i386.opt (mindirect-branch=): New option.
4946 (indirect_branch): New.
4947 (keep): Likewise.
4948 (thunk): Likewise.
4949 (thunk-inline): Likewise.
4950 (thunk-extern): Likewise.
4951 * doc/extend.texi: Document indirect_branch function attribute.
4952 * doc/invoke.texi: Document -mindirect-branch= option.
4954 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4956 PR ipa/83051
4957 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4959 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4961 * ipa-inline.c (want_inline_small_function_p): Return false if
4962 inlining has already failed with CIF_FINAL_ERROR.
4963 (update_caller_keys): Call want_inline_small_function_p before
4964 can_inline_edge_p.
4965 (update_callee_keys): Likewise.
4967 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4969 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4970 New function.
4971 (rs6000_quadword_masked_address_p): Likewise.
4972 (quad_aligned_load_p): Likewise.
4973 (quad_aligned_store_p): Likewise.
4974 (const_load_sequence_p): Add comment to describe the outer-most loop.
4975 (mimic_memory_attributes_and_flags): New function.
4976 (rs6000_gen_stvx): Likewise.
4977 (replace_swapped_aligned_store): Likewise.
4978 (rs6000_gen_lvx): Likewise.
4979 (replace_swapped_aligned_load): Likewise.
4980 (replace_swapped_load_constant): Capitalize argument name in
4981 comment describing this function.
4982 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4983 and stores that access quad-word aligned addresses and replace
4984 with stvx or lvx instructions when appropriate.
4985 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4986 New function prototype.
4987 (rs6000_quadword_masked_address_p): Likewise.
4988 (rs6000_gen_lvx): Likewise.
4989 (rs6000_gen_stvx): Likewise.
4990 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4991 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4992 when memory address is aligned.
4993 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4994 this split to select lvx instruction when memory address is aligned.
4995 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4996 instruction when memory address is aligned.
4997 (*vsx_le_perm_load_v16qi): Likewise.
4998 (four unnamed splitters): Modify to select the stvx instruction
4999 when memory is aligned.
5001 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5003 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5004 which appears in the queue multiple times.
5006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5007 Alan Hayward <alan.hayward@arm.com>
5008 David Sherwood <david.sherwood@arm.com>
5010 * tree-vectorizer.h (vec_lower_bound): New structure.
5011 (_loop_vec_info): Add check_nonzero and lower_bounds.
5012 (LOOP_VINFO_CHECK_NONZERO): New macro.
5013 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5014 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5015 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5016 fields. Make seg_len the distance travelled, not including the
5017 access size.
5018 (dr_direction_indicator): Declare.
5019 (dr_zero_step_indicator): Likewise.
5020 (dr_known_forward_stride_p): Likewise.
5021 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5022 tree-ssanames.h.
5023 (runtime_alias_check_p): Allow runtime alias checks with
5024 variable strides.
5025 (operator ==): Compare access_size and align.
5026 (prune_runtime_alias_test_list): Rework for new distinction between
5027 the access_size and seg_len.
5028 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5029 segment lengths.
5030 (get_segment_min_max): New function.
5031 (create_intersect_range_checks): Use it.
5032 (dr_step_indicator): New function.
5033 (dr_direction_indicator): Likewise.
5034 (dr_zero_step_indicator): Likewise.
5035 (dr_known_forward_stride_p): Likewise.
5036 * tree-loop-distribution.c (data_ref_segment_size): Return
5037 DR_STEP * (niters - 1).
5038 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5039 constructor.
5040 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5041 (vect_preserves_scalar_order_p): New function, split out from...
5042 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5043 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5044 (vect_vfa_access_size): New function.
5045 (vect_vfa_align): Likewise.
5046 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5047 (dump_lower_bound): New function.
5048 (vect_check_lower_bound): Likewise.
5049 (vect_small_gap_p): Likewise.
5050 (vectorizable_with_step_bound_p): Likewise.
5051 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5052 depencies if the vectorization factor is 1. Convert the checks
5053 for nonzero steps into checks on the bounds of DR_STEP. Try using
5054 a bunds check for variable steps if the minimum required step is
5055 relatively small. Update calls to the dr_with_seg_len
5056 constructor and to vect_compile_time_alias.
5057 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5058 function.
5059 (vect_loop_versioning): Call it.
5060 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5061 when retrying.
5062 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5065 Alan Hayward <alan.hayward@arm.com>
5066 David Sherwood <david.sherwood@arm.com>
5068 * doc/sourcebuild.texi (vect_scatter_store): Document.
5069 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5070 optabs.
5071 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5072 Document.
5073 * genopinit.c (main): Add supports_vec_scatter_store and
5074 supports_vec_scatter_store_cached to target_optabs.
5075 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5076 IFN_MASK_SCATTER_STORE.
5077 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5078 functions.
5079 * internal-fn.h (internal_store_fn_p): Declare.
5080 (internal_fn_stored_value_index): Likewise.
5081 * internal-fn.c (scatter_store_direct): New macro.
5082 (expand_scatter_store_optab_fn): New function.
5083 (direct_scatter_store_optab_supported_p): New macro.
5084 (internal_store_fn_p): New function.
5085 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5086 IFN_MASK_SCATTER_STORE.
5087 (internal_fn_mask_index): Likewise.
5088 (internal_fn_stored_value_index): New function.
5089 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5090 for scatter stores.
5091 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5092 * optabs-query.c (supports_vec_scatter_store_p): New function.
5093 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5094 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5095 true for scatter stores.
5096 (vect_gather_scatter_fn_p): Handle scatter stores too.
5097 (vect_check_gather_scatter): Consider using scatter stores if
5098 supports_vec_scatter_store_p.
5099 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5100 scatter stores too.
5101 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5102 internal_fn_stored_value_index.
5103 (check_load_store_masking): Handle scatter stores too.
5104 (vect_get_store_rhs): Make public.
5105 (vectorizable_call): Use internal_store_fn_p.
5106 (vectorizable_store): Handle scatter store internal functions.
5107 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5108 when deciding whether the end of the group has been reached.
5109 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5110 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5111 (mask_scatter_store<mode>): New insns.
5113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5114 Alan Hayward <alan.hayward@arm.com>
5115 David Sherwood <david.sherwood@arm.com>
5117 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5118 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5119 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5120 function.
5121 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5122 Use vect_truncate_gather_scatter_offset if we can't treat the
5123 operation as a normal gather load or scatter store.
5124 (get_group_load_store_type): Take the gather_scatter_info
5125 as argument. Try using a gather load or scatter store for
5126 single-element groups.
5127 (get_load_store_type): Update calls to get_group_load_store_type
5128 and vect_use_strided_gather_scatters_p.
5130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5131 Alan Hayward <alan.hayward@arm.com>
5132 David Sherwood <david.sherwood@arm.com>
5134 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5135 optional tree argument.
5136 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5137 null target hooks.
5138 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5139 but continue to use the current value as a fallback.
5140 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5141 to compare the updates.
5142 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5143 (get_load_store_type): Use it when handling a strided access.
5144 (vect_get_strided_load_store_ops): New function.
5145 (vect_get_data_ptr_increment): Likewise.
5146 (vectorizable_load): Handle strided gather loads. Always pass
5147 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5150 Alan Hayward <alan.hayward@arm.com>
5151 David Sherwood <david.sherwood@arm.com>
5153 * doc/md.texi (gather_load@var{m}): Document.
5154 (mask_gather_load@var{m}): Likewise.
5155 * genopinit.c (main): Add supports_vec_gather_load and
5156 supports_vec_gather_load_cached to target_optabs.
5157 * optabs-tree.c (init_tree_optimization_optabs): Use
5158 ggc_cleared_alloc to allocate target_optabs.
5159 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5160 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5161 functions.
5162 * internal-fn.h (internal_load_fn_p): Declare.
5163 (internal_gather_scatter_fn_p): Likewise.
5164 (internal_fn_mask_index): Likewise.
5165 (internal_gather_scatter_fn_supported_p): Likewise.
5166 * internal-fn.c (gather_load_direct): New macro.
5167 (expand_gather_load_optab_fn): New function.
5168 (direct_gather_load_optab_supported_p): New macro.
5169 (direct_internal_fn_optab): New function.
5170 (internal_load_fn_p): Likewise.
5171 (internal_gather_scatter_fn_p): Likewise.
5172 (internal_fn_mask_index): Likewise.
5173 (internal_gather_scatter_fn_supported_p): Likewise.
5174 * optabs-query.c (supports_at_least_one_mode_p): New function.
5175 (supports_vec_gather_load_p): Likewise.
5176 * optabs-query.h (supports_vec_gather_load_p): Declare.
5177 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5178 and memory_type field.
5179 (NUM_PATTERNS): Bump to 15.
5180 * tree-vect-data-refs.c: Include internal-fn.h.
5181 (vect_gather_scatter_fn_p): New function.
5182 (vect_describe_gather_scatter_call): Likewise.
5183 (vect_check_gather_scatter): Try using internal functions for
5184 gather loads. Recognize existing calls to a gather load function.
5185 (vect_analyze_data_refs): Consider using gather loads if
5186 supports_vec_gather_load_p.
5187 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5188 (vect_get_gather_scatter_offset_type): Likewise.
5189 (vect_convert_mask_for_vectype): Likewise.
5190 (vect_add_conversion_to_patterm): Likewise.
5191 (vect_try_gather_scatter_pattern): Likewise.
5192 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5193 (vect_vect_recog_func_ptrs): Add it.
5194 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5195 internal_fn_mask_index and internal_gather_scatter_fn_p.
5196 (check_load_store_masking): Take the gather_scatter_info as an
5197 argument and handle gather loads.
5198 (vect_get_gather_scatter_ops): New function.
5199 (vectorizable_call): Check internal_load_fn_p.
5200 (vectorizable_load): Likewise. Handle gather load internal
5201 functions.
5202 (vectorizable_store): Update call to check_load_store_masking.
5203 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5204 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5205 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5206 (aarch64_gather_scale_operand_d): New predicates.
5207 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5208 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
5215 * doc/md.texi (fold_left_plus_@var{m}): Document.
5216 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5217 * internal-fn.c (fold_left_direct): Define.
5218 (expand_fold_left_optab_fn): Likewise.
5219 (direct_fold_left_optab_supported_p): Likewise.
5220 * fold-const-call.c (fold_const_fold_left): New function.
5221 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5222 * tree-parloops.c (valid_reduction_p): New function.
5223 (gather_scalar_reductions): Use it.
5224 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5225 (vect_finish_replace_stmt): Declare.
5226 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5227 (needs_fold_left_reduction_p): New function, split out from...
5228 (vect_is_simple_reduction): ...here. Accept reductions that
5229 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5230 (vect_force_simple_reduction): Also store the reduction type in
5231 the assignment's STMT_VINFO_REDUC_TYPE.
5232 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5233 (merge_with_identity): New function.
5234 (vect_expand_fold_left): Likewise.
5235 (vectorize_fold_left_reduction): Likewise.
5236 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5237 scalar phi in place for it. Check for target support and reject
5238 cases that would reassociate the operation. Defer the transform
5239 phase to vectorize_fold_left_reduction.
5240 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5241 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5242 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5246 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5247 call to ifc_temp_var.
5249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5250 Alan Hayward <alan.hayward@arm.com>
5251 David Sherwood <david.sherwood@arm.com>
5253 * target.def (legitimize_address_displacement): Take the original
5254 offset as a poly_int.
5255 * targhooks.h (default_legitimize_address_displacement): Update
5256 accordingly.
5257 * targhooks.c (default_legitimize_address_displacement): Likewise.
5258 * doc/tm.texi: Regenerate.
5259 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5260 as an argument, moving assert of ad->disp == ad->disp_term to...
5261 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5262 Try calling targetm.legitimize_address_displacement before expanding
5263 the address rather than afterwards, and adjust for the new interface.
5264 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5265 Match the new hook interface. Handle SVE addresses.
5266 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5267 new hook interface.
5269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5271 * Makefile.in (OBJS): Add early-remat.o.
5272 * target.def (select_early_remat_modes): New hook.
5273 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5274 * doc/tm.texi: Regenerate.
5275 * targhooks.h (default_select_early_remat_modes): Declare.
5276 * targhooks.c (default_select_early_remat_modes): New function.
5277 * timevar.def (TV_EARLY_REMAT): New timevar.
5278 * passes.def (pass_early_remat): New pass.
5279 * tree-pass.h (make_pass_early_remat): Declare.
5280 * early-remat.c: New file.
5281 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5282 function.
5283 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5286 Alan Hayward <alan.hayward@arm.com>
5287 David Sherwood <david.sherwood@arm.com>
5289 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5290 vfm1 with a bound_epilog parameter.
5291 (vect_do_peeling): Update calls accordingly, and move the prologue
5292 call earlier in the function. Treat the base bound_epilog as 0 for
5293 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5294 this base when peeling for gaps.
5295 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5296 with fully-masked loops.
5297 (vect_estimate_min_profitable_iters): Handle the single peeled
5298 iteration in that case.
5300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5301 Alan Hayward <alan.hayward@arm.com>
5302 David Sherwood <david.sherwood@arm.com>
5304 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5305 single-element interleaving even if the size is not a power of 2.
5306 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5307 accesses for single-element interleaving if the group size is
5308 not a power of 2.
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
5315 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5316 * optabs.def (fold_extract_last_optab): New optab.
5317 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5318 * internal-fn.c (fold_extract_direct): New macro.
5319 (expand_fold_extract_optab_fn): Likewise.
5320 (direct_fold_extract_optab_supported_p): Likewise.
5321 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5322 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5323 EXTRACT_LAST_REDUCTION.
5324 (get_initial_def_for_reduction): Do not create an initial vector
5325 for EXTRACT_LAST_REDUCTION reductions.
5326 (vectorizable_reduction): Leave the scalar phi in place for
5327 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5328 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5329 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5330 transform phase to vectorizable_condition.
5331 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5332 split out from...
5333 (vect_finish_stmt_generation): ...here.
5334 (vect_finish_replace_stmt): New function.
5335 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5336 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5337 pattern.
5338 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5341 Alan Hayward <alan.hayward@arm.com>
5342 David Sherwood <david.sherwood@arm.com>
5344 * doc/md.texi (extract_last_@var{m}): Document.
5345 * optabs.def (extract_last_optab): New optab.
5346 * internal-fn.def (EXTRACT_LAST): New internal function.
5347 * internal-fn.c (cond_unary_direct): New macro.
5348 (expand_cond_unary_optab_fn): Likewise.
5349 (direct_cond_unary_optab_supported_p): Likewise.
5350 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5351 loops using EXTRACT_LAST.
5352 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5353 (extract_last_<mode>): ...this optab.
5354 (vec_extract<mode><Vel>): Update accordingly.
5356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5357 Alan Hayward <alan.hayward@arm.com>
5358 David Sherwood <david.sherwood@arm.com>
5360 * target.def (empty_mask_is_expensive): New hook.
5361 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5362 * doc/tm.texi: Regenerate.
5363 * targhooks.h (default_empty_mask_is_expensive): Declare.
5364 * targhooks.c (default_empty_mask_is_expensive): New function.
5365 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5366 if the target says that empty masks are expensive.
5367 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5368 New function.
5369 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5376 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5377 (vect_use_loop_mask_for_alignment_p): New function.
5378 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5379 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5380 niters_skip argument. Make sure that the first niters_skip elements
5381 of the first iteration are inactive.
5382 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5383 Update call to vect_set_loop_masks_directly.
5384 (get_misalign_in_elems): New function, split out from...
5385 (vect_gen_prolog_loop_niters): ...here.
5386 (vect_update_init_of_dr): Take a code argument that specifies whether
5387 the adjustment should be added or subtracted.
5388 (vect_update_init_of_drs): Likewise.
5389 (vect_prepare_for_masked_peels): New function.
5390 (vect_do_peeling): Skip prologue peeling if we're using a mask
5391 instead. Update call to vect_update_inits_of_drs.
5392 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5393 mask_skip_niters.
5394 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5395 alignment. Do not include the number of peeled iterations in
5396 the minimum threshold in that case.
5397 (vectorizable_induction): Adjust the start value down by
5398 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5399 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5400 Take the number of skipped iterations into account when calculating
5401 the loop bounds.
5402 * tree-vect-stmts.c (vect_gen_while_not): New function.
5404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5405 Alan Hayward <alan.hayward@arm.com>
5406 David Sherwood <david.sherwood@arm.com>
5408 * doc/sourcebuild.texi (vect_fully_masked): Document.
5409 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5410 default value to 0.
5411 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5412 split out from...
5413 (vect_analyze_loop_2): ...here. Don't check the vectorization
5414 factor against the number of loop iterations if the loop is
5415 fully-masked.
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5422 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5423 (dump_groups): Update accordingly.
5424 (iv_use::mem_type): New member variable.
5425 (address_p): New function.
5426 (record_use): Add a mem_type argument and initialize the new
5427 mem_type field.
5428 (record_group_use): Add a mem_type argument. Use address_p.
5429 Remove obsolete null checks of base_object. Update call to record_use.
5430 (find_interesting_uses_op): Update call to record_group_use.
5431 (find_interesting_uses_cond): Likewise.
5432 (find_interesting_uses_address): Likewise.
5433 (get_mem_type_for_internal_fn): New function.
5434 (find_address_like_use): Likewise.
5435 (find_interesting_uses_stmt): Try find_address_like_use before
5436 calling find_interesting_uses_op.
5437 (addr_offset_valid_p): Use the iv mem_type field as the type
5438 of the addressed memory.
5439 (add_autoinc_candidates): Likewise.
5440 (get_address_cost): Likewise.
5441 (split_small_address_groups_p): Use address_p.
5442 (split_address_groups): Likewise.
5443 (add_iv_candidate_for_use): Likewise.
5444 (autoinc_possible_for_pair): Likewise.
5445 (rewrite_groups): Likewise.
5446 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5447 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5448 (get_alias_ptr_type_for_ptr_address): New function.
5449 (rewrite_use_address): Rewrite address uses in calls that were
5450 identified by find_address_like_use.
5452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5453 Alan Hayward <alan.hayward@arm.com>
5454 David Sherwood <david.sherwood@arm.com>
5456 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5457 TARGET_MEM_REFs.
5458 * gimple-expr.h (is_gimple_addressable: Likewise.
5459 * gimple-expr.c (is_gimple_address): Likewise.
5460 * internal-fn.c (expand_call_mem_ref): New function.
5461 (expand_mask_load_optab_fn): Use it.
5462 (expand_mask_store_optab_fn): Likewise.
5464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5465 Alan Hayward <alan.hayward@arm.com>
5466 David Sherwood <david.sherwood@arm.com>
5468 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5469 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5470 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5471 (cond_umax@var{mode}): Document.
5472 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5473 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5474 (cond_umin_optab, cond_umax_optab): New optabs.
5475 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5476 (COND_IOR, COND_XOR): New internal functions.
5477 * internal-fn.h (get_conditional_internal_fn): Declare.
5478 * internal-fn.c (cond_binary_direct): New macro.
5479 (expand_cond_binary_optab_fn): Likewise.
5480 (direct_cond_binary_optab_supported_p): Likewise.
5481 (get_conditional_internal_fn): New function.
5482 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5483 Cope with reduction statements that are vectorized as calls rather
5484 than assignments.
5485 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5486 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5487 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5488 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5489 (UNSPEC_COND_EOR): New unspecs.
5490 (optab): Add mappings for them.
5491 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5492 (sve_int_op, sve_fp_op): New int attributes.
5494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5495 Alan Hayward <alan.hayward@arm.com>
5496 David Sherwood <david.sherwood@arm.com>
5498 * optabs.def (while_ult_optab): New optab.
5499 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5500 * internal-fn.def (WHILE_ULT): New internal function.
5501 * internal-fn.h (direct_internal_fn_supported_p): New override
5502 that takes two types as argument.
5503 * internal-fn.c (while_direct): New macro.
5504 (expand_while_optab_fn): New function.
5505 (convert_optab_supported_p): Likewise.
5506 (direct_while_optab_supported_p): New macro.
5507 * wide-int.h (wi::udiv_ceil): New function.
5508 * tree-vectorizer.h (rgroup_masks): New structure.
5509 (vec_loop_masks): New typedef.
5510 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5511 and fully_masked_p.
5512 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5513 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5514 (vect_max_vf): New function.
5515 (slpeel_make_loop_iterate_ntimes): Delete.
5516 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5517 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5518 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5519 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5520 internal-fn.h, stor-layout.h and optabs-query.h.
5521 (vect_set_loop_mask): New function.
5522 (add_preheader_seq): Likewise.
5523 (add_header_seq): Likewise.
5524 (interleave_supported_p): Likewise.
5525 (vect_maybe_permute_loop_masks): Likewise.
5526 (vect_set_loop_masks_directly): Likewise.
5527 (vect_set_loop_condition_masked): Likewise.
5528 (vect_set_loop_condition_unmasked): New function, split out from
5529 slpeel_make_loop_iterate_ntimes.
5530 (slpeel_make_loop_iterate_ntimes): Rename to..
5531 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5532 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5533 (vect_do_peeling): Update call accordingly.
5534 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5535 loops.
5536 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5537 mask_compare_type, can_fully_mask_p and fully_masked_p.
5538 (release_vec_loop_masks): New function.
5539 (_loop_vec_info): Use it to free the loop masks.
5540 (can_produce_all_loop_masks_p): New function.
5541 (vect_get_max_nscalars_per_iter): Likewise.
5542 (vect_verify_full_masking): Likewise.
5543 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5544 retries, and free the mask rgroups before retrying. Check loop-wide
5545 reasons for disallowing fully-masked loops. Make the final decision
5546 about whether use a fully-masked loop or not.
5547 (vect_estimate_min_profitable_iters): Do not assume that peeling
5548 for the number of iterations will be needed for fully-masked loops.
5549 (vectorizable_reduction): Disable fully-masked loops.
5550 (vectorizable_live_operation): Likewise.
5551 (vect_halve_mask_nunits): New function.
5552 (vect_double_mask_nunits): Likewise.
5553 (vect_record_loop_mask): Likewise.
5554 (vect_get_loop_mask): Likewise.
5555 (vect_transform_loop): Handle the case in which the final loop
5556 iteration might handle a partial vector. Call vect_set_loop_condition
5557 instead of slpeel_make_loop_iterate_ntimes.
5558 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5559 (check_load_store_masking): New function.
5560 (prepare_load_store_mask): Likewise.
5561 (vectorizable_store): Handle fully-masked loops.
5562 (vectorizable_load): Likewise.
5563 (supportable_widening_operation): Use vect_halve_mask_nunits for
5564 booleans.
5565 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5566 (vect_gen_while): New function.
5567 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5568 (aarch64_uqdec<mode>): New insn.
5570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5571 Alan Hayward <alan.hayward@arm.com>
5572 David Sherwood <david.sherwood@arm.com>
5574 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5575 (reduc_xor_scal_optab): New optabs.
5576 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5577 (reduc_xor_scal_@var{m}): Document.
5578 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5579 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5580 internal functions.
5581 * fold-const-call.c (fold_const_call): Handle them.
5582 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5583 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5584 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5585 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5586 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5587 (UNSPEC_XORV): New unspecs.
5588 (optab): Add entries for them.
5589 (BITWISEV): New int iterator.
5590 (bit_reduc_op): New int attributes.
5592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5593 Alan Hayward <alan.hayward@arm.com>
5594 David Sherwood <david.sherwood@arm.com>
5596 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5597 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5598 * optabs.def (vec_shl_insert_optab): New optab.
5599 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5600 (duplicate_and_interleave): Likewise.
5601 * tree-vect-loop.c: Include internal-fn.h.
5602 (neutral_op_for_slp_reduction): New function, split out from
5603 get_initial_defs_for_reduction.
5604 (get_initial_def_for_reduction): Handle option 2 for variable-length
5605 vectors by loading the neutral value into a vector and then shifting
5606 the initial value into element 0.
5607 (get_initial_defs_for_reduction): Replace the code argument with
5608 the neutral value calculated by neutral_op_for_slp_reduction.
5609 Use gimple_build_vector for constant-length vectors.
5610 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5611 but the first group_size elements have a neutral value.
5612 Use duplicate_and_interleave otherwise.
5613 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5614 Update call to get_initial_defs_for_reduction. Handle SLP
5615 reductions for variable-length vectors by creating one vector
5616 result for each scalar result, with the elements associated
5617 with other scalar results stubbed out with the neutral value.
5618 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5619 Require IFN_VEC_SHL_INSERT for double reductions on
5620 variable-length vectors, or SLP reductions that have
5621 a neutral value. Require can_duplicate_and_interleave_p
5622 support for variable-length unchained SLP reductions if there
5623 is no neutral value, such as for MIN/MAX reductions. Also require
5624 the number of vector elements to be a multiple of the number of
5625 SLP statements when doing variable-length unchained SLP reductions.
5626 Update call to vect_create_epilog_for_reduction.
5627 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5628 and remove initial values.
5629 (duplicate_and_interleave): Make public.
5630 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5631 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5634 Alan Hayward <alan.hayward@arm.com>
5635 David Sherwood <david.sherwood@arm.com>
5637 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5638 (can_duplicate_and_interleave_p): New function.
5639 (vect_get_and_check_slp_defs): Take the vector of statements
5640 rather than just the current one. Remove excess parentheses.
5641 Restriction rejectinon of vect_constant_def and vect_external_def
5642 for variable-length vectors to boolean types, or types for which
5643 can_duplicate_and_interleave_p is false.
5644 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5645 (duplicate_and_interleave): New function.
5646 (vect_get_constant_vectors): Use gimple_build_vector for
5647 constant-length vectors and suitable variable-length constant
5648 vectors. Use duplicate_and_interleave for other variable-length
5649 vectors. Don't defer the update when inserting new statements.
5651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5652 Alan Hayward <alan.hayward@arm.com>
5653 David Sherwood <david.sherwood@arm.com>
5655 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5656 min_profitable_iters doesn't go negative.
5658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5659 Alan Hayward <alan.hayward@arm.com>
5660 David Sherwood <david.sherwood@arm.com>
5662 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5663 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5664 * optabs.def (vec_mask_load_lanes_optab): New optab.
5665 (vec_mask_store_lanes_optab): Likewise.
5666 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5667 (MASK_STORE_LANES): Likewise.
5668 * internal-fn.c (mask_load_lanes_direct): New macro.
5669 (mask_store_lanes_direct): Likewise.
5670 (expand_mask_load_optab_fn): Handle masked operations.
5671 (expand_mask_load_lanes_optab_fn): New macro.
5672 (expand_mask_store_optab_fn): Handle masked operations.
5673 (expand_mask_store_lanes_optab_fn): New macro.
5674 (direct_mask_load_lanes_optab_supported_p): Likewise.
5675 (direct_mask_store_lanes_optab_supported_p): Likewise.
5676 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5677 parameter.
5678 (vect_load_lanes_supported): Likewise.
5679 * tree-vect-data-refs.c (strip_conversion): New function.
5680 (can_group_stmts_p): Likewise.
5681 (vect_analyze_data_ref_accesses): Use it instead of checking
5682 for a pair of assignments.
5683 (vect_store_lanes_supported): Take a masked_p parameter.
5684 (vect_load_lanes_supported): Likewise.
5685 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5686 vect_store_lanes_supported and vect_load_lanes_supported.
5687 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5688 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5689 parameter. Don't allow gaps for masked accesses.
5690 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5691 and vect_load_lanes_supported.
5692 (get_load_store_type): Take a masked_p parameter and update
5693 call to get_group_load_store_type.
5694 (vectorizable_store): Update call to get_load_store_type.
5695 Handle IFN_MASK_STORE_LANES.
5696 (vectorizable_load): Update call to get_load_store_type.
5697 Handle IFN_MASK_LOAD_LANES.
5699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5700 Alan Hayward <alan.hayward@arm.com>
5701 David Sherwood <david.sherwood@arm.com>
5703 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5704 modes for SVE.
5705 * config/aarch64/aarch64-protos.h
5706 (aarch64_sve_struct_memory_operand_p): Declare.
5707 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5708 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5709 (VPRED, vpred): Handle SVE structure modes.
5710 * config/aarch64/constraints.md (Utx): New constraint.
5711 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5712 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5713 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5714 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5715 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5716 structure modes. Split into pieces after RA.
5717 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5718 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5719 New patterns.
5720 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5721 SVE structure modes.
5722 (aarch64_classify_address): Likewise.
5723 (sizetochar): Move earlier in file.
5724 (aarch64_print_operand): Handle SVE register lists.
5725 (aarch64_array_mode): New function.
5726 (aarch64_sve_struct_memory_operand_p): Likewise.
5727 (TARGET_ARRAY_MODE): Redefine.
5729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5730 Alan Hayward <alan.hayward@arm.com>
5731 David Sherwood <david.sherwood@arm.com>
5733 * target.def (array_mode): New target hook.
5734 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5735 * doc/tm.texi: Regenerate.
5736 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5737 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5738 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5739 targetm.array_mode.
5740 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5741 type sizes.
5743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5744 Alan Hayward <alan.hayward@arm.com>
5745 David Sherwood <david.sherwood@arm.com>
5747 * fold-const.c (fold_binary_loc): Check the argument types
5748 rather than the result type when testing for a vector operation.
5750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5752 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5753 * doc/tm.texi: Regenerate.
5755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5756 Alan Hayward <alan.hayward@arm.com>
5757 David Sherwood <david.sherwood@arm.com>
5759 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5760 (sve): Document new AArch64 extension.
5761 * doc/md.texi (w): Extend the description of the AArch64
5762 constraint to include SVE vectors.
5763 (Upl, Upa): Document new AArch64 predicate constraints.
5764 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5765 enum.
5766 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5767 (msve-vector-bits=): New option.
5768 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5769 SVE when these are disabled.
5770 (sve): New extension.
5771 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5772 modes. Adjust their number of units based on aarch64_sve_vg.
5773 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5774 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5775 aarch64_addr_query_type.
5776 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5777 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5778 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5779 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5780 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5781 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5782 (aarch64_simd_imm_zero_p): Delete.
5783 (aarch64_check_zero_based_sve_index_immediate): Declare.
5784 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5785 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5786 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5787 (aarch64_sve_float_mul_immediate_p): Likewise.
5788 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5789 rather than an rtx.
5790 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5791 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5792 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5793 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5794 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5795 (aarch64_regmode_natural_size): Likewise.
5796 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5797 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5798 left one place.
5799 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5800 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5801 for VG and the SVE predicate registers.
5802 (V_ALIASES): Add a "z"-prefixed alias.
5803 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5804 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5805 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5806 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5807 (REG_CLASS_NAMES): Add entries for them.
5808 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5809 and the predicate registers.
5810 (aarch64_sve_vg): Declare.
5811 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5812 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5813 (REGMODE_NATURAL_SIZE): Define.
5814 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5815 SVE macros.
5816 * config/aarch64/aarch64.c: Include cfgrtl.h.
5817 (simd_immediate_info): Add a constructor for series vectors,
5818 and an associated step field.
5819 (aarch64_sve_vg): New variable.
5820 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5821 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5822 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5823 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5824 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5825 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5826 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5827 (aarch64_get_mask_mode): New functions.
5828 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5829 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5830 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5831 predicate modes and predicate registers. Explicitly restrict
5832 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5833 to store a vector mode if it is recognized by
5834 aarch64_classify_vector_mode.
5835 (aarch64_regmode_natural_size): New function.
5836 (aarch64_hard_regno_caller_save_mode): Return the original mode
5837 for predicates.
5838 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5839 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5840 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5841 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5842 functions.
5843 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5844 does not overlap dest if the function is frame-related. Handle
5845 SVE constants.
5846 (aarch64_split_add_offset): New function.
5847 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5848 them aarch64_add_offset.
5849 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5850 and update call to aarch64_sub_sp.
5851 (aarch64_add_cfa_expression): New function.
5852 (aarch64_expand_prologue): Pass extra temporary registers to the
5853 functions above. Handle the case in which we need to emit new
5854 DW_CFA_expressions for registers that were originally saved
5855 relative to the stack pointer, but now have to be expressed
5856 relative to the frame pointer.
5857 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5858 functions above.
5859 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5860 IP0 and IP1 values for SVE frames.
5861 (aarch64_expand_vec_series): New function.
5862 (aarch64_expand_sve_widened_duplicate): Likewise.
5863 (aarch64_expand_sve_const_vector): Likewise.
5864 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5865 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5866 into the register, rather than emitting a SET directly.
5867 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5868 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5869 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5870 (offset_9bit_signed_scaled_p): New functions.
5871 (aarch64_replicate_bitmask_imm): New function.
5872 (aarch64_bitmask_imm): Use it.
5873 (aarch64_cannot_force_const_mem): Reject expressions involving
5874 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5875 (aarch64_classify_index): Handle SVE indices, by requiring
5876 a plain register index with a scale that matches the element size.
5877 (aarch64_classify_address): Handle SVE addresses. Assert that
5878 the mode of the address is VOIDmode or an integer mode.
5879 Update call to aarch64_classify_symbol.
5880 (aarch64_classify_symbolic_expression): Update call to
5881 aarch64_classify_symbol.
5882 (aarch64_const_vec_all_in_range_p): New function.
5883 (aarch64_print_vector_float_operand): Likewise.
5884 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5885 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5886 and the FP immediates 1.0 and 0.5.
5887 (aarch64_print_address_internal): Handle SVE addresses.
5888 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5889 (aarch64_regno_regclass): Handle predicate registers.
5890 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5891 data modes.
5892 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5893 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5894 (aarch64_convert_sve_vector_bits): New function.
5895 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5896 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5897 rather than an rtx.
5898 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5899 Handle SVE vector and predicate modes. Accept VL-based constants
5900 that need only one temporary register, and VL offsets that require
5901 no temporary registers.
5902 (aarch64_conditional_register_usage): Mark the predicate registers
5903 as fixed if SVE isn't available.
5904 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5905 Return true for SVE vector and predicate modes.
5906 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5907 rather than an unsigned int. Handle SVE modes.
5908 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5909 SVE modes.
5910 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5911 if SVE is enabled.
5912 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5913 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5914 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5915 (aarch64_sve_float_mul_immediate_p): New functions.
5916 (aarch64_sve_valid_immediate): New function.
5917 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5918 Explicitly reject structure modes. Check for INDEX constants.
5919 Handle PTRUE and PFALSE constants.
5920 (aarch64_check_zero_based_sve_index_immediate): New function.
5921 (aarch64_simd_imm_zero_p): Delete.
5922 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5923 vector modes. Accept constants in the range of CNT[BHWD].
5924 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5925 ask for an Advanced SIMD mode.
5926 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5927 (aarch64_simd_vector_alignment): Handle SVE predicates.
5928 (aarch64_vectorize_preferred_vector_alignment): New function.
5929 (aarch64_simd_vector_alignment_reachable): Use it instead of
5930 the vector size.
5931 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5932 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5933 functions.
5934 (MAX_VECT_LEN): Delete.
5935 (expand_vec_perm_d): Add a vec_flags field.
5936 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5937 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5938 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5939 for SVE modes.
5940 (aarch64_evpc_rev): Rename to...
5941 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5942 (aarch64_evpc_rev_global): New function.
5943 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5944 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5945 MAX_VECT_LEN.
5946 (aarch64_evpc_sve_tbl): New function.
5947 (aarch64_expand_vec_perm_const_1): Update after rename of
5948 aarch64_evpc_rev. Handle SVE permutes too, trying
5949 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5950 than aarch64_evpc_tbl.
5951 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5952 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5953 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5954 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5955 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5956 (aarch64_expand_sve_vcond): New functions.
5957 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5958 of aarch64_vector_mode_p.
5959 (aarch64_dwarf_poly_indeterminate_value): New function.
5960 (aarch64_compute_pressure_classes): Likewise.
5961 (aarch64_can_change_mode_class): Likewise.
5962 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5963 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5964 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5965 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5966 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5967 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5968 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5969 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5970 constraints.
5971 (Dn, Dl, Dr): Accept const as well as const_vector.
5972 (Dz): Likewise. Compare against CONST0_RTX.
5973 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5974 of "vector" where appropriate.
5975 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5976 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5977 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5978 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5979 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5980 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5981 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5982 (v_int_equiv): Extend to SVE modes.
5983 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5984 mode attributes.
5985 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5986 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5987 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5988 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5989 (SVE_COND_FP_CMP): New int iterators.
5990 (perm_hilo): Handle the new unpack unspecs.
5991 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5992 attributes.
5993 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5994 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5995 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5996 (aarch64_equality_operator, aarch64_constant_vector_operand)
5997 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5998 (aarch64_sve_nonimmediate_operand): Likewise.
5999 (aarch64_sve_general_operand): Likewise.
6000 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6001 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6002 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6003 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6004 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6005 (aarch64_sve_float_arith_immediate): Likewise.
6006 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6007 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6008 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6009 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6010 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6011 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6012 (aarch64_sve_float_arith_operand): Likewise.
6013 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6014 (aarch64_sve_float_mul_operand): Likewise.
6015 (aarch64_sve_vec_perm_operand): Likewise.
6016 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6017 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6018 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6019 as well as const_vector.
6020 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6021 in file. Use CONST0_RTX and CONSTM1_RTX.
6022 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6023 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6024 Use aarch64_simd_imm_zero.
6025 * config/aarch64/aarch64-sve.md: New file.
6026 * config/aarch64/aarch64.md: Include it.
6027 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6028 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6029 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6030 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6031 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6032 (sve): New attribute.
6033 (enabled): Disable instructions with the sve attribute unless
6034 TARGET_SVE.
6035 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6036 aarch64_expand_mov_immediate.
6037 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6038 CNT[BHSD] immediates.
6039 (movti): Split CONST_POLY_INT moves into two halves.
6040 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6041 Split additions that need a temporary here if the destination
6042 is the stack pointer.
6043 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6044 (*add<mode>3_poly_1): New instruction.
6045 (set_clobber_cc): New expander.
6047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6049 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6050 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6051 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6052 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6053 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6054 Change innermode from fixed_mode_size to machine_mode.
6055 (simplify_subreg): Update call accordingly. Handle a constant-sized
6056 subreg of a variable-length CONST_VECTOR.
6058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6059 Alan Hayward <alan.hayward@arm.com>
6060 David Sherwood <david.sherwood@arm.com>
6062 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6063 (add_offset_to_base): New function, split out from...
6064 (create_mem_ref): ...here. When handling a scale other than 1,
6065 check first whether the address is valid without the offset.
6066 Add it into the base if so, leaving the index and scale as-is.
6068 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6070 PR c++/83778
6071 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6072 fold_for_warn before checking if arg2 is INTEGER_CST.
6074 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6076 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6077 (store_multiple_operation): Delete.
6078 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6079 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6080 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6081 guarded by TARGET_STRING.
6082 (rs6000_output_load_multiple): Delete.
6083 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6084 OPTION_MASK_STRING / TARGET_STRING handling.
6085 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6086 (const rs6000_opt_masks) <"string">: Change mask to 0.
6087 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6088 (MASK_STRING): Delete.
6089 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6090 parts. Simplify.
6091 (load_multiple): Delete.
6092 (*ldmsi8): Delete.
6093 (*ldmsi7): Delete.
6094 (*ldmsi6): Delete.
6095 (*ldmsi5): Delete.
6096 (*ldmsi4): Delete.
6097 (*ldmsi3): Delete.
6098 (store_multiple): Delete.
6099 (*stmsi8): Delete.
6100 (*stmsi7): Delete.
6101 (*stmsi6): Delete.
6102 (*stmsi5): Delete.
6103 (*stmsi4): Delete.
6104 (*stmsi3): Delete.
6105 (movmemsi_8reg): Delete.
6106 (corresponding unnamed define_insn): Delete.
6107 (movmemsi_6reg): Delete.
6108 (corresponding unnamed define_insn): Delete.
6109 (movmemsi_4reg): Delete.
6110 (corresponding unnamed define_insn): Delete.
6111 (movmemsi_2reg): Delete.
6112 (corresponding unnamed define_insn): Delete.
6113 (movmemsi_1reg): Delete.
6114 (corresponding unnamed define_insn): Delete.
6115 * config/rs6000/rs6000.opt (mno-string): New.
6116 (mstring): Replace by deprecation warning stub.
6117 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6119 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6121 * regrename.c (regrename_do_replace): If replacing the same
6122 reg multiple times, try to reuse last created gen_raw_REG.
6124 PR debug/81155
6125 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6126 main to workaround a bug in GDB.
6128 2018-01-12 Tom de Vries <tom@codesourcery.com>
6130 PR target/83737
6131 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6133 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6135 PR rtl-optimization/80481
6136 * ira-color.c (get_cap_member): New function.
6137 (allocnos_conflict_by_live_ranges_p): Use it.
6138 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6139 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6141 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6143 PR target/83628
6144 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6145 (*saddl_se_1): Ditto.
6146 (*ssubsi_1): Ditto.
6147 (*ssubl_se_1): Ditto.
6149 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6151 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6152 rather than wi::to_widest for DR_INITs.
6153 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6154 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6155 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6156 INTEGER_CSTs.
6157 (vect_analyze_group_access_1): Note that here.
6159 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6161 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6162 polynomial type sizes.
6164 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6166 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6167 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6168 (gimple_add_tmp_var): Likewise.
6170 2018-01-12 Martin Liska <mliska@suse.cz>
6172 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6173 (gimple_alloc_sizes): Likewise.
6174 (dump_gimple_statistics): Use PRIu64 in printf format.
6175 * gimple.h: Change uint64_t to int.
6177 2018-01-12 Martin Liska <mliska@suse.cz>
6179 * tree-core.h: Use uint64_t instead of int.
6180 * tree.c (tree_node_counts): Likewise.
6181 (tree_node_sizes): Likewise.
6182 (dump_tree_statistics): Use PRIu64 in printf format.
6184 2018-01-12 Martin Liska <mliska@suse.cz>
6186 * Makefile.in: As qsort_chk is implemented in vec.c, add
6187 vec.o to linkage of gencfn-macros.
6188 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6189 passing the info to record_node_allocation_statistics.
6190 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6191 and pass the info.
6192 * ggc-common.c (struct ggc_usage): Add operator== and use
6193 it in operator< and compare function.
6194 * mem-stats.h (struct mem_usage): Likewise.
6195 * vec.c (struct vec_usage): Remove operator< and compare
6196 function. Can be simply inherited.
6198 2018-01-12 Martin Jambor <mjambor@suse.cz>
6200 PR target/81616
6201 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6202 * tree-ssa-math-opts.c: Include domwalk.h.
6203 (convert_mult_to_fma_1): New function.
6204 (fma_transformation_info): New type.
6205 (fma_deferring_state): Likewise.
6206 (cancel_fma_deferring): New function.
6207 (result_of_phi): Likewise.
6208 (last_fma_candidate_feeds_initial_phi): Likewise.
6209 (convert_mult_to_fma): Added deferring logic, split actual
6210 transformation to convert_mult_to_fma_1.
6211 (math_opts_dom_walker): New type.
6212 (math_opts_dom_walker::after_dom_children): New method, body moved
6213 here from pass_optimize_widening_mul::execute, added deferring logic
6214 bits.
6215 (pass_optimize_widening_mul::execute): Moved most of code to
6216 math_opts_dom_walker::after_dom_children.
6217 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6218 * config/i386/i386.c (ix86_option_override_internal): Added
6219 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6221 2018-01-12 Richard Biener <rguenther@suse.de>
6223 PR debug/83157
6224 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6225 inline instance vars.
6227 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6229 PR target/81819
6230 * config/rx/rx.c (rx_is_restricted_memory_address):
6231 Handle SUBREG case.
6233 2018-01-12 Richard Biener <rguenther@suse.de>
6235 PR tree-optimization/80846
6236 * target.def (split_reduction): New target hook.
6237 * targhooks.c (default_split_reduction): New function.
6238 * targhooks.h (default_split_reduction): Declare.
6239 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6240 target requests first reduce vectors by combining low and high
6241 parts.
6242 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6243 (get_vectype_for_scalar_type_and_size): Export.
6244 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6245 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6246 * doc/tm.texi: Regenerate.
6247 * config/i386/i386.c (ix86_split_reduction): Implement
6248 TARGET_VECTORIZE_SPLIT_REDUCTION.
6250 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6252 PR target/83368
6253 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6254 in PIC mode except for TARGET_VXWORKS_RTP.
6255 * config/sparc/sparc.c: Include cfgrtl.h.
6256 (TARGET_INIT_PIC_REG): Define.
6257 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6258 (sparc_pic_register_p): New predicate.
6259 (sparc_legitimate_address_p): Use it.
6260 (sparc_legitimize_pic_address): Likewise.
6261 (sparc_delegitimize_address): Likewise.
6262 (sparc_mode_dependent_address_p): Likewise.
6263 (gen_load_pcrel_sym): Remove 4th parameter.
6264 (load_got_register): Adjust call to above. Remove obsolete stuff.
6265 (sparc_expand_prologue): Do not call load_got_register here.
6266 (sparc_flat_expand_prologue): Likewise.
6267 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6268 (sparc_use_pseudo_pic_reg): New function.
6269 (sparc_init_pic_reg): Likewise.
6270 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6271 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6273 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6275 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6276 Add item for branch_cost.
6278 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6280 PR rtl-optimization/83565
6281 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6282 not extend the result to a larger mode for rotate operations.
6283 (num_sign_bit_copies1): Likewise.
6285 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6287 PR target/40411
6288 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6289 -symbolic.
6290 Use values-Xc.o for -pedantic.
6291 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6293 2018-01-12 Martin Liska <mliska@suse.cz>
6295 PR ipa/83054
6296 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6297 New function.
6298 (possible_polymorphic_call_targets): Use it.
6299 (ipa_devirt): Likewise.
6301 2018-01-12 Martin Liska <mliska@suse.cz>
6303 * profile-count.h (enum profile_quality): Use 0 as invalid
6304 enum value of profile_quality.
6306 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6308 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6309 -mext-string options.
6311 2018-01-12 Richard Biener <rguenther@suse.de>
6313 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6314 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6315 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6316 Likewise.
6317 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6319 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6321 * configure.ac (--with-long-double-format): Add support for the
6322 configuration option to change the default long double format on
6323 PowerPC systems.
6324 * config.gcc (powerpc*-linux*-*): Likewise.
6325 * configure: Regenerate.
6326 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6327 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6328 used without modification.
6330 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6332 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6333 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6334 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6335 MISC_BUILTIN_SPEC_BARRIER.
6336 (rs6000_init_builtins): Likewise.
6337 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6338 enum value.
6339 (speculation_barrier): New define_insn.
6340 * doc/extend.texi: Document __builtin_speculation_barrier.
6342 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6344 PR target/83203
6345 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6346 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6347 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6348 iterators.
6349 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6350 integral modes instead of "ss" and "sd".
6351 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6352 vectors with 32-bit and 64-bit elements.
6353 (vecdupssescalarmodesuffix): New mode attribute.
6354 (vec_dup<mode>): Use it.
6356 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6358 PR target/83330
6359 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6360 frame if argument is passed on stack.
6362 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6364 PR target/82682
6365 * ree.c (combine_reaching_defs): Optimize also
6366 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6367 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6369 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6371 PR middle-end/83189
6372 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6374 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6376 PR middle-end/83718
6377 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6378 after they are computed.
6380 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6382 PR tree-optimization/83695
6383 * gimple-loop-linterchange.cc
6384 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6385 reset cached scev information after interchange.
6386 (pass_linterchange::execute): Remove call to scev_reset_htab.
6388 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6390 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6391 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6392 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6393 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6394 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6395 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6396 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6397 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6398 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6399 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6400 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6401 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6402 (V_lane_reg): Likewise.
6403 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6404 New define_expand.
6405 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6406 (vfmal_lane_low<mode>_intrinsic,
6407 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6408 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6409 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6410 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6411 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6412 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6414 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6416 * config/arm/arm-cpus.in (fp16fml): New feature.
6417 (ALL_SIMD): Add fp16fml.
6418 (armv8.2-a): Add fp16fml as an option.
6419 (armv8.3-a): Likewise.
6420 (armv8.4-a): Add fp16fml as part of fp16.
6421 * config/arm/arm.h (TARGET_FP16FML): Define.
6422 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6423 when appropriate.
6424 * config/arm/arm-modes.def (V2HF): Define.
6425 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6426 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6427 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6428 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6429 vfmsl_low, vfmsl_high): New set of builtins.
6430 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6431 (vfml_op): New code attribute.
6432 (VFMLHALVES): New int iterator.
6433 (VFML, VFMLSEL): New mode attributes.
6434 (V_reg): Define mapping for V2HF.
6435 (V_hi, V_lo): New mode attributes.
6436 (VF_constraint): Likewise.
6437 (vfml_half, vfml_half_selector): New int attributes.
6438 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6439 define_expand.
6440 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6441 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6442 New define_insn.
6443 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6444 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6445 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6446 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6447 documentation.
6448 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6449 Document new effective target and option set.
6451 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6453 * config/arm/arm-cpus.in (armv8_4): New feature.
6454 (ARMv8_4a): New fgroup.
6455 (armv8.4-a): New arch.
6456 * config/arm/arm-tables.opt: Regenerate.
6457 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6458 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6459 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6460 Add matching rules for -march=armv8.4-a and extensions.
6461 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6463 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6465 PR target/81821
6466 * config/rx/rx.md (BW): New mode attribute.
6467 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6469 2018-01-11 Richard Biener <rguenther@suse.de>
6471 PR tree-optimization/83435
6472 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6473 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6474 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6476 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6477 Alan Hayward <alan.hayward@arm.com>
6478 David Sherwood <david.sherwood@arm.com>
6480 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6481 field.
6482 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6483 (aarch64_print_address_internal): Use it to check for a zero offset.
6485 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6486 Alan Hayward <alan.hayward@arm.com>
6487 David Sherwood <david.sherwood@arm.com>
6489 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6490 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6491 Return a poly_int64 rather than a HOST_WIDE_INT.
6492 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6493 rather than a HOST_WIDE_INT.
6494 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6495 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6496 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6497 final_offset from HOST_WIDE_INT to poly_int64.
6498 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6499 to_constant when getting the number of units in an Advanced SIMD
6500 mode.
6501 (aarch64_builtin_vectorized_function): Check for a constant number
6502 of units.
6503 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6504 GET_MODE_SIZE.
6505 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6506 attribute instead of GET_MODE_NUNITS.
6507 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6508 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6509 GET_MODE_SIZE for fixed-size registers.
6510 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6511 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6512 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6513 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6514 (aarch64_print_operand, aarch64_print_address_internal)
6515 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6516 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6517 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6518 Handle polynomial GET_MODE_SIZE.
6519 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6520 wider than SImode without modification.
6521 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6522 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6523 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6524 passing and returning SVE modes.
6525 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6526 rather than GEN_INT.
6527 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6528 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6529 (aarch64_allocate_and_probe_stack_space): Likewise.
6530 (aarch64_layout_frame): Cope with polynomial offsets.
6531 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6532 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6533 polynomial offsets.
6534 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6535 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6536 poly_int64 rather than a HOST_WIDE_INT.
6537 (aarch64_get_separate_components, aarch64_process_components)
6538 (aarch64_expand_prologue, aarch64_expand_epilogue)
6539 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6540 (aarch64_anchor_offset): New function, split out from...
6541 (aarch64_legitimize_address): ...here.
6542 (aarch64_builtin_vectorization_cost): Handle polynomial
6543 TYPE_VECTOR_SUBPARTS.
6544 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6545 GET_MODE_NUNITS.
6546 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6547 number of elements from the PARALLEL rather than the mode.
6548 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6549 rather than GET_MODE_BITSIZE.
6550 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6551 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6552 (aarch64_expand_vec_perm_const_1): Handle polynomial
6553 d->perm.length () and d->perm elements.
6554 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6555 Apply to_constant to d->perm elements.
6556 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6557 polynomial CONST_VECTOR_NUNITS.
6558 (aarch64_move_pointer): Take amount as a poly_int64 rather
6559 than an int.
6560 (aarch64_progress_pointer): Avoid temporary variable.
6561 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6562 the mode attribute instead of GET_MODE.
6564 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6565 Alan Hayward <alan.hayward@arm.com>
6566 David Sherwood <david.sherwood@arm.com>
6568 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6569 x exists before using it.
6570 (aarch64_add_constant_internal): Rename to...
6571 (aarch64_add_offset_1): ...this. Replace regnum with separate
6572 src and dest rtxes. Handle the case in which they're different,
6573 including when the offset is zero. Replace scratchreg with an rtx.
6574 Use 2 additions if there is no spare register into which we can
6575 move a 16-bit constant.
6576 (aarch64_add_constant): Delete.
6577 (aarch64_add_offset): Replace reg with separate src and dest
6578 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6579 Use aarch64_add_offset_1.
6580 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6581 an rtx rather than an int. Take the delta as a poly_int64
6582 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6583 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6584 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6585 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6586 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6587 and aarch64_add_sp.
6588 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6589 aarch64_add_constant.
6591 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6593 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6594 Use scalar_float_mode.
6596 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6598 * config/aarch64/aarch64-simd.md
6599 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6600 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6601 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6602 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6603 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6604 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6605 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6606 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6607 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6608 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6610 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6612 PR target/83514
6613 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6614 targ_options->x_arm_arch_string is non NULL.
6616 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6618 * config/aarch64/aarch64.h
6619 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6621 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6623 PR target/82096
6624 * expmed.c (emit_store_flag_force): Swap if const op0
6625 and change VOIDmode to mode of op0.
6627 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6629 PR rtl-optimization/83761
6630 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6631 than bytes to mode_for_size.
6633 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6635 PR middle-end/83189
6636 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6637 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6638 profile.
6640 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6642 PR middle-end/83575
6643 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6644 when in layout mode.
6645 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6646 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6647 partition fixup.
6649 2018-01-10 Michael Collison <michael.collison@arm.com>
6651 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6652 * config/aarch64/aarch64-option-extension.def: Add
6653 AARCH64_OPT_EXTENSION of 'fp16fml'.
6654 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6655 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6656 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6657 * config/aarch64/constraints.md (Ui7): New constraint.
6658 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6659 (VFMLA_SEL_W): Ditto.
6660 (f16quad): Ditto.
6661 (f16mac1): Ditto.
6662 (VFMLA16_LOW): New int iterator.
6663 (VFMLA16_HIGH): Ditto.
6664 (UNSPEC_FMLAL): New unspec.
6665 (UNSPEC_FMLSL): Ditto.
6666 (UNSPEC_FMLAL2): Ditto.
6667 (UNSPEC_FMLSL2): Ditto.
6668 (f16mac): New code attribute.
6669 * config/aarch64/aarch64-simd-builtins.def
6670 (aarch64_fmlal_lowv2sf): Ditto.
6671 (aarch64_fmlsl_lowv2sf): Ditto.
6672 (aarch64_fmlalq_lowv4sf): Ditto.
6673 (aarch64_fmlslq_lowv4sf): Ditto.
6674 (aarch64_fmlal_highv2sf): Ditto.
6675 (aarch64_fmlsl_highv2sf): Ditto.
6676 (aarch64_fmlalq_highv4sf): Ditto.
6677 (aarch64_fmlslq_highv4sf): Ditto.
6678 (aarch64_fmlal_lane_lowv2sf): Ditto.
6679 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6680 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6681 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6682 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6683 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6684 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6685 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6686 (aarch64_fmlal_lane_highv2sf): Ditto.
6687 (aarch64_fmlsl_lane_highv2sf): Ditto.
6688 (aarch64_fmlal_laneq_highv2sf): Ditto.
6689 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6690 (aarch64_fmlalq_lane_highv4sf): Ditto.
6691 (aarch64_fmlsl_lane_highv4sf): Ditto.
6692 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6693 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6694 * config/aarch64/aarch64-simd.md:
6695 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6696 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6697 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6698 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6699 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6700 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6701 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6702 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6703 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6704 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6705 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6706 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6707 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6708 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6709 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6710 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6711 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6712 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6713 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6714 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6715 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6716 (vfmlsl_low_u32): Ditto.
6717 (vfmlalq_low_u32): Ditto.
6718 (vfmlslq_low_u32): Ditto.
6719 (vfmlal_high_u32): Ditto.
6720 (vfmlsl_high_u32): Ditto.
6721 (vfmlalq_high_u32): Ditto.
6722 (vfmlslq_high_u32): Ditto.
6723 (vfmlal_lane_low_u32): Ditto.
6724 (vfmlsl_lane_low_u32): Ditto.
6725 (vfmlal_laneq_low_u32): Ditto.
6726 (vfmlsl_laneq_low_u32): Ditto.
6727 (vfmlalq_lane_low_u32): Ditto.
6728 (vfmlslq_lane_low_u32): Ditto.
6729 (vfmlalq_laneq_low_u32): Ditto.
6730 (vfmlslq_laneq_low_u32): Ditto.
6731 (vfmlal_lane_high_u32): Ditto.
6732 (vfmlsl_lane_high_u32): Ditto.
6733 (vfmlal_laneq_high_u32): Ditto.
6734 (vfmlsl_laneq_high_u32): Ditto.
6735 (vfmlalq_lane_high_u32): Ditto.
6736 (vfmlslq_lane_high_u32): Ditto.
6737 (vfmlalq_laneq_high_u32): Ditto.
6738 (vfmlslq_laneq_high_u32): Ditto.
6739 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6740 (AARCH64_FL_FOR_ARCH8_4): New.
6741 (AARCH64_ISA_F16FML): New ISA flag.
6742 (TARGET_F16FML): New feature flag for fp16fml.
6743 (doc/invoke.texi): Document new fp16fml option.
6745 2018-01-10 Michael Collison <michael.collison@arm.com>
6747 * config/aarch64/aarch64-builtins.c:
6748 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6749 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6750 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6751 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6752 (AARCH64_ISA_SHA3): New ISA flag.
6753 (TARGET_SHA3): New feature flag for sha3.
6754 * config/aarch64/iterators.md (sha512_op): New int attribute.
6755 (CRYPTO_SHA512): New int iterator.
6756 (UNSPEC_SHA512H): New unspec.
6757 (UNSPEC_SHA512H2): Ditto.
6758 (UNSPEC_SHA512SU0): Ditto.
6759 (UNSPEC_SHA512SU1): Ditto.
6760 * config/aarch64/aarch64-simd-builtins.def
6761 (aarch64_crypto_sha512hqv2di): New builtin.
6762 (aarch64_crypto_sha512h2qv2di): Ditto.
6763 (aarch64_crypto_sha512su0qv2di): Ditto.
6764 (aarch64_crypto_sha512su1qv2di): Ditto.
6765 (aarch64_eor3qv8hi): Ditto.
6766 (aarch64_rax1qv2di): Ditto.
6767 (aarch64_xarqv2di): Ditto.
6768 (aarch64_bcaxqv8hi): Ditto.
6769 * config/aarch64/aarch64-simd.md:
6770 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6771 (aarch64_crypto_sha512su0qv2di): Ditto.
6772 (aarch64_crypto_sha512su1qv2di): Ditto.
6773 (aarch64_eor3qv8hi): Ditto.
6774 (aarch64_rax1qv2di): Ditto.
6775 (aarch64_xarqv2di): Ditto.
6776 (aarch64_bcaxqv8hi): Ditto.
6777 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6778 (vsha512h2q_u64): Ditto.
6779 (vsha512su0q_u64): Ditto.
6780 (vsha512su1q_u64): Ditto.
6781 (veor3q_u16): Ditto.
6782 (vrax1q_u64): Ditto.
6783 (vxarq_u64): Ditto.
6784 (vbcaxq_u16): Ditto.
6785 * config/arm/types.md (crypto_sha512): New type attribute.
6786 (crypto_sha3): Ditto.
6787 (doc/invoke.texi): Document new sha3 option.
6789 2018-01-10 Michael Collison <michael.collison@arm.com>
6791 * config/aarch64/aarch64-builtins.c:
6792 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6793 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6794 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6795 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6796 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6797 (AARCH64_ISA_SM4): New ISA flag.
6798 (TARGET_SM4): New feature flag for sm4.
6799 * config/aarch64/aarch64-simd-builtins.def
6800 (aarch64_sm3ss1qv4si): Ditto.
6801 (aarch64_sm3tt1aq4si): Ditto.
6802 (aarch64_sm3tt1bq4si): Ditto.
6803 (aarch64_sm3tt2aq4si): Ditto.
6804 (aarch64_sm3tt2bq4si): Ditto.
6805 (aarch64_sm3partw1qv4si): Ditto.
6806 (aarch64_sm3partw2qv4si): Ditto.
6807 (aarch64_sm4eqv4si): Ditto.
6808 (aarch64_sm4ekeyqv4si): Ditto.
6809 * config/aarch64/aarch64-simd.md:
6810 (aarch64_sm3ss1qv4si): Ditto.
6811 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6812 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6813 (aarch64_sm4eqv4si): Ditto.
6814 (aarch64_sm4ekeyqv4si): Ditto.
6815 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6816 (sm3part_op): Ditto.
6817 (CRYPTO_SM3TT): Ditto.
6818 (CRYPTO_SM3PART): Ditto.
6819 (UNSPEC_SM3SS1): New unspec.
6820 (UNSPEC_SM3TT1A): Ditto.
6821 (UNSPEC_SM3TT1B): Ditto.
6822 (UNSPEC_SM3TT2A): Ditto.
6823 (UNSPEC_SM3TT2B): Ditto.
6824 (UNSPEC_SM3PARTW1): Ditto.
6825 (UNSPEC_SM3PARTW2): Ditto.
6826 (UNSPEC_SM4E): Ditto.
6827 (UNSPEC_SM4EKEY): Ditto.
6828 * config/aarch64/constraints.md (Ui2): New constraint.
6829 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6830 * config/arm/types.md (crypto_sm3): New type attribute.
6831 (crypto_sm4): Ditto.
6832 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6833 (vsm3tt1aq_u32): Ditto.
6834 (vsm3tt1bq_u32): Ditto.
6835 (vsm3tt2aq_u32): Ditto.
6836 (vsm3tt2bq_u32): Ditto.
6837 (vsm3partw1q_u32): Ditto.
6838 (vsm3partw2q_u32): Ditto.
6839 (vsm4eq_u32): Ditto.
6840 (vsm4ekeyq_u32): Ditto.
6841 (doc/invoke.texi): Document new sm4 option.
6843 2018-01-10 Michael Collison <michael.collison@arm.com>
6845 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6846 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6847 (AARCH64_FL_FOR_ARCH8_4): New.
6848 (AARCH64_FL_V8_4): New flag.
6849 (doc/invoke.texi): Document new armv8.4-a option.
6851 2018-01-10 Michael Collison <michael.collison@arm.com>
6853 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6854 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6855 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6856 * config/aarch64/aarch64-option-extension.def: Add
6857 AARCH64_OPT_EXTENSION of 'sha2'.
6858 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6859 (crypto): Disable sha2 and aes if crypto disabled.
6860 (crypto): Enable aes and sha2 if enabled.
6861 (simd): Disable sha2 and aes if simd disabled.
6862 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6863 New flags.
6864 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6865 (TARGET_SHA2): New feature flag for sha2.
6866 (TARGET_AES): New feature flag for aes.
6867 * config/aarch64/aarch64-simd.md:
6868 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6869 conditional on TARGET_AES.
6870 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6871 (aarch64_crypto_sha1hsi): Make pattern conditional
6872 on TARGET_SHA2.
6873 (aarch64_crypto_sha1hv4si): Ditto.
6874 (aarch64_be_crypto_sha1hv4si): Ditto.
6875 (aarch64_crypto_sha1su1v4si): Ditto.
6876 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6877 (aarch64_crypto_sha1su0v4si): Ditto.
6878 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6879 (aarch64_crypto_sha256su0v4si): Ditto.
6880 (aarch64_crypto_sha256su1v4si): Ditto.
6881 (doc/invoke.texi): Document new aes and sha2 options.
6883 2018-01-10 Martin Sebor <msebor@redhat.com>
6885 PR tree-optimization/83781
6886 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6887 as string arrays.
6889 2018-01-11 Martin Sebor <msebor@gmail.com>
6890 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6892 PR tree-optimization/83501
6893 PR tree-optimization/81703
6895 * tree-ssa-strlen.c (get_string_cst): Rename...
6896 (get_string_len): ...to this. Handle global constants.
6897 (handle_char_store): Adjust.
6899 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6900 Jim Wilson <jimw@sifive.com>
6902 * config/riscv/riscv-protos.h (riscv_output_return): New.
6903 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6904 (riscv_attribute_table, riscv_output_return),
6905 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6906 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6907 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6908 function.
6909 (riscv_expand_prologue): Add early return for naked function.
6910 (riscv_expand_epilogue): Likewise.
6911 (riscv_function_ok_for_sibcall): Return false for naked function.
6912 (riscv_set_current_function): New.
6913 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6914 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6915 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6916 * doc/extend.texi (RISC-V Function Attributes): New.
6918 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6920 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6921 check for 128-bit long double before checking TCmode.
6922 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6923 128-bit long doubles before checking TFmode or TCmode.
6924 (FLOAT128_IBM_P): Likewise.
6926 2018-01-10 Martin Sebor <msebor@redhat.com>
6928 PR tree-optimization/83671
6929 * builtins.c (c_strlen): Unconditionally return zero for the empty
6930 string.
6931 Use -Warray-bounds for warnings.
6932 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6933 for non-constant array indices with COMPONENT_REF, arrays of
6934 arrays, and pointers to arrays.
6935 (gimple_fold_builtin_strlen): Determine and set length range for
6936 non-constant character arrays.
6938 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6940 PR middle-end/81897
6941 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6942 empty blocks.
6944 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6946 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6948 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6950 PR target/83399
6951 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6952 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6953 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6954 indexed_or_indirect_operand predicate.
6955 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6956 (*vsx_le_perm_load_v8hi): Likewise.
6957 (*vsx_le_perm_load_v16qi): Likewise.
6958 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6959 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6960 (*vsx_le_perm_store_v8hi): Likewise.
6961 (*vsx_le_perm_store_v16qi): Likewise.
6962 (eight unnamed splitters): Likewise.
6964 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6966 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6967 * config/rs6000/emmintrin.h: Likewise.
6968 * config/rs6000/mmintrin.h: Likewise.
6969 * config/rs6000/xmmintrin.h: Likewise.
6971 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6973 PR c++/43486
6974 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6975 "public_flag".
6976 * tree.c (tree_nop_conversion): Return true for location wrapper
6977 nodes.
6978 (maybe_wrap_with_location): New function.
6979 (selftest::check_strip_nops): New function.
6980 (selftest::test_location_wrappers): New function.
6981 (selftest::tree_c_tests): Call it.
6982 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6983 (maybe_wrap_with_location): New decl.
6984 (EXPR_LOCATION_WRAPPER_P): New macro.
6985 (location_wrapper_p): New inline function.
6986 (tree_strip_any_location_wrapper): New inline function.
6988 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6990 PR target/83735
6991 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6992 stack_realign_offset for the largest alignment of stack slot
6993 actually used.
6994 (ix86_find_max_used_stack_alignment): New function.
6995 (ix86_finalize_stack_frame_flags): Use it. Set
6996 max_used_stack_alignment if we don't realign stack.
6997 * config/i386/i386.h (machine_function): Add
6998 max_used_stack_alignment.
7000 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7002 * config/arm/arm.opt (-mbranch-cost): New option.
7003 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7004 account.
7006 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7008 PR target/83629
7009 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7010 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7012 2018-01-10 Richard Biener <rguenther@suse.de>
7014 PR debug/83765
7015 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7016 early out so it also covers the case where we have a non-NULL
7017 origin.
7019 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7021 PR tree-optimization/83753
7022 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7023 for non-strided grouped accesses if the number of elements is 1.
7025 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7027 PR target/81616
7028 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7029 * i386.h (TARGET_USE_GATHER): Define.
7030 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7032 2018-01-10 Martin Liska <mliska@suse.cz>
7034 PR bootstrap/82831
7035 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7036 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7037 partitioning.
7038 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7039 CLEANUP_NO_PARTITIONING is not set.
7041 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7043 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7044 for vectors, as a partial revert of r254296.
7045 * rtl.h (const_vec_p): Delete.
7046 (const_vec_duplicate_p): Don't test for vector CONSTs.
7047 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7048 * expmed.c (make_tree): Likewise.
7050 Revert:
7051 * common.md (E, F): Use CONSTANT_P instead of checking for
7052 CONST_VECTOR.
7053 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7054 checking for CONST_VECTOR.
7056 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7058 PR middle-end/83575
7059 * predict.c (force_edge_cold): Handle in more sane way edges
7060 with no prediction.
7062 2018-01-09 Carl Love <cel@us.ibm.com>
7064 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7065 V4SI, V4SF types.
7066 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7067 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7068 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7069 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7070 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7071 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7072 * config/rs6000/rs6000-protos.h: Add extern defition for
7073 rs6000_generate_float2_double_code.
7074 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7075 function.
7076 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7077 (float2_v2df): Add define_expand.
7079 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7081 PR target/83628
7082 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7083 op_mode in the force_to_mode call.
7085 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7087 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7088 instead of checking each element individually.
7089 (aarch64_evpc_uzp): Likewise.
7090 (aarch64_evpc_zip): Likewise.
7091 (aarch64_evpc_ext): Likewise.
7092 (aarch64_evpc_rev): Likewise.
7093 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7094 instead of checking each element individually. Return true without
7095 generating rtl if
7096 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7097 whether all selected elements come from the same input, instead of
7098 checking each element individually. Remove calls to gen_rtx_REG,
7099 start_sequence and end_sequence and instead assert that no rtl is
7100 generated.
7102 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7104 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7105 order of HIGH and CONST checks.
7107 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7109 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7110 if the destination isn't an SSA_NAME.
7112 2018-01-09 Richard Biener <rguenther@suse.de>
7114 PR tree-optimization/83668
7115 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7116 move prologue...
7117 (canonicalize_loop_form): ... here, renamed from ...
7118 (canonicalize_loop_closed_ssa_form): ... this and amended to
7119 swap successor edges for loop exit blocks to make us use
7120 the RPO order we need for initial schedule generation.
7122 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7124 PR tree-optimization/64811
7125 * match.pd: When optimizing comparisons with Inf, avoid
7126 introducing or losing exceptions from comparisons with NaN.
7128 2018-01-09 Martin Liska <mliska@suse.cz>
7130 PR sanitizer/82517
7131 * asan.c (shadow_mem_size): Add gcc_assert.
7133 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7135 Don't save registers in main().
7137 PR target/83738
7138 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7139 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7140 * config/avr/avr.c (avr_set_current_function): Don't error if
7141 naked, OS_task or OS_main are specified at the same time.
7142 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7143 OS_main.
7144 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7145 attribute.
7146 * common/config/avr/avr-common.c (avr_option_optimization_table):
7147 Switch on -mmain-is-OS_task for optimizing compilations.
7149 2018-01-09 Richard Biener <rguenther@suse.de>
7151 PR tree-optimization/83572
7152 * graphite.c: Include cfganal.h.
7153 (graphite_transform_loops): Connect infinite loops to exit
7154 and remove fake edges at the end.
7156 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7158 * ipa-inline.c (edge_badness): Revert accidental checkin.
7160 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7162 PR ipa/80763
7163 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7164 symbols; not inline clones.
7166 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7168 PR target/83507
7169 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7170 hard registers. Formatting fixes.
7172 PR preprocessor/83722
7173 * gcc.c (try_generate_repro): Pass
7174 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7175 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7176 do_report_bug.
7178 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7179 Kito Cheng <kito.cheng@gmail.com>
7181 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7182 (riscv_leaf_function_p): Delete.
7183 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7185 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7187 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7188 function.
7189 (do_ifelse): New function.
7190 (do_isel): New function.
7191 (do_sub3): New function.
7192 (do_add3): New function.
7193 (do_load_mask_compare): New function.
7194 (do_overlap_load_compare): New function.
7195 (expand_compare_loop): New function.
7196 (expand_block_compare): Call expand_compare_loop() when appropriate.
7197 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7198 option description.
7199 (-mblock-compare-inline-loop-limit): New option.
7201 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7203 PR target/83677
7204 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7205 Reverse order of second and third operands in first alternative.
7206 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7207 of first and second elements in UNSPEC_VPERMR vector.
7208 (altivec_expand_vec_perm_le): Likewise.
7210 2018-01-08 Jeff Law <law@redhat.com>
7212 PR rtl-optimizatin/81308
7213 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7214 (process_switch): If group_case_labels makes a change, then set
7215 cfg_altered.
7216 (pass_convert_switch::execute): If a switch is converted, then
7217 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7219 PR rtl-optimization/81308
7220 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7221 splitting insns.
7223 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7225 PR target/83663 - Revert r255946
7226 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7227 generation for cases where splatting a value is not useful.
7228 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7229 across a vec_duplicate and a paradoxical subreg forming a vector
7230 mode to a vec_concat.
7232 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7234 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7235 -march=armv8.3-a variants.
7236 * config/arm/t-multilib: Likewise.
7237 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7239 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7241 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7242 to generate rtl.
7243 (cceq_ior_compare_complement): Give it a name so I can use it, and
7244 change boolean_or_operator predicate to boolean_operator so it can
7245 be used to generate a crand.
7246 (eqne): New code iterator.
7247 (bd/bd_neg): New code_attrs.
7248 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7249 a single define_insn.
7250 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7251 decrement (bdnzt/bdnzf/bdzt/bdzf).
7252 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7253 with the new names of the branch decrement patterns, and added the
7254 names of the branch decrement conditional patterns.
7256 2018-01-08 Richard Biener <rguenther@suse.de>
7258 PR tree-optimization/83563
7259 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7260 cache.
7262 2018-01-08 Richard Biener <rguenther@suse.de>
7264 PR middle-end/83713
7265 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7267 2018-01-08 Richard Biener <rguenther@suse.de>
7269 PR tree-optimization/83685
7270 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7271 references to abnormals.
7273 2018-01-08 Richard Biener <rguenther@suse.de>
7275 PR lto/83719
7276 * dwarf2out.c (output_indirect_strings): Handle empty
7277 skeleton_debug_str_hash.
7278 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7280 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7282 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7283 (emit_store_direct): Likewise.
7284 (arc_trampoline_adjust_address): Likewise.
7285 (arc_asm_trampoline_template): New function.
7286 (arc_initialize_trampoline): Use asm_trampoline_template.
7287 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7288 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7289 * config/arc/arc.md (flush_icache): Delete pattern.
7291 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7293 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7294 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7295 munaligned-access.
7297 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7299 PR target/83681
7300 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7301 by not USED_FOR_TARGET.
7302 (make_pass_resolve_sw_modes): Likewise.
7304 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7306 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7307 USED_FOR_TARGET.
7309 2018-01-08 Richard Biener <rguenther@suse.de>
7311 PR middle-end/83580
7312 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7314 2018-01-08 Richard Biener <rguenther@suse.de>
7316 PR middle-end/83517
7317 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7319 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7321 PR middle-end/81897
7322 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7323 basic blocks with a small number of successors.
7324 (convert_control_dep_chain_into_preds): Improve handling of
7325 forwarder blocks.
7326 (dump_predicates): Split apart into...
7327 (dump_pred_chain): ...here...
7328 (dump_pred_info): ...and here.
7329 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7330 (can_chain_union_be_invalidated_p): Improve check for invalidation
7331 of paths.
7332 (uninit_uses_cannot_happen): Avoid unnecessary if
7333 convert_control_dep_chain_into_preds yielded nothing.
7335 2018-01-06 Martin Sebor <msebor@redhat.com>
7337 PR tree-optimization/83640
7338 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7339 subtracting negative offset from size.
7340 (builtin_access::overlap): Adjust offset bounds of the access to fall
7341 within the size of the object if possible.
7343 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7345 PR rtl-optimization/83699
7346 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7347 extract_bit_field_as_subreg to cases in which the extracted
7348 value is also a vector.
7350 * lra-constraints.c (process_alt_operands): Test for the equivalence
7351 substitutions when detecting a possible reload cycle.
7353 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7355 PR debug/83480
7356 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7357 by default if flag_selective_schedling{,2}. Formatting fixes.
7359 PR rtl-optimization/83682
7360 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7361 if it has non-VECTOR_MODE element mode.
7362 (vec_duplicate_p): Likewise.
7364 PR middle-end/83694
7365 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7366 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7368 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7370 PR target/83604
7371 * config/i386/i386-builtin.def
7372 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7373 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7374 Require also OPTION_MASK_ISA_AVX512F in addition to
7375 OPTION_MASK_ISA_GFNI.
7376 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7377 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7378 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7379 to OPTION_MASK_ISA_GFNI.
7380 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7381 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7382 OPTION_MASK_ISA_AVX512BW.
7383 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7384 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7385 addition to OPTION_MASK_ISA_GFNI.
7386 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7387 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7388 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7389 to OPTION_MASK_ISA_GFNI.
7390 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7391 a requirement for all ISAs rather than any of them with a few
7392 exceptions.
7393 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7394 processing.
7395 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7396 bitmasks to be enabled with 3 exceptions, instead of requiring any
7397 enabled ISA with lots of exceptions.
7398 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7399 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7400 Change avx512bw in isa attribute to avx512f.
7401 * config/i386/sgxintrin.h: Add license boilerplate.
7402 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7403 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7404 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7405 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7406 defined.
7407 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7408 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7409 temporarily sse2 rather than sse if not enabled already.
7411 PR target/83604
7412 * config/i386/sse.md (VI248_VLBW): Rename to ...
7413 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7414 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7415 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7416 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7417 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7418 mode iterator instead of VI248_VLBW.
7420 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7422 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7423 (record_modified): Skip clobbers; add debug output.
7424 (param_change_prob): Use sreal frequencies.
7426 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7428 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7429 punt for user-aligned variables.
7431 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7433 * tree-chrec.c (chrec_contains_symbols): Return true for
7434 POLY_INT_CST.
7436 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7438 PR target/82439
7439 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7440 of (x|y) == x for BICS pattern.
7442 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7444 PR tree-optimization/83605
7445 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7446 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7447 can throw.
7449 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7451 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7452 * config/epiphany/rtems.h: New file.
7454 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7455 Uros Bizjak <ubizjak@gmail.com>
7457 PR target/83554
7458 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7459 QIreg_operand instead of register_operand predicate.
7460 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7461 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7462 comments instead of -fmitigate[-_]rop.
7464 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7466 PR bootstrap/81926
7467 * cgraphunit.c (symbol_table::compile): Switch to text_section
7468 before calling assembly_start debug hook.
7469 * run-rtl-passes.c (run_rtl_passes): Likewise.
7470 Include output.h.
7472 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7474 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7475 range_int_cst_p rather than !symbolic_range_p before calling
7476 extract_range_from_multiplicative_op_1.
7478 2018-01-04 Jeff Law <law@redhat.com>
7480 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7481 redundant test in assertion.
7483 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7485 * doc/rtl.texi: Document machine_mode wrapper classes.
7487 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7489 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7490 using tree_to_uhwi.
7492 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7494 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7495 the VEC_PERM_EXPR fold to fail.
7497 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7499 PR debug/83585
7500 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7501 to switched_sections.
7503 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7505 PR target/83680
7506 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7507 test for d.testing.
7509 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7511 PR target/83387
7512 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7513 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7515 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7517 PR debug/83666
7518 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7519 is BLKmode and bitpos not zero or mode change is needed.
7521 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7523 PR target/83675
7524 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7525 TARGET_VIS2.
7527 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7529 PR target/83628
7530 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7531 instead of MULT rtx. Update all corresponding splitters.
7532 (*saddl_se): Ditto.
7533 (*ssub<modesuffix>): Ditto.
7534 (*ssubl_se): Ditto.
7535 (*cmp_sadd_di): Update split patterns.
7536 (*cmp_sadd_si): Ditto.
7537 (*cmp_sadd_sidi): Ditto.
7538 (*cmp_ssub_di): Ditto.
7539 (*cmp_ssub_si): Ditto.
7540 (*cmp_ssub_sidi): Ditto.
7541 * config/alpha/predicates.md (const23_operand): New predicate.
7542 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7543 Look for ASHIFT, not MULT inner operand.
7544 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7546 2018-01-04 Martin Liska <mliska@suse.cz>
7548 PR gcov-profile/83669
7549 * gcov.c (output_intermediate_file): Add version to intermediate
7550 gcov file.
7551 * doc/gcov.texi: Document new field 'version' in intermediate
7552 file format. Fix location of '-k' option of gcov command.
7554 2018-01-04 Martin Liska <mliska@suse.cz>
7556 PR ipa/82352
7557 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7559 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7561 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7563 2018-01-03 Martin Sebor <msebor@redhat.com>
7565 PR tree-optimization/83655
7566 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7567 checking calls with invalid arguments.
7569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7571 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7572 (vectorizable_mask_load_store): Delete.
7573 (vectorizable_call): Return false for masked loads and stores.
7574 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7575 instead of gimple_assign_rhs1.
7576 (vectorizable_load): Handle IFN_MASK_LOAD.
7577 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7581 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7582 split out from..,
7583 (vectorizable_mask_load_store): ...here.
7584 (vectorizable_load): ...and here.
7586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7588 * tree-vect-stmts.c (vect_build_all_ones_mask)
7589 (vect_build_zero_merge_argument): New functions, split out from...
7590 (vectorizable_load): ...here.
7592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7594 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7595 split out from...
7596 (vectorizable_mask_load_store): ...here.
7597 (vectorizable_store): ...and here.
7599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7601 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7602 split out from...
7603 (vectorizable_mask_load_store): ...here.
7605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7607 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7608 (vect_model_store_cost): Take a vec_load_store_type instead of a
7609 vect_def_type.
7610 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7611 (vect_model_store_cost): Take a vec_load_store_type instead of a
7612 vect_def_type.
7613 (vectorizable_mask_load_store): Update accordingly.
7614 (vectorizable_store): Likewise.
7615 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7619 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7620 IFN_MASK_LOAD calls here rather than...
7621 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7624 Alan Hayward <alan.hayward@arm.com>
7625 David Sherwood <david.sherwood@arm.com>
7627 * expmed.c (extract_bit_field_1): For vector extracts,
7628 fall back to extract_bit_field_as_subreg if vec_extract
7629 isn't available.
7631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7636 they are variable or constant sized.
7637 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7638 slots for constant-sized data.
7640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7641 Alan Hayward <alan.hayward@arm.com>
7642 David Sherwood <david.sherwood@arm.com>
7644 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7645 handling COND_EXPRs with boolean comparisons, try to find a better
7646 basis for the mask type than the boolean itself.
7648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7650 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7651 is calculated and how it can be overridden.
7652 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7653 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7654 if defined.
7655 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7656 if nonzero.
7658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7663 Remove the mode argument.
7664 (aarch64_simd_valid_immediate): Remove the mode and inverse
7665 arguments.
7666 * config/aarch64/iterators.md (bitsize): New iterator.
7667 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7668 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7669 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7670 aarch64_simd_valid_immediate.
7671 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7672 (aarch64_reg_or_bic_imm): Likewise.
7673 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7674 with an insn_type enum and msl with a modifier_type enum.
7675 Replace element_width with a scalar_mode. Change the shift
7676 to unsigned int. Add constructors for scalar_float_mode and
7677 scalar_int_mode elements.
7678 (aarch64_vect_float_const_representable_p): Delete.
7679 (aarch64_can_const_movi_rtx_p)
7680 (aarch64_simd_scalar_immediate_valid_for_move)
7681 (aarch64_simd_make_constant): Update call to
7682 aarch64_simd_valid_immediate.
7683 (aarch64_advsimd_valid_immediate_hs): New function.
7684 (aarch64_advsimd_valid_immediate): Likewise.
7685 (aarch64_simd_valid_immediate): Remove mode and inverse
7686 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7687 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7688 and aarch64_float_const_representable_p on the result.
7689 (aarch64_output_simd_mov_immediate): Remove mode argument.
7690 Update call to aarch64_simd_valid_immediate and use of
7691 simd_immediate_info.
7692 (aarch64_output_scalar_simd_mov_immediate): Update call
7693 accordingly.
7695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7696 Alan Hayward <alan.hayward@arm.com>
7697 David Sherwood <david.sherwood@arm.com>
7699 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7700 (mode_nunits): Likewise CONST_MODE_NUNITS.
7701 * machmode.def (ADJUST_NUNITS): Document.
7702 * genmodes.c (mode_data::need_nunits_adj): New field.
7703 (blank_mode): Update accordingly.
7704 (adj_nunits): New variable.
7705 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7706 parameter.
7707 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7708 listed in adj_nunits.
7709 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7710 listed in adj_nunits. Don't emit case statements for such modes.
7711 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7712 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7713 nothing if adj_nunits is nonnull.
7714 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7715 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7716 (emit_mode_fbit): Update use of print_maybe_const_decl.
7717 (emit_move_size): Likewise. Treat the array as non-const
7718 if adj_nunits.
7719 (emit_mode_adjustments): Handle adj_nunits.
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7723 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7724 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7725 (VECTOR_MODES): Use it.
7726 (make_vector_modes): Take the prefix as an argument.
7728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7729 Alan Hayward <alan.hayward@arm.com>
7730 David Sherwood <david.sherwood@arm.com>
7732 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7733 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7734 for MODE_VECTOR_BOOL.
7735 * machmode.def (VECTOR_BOOL_MODE): Document.
7736 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7737 (make_vector_bool_mode): New function.
7738 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7739 MODE_VECTOR_BOOL.
7740 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7741 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7742 Likewise.
7743 * stor-layout.c (int_mode_for_mode): Likewise.
7744 * tree.c (build_vector_type_for_mode): Likewise.
7745 * varasm.c (output_constant_pool_2): Likewise.
7746 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7747 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7748 for MODE_VECTOR_BOOL.
7749 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7750 of mode class checks.
7751 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7752 instead of a list of mode class checks.
7753 (expand_vector_scalar_condition): Likewise.
7754 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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_size): Change from unsigned short to
7761 poly_uint16_pod.
7762 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7763 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7764 or if measurement_type is not polynomial.
7765 (fixed_size_mode::includes_p): Check for constant-sized modes.
7766 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7767 return a poly_uint16 rather than an unsigned short.
7768 (emit_mode_size): Change the type of mode_size from unsigned short
7769 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7770 (emit_mode_adjustments): Cope with polynomial vector sizes.
7771 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7772 for GET_MODE_SIZE.
7773 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7774 for GET_MODE_SIZE.
7775 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7776 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7777 * caller-save.c (setup_save_areas): Likewise.
7778 (replace_reg_with_saved_mem): Likewise.
7779 * calls.c (emit_library_call_value_1): Likewise.
7780 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7781 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7782 (gen_lowpart_for_combine): Likewise.
7783 * convert.c (convert_to_integer_1): Likewise.
7784 * cse.c (equiv_constant, cse_insn): Likewise.
7785 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7786 (cselib_subst_to_values): Likewise.
7787 * dce.c (word_dce_process_block): Likewise.
7788 * df-problems.c (df_word_lr_mark_ref): Likewise.
7789 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7790 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7791 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7792 (rtl_for_decl_location): Likewise.
7793 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7794 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7795 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7796 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7797 (expand_expr_real_1): Likewise.
7798 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7799 (pad_below): Likewise.
7800 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7801 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7802 * ira.c (get_subreg_tracking_sizes): Likewise.
7803 * ira-build.c (ira_create_allocno_objects): Likewise.
7804 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7805 (ira_sort_regnos_for_alter_reg): Likewise.
7806 * ira-costs.c (record_operand_costs): Likewise.
7807 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7808 (resolve_simple_move): Likewise.
7809 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7810 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7811 (lra_constraints): Likewise.
7812 (CONST_POOL_OK_P): Reject variable-sized modes.
7813 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7814 (add_pseudo_to_slot, lra_spill): Likewise.
7815 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7816 * optabs-query.c (get_best_extraction_insn): Likewise.
7817 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7818 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7819 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7820 * recog.c (offsettable_address_addr_space_p): Likewise.
7821 * regcprop.c (maybe_mode_change): Likewise.
7822 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7823 * regrename.c (build_def_use): Likewise.
7824 * regstat.c (dump_reg_info): Likewise.
7825 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7826 (find_reloads, find_reloads_subreg_address): Likewise.
7827 * reload1.c (eliminate_regs_1): Likewise.
7828 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7829 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7830 (simplify_binary_operation_1, simplify_subreg): Likewise.
7831 * targhooks.c (default_function_arg_padding): Likewise.
7832 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7833 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7834 (verify_gimple_assign_ternary): Likewise.
7835 * tree-inline.c (estimate_move_cost): Likewise.
7836 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7837 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7838 (get_address_cost_ainc): Likewise.
7839 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7840 (vect_supportable_dr_alignment): Likewise.
7841 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7842 (vectorizable_reduction): Likewise.
7843 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7844 (vectorizable_operation, vectorizable_load): Likewise.
7845 * tree.c (build_same_sized_truth_vector_type): Likewise.
7846 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7847 * var-tracking.c (emit_note_insn_var_location): Likewise.
7848 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7849 (ADDR_VEC_ALIGN): Likewise.
7851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7852 Alan Hayward <alan.hayward@arm.com>
7853 David Sherwood <david.sherwood@arm.com>
7855 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7856 unsigned short.
7857 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7858 or if measurement_type is polynomial.
7859 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7860 * combine.c (make_extraction): Likewise.
7861 * dse.c (find_shift_sequence): Likewise.
7862 * dwarf2out.c (mem_loc_descriptor): Likewise.
7863 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7864 (extract_bit_field, extract_low_bits): Likewise.
7865 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7866 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7867 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7868 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7869 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7870 * reload.c (find_reloads): Likewise.
7871 * reload1.c (alter_reg): Likewise.
7872 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7873 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7874 * tree-if-conv.c (predicate_mem_writes): Likewise.
7875 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7876 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7877 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7878 * valtrack.c (dead_debug_insert_temp): Likewise.
7879 * varasm.c (mergeable_constant_section): Likewise.
7880 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7883 Alan Hayward <alan.hayward@arm.com>
7884 David Sherwood <david.sherwood@arm.com>
7886 * expr.c (expand_assignment): Cope with polynomial mode sizes
7887 when assigning to a CONCAT.
7889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7890 Alan Hayward <alan.hayward@arm.com>
7891 David Sherwood <david.sherwood@arm.com>
7893 * machmode.h (mode_precision): Change from unsigned short to
7894 poly_uint16_pod.
7895 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7896 short.
7897 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7898 or if measurement_type is not polynomial.
7899 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7900 in which the mode is already known to be a scalar_int_mode.
7901 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7902 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7903 initializer.
7904 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7905 for GET_MODE_PRECISION.
7906 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7907 for GET_MODE_PRECISION.
7908 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7909 as polynomial.
7910 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7911 (expand_field_assignment, make_extraction): Likewise.
7912 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7913 (get_last_value): Likewise.
7914 * convert.c (convert_to_integer_1): Likewise.
7915 * cse.c (cse_insn): Likewise.
7916 * expr.c (expand_expr_real_1): Likewise.
7917 * lra-constraints.c (simplify_operand_subreg): Likewise.
7918 * optabs-query.c (can_atomic_load_p): Likewise.
7919 * optabs.c (expand_atomic_load): Likewise.
7920 (expand_atomic_store): Likewise.
7921 * ree.c (combine_reaching_defs): Likewise.
7922 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7923 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7924 * tree.h (type_has_mode_precision_p): Likewise.
7925 * ubsan.c (instrument_si_overflow): Likewise.
7927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7932 polynomial numbers of units.
7933 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7934 (valid_vector_subparts_p): New function.
7935 (build_vector_type): Remove temporary shim and take the number
7936 of units as a poly_uint64 rather than an int.
7937 (build_opaque_vector_type): Take the number of units as a
7938 poly_uint64 rather than an int.
7939 * tree.c (build_vector_from_ctor): Handle polynomial
7940 TYPE_VECTOR_SUBPARTS.
7941 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7942 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7943 (build_vector_from_val): If the number of units is variable,
7944 use build_vec_duplicate_cst for constant operands and
7945 VEC_DUPLICATE_EXPR otherwise.
7946 (make_vector_type): Remove temporary is_constant ().
7947 (build_vector_type, build_opaque_vector_type): Take the number of
7948 units as a poly_uint64 rather than an int.
7949 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7950 VECTOR_CST_NELTS.
7951 * cfgexpand.c (expand_debug_expr): Likewise.
7952 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7953 (store_constructor, expand_expr_real_1): Likewise.
7954 (const_scalar_mask_from_tree): Likewise.
7955 * fold-const-call.c (fold_const_reduction): Likewise.
7956 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7957 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7958 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7959 (fold_relational_const): Likewise.
7960 (native_interpret_vector): Likewise. Change the size from an
7961 int to an unsigned int.
7962 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7963 TYPE_VECTOR_SUBPARTS.
7964 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7965 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7966 duplicating a non-constant operand into a variable-length vector.
7967 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7968 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7969 * ipa-icf.c (sem_variable::equals): Likewise.
7970 * match.pd: Likewise.
7971 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7972 * print-tree.c (print_node): Likewise.
7973 * stor-layout.c (layout_type): Likewise.
7974 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7975 * tree-cfg.c (verify_gimple_comparison): Likewise.
7976 (verify_gimple_assign_binary): Likewise.
7977 (verify_gimple_assign_ternary): Likewise.
7978 (verify_gimple_assign_single): Likewise.
7979 * tree-pretty-print.c (dump_generic_node): Likewise.
7980 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7981 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7982 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7983 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7984 (vect_shift_permute_load_chain): Likewise.
7985 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7986 (expand_vector_condition, optimize_vector_constructor): Likewise.
7987 (lower_vec_perm, get_compute_type): Likewise.
7988 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7989 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7990 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7991 (vect_recog_mask_conversion_pattern): Likewise.
7992 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7993 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7994 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7995 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7996 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7997 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7998 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7999 (supportable_widening_operation): Likewise.
8000 (supportable_narrowing_operation): Likewise.
8001 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8002 Likewise.
8003 * varasm.c (output_constant): Likewise.
8005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8006 Alan Hayward <alan.hayward@arm.com>
8007 David Sherwood <david.sherwood@arm.com>
8009 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8010 so that both the length == 3 and length != 3 cases set up their
8011 own permute vectors. Add comments explaining why we know the
8012 number of elements is constant.
8013 (vect_permute_load_chain): Likewise.
8015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8016 Alan Hayward <alan.hayward@arm.com>
8017 David Sherwood <david.sherwood@arm.com>
8019 * machmode.h (mode_nunits): Change from unsigned char to
8020 poly_uint16_pod.
8021 (ONLY_FIXED_SIZE_MODES): New macro.
8022 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8023 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8024 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8025 New typedefs.
8026 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8027 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8028 or if measurement_type is not polynomial.
8029 * genmodes.c (ZERO_COEFFS): New macro.
8030 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8031 poly_uint16.
8032 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8033 Use ZERO_COEFFS when emitting initializers.
8034 * data-streamer.h (bp_pack_poly_value): New function.
8035 (bp_unpack_poly_value): Likewise.
8036 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8037 for GET_MODE_NUNITS.
8038 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8039 for GET_MODE_NUNITS.
8040 * tree.c (make_vector_type): Remove temporary shim and make
8041 the real function take the number of units as a poly_uint64
8042 rather than an int.
8043 (build_vector_type_for_mode): Handle polynomial nunits.
8044 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8045 * emit-rtl.c (const_vec_series_p_1): Likewise.
8046 (gen_rtx_CONST_VECTOR): Likewise.
8047 * fold-const.c (test_vec_duplicate_folding): Likewise.
8048 * genrecog.c (validate_pattern): Likewise.
8049 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8050 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8051 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8052 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8053 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8054 * rtlanal.c (subreg_get_info): Likewise.
8055 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8056 (vect_grouped_load_supported): Likewise.
8057 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8058 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8059 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8060 (simplify_const_unary_operation, simplify_binary_operation_1)
8061 (simplify_const_binary_operation, simplify_ternary_operation)
8062 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8063 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8064 instead of CONST_VECTOR_NUNITS.
8065 * varasm.c (output_constant_pool_2): Likewise.
8066 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8067 explicit-encoded elements in the XVEC for variable-length vectors.
8069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8071 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8074 Alan Hayward <alan.hayward@arm.com>
8075 David Sherwood <david.sherwood@arm.com>
8077 * coretypes.h (fixed_size_mode): Declare.
8078 (fixed_size_mode_pod): New typedef.
8079 * builtins.h (target_builtins::x_apply_args_mode)
8080 (target_builtins::x_apply_result_mode): Change type to
8081 fixed_size_mode_pod.
8082 * builtins.c (apply_args_size, apply_result_size, result_vector)
8083 (expand_builtin_apply_args_1, expand_builtin_apply)
8084 (expand_builtin_return): Update accordingly.
8086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8088 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8089 * cselib.c (cselib_hash_rtx): Likewise.
8090 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8091 CONST_VECTOR encoding.
8093 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8094 Jeff Law <law@redhat.com>
8096 PR target/83641
8097 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8098 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8099 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8100 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8102 PR target/83641
8103 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8104 explicitly probe *sp in a noreturn function if there were any callee
8105 register saves or frame pointer is needed.
8107 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8109 PR debug/83621
8110 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8111 BLKmode for ternary, binary or unary expressions.
8113 PR debug/83645
8114 * var-tracking.c (delete_vta_debug_insn): New inline function.
8115 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8116 insns from get_insns () to NULL instead of each bb separately.
8117 Use delete_vta_debug_insn. No longer static.
8118 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8119 delete_vta_debug_insns callers.
8120 * rtl.h (delete_vta_debug_insns): Declare.
8121 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8122 instead of variable_tracking_main.
8124 2018-01-03 Martin Sebor <msebor@redhat.com>
8126 PR tree-optimization/83603
8127 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8128 arguments past the endof the argument list in functions declared
8129 without a prototype.
8130 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8131 Avoid checking when arguments are null.
8133 2018-01-03 Martin Sebor <msebor@redhat.com>
8135 PR c/83559
8136 * doc/extend.texi (attribute const): Fix a typo.
8137 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8138 issuing -Wsuggest-attribute for void functions.
8140 2018-01-03 Martin Sebor <msebor@redhat.com>
8142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8143 offset_int::from instead of wide_int::to_shwi.
8144 (maybe_diag_overlap): Remove assertion.
8145 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8146 * gimple-ssa-sprintf.c (format_directive): Same.
8147 (parse_directive): Same.
8148 (sprintf_dom_walker::compute_format_length): Same.
8149 (try_substitute_return_value): Same.
8151 2018-01-03 Jeff Law <law@redhat.com>
8153 PR middle-end/83654
8154 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8155 non-constant residual for zero at runtime and avoid probing in
8156 that case. Reorganize code for trailing problem to mirror handling
8157 of the residual.
8159 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8161 PR tree-optimization/83501
8162 * tree-ssa-strlen.c (get_string_cst): New.
8163 (handle_char_store): Call get_string_cst.
8165 2018-01-03 Martin Liska <mliska@suse.cz>
8167 PR tree-optimization/83593
8168 * tree-ssa-strlen.c: Include tree-cfg.h.
8169 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8170 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8171 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8172 to false.
8173 (strlen_dom_walker::before_dom_children): Call
8174 gimple_purge_dead_eh_edges. Dump tranformation with details
8175 dump flags.
8176 (strlen_dom_walker::before_dom_children): Update call by adding
8177 new argument cleanup_eh.
8178 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8180 2018-01-03 Martin Liska <mliska@suse.cz>
8182 PR ipa/83549
8183 * cif-code.def (VARIADIC_THUNK): New enum value.
8184 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8185 thunks.
8187 2018-01-03 Jan Beulich <jbeulich@suse.com>
8189 * sse.md (mov<mode>_internal): Tighten condition for when to use
8190 vmovdqu<ssescalarsize> for TI and OI modes.
8192 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8194 Update copyright years.
8196 2018-01-03 Martin Liska <mliska@suse.cz>
8198 PR ipa/83594
8199 * ipa-visibility.c (function_and_variable_visibility): Skip
8200 functions with noipa attribure.
8202 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8204 * gcc.c (process_command): Update copyright notice dates.
8205 * gcov-dump.c (print_version): Ditto.
8206 * gcov.c (print_version): Ditto.
8207 * gcov-tool.c (print_version): Ditto.
8208 * gengtype.c (create_file): Ditto.
8209 * doc/cpp.texi: Bump @copying's copyright year.
8210 * doc/cppinternals.texi: Ditto.
8211 * doc/gcc.texi: Ditto.
8212 * doc/gccint.texi: Ditto.
8213 * doc/gcov.texi: Ditto.
8214 * doc/install.texi: Ditto.
8215 * doc/invoke.texi: Ditto.
8217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8219 * vector-builder.h (vector_builder::m_full_nelts): Change from
8220 unsigned int to poly_uint64.
8221 (vector_builder::full_nelts): Update prototype accordingly.
8222 (vector_builder::new_vector): Likewise.
8223 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8224 (vector_builder::operator ==): Likewise.
8225 (vector_builder::finalize): Likewise.
8226 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8227 Take the number of elements as a poly_uint64 rather than an
8228 unsigned int.
8229 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8230 from unsigned int to poly_uint64.
8231 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8232 (vec_perm_indices::new_vector): Likewise.
8233 (vec_perm_indices::length): Likewise.
8234 (vec_perm_indices::nelts_per_input): Likewise.
8235 (vec_perm_indices::input_nelts): Likewise.
8236 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8237 number of elements per input as a poly_uint64 rather than an
8238 unsigned int. Use the original encoding for variable-length
8239 vectors, rather than clamping each individual element.
8240 For the second and subsequent elements in each pattern,
8241 clamp the step and base before clamping their sum.
8242 (vec_perm_indices::series_p): Handle polynomial element counts.
8243 (vec_perm_indices::all_in_range_p): Likewise.
8244 (vec_perm_indices_to_tree): Likewise.
8245 (vec_perm_indices_to_rtx): Likewise.
8246 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8247 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8248 (tree_vector_builder::new_binary_operation): Handle polynomial
8249 element counts. Return false if we need to know the number
8250 of elements at compile time.
8251 * fold-const.c (fold_vec_perm): Punt if the number of elements
8252 isn't known at compile time.
8254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8256 * vec-perm-indices.h (vec_perm_builder): Change element type
8257 from HOST_WIDE_INT to poly_int64.
8258 (vec_perm_indices::element_type): Update accordingly.
8259 (vec_perm_indices::clamp): Handle polynomial element_types.
8260 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8261 (vec_perm_indices::all_in_range_p): Likewise.
8262 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8263 than shwi trees.
8264 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8265 polynomial vec_perm_indices element types.
8266 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8267 * fold-const.c (fold_vec_perm): Likewise.
8268 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8269 * tree-vect-generic.c (lower_vec_perm): Likewise.
8270 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8271 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8272 element type to HOST_WIDE_INT.
8274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8275 Alan Hayward <alan.hayward@arm.com>
8276 David Sherwood <david.sherwood@arm.com>
8278 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8279 rather than an int. Use plus_constant.
8280 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8281 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8284 Alan Hayward <alan.hayward@arm.com>
8285 David Sherwood <david.sherwood@arm.com>
8287 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8288 a HOST_WIDE_INT to a poly_int64.
8290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8291 Alan Hayward <alan.hayward@arm.com>
8292 David Sherwood <david.sherwood@arm.com>
8294 * calls.c (load_register_parameters): Cope with polynomial
8295 mode sizes. Require a constant size for BLKmode parameters
8296 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8297 forces a parameter to be padded at the lsb end in order to
8298 fill a complete number of words, require the parameter size
8299 to be ordered wrt UNITS_PER_WORD.
8301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8302 Alan Hayward <alan.hayward@arm.com>
8303 David Sherwood <david.sherwood@arm.com>
8305 * reload1.c (spill_stack_slot_width): Change element type
8306 from unsigned int to poly_uint64_pod.
8307 (alter_reg): Treat mode sizes as polynomial.
8309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8310 Alan Hayward <alan.hayward@arm.com>
8311 David Sherwood <david.sherwood@arm.com>
8313 * reload.c (complex_word_subreg_p): New function.
8314 (reload_inner_reg_of_subreg, push_reload): Use it.
8316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8317 Alan Hayward <alan.hayward@arm.com>
8318 David Sherwood <david.sherwood@arm.com>
8320 * lra-constraints.c (process_alt_operands): Reject matched
8321 operands whose sizes aren't ordered.
8322 (match_reload): Refer to this check here.
8324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8325 Alan Hayward <alan.hayward@arm.com>
8326 David Sherwood <david.sherwood@arm.com>
8328 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8329 that the mode size is in the set {1, 2, 4, 8, 16}.
8331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8332 Alan Hayward <alan.hayward@arm.com>
8333 David Sherwood <david.sherwood@arm.com>
8335 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8336 Use plus_constant instead of gen_rtx_PLUS.
8338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8339 Alan Hayward <alan.hayward@arm.com>
8340 David Sherwood <david.sherwood@arm.com>
8342 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8343 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8344 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8345 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8346 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8347 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8348 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8349 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8350 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8351 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8352 a poly_int64.
8353 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8354 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8355 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8356 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8357 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8358 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8359 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8360 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8361 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8362 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8363 function.
8364 * expr.c (emit_move_resolve_push): Treat the input and result
8365 of PUSH_ROUNDING as a poly_int64.
8366 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8367 (emit_push_insn): Likewise.
8368 * lra-eliminations.c (mark_not_eliminable): Likewise.
8369 * recog.c (push_operand): Likewise.
8370 * reload1.c (elimination_effects): Likewise.
8371 * rtlanal.c (nonzero_bits1): Likewise.
8372 * calls.c (store_one_arg): Likewise. Require the padding to be
8373 known at compile time.
8375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8376 Alan Hayward <alan.hayward@arm.com>
8377 David Sherwood <david.sherwood@arm.com>
8379 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8380 Use plus_constant instead of gen_rtx_PLUS.
8382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8383 Alan Hayward <alan.hayward@arm.com>
8384 David Sherwood <david.sherwood@arm.com>
8386 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8387 rather than an int.
8389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8390 Alan Hayward <alan.hayward@arm.com>
8391 David Sherwood <david.sherwood@arm.com>
8393 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8394 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8395 via stack temporaries. Treat the mode size as polynomial too.
8397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8398 Alan Hayward <alan.hayward@arm.com>
8399 David Sherwood <david.sherwood@arm.com>
8401 * expr.c (expand_expr_real_2): When handling conversions involving
8402 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8403 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8404 as a poly_uint64 too.
8406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8407 Alan Hayward <alan.hayward@arm.com>
8408 David Sherwood <david.sherwood@arm.com>
8410 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8413 Alan Hayward <alan.hayward@arm.com>
8414 David Sherwood <david.sherwood@arm.com>
8416 * combine.c (can_change_dest_mode): Handle polynomial
8417 REGMODE_NATURAL_SIZE.
8418 * expmed.c (store_bit_field_1): Likewise.
8419 * expr.c (store_constructor): Likewise.
8420 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8421 and polynomial REGMODE_NATURAL_SIZE.
8422 (gen_lowpart_common): Likewise.
8423 * reginfo.c (record_subregs_of_mode): Likewise.
8424 * rtlanal.c (read_modify_subreg_p): Likewise.
8426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8427 Alan Hayward <alan.hayward@arm.com>
8428 David Sherwood <david.sherwood@arm.com>
8430 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8431 numbers of elements.
8433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8434 Alan Hayward <alan.hayward@arm.com>
8435 David Sherwood <david.sherwood@arm.com>
8437 * match.pd: Cope with polynomial numbers of vector elements.
8439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8440 Alan Hayward <alan.hayward@arm.com>
8441 David Sherwood <david.sherwood@arm.com>
8443 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8444 in a POINTER_PLUS_EXPR.
8446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8447 Alan Hayward <alan.hayward@arm.com>
8448 David Sherwood <david.sherwood@arm.com>
8450 * omp-simd-clone.c (simd_clone_subparts): New function.
8451 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8452 (ipa_simd_modify_function_body): Likewise.
8454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8455 Alan Hayward <alan.hayward@arm.com>
8456 David Sherwood <david.sherwood@arm.com>
8458 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8459 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8460 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8461 (expand_vector_condition, vector_element): Likewise.
8462 (subparts_gt): New function.
8463 (get_compute_type): Use subparts_gt.
8464 (count_type_subparts): Delete.
8465 (expand_vector_operations_1): Use subparts_gt instead of
8466 count_type_subparts.
8468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8469 Alan Hayward <alan.hayward@arm.com>
8470 David Sherwood <david.sherwood@arm.com>
8472 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8473 (vect_compile_time_alias): ...this new function. Do the calculation
8474 on poly_ints rather than trees.
8475 (vect_prune_runtime_alias_test_list): Update call accordingly.
8477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8478 Alan Hayward <alan.hayward@arm.com>
8479 David Sherwood <david.sherwood@arm.com>
8481 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8482 numbers of units.
8483 (vect_schedule_slp_instance): Likewise.
8485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8486 Alan Hayward <alan.hayward@arm.com>
8487 David Sherwood <david.sherwood@arm.com>
8489 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8490 constant and extern definitions for variable-length vectors.
8491 (vect_get_constant_vectors): Note that the number of units
8492 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8499 of units as polynomial. Choose between WIDE and NARROW based
8500 on multiple_p.
8502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8503 Alan Hayward <alan.hayward@arm.com>
8504 David Sherwood <david.sherwood@arm.com>
8506 * tree-vect-stmts.c (simd_clone_subparts): New function.
8507 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8510 Alan Hayward <alan.hayward@arm.com>
8511 David Sherwood <david.sherwood@arm.com>
8513 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8514 vectors as polynomial. Use build_index_vector for
8515 IFN_GOMP_SIMD_LANE.
8517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8518 Alan Hayward <alan.hayward@arm.com>
8519 David Sherwood <david.sherwood@arm.com>
8521 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8522 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8523 for variable-length vectors.
8524 (vectorizable_mask_load_store): Treat the number of units as
8525 polynomial, asserting that it is constant if the condition has
8526 already been enforced.
8527 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
8534 of units as polynomial. Punt if we can't tell at compile time
8535 which vector contains the final result.
8537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8538 Alan Hayward <alan.hayward@arm.com>
8539 David Sherwood <david.sherwood@arm.com>
8541 * tree-vect-loop.c (vectorizable_induction): Treat the number
8542 of units as polynomial. Punt on SLP inductions. Use an integer
8543 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8544 cast of such a series for variable-length floating-point
8545 reductions.
8547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8548 Alan Hayward <alan.hayward@arm.com>
8549 David Sherwood <david.sherwood@arm.com>
8551 * tree.h (build_index_vector): Declare.
8552 * tree.c (build_index_vector): New function.
8553 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8554 of units as polynomial, forcibly converting it to a constant if
8555 vectorizable_reduction has already enforced the condition.
8556 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8557 to create a {1,2,3,...} vector.
8558 (vectorizable_reduction): Treat the number of units as polynomial.
8559 Choose vectype_in based on the largest scalar element size rather
8560 than the smallest number of units. Enforce the restrictions
8561 relied on above.
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-data-refs.c (vector_alignment_reachable_p): Treat the
8568 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8575 * target.def (autovectorize_vector_sizes): Return the vector sizes
8576 by pointer, using vector_sizes rather than a bitmask.
8577 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8578 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8579 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8580 Likewise.
8581 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8582 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8583 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8584 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8585 * omp-general.c (omp_max_vf): Likewise.
8586 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8587 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8588 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8589 * tree-vect-slp.c (vect_slp_bb): Likewise.
8590 * doc/tm.texi: Regenerate.
8591 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8592 to a poly_uint64.
8593 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8594 the vector size as a poly_uint64 rather than an unsigned int.
8595 (current_vector_size): Change from an unsigned int to a poly_uint64.
8596 (get_vectype_for_scalar_type): Update accordingly.
8597 * tree.h (build_truth_vector_type): Take the size and number of
8598 units as a poly_uint64 rather than an unsigned int.
8599 (build_vector_type): Add a temporary overload that takes
8600 the number of units as a poly_uint64 rather than an unsigned int.
8601 * tree.c (make_vector_type): Likewise.
8602 (build_truth_vector_type): Take the number of units as a poly_uint64
8603 rather than an unsigned int.
8605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8606 Alan Hayward <alan.hayward@arm.com>
8607 David Sherwood <david.sherwood@arm.com>
8609 * target.def (get_mask_mode): Take the number of units and length
8610 as poly_uint64s rather than unsigned ints.
8611 * targhooks.h (default_get_mask_mode): Update accordingly.
8612 * targhooks.c (default_get_mask_mode): Likewise.
8613 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8614 * doc/tm.texi: Regenerate.
8616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8617 Alan Hayward <alan.hayward@arm.com>
8618 David Sherwood <david.sherwood@arm.com>
8620 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8621 * omp-general.c (omp_max_vf): Likewise.
8622 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8623 (expand_omp_simd): Handle polynomial safelen.
8624 * omp-low.c (omplow_simd_context): Add a default constructor.
8625 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8626 (lower_rec_simd_input_clauses): Update accordingly.
8627 (lower_rec_input_clauses): Likewise.
8629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8630 Alan Hayward <alan.hayward@arm.com>
8631 David Sherwood <david.sherwood@arm.com>
8633 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8634 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8635 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8636 (vect_analyze_slp_cost): Likewise.
8637 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8638 (vect_model_load_cost): Likewise.
8640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8641 Alan Hayward <alan.hayward@arm.com>
8642 David Sherwood <david.sherwood@arm.com>
8644 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8645 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8646 from an unsigned int * to a poly_uint64_pod *.
8647 (calculate_unrolling_factor): New function.
8648 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8651 Alan Hayward <alan.hayward@arm.com>
8652 David Sherwood <david.sherwood@arm.com>
8654 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8655 from an unsigned int to a poly_uint64.
8656 (_loop_vec_info::slp_unrolling_factor): Likewise.
8657 (_loop_vec_info::vectorization_factor): Change from an int
8658 to a poly_uint64.
8659 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8660 (vect_get_num_vectors): New function.
8661 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8662 (vect_get_num_copies): Use vect_get_num_vectors.
8663 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8664 to an unsigned int *.
8665 (vect_analyze_data_refs): Change min_vf from an int * to a
8666 poly_uint64 *.
8667 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8668 than an unsigned HOST_WIDE_INT.
8669 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8670 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8671 to an unsigned int *.
8672 (vect_analyze_data_ref_dependences): Likewise.
8673 (vect_compute_data_ref_alignment): Handle polynomial vf.
8674 (vect_enhance_data_refs_alignment): Likewise.
8675 (vect_prune_runtime_alias_test_list): Likewise.
8676 (vect_shift_permute_load_chain): Likewise.
8677 (vect_supportable_dr_alignment): Likewise.
8678 (dependence_distance_ge_vf): Take the vectorization factor as a
8679 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8680 (vect_analyze_data_refs): Change min_vf from an int * to a
8681 poly_uint64 *.
8682 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8683 vfm1 as a poly_uint64 rather than an int. Make the same change
8684 for the returned bound_scalar.
8685 (vect_gen_vector_loop_niters): Handle polynomial vf.
8686 (vect_do_peeling): Likewise. Update call to
8687 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8688 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8689 be constant.
8690 * tree-vect-loop.c (vect_determine_vectorization_factor)
8691 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8692 (vect_get_known_peeling_cost): Likewise.
8693 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8694 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8695 (vect_transform_loop): Likewise. Use the lowest possible VF when
8696 updating the upper bounds of the loop.
8697 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8698 rather than an int.
8699 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8700 polynomial unroll factors.
8701 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8702 (vect_make_slp_decision): Likewise.
8703 (vect_supported_load_permutation_p): Likewise, and polynomial
8704 vf too.
8705 (vect_analyze_slp_cost): Handle polynomial vf.
8706 (vect_slp_analyze_node_operations): Likewise.
8707 (vect_slp_analyze_bb_1): Likewise.
8708 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8709 than an unsigned HOST_WIDE_INT.
8710 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8711 (vectorizable_load): Handle polynomial vf.
8712 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8713 a poly_uint64.
8714 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8717 Alan Hayward <alan.hayward@arm.com>
8718 David Sherwood <david.sherwood@arm.com>
8720 * match.pd: Handle bit operations involving three constants
8721 and try to fold one pair.
8723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8725 * tree-vect-loop-manip.c: Include gimple-fold.h.
8726 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8727 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8728 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8729 Add a path that uses a step of VF instead of 1, but disable it
8730 for now.
8731 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8732 and niters_no_overflow parameters. Update calls to
8733 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8734 Create a new SSA name if the latter choses to use a ste other
8735 than zero, and return it via niters_vector_mult_vf_var.
8736 * tree-vect-loop.c (vect_transform_loop): Update calls to
8737 vect_do_peeling, vect_gen_vector_loop_niters and
8738 slpeel_make_loop_iterate_ntimes.
8739 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8740 (vect_gen_vector_loop_niters): Update declarations after above changes.
8742 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8744 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8745 128-bit round to integer instructions.
8746 (ceil<mode>2): Likewise.
8747 (btrunc<mode>2): Likewise.
8748 (round<mode>2): Likewise.
8750 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8752 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8753 unaligned VSX load/store on P8/P9.
8754 (expand_block_clear): Allow the use of unaligned VSX
8755 load/store on P8/P9.
8757 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8759 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8760 New function.
8761 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8762 swap associated with both a load and a store.
8764 2018-01-02 Andrew Waterman <andrew@sifive.com>
8766 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8767 * config/riscv/riscv.md (clear_cache): Use it.
8769 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8771 * web.c: Remove out-of-date comment.
8773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8775 * expr.c (fixup_args_size_notes): Check that any existing
8776 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8777 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8778 (emit_single_push_insn): ...here.
8780 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8782 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8783 (const_vector_encoded_nelts): New function.
8784 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8785 (const_vector_int_elt, const_vector_elt): Declare.
8786 * emit-rtl.c (const_vector_int_elt_1): New function.
8787 (const_vector_elt): Likewise.
8788 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8789 of CONST_VECTOR_ELT.
8791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8793 * expr.c: Include rtx-vector-builder.h.
8794 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8795 directly on the tree encoding.
8796 (const_vector_from_tree): Likewise.
8797 * optabs.c: Include rtx-vector-builder.h.
8798 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8799 sequence of "u" values.
8800 * vec-perm-indices.c: Include rtx-vector-builder.h.
8801 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8802 directly on the vec_perm_indices encoding.
8804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8806 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8807 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8808 * rtx-vector-builder.h: New file.
8809 * rtx-vector-builder.c: Likewise.
8810 * rtl.h (rtx_def::u2): Add a const_vector field.
8811 (CONST_VECTOR_NPATTERNS): New macro.
8812 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8813 (CONST_VECTOR_DUPLICATE_P): Likewise.
8814 (CONST_VECTOR_STEPPED_P): Likewise.
8815 (CONST_VECTOR_ENCODED_ELT): Likewise.
8816 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8817 (unwrap_const_vec_duplicate): Likewise.
8818 (const_vec_series_p): Check for a non-duplicated vector encoding.
8819 Say that the function only returns true for integer vectors.
8820 * emit-rtl.c: Include rtx-vector-builder.h.
8821 (gen_const_vec_duplicate_1): Delete.
8822 (gen_const_vector): Call gen_const_vec_duplicate instead of
8823 gen_const_vec_duplicate_1.
8824 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8825 (gen_const_vec_duplicate): Use rtx_vector_builder.
8826 (gen_const_vec_series): Likewise.
8827 (gen_rtx_CONST_VECTOR): Likewise.
8828 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8829 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8830 Build a new vector rather than modifying a CONST_VECTOR in-place.
8831 (handle_special_swappables): Update call accordingly.
8832 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8833 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8834 Build a new vector rather than modifying a CONST_VECTOR in-place.
8835 (handle_special_swappables): Update call accordingly.
8837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8839 * simplify-rtx.c (simplify_const_binary_operation): Use
8840 CONST_VECTOR_ELT instead of XVECEXP.
8842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8844 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8845 the selector elements to be different from the data elements
8846 if the selector is a VECTOR_CST.
8847 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8848 ssizetype for the selector.
8850 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8852 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8853 before testing each element individually.
8854 * tree-vect-generic.c (lower_vec_perm): Likewise.
8856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8858 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8859 * selftest-run-tests.c (selftest::run_tests): Call it.
8860 * vector-builder.h (vector_builder::operator ==): New function.
8861 (vector_builder::operator !=): Likewise.
8862 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8863 (vec_perm_indices::all_from_input_p): New function.
8864 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8865 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8866 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8867 instead of reading the VECTOR_CST directly. Detect whether both
8868 vector inputs are the same before constructing the vec_perm_indices,
8869 and update the number of inputs argument accordingly. Use the
8870 utility functions added above. Only construct sel2 if we need to.
8872 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8874 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8875 the broadcast of the low byte.
8876 (expand_mult_highpart): Use an explicit encoding for the permutes.
8877 * optabs-query.c (can_mult_highpart_p): Likewise.
8878 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8879 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8880 (vectorizable_bswap): Likewise.
8881 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8882 explicit encoding for the power-of-2 permutes.
8883 (vect_permute_store_chain): Likewise.
8884 (vect_grouped_load_supported): Likewise.
8885 (vect_permute_load_chain): Likewise.
8887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8889 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8890 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8891 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8892 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8893 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8894 (vect_gen_perm_mask_any): Likewise.
8896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8898 * int-vector-builder.h: New file.
8899 * vec-perm-indices.h: Include int-vector-builder.h.
8900 (vec_perm_indices): Redefine as an int_vector_builder.
8901 (auto_vec_perm_indices): Delete.
8902 (vec_perm_builder): Redefine as a stand-alone class.
8903 (vec_perm_indices::vec_perm_indices): New function.
8904 (vec_perm_indices::clamp): Likewise.
8905 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8906 (vec_perm_indices::new_vector): New function.
8907 (vec_perm_indices::new_expanded_vector): Update for new
8908 vec_perm_indices class.
8909 (vec_perm_indices::rotate_inputs): New function.
8910 (vec_perm_indices::all_in_range_p): Operate directly on the
8911 encoded form, without computing elided elements.
8912 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8913 encoding. Update for new vec_perm_indices class.
8914 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8915 the given vec_perm_builder.
8916 (expand_vec_perm_var): Update vec_perm_builder constructor.
8917 (expand_mult_highpart): Use vec_perm_builder instead of
8918 auto_vec_perm_indices.
8919 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8920 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8921 or double series encoding as appropriate.
8922 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8923 vec_perm_indices instead of auto_vec_perm_indices.
8924 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8925 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8926 (vect_permute_store_chain): Likewise.
8927 (vect_grouped_load_supported): Likewise.
8928 (vect_permute_load_chain): Likewise.
8929 (vect_shift_permute_load_chain): Likewise.
8930 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8931 (vect_transform_slp_perm_load): Likewise.
8932 (vect_schedule_slp_instance): Likewise.
8933 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8934 (vectorizable_mask_load_store): Likewise.
8935 (vectorizable_bswap): Likewise.
8936 (vectorizable_store): Likewise.
8937 (vectorizable_load): Likewise.
8938 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8939 vec_perm_indices instead of auto_vec_perm_indices. Use
8940 tree_to_vec_perm_builder to read the vector from a tree.
8941 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8942 vec_perm_builder instead of a vec_perm_indices.
8943 (have_whole_vector_shift): Use vec_perm_builder and
8944 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8945 truncation to calc_vec_perm_mask_for_shift.
8946 (vect_create_epilog_for_reduction): Likewise.
8947 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8948 from auto_vec_perm_indices to vec_perm_indices.
8949 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8950 instead of changing individual elements.
8951 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8952 the vector in d.perm.
8953 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8954 from auto_vec_perm_indices to vec_perm_indices.
8955 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8956 instead of changing individual elements.
8957 (arm_vectorize_vec_perm_const): Use new_vector to install
8958 the vector in d.perm.
8959 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8960 Update vec_perm_builder constructor.
8961 (rs6000_expand_interleave): Likewise.
8962 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8963 (rs6000_expand_interleave): Likewise.
8965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8967 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8968 to qimode could truncate the indices.
8969 * optabs.c (expand_vec_perm_var): Likewise.
8971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8973 * Makefile.in (OBJS): Add vec-perm-indices.o.
8974 * vec-perm-indices.h: New file.
8975 * vec-perm-indices.c: Likewise.
8976 * target.h (vec_perm_indices): Replace with a forward class
8977 declaration.
8978 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8979 * optabs.h: Include vec-perm-indices.h.
8980 (expand_vec_perm): Delete.
8981 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8982 (expand_vec_perm_const): Declare.
8983 * target.def (vec_perm_const_ok): Replace with...
8984 (vec_perm_const): ...this new hook.
8985 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8986 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8987 * doc/tm.texi: Regenerate.
8988 * optabs.def (vec_perm_const): Delete.
8989 * doc/md.texi (vec_perm_const): Likewise.
8990 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8991 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8992 expand_vec_perm for constant permutation vectors. Assert that
8993 the mode of variable permutation vectors is the integer equivalent
8994 of the mode that is being permuted.
8995 * optabs-query.h (selector_fits_mode_p): Declare.
8996 * optabs-query.c: Include vec-perm-indices.h.
8997 (selector_fits_mode_p): New function.
8998 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8999 is defined, instead of checking whether the vec_perm_const_optab
9000 exists. Use targetm.vectorize.vec_perm_const instead of
9001 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9002 fit in the vector mode before using a variable permute.
9003 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9004 vec_perm_indices instead of an rtx.
9005 (expand_vec_perm): Replace with...
9006 (expand_vec_perm_const): ...this new function. Take the selector
9007 as a vec_perm_indices rather than an rtx. Also take the mode of
9008 the selector. Update call to shift_amt_for_vec_perm_mask.
9009 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9010 Use vec_perm_indices::new_expanded_vector to expand the original
9011 selector into bytes. Check whether the indices fit in the vector
9012 mode before using a variable permute.
9013 (expand_vec_perm_var): Make global.
9014 (expand_mult_highpart): Use expand_vec_perm_const.
9015 * fold-const.c: Includes vec-perm-indices.h.
9016 * tree-ssa-forwprop.c: Likewise.
9017 * tree-vect-data-refs.c: Likewise.
9018 * tree-vect-generic.c: Likewise.
9019 * tree-vect-loop.c: Likewise.
9020 * tree-vect-slp.c: Likewise.
9021 * tree-vect-stmts.c: Likewise.
9022 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9023 Delete.
9024 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9025 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9026 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9027 (aarch64_vectorize_vec_perm_const): ...this new function.
9028 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9029 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9030 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9031 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9032 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9033 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9034 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9035 into...
9036 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9037 check for NEON modes.
9038 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9039 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9040 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9041 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9042 into...
9043 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9044 the old VEC_PERM_CONST conditions.
9045 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9046 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9047 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9048 (ia64_vectorize_vec_perm_const_ok): Merge into...
9049 (ia64_vectorize_vec_perm_const): ...this new function.
9050 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9051 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9052 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9053 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9054 * config/mips/mips.c (mips_expand_vec_perm_const)
9055 (mips_vectorize_vec_perm_const_ok): Merge into...
9056 (mips_vectorize_vec_perm_const): ...this new function.
9057 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9058 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9059 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9060 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9061 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9062 (rs6000_expand_vec_perm_const): Delete.
9063 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9064 Delete.
9065 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9066 (altivec_expand_vec_perm_const_le): Take each operand individually.
9067 Operate on constant selectors rather than rtxes.
9068 (altivec_expand_vec_perm_const): Likewise. Update call to
9069 altivec_expand_vec_perm_const_le.
9070 (rs6000_expand_vec_perm_const): Delete.
9071 (rs6000_vectorize_vec_perm_const_ok): Delete.
9072 (rs6000_vectorize_vec_perm_const): New function.
9073 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9074 an element count and rtx array.
9075 (rs6000_expand_extract_even): Update call accordingly.
9076 (rs6000_expand_interleave): Likewise.
9077 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9078 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9079 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9080 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9081 (rs6000_expand_vec_perm_const): Delete.
9082 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9083 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9084 (altivec_expand_vec_perm_const_le): Take each operand individually.
9085 Operate on constant selectors rather than rtxes.
9086 (altivec_expand_vec_perm_const): Likewise. Update call to
9087 altivec_expand_vec_perm_const_le.
9088 (rs6000_expand_vec_perm_const): Delete.
9089 (rs6000_vectorize_vec_perm_const_ok): Delete.
9090 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9091 reference to the SPE evmerge intructions.
9092 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9093 an element count and rtx array.
9094 (rs6000_expand_extract_even): Update call accordingly.
9095 (rs6000_expand_interleave): Likewise.
9096 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9097 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9098 new function.
9099 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9103 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9104 vector mode and that that mode matches the mode of the data
9105 being permuted.
9106 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9107 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9108 directly using expand_vec_perm_1 when forcing selectors into
9109 registers.
9110 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9114 * optabs-query.h (can_vec_perm_p): Delete.
9115 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9116 * optabs-query.c (can_vec_perm_p): Split into...
9117 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9118 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9119 particular selector is valid.
9120 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9121 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9122 (vect_grouped_load_supported): Likewise.
9123 (vect_shift_permute_load_chain): Likewise.
9124 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9125 (vect_transform_slp_perm_load): Likewise.
9126 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9127 (vectorizable_bswap): Likewise.
9128 (vect_gen_perm_mask_checked): Likewise.
9129 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9130 implementations of variable permutation vectors into account
9131 when deciding which selector to use.
9132 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9133 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9134 with a false third argument.
9135 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9136 to test whether the constant selector is valid and can_vec_perm_var_p
9137 to test whether a variable selector is valid.
9139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9141 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9142 * optabs-query.c (can_vec_perm_p): Likewise.
9143 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9144 instead of vec_perm_indices.
9145 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9146 (vect_gen_perm_mask_checked): Likewise,
9147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9148 (vect_gen_perm_mask_checked): Likewise,
9150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9152 * optabs-query.h (qimode_for_vec_perm): Declare.
9153 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9154 (qimode_for_vec_perm): ...this new function.
9155 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9157 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9159 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9160 does not have a conditional at the top.
9162 2018-01-02 Richard Biener <rguenther@suse.de>
9164 * ipa-inline.c (big_speedup_p): Fix expression.
9166 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9168 PR target/81616
9169 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9170 for generic 4->6.
9172 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9174 PR target/81616
9175 Generic tuning.
9176 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9177 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9178 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9179 cond_taken_branch_cost 3->4.
9181 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9183 PR tree-optimization/83581
9184 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9185 TODO_cleanup_cfg if any changes have been made.
9187 PR middle-end/83608
9188 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9189 convert_modes if target mode has the right side, but different mode
9190 class.
9192 PR middle-end/83609
9193 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9194 last argument when extracting from CONCAT. If either from_real or
9195 from_imag is NULL, use expansion through memory. If result is not
9196 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9197 the parts directly to inner mode, if even that fails, use expansion
9198 through memory.
9200 PR middle-end/83623
9201 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9202 check for bswap in mode rather than HImode and use that in expand_unop
9203 too.
9204 \f
9205 Copyright (C) 2018 Free Software Foundation, Inc.
9207 Copying and distribution of this file, with or without modification,
9208 are permitted in any medium without royalty provided the copyright
9209 notice and this notice are preserved.