1 2018-03-18 Martin Liska <mliska@suse.cz>
3 PR rtl-optimization/84635
4 * regrename.c (build_def_use): Use matches_mode only when
7 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9 PR tree-optimization/84913
10 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11 vectorize chains of COND_EXPRs.
13 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
25 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
26 Kito Cheng <kito.cheng@gmail.com>
28 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
29 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
30 (nds32_adjust_reg_alloc_order): New function.
31 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
33 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
35 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
36 nds32_print_operand, nds32_print_operand_address): Use
37 HOST_WIDE_INT_PRINT_DEC instead.
39 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
41 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
43 2018-03-17 Jakub Jelinek <jakub@redhat.com>
46 * config/i386/i386.c (initial_ix86_tune_features,
47 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
49 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
50 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
51 rather than 1u << ix86_tune. Formatting fix.
52 (ix86_option_override_internal): Change ix86_arch_mask from
53 unsigned int to unsigned HOST_WIDE_INT, initialize to
54 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
55 (ix86_function_specific_restore): Likewise.
57 2018-03-16 Jakub Jelinek <jakub@redhat.com>
60 * postreload.c (reload_combine_recognize_pattern): Perform
61 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
62 truncate_int_for_mode the result for the destination's mode.
65 * hsa-gen.c (mem_type_for_type): Fix comment typo.
66 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
68 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
71 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
74 * lra-assigns.c (lra_split_hard_reg_for): Don't use
75 regno_allocno_class_array and sorted_pseudos.
76 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
77 insns where regno is used.
79 2018-03-16 Martin Liska <mliska@suse.cz>
82 * multiple_target.c (create_dispatcher_calls): Redirect
83 reference in the symbol table.
85 2018-03-16 Martin Liska <mliska@suse.cz>
88 * multiple_target.c (create_dispatcher_calls): Redirect also
91 2018-03-16 Jakub Jelinek <jakub@redhat.com>
95 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
96 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
97 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
99 2018-03-16 Julia Koval <julia.koval@intel.com>
101 * doc/invoke.texi (Skylake Server): Add CLWB.
102 Cannonlake): Remove CLWB.
104 2018-03-16 Jakub Jelinek <jakub@redhat.com>
106 PR tree-optimization/84841
107 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
109 (FLOAT_ONE_CONST_TYPE): Define.
110 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
111 (sort_by_operand_rank): Put entries with higher constant_type last
112 rather than first to match comments.
114 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
116 * config/nios2/nios2.md (movsi_internal): Fix thinko in
119 2018-03-15 Jakub Jelinek <jakub@redhat.com>
122 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
123 check and use address of target always.
125 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
128 * config/i386/i386.c (indirect_thunk_needed): Update comments.
129 (indirect_thunk_bnd_needed): Likewise.
130 (indirect_thunks_used): Likewise.
131 (indirect_thunks_bnd_used): Likewise.
132 (indirect_return_needed): New.
133 (indirect_return_bnd_needed): Likewise.
134 (output_indirect_thunk_function): Add a bool argument for
136 (output_indirect_thunk_function): Don't generate alias for
137 function return thunk.
138 (ix86_code_end): Call output_indirect_thunk_function to generate
139 function return thunks.
140 (ix86_output_function_return): Set indirect_return_bnd_needed
141 and indirect_return_needed instead of indirect_thunk_bnd_needed
142 and indirect_thunk_needed.
144 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
146 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
147 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
148 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
150 2018-03-15 David Malcolm <dmalcolm@redhat.com>
151 Paul Hua <paul.hua.gm@gmail.com>
154 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
156 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
158 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
159 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
162 2018-03-15 Tamar Christina <tamar.christina@arm.com>
165 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
166 instead of GET_MODE_SIZE when comparing Units.
168 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
171 * varasm.c (hash_section): Return an unchangeble hash value
172 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
173 Return !aarch64_can_use_per_function_literal_pools_p ().
175 2018-03-15 Jakub Jelinek <jakub@redhat.com>
178 * optabs.c (emit_conditional_move): Pass address of cmode's copy
179 rather than address of cmode as last argument to prepare_cmp_insn.
181 2018-03-15 Julia Koval <julia.koval@intel.com>
183 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
184 F_AVX512VNNI, F_AVX512BITALG): New.
186 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
189 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
190 insn for floating-point loads and stores.
192 2018-03-14 Carl Love <cel@us.ibm.com>
194 * config/rs6000/rs6000-c.c: Add macro definitions for
195 ALTIVEC_BUILTIN_VEC_PERMXOR.
196 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
197 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
198 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
199 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
201 * config/doc/extend.texi: Add prototypes for vec_permxor.
203 2018-03-14 David Malcolm <dmalcolm@redhat.com>
206 * diagnostic-show-locus.c (class layout_point): Convert m_line
207 from int to linenum_type.
208 (line_span::comparator): Use linenum "compare" function when
209 comparing line numbers.
210 (test_line_span): New function.
211 (layout_range::contains_point): Convert param "row" from int to
213 (layout_range::intersects_line_p): Likewise.
214 (layout::will_show_line_p): Likewise.
215 (layout::print_source_line): Likewise.
216 (layout::should_print_annotation_line_p): Likewise.
217 (layout::print_annotation_line): Likewise.
218 (layout::print_leading_fixits): Likewise.
219 (layout::annotation_line_showed_range_p): Likewise.
220 (struct line_corrections): Likewise for field m_row.
221 (line_corrections::line_corrections): Likewise for param "row".
222 (layout::print_trailing_fixits): Likewise.
223 (layout::get_state_at_point): Likewise.
224 (layout::get_x_bound_for_row): Likewise.
225 (layout::print_line): Likewise.
226 (diagnostic_show_locus): Likewise for locals "last_line" and
228 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
229 * input.c (selftest::test_linenum_comparisons): New function.
230 (selftest::input_c_tests): Call it.
231 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
232 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
233 * selftest.h (ASSERT_GT): New macro.
234 (ASSERT_GT_AT): New macro.
235 (ASSERT_LT): New macro.
236 (ASSERT_LT_AT): New macro.
238 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
240 PR rtl-optimization/84780
241 * combine.c (distribute_links): Don't make a link based on pc_rtx.
243 2018-03-14 Martin Liska <mliska@suse.cz>
245 * tree.c (record_node_allocation_statistics): Use
247 (get_stats_node_kind): New function extracted from
248 record_node_allocation_statistics.
249 (free_node): Use get_stats_node_kind.
251 2018-03-14 Richard Biener <rguenther@suse.de>
253 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
254 that the value-set of ANTIC_IN doesn't grow.
257 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
259 (BB_VISITED_WITH_VISITED_SUCCS): New define.
260 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
262 2018-03-14 Julia Koval <julia.koval@intel.com>
264 * config.gcc (icelake-client, icelake-server): New.
266 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
267 (initial_ix86_arch_features): Ditto.
268 (PTA_SKYLAKE): Add SGX.
269 (PTA_ICELAKE): Remove.
270 (PTA_ICELAKE_CLIENT): New.
271 (PTA_ICELAKE_SERVER): New.
272 (ix86_option_override_internal): Split up icelake on icelake client and
274 (get_builtin_code_for_version): Ditto.
275 (fold_builtin_cpu): Ditto.
276 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
277 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
278 * config/i386/i386.h (processor_type): Ditto.
279 * doc/invoke.texi: Ditto.
281 2018-03-14 Jakub Jelinek <jakub@redhat.com>
284 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
285 INTEGER_CST offset, add it together with bitpos / 8 and
286 sign extend based on POINTER_SIZE.
290 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
293 * config/i386/constraints.md (Yc): New register constraint.
294 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
295 Use Yc constraint for alternative 2 of operand 0. Remove
296 preferred_for_speed attribute.
298 2018-03-14 Richard Biener <rguenther@suse.de>
300 PR tree-optimization/84830
301 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
302 with the old one to avoid oscillations.
304 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
307 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
309 (assign_by_spills): Return a flag of reload assignment failure.
310 Do not process the reload assignment failures. Do not spill other
311 reload pseudos if they has the same reg class. Update n if
313 (lra_assign): Add a return arg. Set up from the result of
314 assign_by_spills call.
315 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
316 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
317 usage_insns if it is not NULL.
318 (spill_hard_reg_in_range): New function.
319 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
320 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
322 (lra_assign): Change prototype.
323 * lra.c (lra): Add code to deal with fails by splitting hard reg
326 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
328 * config/riscv/riscv.opt (mrelax): New option.
329 * config/riscv/riscv.c (riscv_file_start): Emit ".option
330 "norelax" when riscv_mrelax is disabled.
331 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
333 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
336 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
337 reassociation for int modes.
339 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
341 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
342 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
344 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
345 * config/aarch64/aarch64-sve.md
346 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
347 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
348 (*extend<mode><Vwide>2): Rename to...
349 (aarch64_sve_extend<mode><Vwide>2): ...this.
350 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
351 renaming the old pattern to...
352 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
354 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
355 define_expand, renaming the old pattern to...
356 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
357 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
358 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
359 account when deciding which SVE instruction the optab should use.
360 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
362 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
364 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
365 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
366 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
367 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
368 (tlsdesc_small_<mode>): Turn a define_expand and use
369 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
370 (tlsdesc_small_advsimd_<mode>): ...this.
371 (tlsdesc_small_sve_<mode>): New pattern.
373 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
375 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
376 (UNSPEC_UMUL_HIGHPART): New constants.
377 (MUL_HIGHPART): New int iteraor.
378 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
379 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
381 (*<su>mul<mode>3_highpart): New define_insn.
383 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
386 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
389 2018-03-13 Martin Liska <mliska@suse.cz>
392 * (sem_item_optimizer::sem_item_optimizer): Initialize new
394 (sem_item_optimizer::~sem_item_optimizer): Release it.
395 (sem_item_optimizer::merge_classes): Register variable aliases.
396 (sem_item_optimizer::fixup_pt_set): New function.
397 (sem_item_optimizer::fixup_points_to_sets): Likewise.
398 * ipa-icf.h: Declare new variables and functions.
400 2018-03-13 Jakub Jelinek <jakub@redhat.com>
403 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
404 integer_pow2p@2 and test integer_pow2p in condition.
405 (A < 0 ? C : 0): Similarly for @1.
408 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
409 characters starting at p contain '\0' character, don't look beyond
413 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
414 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
417 * reg-stack.c (change_stack): Change update_end var from int to
418 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
419 also call set_block_for_insn on the newly added insns and rescan.
422 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
426 * tree.c (stabilize_reference_1): Return save_expr (e) for
427 STATEMENT_LIST even if it doesn't have side-effects.
429 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
431 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
433 2018-03-12 Renlin Li <renlin.li@arm.com>
435 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
436 aarch64_output_scalar_simd_mov_immediate.
438 2018-03-12 Martin Sebor <msebor@redhat.com>
440 PR tree-optimization/83456
441 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
442 for perfectly overlapping calls to memcpy.
443 (gimple_fold_builtin_memory_chk): Same.
444 (gimple_fold_builtin_strcpy): Handle no-warning.
445 (gimple_fold_builtin_stxcpy_chk): Same.
446 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
448 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
450 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
451 parameter. Use it for SFmode.
452 (rs6000_function_arg_advance_1): Adjust.
453 (rs6000_function_arg): Adjust.
454 (rs6000_gimplify_va_arg): Pass false for that new parameter.
456 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
458 PR rtl-optimization/84169
459 PR rtl-optimization/84780
460 * combine.c (can_combine_p): Check for a 2-insn combination whether
461 the destination register is used between the two insns, too.
463 2018-03-12 Richard Biener <rguenther@suse.de>
465 PR tree-optimization/84803
466 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
467 for refs DR analysis didn't process.
469 2018-03-12 Richard Biener <rguenther@suse.de>
471 PR tree-optimization/84777
472 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
473 force-vectorize loops ignore whether we are optimizing for size.
475 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
477 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
478 (TARGET_MD_ASM_ADJUST): Define.
480 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
481 Kito Cheng <kito.cheng@gmail.com>
482 Chung-Ju Wu <jasonwucj@gmail.com>
484 * config/nds32/nds32.c (nds32_compute_stack_frame,
485 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
486 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
487 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
488 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
489 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
490 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
491 * config/nds32/nds32.md (prologue, epilogue): Use macro
492 NDS32_V3PUSH_AVAILABLE_P to do checking.
494 2018-03-11 Jakub Jelinek <jakub@redhat.com>
497 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
498 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
499 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
500 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
501 addition of most attributes on !orig_type_die or the attribute not
502 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
504 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
505 Chung-Ju Wu <jasonwucj@gmail.com>
507 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
509 * config/nds32/nds32.opt (mvh): New option.
511 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
512 Chung-Ju Wu <jasonwucj@gmail.com>
514 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
516 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
517 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
520 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
521 Chung-Ju Wu <jasonwucj@gmail.com>
523 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
525 * config/nds32/nds32-multiple.md (strlensi): New pattern.
526 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
528 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
529 Kito Cheng <kito.cheng@gmail.com>
530 Chung-Ju Wu <jasonwucj@gmail.com>
532 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
533 UNSPEC_FFMISM and UNSPEC_FLMISM.
534 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
535 for ffb, ffmism and flmism.
536 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
537 (unspec_ffmism): Ditto.
538 (unspec_flmism): Ditto.
539 (nds32_expand_builtin_impl): Check if string extension is available.
540 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
541 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
543 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
546 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
549 * lra-assigns.c (assign_by_spills): Return a flag of reload
550 assignment failure. Do not process the reload assignment
551 failures. Do not spill other reload pseudos if they has the same
553 (lra_assign): Add a return arg. Set up from the result of
554 assign_by_spills call.
555 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
556 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
557 usage_insns if it is not NULL.
558 (spill_hard_reg_in_range): New function.
559 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
560 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
562 (lra_assign): Change prototype.
563 * lra.c (lra): Add code to deal with fails by splitting hard reg
566 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
569 * config/i386/i386.opt: Replace Enforcment with Enforcement.
571 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
574 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
575 (dw_val_node): Add val_symbolic_view.
576 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
577 (symview_upper_bound): New.
578 (new_line_info_table): Initialize symviews_since_reset.
579 (dwarf2out_source_line): Count symviews_since_reset and set
581 (dw_val_equal_p): Handle symview.
582 (add_AT_symview): New.
583 (print_dw_val): Handle symview.
584 (attr_checksum, attr_checksum_ordered): Likewise.
585 (same_dw_val_p, size_of_die): Likewise.
586 (value_format, output_die): Likewise.
587 (add_high_low_attributes): Use add_AT_symview for entry_view.
588 (dwarf2out_finish): Reset symview_upper_bound, clear
591 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
594 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
595 Add strict argument and use it.
596 (rs6000_split_multireg_move): Update for new strict argument.
597 (mem_operand_gpr): Disallow all non-offsettable addresses.
598 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
600 2018-03-09 Jakub Jelinek <jakub@redhat.com>
603 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
604 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
605 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
608 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
609 decl, use remap_type if we want to use the type.
611 2018-03-09 Martin Sebor <msebor@redhat.com>
613 PR tree-optimization/84526
614 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
616 (builtin_access::generic_overlap): Be prepared to handle non-array
619 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
621 PR rtl-optimization/84682
622 * lra-constraints.c (process_address_1): Check is_address flag
623 for address constraints.
624 (process_alt_operands): Likewise.
625 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
626 preprocess_constraints.
627 * recog.h (preprocess_constraints): Add oploc parameter.
629 * recog.c (preprocess_constraints): Test address_operand for
630 CT_ADDRESS constraints.
632 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
635 * lra-assigns.c (assign_by_spills): Return a flag of reload
636 assignment failure. Do not process the reload assignment
637 failures. Do not spill other reload pseudos if they has the same
639 (lra_assign): Add a return arg. Set up from the result of
640 assign_by_spills call.
641 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
642 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
643 usage_insns if it is not NULL.
644 (spill_hard_reg_in_range): New function.
645 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
646 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
648 (lra_assign): Change prototype.
649 * lra.c (lra): Add code to deal with fails by splitting hard reg
652 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
655 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
656 Accept complain bool parameter. Only emit errors if it is true.
657 (arm_parse_cpu_option_name): Likewise.
658 (arm_target_thumb_only): Adjust callers of the above.
659 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
660 prototype to take a default true bool parameter.
661 (arm_parse_arch_option_name): Likewise.
663 2018-03-09 David Malcolm <dmalcolm@redhat.com>
664 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
668 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
669 * configure: Regenerate.
670 * configure.ac ("linker --version-script option"): New.
671 ("linker soname option"): New.
673 2018-03-09 Richard Biener <rguenther@suse.de>
675 PR tree-optimization/84775
676 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
677 immediate uses of predicate stmts and mark them modified.
680 PR tree-optimization/84178
681 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
683 (version_loop_for_if_conversion): Delay update_ssa call.
684 (tree_if_conversion): Delay update_ssa until after predicate
687 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
690 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
691 when the function accesses prior frames.
693 2018-03-08 Jakub Jelinek <jakub@redhat.com>
696 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
697 gen_llsym, otherwise call maybe_gen_llsym.
700 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
701 has ',' character inside of it.
703 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
706 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
707 as clobbering CC_REGNUM.
709 2018-03-08 Richard Biener <rguenther@suse.de>
712 * tree-scalar-evolution.c: Include tree-into-ssa.h.
713 (follow_copies_to_constant): Do not follow SSA names registered
716 2018-03-08 Richard Biener <rguenther@suse.de>
718 PR tree-optimization/84178
719 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
721 (version_loop_for_if_conversion): Delay update_ssa call.
722 (tree_if_conversion): Delay update_ssa until after predicate
725 2018-03-08 David Malcolm <dmalcolm@redhat.com>
727 PR tree-optimization/84178
728 * tree-if-conv.c (release_bb_predicate): Remove the
729 the assertion that the stmts have NULL use_ops.
730 Discard the statements, asserting that they haven't
731 yet been added to a BB.
733 2018-03-08 Richard Biener <rguenther@suse.de>
735 PR tree-optimization/84746
736 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
737 (phi_translate): Pass in destination ANTIC_OUT set.
738 (phi_translate_1): Likewise. For a simplified result lookup
739 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
740 (phi_translate_set): Adjust.
741 (do_pre_regular_insertion): Likewise.
742 (do_pre_partial_partial_insertion): Likewise.
744 2018-03-08 Martin Liska <mliska@suse.cz>
746 PR gcov-profile/84735
747 * doc/gcov.texi: Document usage of profile files.
748 * gcov-io.h: Document changes in the format.
750 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
754 * dwarf2out.c (struct dw_line_info_table): Update comments for
756 (FORCE_RESET_NEXT_VIEW): New.
757 (FORCE_RESETTING_VIEW_P): New.
758 (RESETTING_VIEW_P): Check for -1 too.
759 (ZERO_VIEW_P): Likewise.
760 (new_line_info_table): Force-reset next view.
761 (dwarf2out_begin_function): Likewise.
762 (dwarf2out_source_line): Simplify zero_view_p initialization.
763 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
764 view directly. Omit view when omitting .loc at line 0.
766 2018-03-08 Jakub Jelinek <jakub@redhat.com>
768 PR tree-optimization/84740
769 * tree-switch-conversion.c (process_switch): Call build_constructors
770 only if info.phi_count is non-zero.
772 PR tree-optimization/84739
773 * tree-tailcall.c (find_tail_calls): Check call arguments against
774 DECL_ARGUMENTS (current_function_decl) rather than
775 DECL_ARGUMENTS (func) when checking for tail recursion.
777 2018-03-07 Jakub Jelinek <jakub@redhat.com>
779 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
780 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
781 Volker Reichelt's entry and add entries for people that perform
782 GCC fuzzy testing and report numerous bugs.
784 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
787 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
788 readonly data in sdata, if that is disabled.
789 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
790 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
791 -mreadonly-in-sdata option.
793 2018-03-07 Martin Sebor <msebor@redhat.com>
795 PR tree-optimization/84468
796 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
797 basic block when looking for nul assignment.
799 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
802 * except.h (output_function_exception_table): Adjust prototype.
803 * except.c (output_function_exception_table): Remove FNNAME parameter
804 and add SECTION parameter. Ouput one part of the table at a time.
805 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
806 the first part of the exception table and emit unwind directives.
807 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
808 (i386_pe_seh_cold_init): Likewise.
809 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
810 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
811 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
812 (ix86_output_call_insn): Emit a nop in one more case for SEH.
813 * config/i386/winnt.c: Include except.h.
814 (struct seh_frame_state): Add reg_offset, after_prologue and
815 in_cold_section fields.
816 (i386_pe_seh_end_prologue): Set seh->after_prologue.
817 (i386_pe_seh_cold_init): New function.
818 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
819 to seh->in_cold_section.
820 (seh_emit_push): Record the offset of the push.
821 (seh_emit_save): Record the offet of the save.
822 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
823 Test seh->after_prologue to disregard the epilogue.
824 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
825 (i386_pe_end_cold_function): New function.
827 2018-03-07 Jakub Jelinek <jakub@redhat.com>
830 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
831 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
834 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
836 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
837 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
840 * multiple_target.c: Include tree-inline.h and intl.h.
841 (expand_target_clones): Diagnose and fail if node->definition and
842 !tree_versionable_function_p (node->decl).
844 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
846 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
848 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
849 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
850 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
852 2018-03-06 Jakub Jelinek <jakub@redhat.com>
855 * combine.c (try_combine): Use reg_or_subregno instead of handling
856 just paradoxical SUBREGs and REGs.
858 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
860 * config/arc/arc.c (arc_finalize_pic): Remove function.
861 (arc_must_save_register): We use single base PIC register, remove
862 checks to save/restore the PIC register.
863 (arc_expand_prologue): Likewise.
864 * config/arc/arc-protos.h (arc_set_default_type_attributes):
866 (arc_verify_short): Likewise.
867 (arc_attr_type): Likewise.
868 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
869 (walk_stores): Likewise.
870 (arc_address_cost): Make it static.
871 (arc_verify_short): Likewise.
872 (branch_dest): Likewise.
873 (arc_attr_type): Likewise.
874 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
875 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
876 (arc_final_prescan_insn): Remove inserting the nops due to
877 hardware hazards. It is done in reorg step.
878 (insn_length_variant_t): Remove.
879 (insn_length_parameters_t): Likewise.
880 (arc_insn_length_parameters): Likewise.
881 (arc_get_insn_variants): Likewise.
882 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
884 2018-03-06 Jakub Jelinek <jakub@redhat.com>
887 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
890 PR tree-optimization/84687
891 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
893 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
895 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
897 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
898 Rename to ppc_speculation_barrier.
899 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
900 __builtin_ppc_speculation_barrier.
902 2018-03-05 Jakub Jelinek <jakub@redhat.com>
905 * combine.c (combine_simplify_rtx): Don't try to simplify if
906 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
909 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
911 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
912 to 32 bytes when compiling for POWER9.
914 2018-03-05 Jakub Jelinek <jakub@redhat.com>
917 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
918 regparm >= 3 with no arg reg available also for calls with
919 flag_force_indirect_call. Pass decl to ix86_function_regparm.
922 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
924 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
926 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
929 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
931 2018-03-05 Richard Biener <rguenther@suse.de>
933 PR tree-optimization/84486
934 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
935 When inserting a __builtin_assume_aligned call set the LHS
936 SSA name alignment info accordingly.
938 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
940 PR tree-optimization/84114
941 * config/aarch64/aarch64.c (aarch64_reassociation_width)
942 Avoid reassociation of FLOAT_MODE addition.
944 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
946 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
947 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
948 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
949 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
950 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
951 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
952 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
954 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
955 __builtin_ia32_wbinvd): New builtins.
956 (SPECIAL_ARGS2): New.
957 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
958 (SPECIAL_ARGS2): New.
959 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
960 (ix86_valid_target_attribute_inner_p): Ditto.
961 (ix86_init_mmx_sse_builtins): Add special_args2.
962 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
963 TARGET_WBNOINVD_P): New.
964 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
965 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
966 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
967 * config/i386/immintrin.h (_wbinvd): New intrinsic.
968 * config/i386/pconfigintrin.h: New file.
969 * config/i386/wbnoinvdintrin.h: Ditto.
970 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
971 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
973 2018-03-05 Richard Biener <rguenther@suse.de>
975 PR tree-optimization/84670
976 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
978 (BB_VISITED_WITH_VISITED_SUCCS): New define.
979 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
980 (compute_antic_aux): Only assert the number of values in ANTIC_IN
981 doesn't grow if all successors (recursively) were visited at least
984 2018-03-05 Richard Biener <rguenther@suse.de>
986 PR tree-optimization/84650
987 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
988 if executed in the loop pipeline.
990 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
992 * doc/configfiles.texi (Configuration Files): Move info about
993 conditionalizing $target-protos.h to...
994 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
995 differs from $target-protos.h.
997 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
998 Chung-Ju Wu <jasonwucj@gmail.com>
1000 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1001 * config/nds32/nds32-multiple.md (setmemsi): Define.
1002 * config/nds32/nds32-memory-manipulation.c
1003 (nds32_gen_dup_4_byte_to_word_value): New.
1004 (emit_setmem_word_loop): New.
1005 (emit_setmem_byte_loop): New.
1006 (nds32_expand_setmem_loop): New.
1007 (nds32_expand_setmem_loop_v3m): New.
1008 (nds32_expand_setmem_unroll): New.
1009 (nds32_expand_setmem): New.
1011 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1012 Chung-Ju Wu <jasonwucj@gmail.com>
1014 * config/nds32/nds32-memory-manipulation.c
1015 (nds32_emit_load_store): New.
1016 (nds32_emit_post_inc_load_store): New.
1017 (nds32_emit_mem_move): New.
1018 (nds32_emit_mem_move_block): New.
1019 (nds32_expand_movmemsi_loop_unknown_size): New.
1020 (nds32_expand_movmemsi_loop_known_size): New.
1021 (nds32_expand_movmemsi_loop): New.
1022 (nds32_expand_movmemsi_unroll): New.
1023 (nds32_expand_movmemqi): Rename ...
1024 (nds32_expand_movmemsi): ... to this.
1025 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1026 (movmemsi): ... to this.
1027 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1028 (nds32_expand_movmemsi): ... to this.
1030 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1031 Monk Chiang <sh.chiang04@gmail.com>
1032 Chung-Ju Wu <jasonwucj@gmail.com>
1034 * config/nds32/nds32-protos.h
1035 (nds32_expand_load_multiple): New arguments.
1036 (nds32_expand_store_multiple): Ditto.
1037 (nds32_valid_multiple_load_store): Rename ...
1038 (nds32_valid_multiple_load_store_p): ... to this.
1039 * config/nds32/nds32-memory-manipulation.c
1040 (nds32_expand_load_multiple): Refine implementation.
1041 (nds32_expand_store_multiple): Ditto.
1042 * config/nds32/nds32-multiple.md
1043 (load_multiple): Update nds32_expand_load_multiple interface.
1044 (store_multiple): Update nds32_expand_store_multiple interface.
1045 * config/nds32/nds32-predicates.c
1046 (nds32_valid_multiple_load_store): Rename ...
1047 (nds32_valid_multiple_load_store_p): ... to this and refine
1049 * config/nds32/predicates.md
1050 (nds32_load_multiple_and_update_address_operation): New predicate.
1051 (nds32_store_multiple_and_update_address_operation): New predicate.
1053 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1054 Chung-Ju Wu <jasonwucj@gmail.com>
1056 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1057 (combo): New attribute.
1058 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1060 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1062 * config/nds32/nds32.opt: Change -mcmodel= default value.
1064 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1065 Monk Chiang <sh.chiang04@gmail.com>
1066 Chung-Ju Wu <jasonwucj@gmail.com>
1068 * config/nds32/constants.md (unspec_element): New enum.
1069 * config/nds32/constraints.md (Umw): New constraint.
1070 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1071 * config/nds32/nds32-intrinsic.md: Likewise.
1072 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1073 (nds32_valid_smw_lwm_base_p): New.
1074 (nds32_output_smw_single_word): New.
1075 (nds32_output_lmw_single_word): New.
1076 (nds32_expand_unaligned_load): New.
1077 (nds32_expand_unaligned_store): New.
1078 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1079 (nds32_output_smw_single_word): Declare.
1080 (nds32_output_lmw_single_word): Declare.
1081 (nds32_expand_unaligned_load): Declare.
1082 (nds32_expand_unaligned_store): Declare.
1083 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1084 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1085 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1086 NDS32_BUILTIN_UASTORE_DW.
1087 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1090 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1091 Kito Cheng <kito.cheng@gmail.com>
1092 Chung-Ju Wu <jasonwucj@gmail.com>
1094 * config/nds32/nds32-intrinsic.c
1095 (nds32_expand_builtin_null_ftype_reg): Delete.
1096 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1097 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1098 (nds32_read_argument): New.
1099 (nds32_legitimize_target): Ditto.
1100 (nds32_legitimize_argument): Ditto.
1101 (nds32_check_constant_argument): Ditto.
1102 (nds32_expand_unop_builtin): Ditto.
1103 (nds32_expand_unopimm_builtin): Ditto.
1104 (nds32_expand_binop_builtin): Ditto.
1105 (nds32_builtin_decl_impl): Ditto.
1106 (builtin_description): Ditto.
1107 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1108 (nds32_init_builtins_impl): Ditto.
1109 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1110 (nds32_builtin_decl): New.
1111 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1112 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1114 2018-03-02 Jeff Law <law@redhat.com>
1116 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1117 (redundant_insn, fill_simple_delay_slots): Likewise.
1118 (fill_slots_from_thread): Likewise.
1119 * resource.c (mark_referenced_resources): Likewise.
1120 (mark_set_resources, find_dead_or_set_registers): Likewise.
1122 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1124 * substring-locations.h (format_warning_va): Formatting fix for
1126 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1128 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1129 * substring-locations.c: Include intl.h.
1130 (format_warning_va): Turned into small wrapper around
1131 format_warning_n_va, renamed to ...
1132 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1133 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1135 (format_warning_at_substring_n): New function.
1136 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1137 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1138 format_warning_at_substring with just a shorter name instead of
1139 const function pointer.
1140 (fmtwarn_n): New function.
1141 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1142 appropriate, get rid of all the fmtstr temporaries, move conditionals
1143 with G_() wrapped string literals directly into fmtwarn arguments,
1144 cast dir.len to (int), formatting fixes.
1146 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1148 * doc/invoke.texi: Remove "Cilk Plus" references.
1150 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1151 Richard Biener <rguenther@suse.de>
1154 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1155 for error or warning attributes if CALL_FROM_THUNK_P is set.
1158 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1161 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1162 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1165 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1166 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1168 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1169 instead of -1U in last predictors element's probability member.
1171 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1174 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1175 arguments if they are comparable.
1177 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1179 PR tree-optimization/84634
1180 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1181 masks and masked_loop_p with a single loop_masks, making sure it's
1182 null for bb vectorization.
1184 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1186 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1187 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1188 loop->force_vectorize to check whether there is no alias.
1190 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1193 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1195 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1197 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1198 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1199 instead of a loop around prev_real_insn.
1200 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1204 * config/i386/i386.c (ix86_print_operand): Use conditional
1205 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1208 2018-03-02 Richard Biener <rguenther@suse.de>
1210 PR tree-optimization/84427
1211 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1212 (bitmap_set_subtract_values): Rewrite to handle multiple
1215 (prune_clobbered_mems): Likewise.
1216 (phi_translate): Take edge instead of pred/phiblock.
1217 (phi_translate_1): Likewise.
1218 (phi_translate_set): Likewise. Insert all translated
1219 exprs for a value into the set, keeping possibly multiple
1220 expressions per value.
1221 (compute_antic_aux): Adjust for phi_translate changes.
1222 When intersecting union the expressions and prune those
1223 not in the final value set, keeping possibly multiple
1224 expressions per value. Do not use value-insertion
1225 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1226 all expressions. Add verification that the value-sets
1227 only shrink during iteration.
1228 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1229 (do_pre_regular_insertion): Likewise.
1230 (do_pre_partial_partial_insertion): Likewise.
1232 2018-03-02 Richard Biener <rguenther@suse.de>
1235 * config/darwin.c (saved_debug_info_level): New static global.
1236 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1237 (darwin_asm_lto_end): Restore debug info generation settings.
1239 2018-03-01 Martin Liska <mliska@suse.cz>
1242 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1245 2018-03-01 Richard Biener <rguenther@suse.de>
1248 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1249 type annotation with LTO when debug was disabled at compile-time.
1251 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1253 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1255 (mips_final_postscan_insn): Likewise.
1257 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1259 PR rtl-optimization/84528
1260 * alias.c (init_alias_target): Add commentary.
1261 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1262 a unique base value if the frame pointer is not eliminated
1263 to the stack pointer.
1265 2018-03-01 Tom de Vries <tom@codesourcery.com>
1267 PR rtl-optimization/83327
1268 * lra-int.h (hard_regs_spilled_into): Declare.
1269 * lra.c (hard_regs_spilled_into): Define.
1270 (init_reg_info): Init hard_regs_spilled_into.
1271 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1272 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1273 (process_bb_lives): Handle hard_regs_spilled_into.
1274 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1275 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1277 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1279 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1280 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1281 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1282 * config/rs6000/aix72.h: New file.
1284 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1286 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1287 instead of warning_at with conditional singular and plural messages
1291 * stor-layout.c (update_alignment_for_field): For
1292 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1293 && !DECL_PACKED (field), do the alignment update, just use
1294 only desired_align instead of MAX (type_align, desired_align)
1296 (place_field): Don't do known_align < desired_align handling
1297 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1298 is non-NULL, instead do it after rli->prev_field handling and
1299 only if not within a bitfield word. For DECL_PACKED (field)
1300 use type_align of BITS_PER_UNIT.
1302 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1304 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1305 superfluous parentheses and trailing spaces.
1307 2018-02-28 Richard Biener <rguenther@suse.de>
1309 PR tree-optimization/84584
1310 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1311 SCoPs with fake exit edge.
1313 2018-02-28 Martin Liska <mliska@suse.cz>
1316 * timevar.c (timer::print): Fix format to properly print 100%
1319 2018-02-28 Richard Biener <rguenther@suse.de>
1322 * genmatch.c (capture_info::walk_match): Do not mark
1323 captured expressions without operands as expr_p given
1324 they act more like predicates and should be subject to
1325 "lost tail" side-effect preserving.
1327 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1329 PR rtl-optimization/81611
1330 * auto-inc-dec.c (attempt_change): Move dead note from
1331 mem_insn if it's the next use of regno
1332 (find_address): Take address use of reg holding
1333 non-incremented value. Add parm to limit search to the named
1335 (merge_in_block): Attempt to use a mem insn that is the next
1336 use of the original regno.
1338 2018-02-27 Martin Sebor <msebor@redhat.com>
1341 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1342 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1344 2018-02-27 Martin Sebor <msebor@redhat.com>
1346 PR translation/84207
1347 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1348 n argument to unsigned HOST_WIDE_INT.
1349 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1350 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1351 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1352 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1354 2018-02-27 Richard Biener <rguenther@suse.de>
1356 PR tree-optimization/84512
1357 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1358 Do not use the estimate returned from record_stmt_cost for
1359 the scalar iteration cost but sum properly using add_stmt_cost.
1361 2018-02-27 Richard Biener <rguenther@suse.de>
1363 PR tree-optimization/84466
1364 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1365 Adjust last change to less strictly validate use operands.
1367 2018-02-27 Martin Liska <mliska@suse.cz>
1369 PR gcov-profile/84548
1370 * gcov.c (process_file): Allow partial overlap and consider it
1371 also as group functions.
1372 (output_lines): Properly calculate range of lines for a group.
1374 2018-02-27 Martin Liska <mliska@suse.cz>
1376 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1377 'ggc' suffixes. Change first column width.
1378 (timer::print): Fix formatting of the column.
1380 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1382 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1383 preserve inline entry blocks for the sake of debug inline
1384 entry point markers alone.
1385 (remove_unused_locals): Suggest in comments a better place to
1386 force the preservation of inline entry blocks that are
1387 otherwise unused, but do not preserve them.
1389 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1391 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1393 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1396 * config/i386/constraints.md (Bs): Replace
1397 ix86_indirect_branch_register with
1398 TARGET_INDIRECT_BRANCH_REGISTER.
1400 * config/i386/i386.md (indirect_jump): Likewise.
1401 (tablejump): Likewise.
1402 (*sibcall_memory): Likewise.
1403 (*sibcall_value_memory): Likewise.
1404 Peepholes of indirect call and jump via memory: Likewise.
1405 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1406 (*sibcall_value_GOT_32): Likewise.
1407 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1408 (GOT_memory_operand): Likewise.
1409 (call_insn_operand): Likewise.
1410 (sibcall_insn_operand): Likewise.
1411 (GOT32_symbol_operand): Likewise.
1412 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1414 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1416 PR rtl-optimization/83496
1417 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1418 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1419 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1420 redundant insn, if any.
1421 (relax_delay_slots): Likewise.
1422 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1424 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1426 PR tree-optimization/83965
1427 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1428 that grouped statements are part of a reduction chain. Return
1429 true if the statement is not marked as a reduction itself but
1431 (vect_recog_dot_prod_pattern): Don't check whether the statement
1432 is part of a group here.
1433 (vect_recog_sad_pattern): Likewise.
1434 (vect_recog_widen_sum_pattern): Likewise.
1436 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1439 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1441 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1444 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1446 (ix86_output_indirect_function_return): New prototype.
1447 (ix86_split_simple_return_pop_internal): Likewise.
1448 * config/i386/i386.c (indirect_return_via_cx): New.
1449 (indirect_return_via_cx_bnd): Likewise.
1450 (indirect_thunk_name): Handle return va CX_REG.
1451 (output_indirect_thunk_function): Create alias for
1452 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1453 (ix86_output_indirect_jmp): Remove the bool argument.
1454 (ix86_output_indirect_function_return): New function.
1455 (ix86_split_simple_return_pop_internal): Likewise.
1456 * config/i386/i386.md (*indirect_jump): Don't pass false
1457 to ix86_output_indirect_jmp.
1458 (*tablejump_1): Likewise.
1459 (simple_return_pop_internal): Change it to define_insn_and_split.
1460 Call ix86_split_simple_return_pop_internal to split it for
1462 (simple_return_indirect_internal): Call
1463 ix86_output_indirect_function_return instead of
1464 ix86_output_indirect_jmp.
1466 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1469 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1470 memset and value initialization afterwards.
1472 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1474 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1476 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1479 * common/config/aarch64/aarch64-common.c
1480 (aarch_option_optimization_table[]): Switch
1481 off fomit-frame-pointer
1483 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1484 Chung-Ju Wu <jasonwucj@gmail.com>
1486 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1488 (store_multiple): Ditto.
1490 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1492 * config.gcc: Add --with-cpu support for nds32 target.
1493 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1494 * config/nds32/nds32.opt: Add -mcpu= option.
1496 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1498 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1499 isel=yes): Warn for these deprecated options.
1501 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1503 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1504 ISA_2_5_MASKS_EMBEDDED.
1506 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1508 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1509 p->max as pointers rather than using iterative_hash_expr.
1511 2018-02-23 Carl Love <cel@us.ibm.com>
1513 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1514 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1516 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1517 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1518 P8V_BUILTIN_VEC_VUNSIGNED2.
1520 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1523 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1524 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1526 * lra-constraints.c (curr_insn_transform): Set up
1527 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1529 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1530 * lra-eliminations.c (spill_pseudos): Ditto.
1531 (process_insn_for_elimination): Ditto.
1532 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1533 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1534 LRA_NON_CLOBBERED_ALT.
1536 2018-02-22 Martin Sebor <msebor@redhat.com>
1538 PR tree-optimization/84480
1539 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1540 to maybe_diag_stxncpy_trunc. Call it.
1541 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1542 from gimple_fold_builtin_strcpy. Print inlining stack.
1543 (handle_builtin_stxncpy): Print inlining stack.
1544 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1546 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1549 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1550 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1551 and -fcheck-pointer-bounds are used together.
1552 (indirect_thunk_prefix): New enum.
1553 (indirect_thunk_need_prefix): New function.
1554 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1555 "_nt" instead of "_bnd" for NOTRACK prefix.
1556 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1557 (output_indirect_thunk_function): Likewise.
1559 (ix86_code_end): Update output_indirect_thunk_function calls.
1560 (ix86_output_indirect_branch_via_reg): Replace
1561 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1562 (ix86_output_indirect_branch_via_push): Likewise.
1563 (ix86_output_function_return): Likewise.
1564 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1565 incompatible with -fcf-protection=branch and
1566 -fcheck-pointer-bounds.
1568 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1571 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1572 Change gcc_assert call to output_operand_lossage.
1574 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1576 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1578 2018-02-22 DJ Delorie <dj@redhat.com>
1579 Sebastian Perta <sebastian.perta@renesas.com>
1580 Oleg Endo <olegendo@gcc.gnu.org>
1582 * config/rx/rx.c (rx_rtx_costs): New function.
1583 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1585 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1587 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1589 2018-02-22 Martin Liska <mliska@suse.cz>
1592 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1593 Add "native" as a possible value.
1595 2018-02-22 Martin Liska <mliska@suse.cz>
1598 * config/i386/i386.c (ix86_option_override_internal):
1599 Add "native" as a possible value for -march and -mtune.
1601 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1604 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1605 to all type variants.
1607 PR tree-optimization/84503
1608 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1609 width as info->bitpos + info->bitsize - start.
1610 (merged_store_group::merge_overlapping): Simplify width computation.
1611 (check_no_overlap): New function.
1612 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1613 start + width and last_order of the group, fail if check_no_overlap
1615 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1616 to group if check_no_overlap fails.
1618 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1620 * config/rs6000/altivec.md: Delete contraint arguments to
1621 define_expand, define_split, and define_peephole2, and in
1622 define_insn_and_split if always unused.
1623 * config/rs6000/darwin.md: Ditto.
1624 * config/rs6000/dfp.md: Ditto.
1625 * config/rs6000/rs6000.md: Ditto.
1626 * config/rs6000/sync.md: Ditto.
1627 * config/rs6000/vector.md: Ditto.
1628 * config/rs6000/vsx.md: Ditto.
1630 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1632 * config/rs6000/altivec.md: Write output control strings as braced
1633 blocks instead of double-quoted strings.
1634 * config/rs6000/darwin.md: Ditto.
1635 * config/rs6000/rs6000.md: Ditto.
1636 * config/rs6000/vector.md: Ditto.
1637 * config/rs6000/vsx.md: Ditto.
1639 2018-02-21 Jason Merrill <jason@redhat.com>
1641 PR c++/84314 - ICE with templates and fastcall attribute.
1642 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1644 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1646 * ipa-cp.c (determine_versionability): Fix comment typos.
1648 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1651 * ipa-cp.c (determine_versionability): Do not version functions caling
1654 2018-02-21 Martin Liska <mliska@suse.cz>
1657 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1658 Add "native" as a possible value.
1659 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1660 the macro when native cpu detection is available.
1662 2018-02-21 Martin Liska <mliska@suse.cz>
1665 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1666 Add "native" as a possible value.
1667 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1668 when native cpu detection is available.
1670 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1671 Martin Sebor <msebor@redhat.com>
1673 PR tree-optimization/84478
1674 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1676 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1677 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1678 support which is conservatively correct, for 2 only stay conservative
1679 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1680 argument to the 2 argument get_range_strlen, adjust 6 arg
1681 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1683 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1684 (gimple_fold_builtin_strlen): Pass true as last argument to
1687 2018-02-20 Martin Sebor <msebor@redhat.com>
1690 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1691 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1692 (builtin_memref::builtin_memref): Factor out parts into
1693 set_base_and_offset and call it.
1695 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1698 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1699 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1700 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1701 search at the associated MODE_INT.
1703 2018-02-20 Jeff Law <law@redhat.com>
1706 PR tree-optimization/81592
1708 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1709 for range data rather than using global data.
1710 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1711 range data rather than using global data.
1712 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1713 pass it to children as needed.
1714 (struct directive::fmtresult): Similarly.
1715 (struct directive::set_width): Similarly.
1716 (struct directive::set_precision): Similarly.
1717 (format_integer, format_directive, parse_directive): Similarly.
1718 (format_none): Accept unnamed vr_values parameter.
1719 (format_percent, format_floating, format_character): Similarly.
1720 (format_string, format_plain): Similarly.
1721 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1722 the EVRP range analyzer for range data rather than using global data.
1723 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1724 gimple-ssa-evrp-analyze.h
1725 (class sprintf_dom_walker): Add after_dom_children member function.
1726 Add evrp_range_analyzer member.
1727 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1728 range analyzer as needed.
1729 (sprintf_dom_walker::after_dom_children): New member function.
1730 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1732 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1733 (evrp_range_analyzer::pop_to_marker): Likewise.
1735 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1737 PR tree-optimization/84419
1738 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1739 with the required type if its current type is compatible but
1742 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1745 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1746 after vectorization.
1748 2018-02-20 Martin Liska <mliska@suse.cz>
1751 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1752 possible values if we don't have a hint.
1754 2018-02-20 Martin Liska <mliska@suse.cz>
1758 * final.c (shorten_branches): Build align_tab array with one
1760 * opts.c (finish_options): Add alignment option limit check.
1761 (MAX_CODE_ALIGN): Likewise.
1762 (MAX_CODE_ALIGN_VALUE): Likewise.
1763 * doc/invoke.texi: Document maximum allowed option value for
1764 all -falign-* options.
1766 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1769 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1770 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1771 * var-tracking.c (emit_note_insn_var_location): Remove all references
1772 to NOTE_INSN_CALL_ARG_LOCATION.
1773 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1774 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1775 Use copy_rtx_if_shared.
1776 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1777 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1778 (dwarf2out_var_location): Remove handling of
1779 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1781 * final.c (final_scan_insn): Remove all references to
1782 NOTE_INSN_CALL_ARG_LOCATION.
1783 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1784 before dumping final insns.
1785 * except.c (emit_note_eh_region_end): Remove all references to
1786 NOTE_INSN_CALL_ARG_LOCATION.
1787 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1788 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1789 * config/arc/arc.c (hwloop_optimize): Likewise.
1790 * config/arm/arm.c (create_fix_barrier): Likewise.
1791 * config/s390/s390.c (s390_chunkify_start): Likewise.
1792 * config/sh/sh.c (find_barrier): Likewise.
1793 * config/i386/i386.c (rest_of_insert_endbranch,
1794 ix86_seh_fixup_eh_fallthru): Likewise.
1795 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1796 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1797 * config/frv/frv.c (frv_function_prologue): Likewise.
1798 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1800 (note_outside_basic_block_p): Remove all references to
1801 NOTE_INSN_CALL_ARG_LOCATION.
1802 * gengtype.c (adjust_field_rtx_def): Likewise.
1803 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1805 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1806 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1809 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1812 PR tree-optimization/84452
1813 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1814 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1817 2018-02-19 Martin Liska <mliska@suse.cz>
1820 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1822 2018-02-19 Martin Liska <mliska@suse.cz>
1823 Richard Sandiford <richard.sandiford@linaro.org>
1825 PR tree-optimization/82491
1826 * gimple-fold.c (get_base_constructor): Make earlier bail out
1829 2018-02-19 Carl Love <cel@us.ibm.com>
1831 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1832 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1834 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1835 P8V_BUILTIN_VEC_NEG.
1837 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1839 * config/rl78/rl78.md (movdf): New define expand.
1841 2018-02-19 Martin Liska <mliska@suse.cz>
1844 * doc/invoke.texi: Fix typo.
1845 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1847 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1849 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1850 handle rs6000_single_float and rs6000_double_float specially for
1853 2018-02-16 Jeff Law <law@redhat.com>
1855 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1858 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1861 * ipa-inline.c (inline_small_functions): Fix a typo.
1863 2018-02-16 Nathan Sidwell <nathan@acm.org>
1865 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1867 2018-02-16 Carl Love <cel@us.ibm.com>
1869 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1870 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1871 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1872 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1873 expansion to P8V_BUILTIN_VEC_FLOAT2.
1875 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1877 PR rtl-optimization/70023
1878 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1879 src_regno into account.
1881 2018-02-16 Carl Love <cel@us.ibm.com>
1883 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1884 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1885 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1886 * config/rs6000/rs6000.c: Remove case statements for
1887 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1888 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1889 and P9V_BUILTIN_VEC_VINSERT4B.
1890 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1891 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1892 * config/rs6000/vsx.md:
1893 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1896 2018-02-16 Carl Love <cel@us.ibm.com>
1898 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1900 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1902 * config/rs6000/rs6000-c.c: Add the definitions for
1903 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1904 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1905 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1906 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1907 definition for insert4b and define insn *insert3b_internal.
1908 * doc/extend.texi: Add documentation for vec_extract4b.
1910 2018-02-16 Nathan Sidwell <nathan@acm.org>
1912 * doc/extend.texi (Backwards Compatibility): Mention friend
1913 injection. Note for-scope is deprecated.
1914 * doc/invoke.texi (-ffriend-injection): Deprecate.
1916 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1918 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1919 that moved to I2, also allow destinations that are a paradoxical
1920 subreg (instead of a normal reg).
1922 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1925 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1928 2018-02-16 Richard Biener <rguenther@suse.de>
1930 PR tree-optimization/84037
1931 PR tree-optimization/84016
1933 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1934 Adjust vec_construct for the fact we need additional higher latency
1935 128bit inserts for AVX256 and AVX512 vector builds.
1936 (ix86_add_stmt_cost): Scale vector construction cost for
1939 2018-02-16 Richard Biener <rguenther@suse.de>
1941 PR tree-optimization/84417
1942 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1943 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1944 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1946 2018-02-16 Martin Liska <mliska@suse.cz>
1949 * internal-fn.def (ASAN_CHECK): Set proper flags.
1950 (ASAN_MARK): Likewise.
1952 2018-02-16 Julia Koval <julia.koval@intel.com>
1954 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1955 from PTA_CANNONLAKE.
1957 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1960 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1961 Use ++iter rather than iter++ for std::list iterators.
1962 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1963 defer deleting them until all nodes in the forest are processed. Do
1964 free even leaf nodes. Change to_process into auto_vec.
1967 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1968 * vec.h (vec_default_construct): Use memset instead of placement new
1969 if BROKEN_VALUE_INITIALIZATION is defined.
1970 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1971 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1974 PR rtl-optimization/83723
1975 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1976 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1977 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1979 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1981 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1983 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1985 PR rtl-optimization/81443
1986 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1987 from inner REGs to paradoxical SUBREGs.
1989 2018-02-16 Richard Biener <rguenther@suse.de>
1991 PR tree-optimization/84399
1992 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1993 For operands we can analyze at their definition make sure we can
1994 analyze them at each use as well.
1996 2018-02-16 Richard Biener <rguenther@suse.de>
1998 PR tree-optimization/84190
1999 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2000 volatile accesses if the decl isn't volatile.
2002 2018-02-15 Jason Merrill <jason@redhat.com>
2004 PR c++/84314 - ICE with templates and fastcall attribute.
2005 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2006 TYPE_CANONICAL on an existing type.
2008 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2010 PR tree-optimization/84383
2011 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2012 dstoff nor call operand_equal_p if dstbase is NULL.
2014 PR tree-optimization/84334
2015 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2016 also a CONSTANT_CLASS_P, punt.
2018 2018-02-14 Jim Wilson <jimw@sifive.com>
2020 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2021 first SMALL_OPERAND check. New local min_second_step. Move assert
2022 to where locals are set. Add TARGET_RVC support.
2023 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2025 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2027 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2029 2018-02-14 Martin Sebor <msebor@redhat.com>
2031 PR tree-optimization/83698
2032 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2033 arrays constrain the offset range to their bounds.
2034 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2035 (builtin_access::overlap): Avoid setting the size of overlap if it's
2037 (maybe_diag_overlap): Also consider arrays when deciding what values
2038 of offsets to include in diagnostics.
2040 2018-02-14 Martin Sebor <msebor@redhat.com>
2043 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2044 that correspond to the kind of a declaration.
2046 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2049 * config/pa/pa.md: Load address of PIC label using the linkage table
2050 if the label is nonlocal.
2052 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2054 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2055 warning message if user requests -maltivec=be.
2056 * doc/invoke.texi: Document deprecation of -maltivec=be.
2058 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2061 * config/rs6000/rs6000-c.c: Update definitions for
2062 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2063 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2065 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2068 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2069 add _get_ssp intrinsics. Remove argument from
2070 __builtin_ia32_rdssp[d|q].
2071 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2072 * config/i386/i386-builtin.def: Remove argument from
2073 __builtin_ia32_rdssp[d|q].
2074 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2075 ix86_expand_special_args_builtin for _rdssp[d|q].
2076 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2077 Clear register before usage.
2078 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2079 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2081 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2083 PR tree-optimization/84357
2084 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2085 operand 1 of an ARRAY_REF too.
2087 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2090 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2091 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2093 (set_of_reg): New struct.
2094 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2095 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2096 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2098 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2099 Split into bitclr, bitset, bitinvert patterns if appropriate.
2100 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2101 use rx_fuse_in_memory_bitop.
2102 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2103 to named insn, correct maximum insn length.
2105 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2108 * machmode.def: Define a complex mode for PARTIAL_INT.
2109 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2111 * doc/rtl.texi: Document CSPImode.
2112 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2114 (msp430_hard_regno_nregs_with_padding): Likewise.
2116 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2119 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2121 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2123 PR rtl-optimization/84169
2124 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2125 we generated a parallel as new i3 and we split that to new i2 and i3
2126 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2127 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2128 those to i2, not i1. Partially rewrite this scan code.
2130 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2133 * stor-layout.c (place_field): For variable length fields, adjust
2134 offset_align afterwards not just based on the field's alignment,
2135 but also on the size.
2138 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2139 of exps and logs in the use_exp2 case.
2141 2018-02-13 Jeff Law <law@redhat.com>
2143 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2146 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2149 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2153 * common.opt (gas-loc-support, gas-locview-support): New.
2154 (ginline-points, ginternal-reset-location-views): New.
2155 * doc/invoke.texi: Document them. Use @itemx where intended.
2156 (gvariable-location-views): Adjust.
2157 * target.def (reset_location_view): New.
2158 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2159 (TARGET_RESET_LOCATION_VIEW): New.
2160 * doc/tm.texi: Rebuilt.
2161 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2162 (dwarf2out_default_as_locview_support): New.
2163 (output_asm_line_debug_info): Use option variables.
2164 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2165 (output_loc_list): Likewise.
2166 (add_high_low_attributes): Check option variables.
2167 Don't output entry view attribute in strict mode.
2168 (gen_inlined_subroutine_die): Check option variables.
2169 (dwarf2out_inline_entry): Likewise.
2170 (init_sections_and_labels): Likewise.
2171 (dwarf2out_early_finish): Likewise.
2172 (maybe_reset_location_view): New, from...
2173 (dwarf2out_var_location): ... here. Call it.
2174 * debug.h (dwarf2out_default_as_loc_support): Declare.
2175 (dwarf2out_default_as_locview_support): Declare.
2176 * hooks.c (hook_int_rtx_insn_0): New.
2177 * hooks.h (hook_int_rtx_insn_0): Declare.
2178 * toplev.c (process_options): Take -gas-loc-support and
2179 -gas-locview-support from dwarf2out. Enable
2180 -gvariable-location-views by default only with locview
2181 assembler support. Enable -ginternal-reset-location-views by
2182 default only if the target defines the corresponding hook.
2183 Enable -ginline-points by default if location views are
2184 enabled; force it disabled if statement frontiers are
2186 * tree-inline.c (expand_call_inline): Check option variables.
2187 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2189 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2191 PR tree-optimization/84321
2192 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2193 handling. Also check whether the anti-range contains any values
2194 that satisfy the mask; switch to a VR_RANGE if not.
2196 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2199 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2201 2018-02-13 Martin Jambor <mjambor@suse.cz>
2204 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2205 of call statements, also set location of a load to a temporary.
2207 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2209 * config/rl78/rl78.c (add_vector_labels): New function.
2210 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2211 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2212 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2213 which checks that no arguments are passed.
2214 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2215 * doc/extend.texi: Documentation for the new attribute.
2217 2018-02-13 Andreas Schwab <schwab@suse.de>
2219 * config/riscv/linux.h (CPP_SPEC): Define.
2221 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2224 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2225 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2226 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2227 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2228 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2229 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2230 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2231 temporarily for AES and PCLMUL builtins.
2233 PR tree-optimization/84339
2234 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2235 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2239 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2240 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2241 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2243 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2245 * omp-simd-clone.h: New file.
2246 * omp-simd-clone.c: Include omp-simd-clone.h.
2247 (expand_simd_clones): No longer static.
2248 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2249 cgraph.h and omp-simd-clone.h.
2250 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2251 (vect_recog_widen_shift_pattern): Formatting fix.
2252 (vect_pattern_recog_1): Don't check optab for calls.
2255 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2256 operands[2] into a REG before using gen_lowpart on it.
2258 2018-02-12 Jeff Law <law@redhat.com>
2261 * config/sh/sh.c (find_barrier): Consider a sibling call
2264 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2265 successfully back substituting a reg.
2267 2018-02-12 Richard Biener <rguenther@suse.de>
2269 PR tree-optimization/84037
2270 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2271 parameter, move visited init to caller.
2272 (vect_slp_analyze_operations): Separate cost from validity
2273 check, initialize visited once for all instances.
2274 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2276 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2278 (vect_model_promotion_demotion_cost): Likewise.
2279 (vectorizable_bswap): Guard cost modeling with !slp_node
2280 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2282 (vectorizable_call): Likewise.
2283 (vectorizable_conversion): Likewise.
2284 (vectorizable_assignment): Likewise.
2285 (vectorizable_shift): Likewise.
2286 (vectorizable_operation): Likewise.
2287 (vectorizable_store): Likewise.
2288 (vectorizable_load): Likewise.
2289 (vectorizable_condition): Likewise.
2290 (vectorizable_comparison): Likewise.
2292 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2295 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2296 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2297 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2299 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2302 * params.def (inline-min-speedup): Increase from 8 to 15.
2303 (max-inline-insns-auto): Decrease from 40 to 30.
2304 * ipa-split.c (consider_split): Add some buffer for function to
2305 be considered inlining candidate.
2306 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2309 2018-02-12 Richard Biener <rguenther@suse.de>
2311 PR tree-optimization/84037
2312 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2313 matched stmts if we cannot swap the non-matched ones.
2315 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2317 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2318 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2319 _mm_maskz_scalef_round_ss): New intrinsics.
2320 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2321 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2322 __builtin_ia32_scalefss_round): Remove.
2323 (__builtin_ia32_scalefsd_mask_round,
2324 __builtin_ia32_scalefss_mask_round): New intrinsics.
2325 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2326 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2327 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2328 "<round_constraint>")): Changed to ...
2329 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2330 "<round_scalar_constraint>")): ... this.
2331 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2332 %0, %1, %2<round_op3>}"): Changed to ...
2333 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2334 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2335 %2<round_scalar_mask_op3>}"): ... this.
2336 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2338 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2340 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2341 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2342 (_mm_maskz_sqrt_round_ss): New intrinsics.
2343 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2344 (__builtin_ia32_sqrtsd_mask_round)
2345 (__builtin_ia32_sqrtss_mask_round): New builtins.
2346 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2347 (__builtin_ia32_sqrtss_round): Remove.
2348 (__builtin_ia32_sqrtsd_mask_round)
2349 (__builtin_ia32_sqrtss_mask_round): New builtins.
2350 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2351 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2352 ((match_operand:VF_128 1 "vector_operand"
2353 "xBm,<round_constraint>")): Changed to ...
2354 ((match_operand:VF_128 1 "vector_operand"
2355 "xBm,<round_scalar_constraint>")): ... this.
2356 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2357 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2358 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2359 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2360 %<iptr>1<round_scalar_mask_op3>}): ... this.
2361 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2362 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2364 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2367 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2368 Cast vec_cmpeq result to correct type.
2369 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2370 Cast vec_cmpgt result to correct type.
2372 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2374 * final.c (final_scan_insn_1): Renamed from...
2375 (final_scan_insn): ... this. New wrapper, to recover
2376 seen from the outermost call in recursive ones.
2377 * config/sparc/sparc.c (output_return): Drop seen from call.
2378 (output_sibcall): Likewise.
2379 * config/visium/visium.c (output_branch): Likewise.
2381 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2383 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2386 2018-02-10 Alan Modra <amodra@gmail.com>
2389 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2390 Specify LR as an input.
2392 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2395 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2396 remove_member_access_dummy_vars): New functions.
2397 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2398 lower_omp_1, execute_lower_omp): Use them.
2400 PR rtl-optimization/84308
2401 * shrink-wrap.c (spread_components): Release todo vector.
2403 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2405 PR rtl-optimization/57193
2406 * ira-color.c (struct allocno_color_data): Add member
2407 conflict_allocno_hard_prefs.
2408 (update_conflict_allocno_hard_prefs): New.
2409 (bucket_allocno_compare_func): Add a preference based on
2410 conflict_allocno_hard_prefs.
2411 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2412 (color_allocnos): Remove a dead code. Initiate
2413 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2415 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2418 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2419 constraint from =wa to wa. Avoid a subreg on the output operand,
2420 instead use a pseudo and subreg it in a move.
2421 (p9_xxbrd_<mode>): Changed to ...
2422 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2423 (p9_xxbrd_v2df): New expander.
2424 (p9_xxbrw_<mode>): Changed to ...
2425 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2426 (p9_xxbrw_v4sf): New expander.
2428 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2430 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2432 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2435 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2436 multiply in 32-bit mode.
2437 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2438 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2441 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2443 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2444 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2445 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2446 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2448 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2451 * dwarf2out.c (is_trivial_indirect_ref): New function.
2452 (dwarf2out_late_global_decl): Do not generate a location
2453 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2454 and that are not defined in the current unit.
2456 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2458 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2459 instead of a libcall for UNORDERED.
2461 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2464 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2465 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2467 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2470 * config/s390/s390.c (s390_set_current_function): Invoke
2471 s390_indirect_branch_settings also if fndecl didn't change.
2473 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2475 * config/rs6000/rs6000.md (blockage): Set length to zero.
2477 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2479 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2481 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2484 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2485 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2489 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2490 PARALLEL incoming that failed vt_get_decl_and_offset check.
2493 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2494 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2496 (get_variable_section): For decls in named .bss* sections pass true as
2497 second argument to bss_initializer_p.
2499 2018-02-09 Marek Polacek <polacek@redhat.com>
2500 Jakub Jelinek <jakub@redhat.com>
2503 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2504 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2505 Sync some changes from cxx_fold_indirect_ref.
2507 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2509 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2511 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2512 (BLOCK_INLINE_ENTRY_LABEL): New.
2513 (dwarf2out_var_location): Disregard inline entry markers.
2514 (inline_entry_data): New struct.
2515 (inline_entry_data_hasher): New hashtable type.
2516 (inline_entry_data_hasher::hash): New.
2517 (inline_entry_data_hasher::equal): New.
2518 (inline_entry_data_table): New variable.
2519 (add_high_low_attributes): Add DW_AT_entry_pc and
2520 DW_AT_GNU_entry_view attributes if a pending entry is found
2521 in inline_entry_data_table. Add old entry_pc attribute only
2522 if debug nonbinding markers are disabled.
2523 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2524 markers are enabled.
2525 (block_within_block_p, dwarf2out_inline_entry): New.
2526 (dwarf2out_finish): Check that no entries remained in
2527 inline_entry_data_table.
2528 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2529 (final_scan_insn, notice_source_line): Likewise.
2530 (rest_of_clean_state): Skip inline entry markers.
2531 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2533 * gimple.c (gimple_build_debug_inline_entry): New.
2534 * gimple.h (enum gimple_debug_subcode): Add
2535 GIMPLE_DEBUG_INLINE_ENTRY.
2536 (gimple_build_debug_inline_entry): Declare.
2537 (gimple_debug_inline_entry_p): New.
2538 (gimple_debug_nonbind_marker_p): Adjust.
2539 * insn-notes.def (INLINE_ENTRY): New.
2540 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2541 inline entry marker notes.
2542 (print_insn): Likewise.
2543 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2544 (INSN_DEBUG_MARKER_KIND): Likewise.
2545 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2546 * tree-inline.c (expand_call_inline): Build and insert
2547 debug_inline_entry stmt.
2548 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2549 inline entry blocks early, if nonbind markers are enabled.
2550 (dump_scope_block): Dump fragment info.
2551 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2552 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2553 (gimple_build_debug_inline_entry): New.
2554 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2555 Enable/disable inline entry points too.
2556 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2557 (DEBUG_INSN): Describe inline entry markers.
2559 * common.opt (gvariable-location-views): New.
2560 (gvariable-location-views=incompat5): New.
2561 * config.in: Rebuilt.
2562 * configure: Rebuilt.
2563 * configure.ac: Test assembler for view support.
2564 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2565 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2566 * dwarf2out.c (var_loc_view): New typedef.
2567 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2568 (dwarf2out_locviews_in_attribute): New.
2569 (dwarf2out_locviews_in_loclist): New.
2570 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2571 (enum dw_line_info_opcode): Add LI_adv_address.
2572 (struct dw_line_info_table): Add view.
2573 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2574 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2575 (zero_view_p): New variable.
2576 (ZERO_VIEW_P): New macro.
2577 (output_asm_line_debug_info): New.
2578 (struct var_loc_node): Add view.
2579 (add_AT_view_list, AT_loc_list): New.
2580 (add_var_loc_to_decl): Add view param. Test it against last.
2581 (new_loc_list): Add view params. Record them.
2582 (AT_loc_list_ptr): Handle loc and view lists.
2583 (view_list_to_loc_list_val_node): New.
2584 (print_dw_val): Handle dw_val_class_view_list.
2585 (size_of_die): Likewise.
2586 (value_format): Likewise.
2587 (loc_list_has_views): New.
2588 (gen_llsym): Set vl_symbol too.
2589 (maybe_gen_llsym, skip_loc_list_entry): New.
2590 (dwarf2out_maybe_output_loclist_view_pair): New.
2591 (output_loc_list): Output view list or entries too.
2592 (output_view_list_offset): New.
2593 (output_die): Handle dw_val_class_view_list.
2594 (output_dwarf_version): New.
2595 (output_compilation_unit_header): Use it.
2596 (output_skeleton_debug_sections): Likewise.
2597 (output_rnglists, output_line_info): Likewise.
2598 (output_pubnames, output_aranges): Update version comments.
2599 (output_one_line_info_table): Output view numbers in asm comments.
2600 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2601 Call maybe_gen_llsym.
2602 (loc_list_from_tree_1): Adjust.
2603 (add_AT_location_description): Create view list attribute if
2604 needed, check it's absent otherwise.
2605 (convert_cfa_to_fb_loc_list): Adjust.
2606 (maybe_emit_file): Call output_asm_line_debug_info for test.
2607 (dwarf2out_var_location): Reset views as needed. Precompute
2608 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2609 attribute. Set view.
2610 (new_line_info_table): Reset next view.
2611 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2612 (dwarf2out_source_line): Likewise. Output view resets and labels to
2613 the assembler, or select appropriate line info opcodes.
2614 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2615 (optimize_string_length): Catch it. Adjust.
2616 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2617 dw_val_class_view_list, and remove it if no longer needed.
2618 (hash_loc_list): Hash view numbers.
2619 (loc_list_hasher::equal): Compare them.
2620 (optimize_location_lists): Check whether a view list symbol is
2621 needed, and whether the locview attribute is present, and
2622 whether they match. Remove the locview attribute if no longer
2624 (index_location_lists): Call skip_loc_list_entry for test.
2625 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2626 Use output_dwarf_version.
2627 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2628 (struct dw_val_node): Add val_view_list.
2629 * final.c (SEEN_NEXT_VIEW): New.
2630 (set_next_view_needed): New.
2631 (clear_next_view_needed): New.
2632 (maybe_output_next_view): New.
2633 (final_start_function): Rename to...
2634 (final_start_function_1): ... this. Take pointer to FIRST,
2635 add SEEN parameter. Emit param bindings in the initial view.
2636 (final_start_function): Reintroduce SEEN-less interface.
2637 (final): Rename to...
2638 (final_1): ... this. Take SEEN parameter. Output final pending
2639 next view at the end.
2640 (final): Reintroduce seen-less interface.
2641 (final_scan_insn): Output pending next view before switching
2642 sections or ending a block. Mark the next view as needed when
2643 outputting variable locations. Notify debug backend of section
2644 changes, and of location view changes.
2645 (rest_of_handle_final): Adjust.
2646 * toplev.c (process_options): Autodetect value for debug variable
2647 location views option. Warn on incompat5 without -gdwarf-5.
2648 * doc/invoke.texi (gvariable-location-views): New.
2649 (gvariable-location-views=incompat5): New.
2650 (gno-variable-location-views): New.
2652 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2654 PR tree-optimization/84136
2655 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2656 that the result of find_edge is non-NULL.
2658 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2661 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2662 storing integer register in SImode. Fix cost of 256 and 512
2663 byte aligned SSE register store.
2665 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2667 * config/i386/i386.c (ix86_multiplication_cost): Fix
2668 multiplication cost for TARGET_AVX512DQ.
2670 2018-02-08 Marek Polacek <polacek@redhat.com>
2672 PR tree-optimization/84238
2673 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2676 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2678 PR tree-optimization/84265
2679 * tree-vect-stmts.c (vectorizable_store): Don't treat
2680 VMAT_CONTIGUOUS accesses as grouped.
2681 (vectorizable_load): Likewise.
2683 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2685 PR tree-optimization/81635
2686 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2687 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2688 (test_round_for_mask): New functions.
2689 (wide_int_cc_tests): Call test_round_for_mask.
2690 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2691 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2692 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2693 range returned by get_range_info.
2695 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2698 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2699 * symtab.c: Include builtins.h
2700 (symtab_node::output_to_lto_symbol_table_p): Move here
2701 from lto-streamer-out.c:output_symbol_p.
2702 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2703 (output_symbol_p): Move all logic to symtab.c
2704 (produce_symtab): Update.
2706 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2708 * config/s390/s390-opts.h (enum indirect_branch): Define.
2709 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2710 (s390_indirect_branch_via_thunk)
2711 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2712 (enum s390_indirect_branch_type): Define.
2713 * config/s390/s390.c (struct s390_frame_layout, struct
2714 machine_function): Remove.
2715 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2716 (indirect_branch_table_label_no, indirect_branch_table_name):
2718 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2719 (enum s390_indirect_branch_option): Define.
2720 (s390_return_addr_from_memory): New function.
2721 (s390_handle_string_attribute): New function.
2722 (s390_attribute_table): Add new attribute handler.
2723 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2724 (s390_indirect_branch_via_thunk): New function.
2725 (s390_indirect_branch_via_inline_thunk): New function.
2726 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2727 sibling call optimization for non z10 compiles.
2728 (s390_emit_call): Force indirect branch target to be a single
2729 register. Add r1 clobber for non-z10 compiles.
2730 (s390_emit_epilogue): Emit return jump via return_use expander.
2731 (s390_reorg): Handle JUMP_INSNs as execute targets.
2732 (s390_option_override_internal): Perform validity checks for the
2733 new command line options.
2734 (s390_indirect_branch_attrvalue): New function.
2735 (s390_indirect_branch_settings): New function.
2736 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2737 (s390_output_indirect_thunk_function): New function.
2738 (s390_code_end): Implement target hook.
2739 (s390_case_values_threshold): Implement target hook.
2740 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2742 * config/s390/s390.h (struct s390_frame_layout)
2743 (struct machine_function): Move here from s390.c.
2744 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2745 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2746 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2747 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2748 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2749 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2750 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2751 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2752 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2753 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2754 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2755 (mnemonic attribute): Add values which aren't recognized
2757 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2758 pattern for branch conversion. Fix mnemonic attribute.
2759 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2760 indirect branch via thunk if requested.
2761 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2762 ("*indirect_jump"): Disable for branch conversion using out of
2764 ("indirect_jump_via_thunk<mode>_z10")
2765 ("indirect_jump_via_thunk<mode>")
2766 ("indirect_jump_via_inlinethunk<mode>_z10")
2767 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2768 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2769 ("casesi_jump_via_inlinethunk<mode>_z10")
2770 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2771 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2772 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2773 ("*indirect2_jump"): Disable for branch conversion.
2774 ("casesi_jump"): Turn into expander and expand patterns for branch
2776 ("return_use"): New expander.
2777 ("*return"): Emit return via thunk and rename it to ...
2778 ("*return<mode>"): ... this one.
2779 * config/s390/s390.opt: Add new options and and enum for the
2782 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2784 * lra-constraints.c (match_reload): Unconditionally use
2785 gen_lowpart_SUBREG, rather than selecting between that
2786 and equivalent gen_rtx_SUBREG code.
2788 2018-02-08 Richard Biener <rguenther@suse.de>
2790 PR tree-optimization/84233
2791 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2792 changed flag instead of boguously re-using phi_inserted.
2794 2018-02-08 Martin Jambor <mjambor@suse.cz>
2796 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2797 static local variables.
2799 2018-02-08 Richard Biener <rguenther@suse.de>
2801 PR tree-optimization/84278
2802 * tree-vect-stmts.c (vectorizable_store): When looking for
2803 smaller vector types to perform grouped strided loads/stores
2804 make sure the mode is supported by the target.
2805 (vectorizable_load): Likewise.
2807 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2809 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2810 Increase LDP/STP opportunities by adding adjacent callee-saves.
2812 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2814 PR rtl-optimization/84068
2815 PR rtl-optimization/83459
2816 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2818 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2820 PR tree-optimization/84224
2821 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2822 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2825 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2828 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2829 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2830 count, remove one USE.
2832 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2834 * doc/install.texi (Configuration): Document the
2835 --with-long-double-format={ibm,ieee} PowerPC configuration
2839 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2840 Convert from define_expand to be define_insn_and_split. Rework
2841 float/double/_Float128 conversions to QI/HI/SImode to work with
2842 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2843 conversions to QI/HImode types did a store and then a load to
2844 truncate the value. For conversions to VSX registers, don't split
2845 the insn, instead emit the code directly. Use the code iterator
2846 any_fix to combine signed and unsigned conversions.
2847 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2848 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2849 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2850 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2851 (fix_<mode>di2_hw): Likewise.
2852 (fixuns_<mode>di2_hw): Likewise.
2853 (fix_<mode>si2_hw): Likewise.
2854 (fixuns_<mode>si2_hw): Likewise.
2855 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2856 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2857 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2858 fix<uns>_trunc<SFDF:mode>si2_p8.
2859 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2861 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2862 (fix<uns>_<mode>_mem): Likewise.
2863 (fctiw<u>z_<mode>_mem): Likewise.
2864 (fix<uns>_<mode>_mem): Likewise.
2865 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2866 the register allocator from doing a direct move to the GPRs to do
2867 a store, and instead use the ISA 3.0 store byte/half-word from
2868 vector register instruction. For IEEE 128-bit floating point,
2869 also optimize stores of 32-bit ints.
2870 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2872 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2874 * genextract.c (push_pathstr_operand): New function to support
2876 (walk_rtx): Call push_pathstr_operand.
2877 (print_path): Support [a-zA-Z].
2879 2018-02-07 Richard Biener <rguenther@suse.de>
2881 PR tree-optimization/84037
2882 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2883 (cse_and_gimplify_to_preheader): Declare.
2884 (vect_get_place_in_interleaving_chain): Likewise.
2885 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2887 (_loop_vec_info::~_loop_vec_info): Delete it.
2888 (cse_and_gimplify_to_preheader): New function.
2889 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2890 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2891 (vectorizable_load): Likewise. For grouped stores always base
2892 the IV on the first element.
2893 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2894 condition before gimplifying.
2896 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2898 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2899 *DIV_EXPR and *MOD_EXPR.
2901 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2904 * config/i386/i386.c (ix86_option_override_internal): Mask out
2905 the CF_SET bit when checking -fcf-protection.
2907 2018-02-07 Tom de Vries <tom@codesourcery.com>
2910 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2913 2018-02-07 Richard Biener <rguenther@suse.de>
2915 PR tree-optimization/84204
2916 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2919 PR tree-optimization/84205
2920 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2921 special-case isl_ast_op_zdiv_r.
2923 PR tree-optimization/84223
2924 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2925 Only add conditions from within the region.
2926 (gather_bbs::after_dom_children): Adjust.
2928 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2931 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2932 * config/avr/avr.md: Only post-reload split REG-REG moves if
2933 either register is GENERAL_REG_P.
2935 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2937 PR tree-optimization/84235
2938 * tree-ssa-scopedtables.c
2939 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2940 if the subtraction is performed in floating point type where NaNs are
2941 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2942 build 1. Formatting fix.
2944 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2947 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2948 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2949 and skip it regardless of bb boundaries. Use CALL_P macro,
2950 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2951 unnecessarily, formatting fix.
2953 2018-02-06 Michael Collison <michael.collison@arm.com>
2955 * config/arm/thumb2.md:
2956 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2957 (*thumb_mov_notscc): Ditto.
2959 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2962 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2963 unsigned_fix, not "s".
2965 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2967 * configure.ac (gcc_fn_eh_frame_ro): New function.
2968 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2969 correct .eh_frame permissions.
2970 * configure: Regenerate.
2972 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2974 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2977 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2979 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2980 Display warning message for -mno-speculate-indirect-jumps.
2982 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2984 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2986 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2988 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2990 PR tree-optimization/84225
2991 * tree-eh.c (find_trapping_overflow): Only call
2992 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2994 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2997 * config/i386/i386.c: Reimplement the check of possible options
2998 -mibt/-mshstk conbination. Change error messages.
2999 * doc/invoke.texi: Fix a typo: remove extra '='.
3001 2018-02-06 Marek Polacek <polacek@redhat.com>
3003 PR tree-optimization/84228
3004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3006 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3009 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3010 emitted arch directives.
3011 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3012 __ARM_FEATURE_COPROC before changing architectures.
3014 2018-02-06 Richard Biener <rguenther@suse.de>
3016 * config/i386/i386.c (print_reg): Fix typo.
3017 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3019 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3021 * configure: Regenerate.
3023 2018-02-05 Martin Sebor <msebor@redhat.com>
3025 PR tree-optimization/83369
3026 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3029 2018-02-05 Martin Liska <mliska@suse.cz>
3031 * doc/invoke.texi: Cherry-pick upstream r323995.
3033 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3035 * ira.c (ira_init_register_move_cost): Adjust comment.
3037 2018-02-05 Martin Liska <mliska@suse.cz>
3039 PR gcov-profile/84137
3040 * doc/gcov.texi: Fix typo in documentation.
3042 2018-02-05 Martin Liska <mliska@suse.cz>
3044 PR gcov-profile/83879
3045 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3046 using dlopen functionality.
3048 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3050 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3051 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3052 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3053 _mm_maskz_range_round_ss): New intrinsics.
3054 (__builtin_ia32_rangesd128_round)
3055 (__builtin_ia32_rangess128_round): Remove.
3056 (__builtin_ia32_rangesd128_mask_round,
3057 __builtin_ia32_rangess128_mask_round): New builtins.
3058 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3059 __builtin_ia32_rangess128_round): Remove.
3060 (__builtin_ia32_rangesd128_mask_round,
3061 __builtin_ia32_rangess128_mask_round): New builtins.
3062 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3063 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3064 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3065 "<round_saeonly_constraint>")): Changed to ...
3066 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3067 "<round_saeonly_scalar_constraint>")): ... this.
3068 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3069 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3070 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3071 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3072 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3074 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3076 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3078 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3079 Remove all values except native, 8540 and 8548.
3081 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3083 * config/i386/i386.c (ix86_output_function_return): Pass
3084 INVALID_REGNUM, instead of -1, as invalid register number to
3085 indirect_thunk_name and output_indirect_thunk.
3087 2018-02-02 Julia Koval <julia.koval@intel.com>
3089 * config.gcc: Add -march=icelake.
3090 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3092 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3093 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3094 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3095 (processor_target_table): Add icelake.
3096 (ix86_option_override_internal): Handle new PTAs.
3097 (get_builtin_code_for_version): Handle icelake.
3098 (M_INTEL_COREI7_ICELAKE): New.
3099 (fold_builtin_cpu): Handle icelake.
3100 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3101 * doc/invoke.texi: Add -march=icelake.
3103 2018-02-02 Julia Koval <julia.koval@intel.com>
3105 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3106 to wide_int_bitmask.
3107 * wide-int-bitmask.h: New.
3109 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3112 * config/i386/i386.md: Replace Pmode with word_mode in
3113 builtin_setjmp_setup and builtin_longjmp to support x32.
3115 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3119 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3121 (rs6000_supported_cpu_names): New static variable.
3122 (linux_cpu_translation_table): Likewise.
3123 (elf_platform) <cpu>: Define new static variable and use it.
3124 Translate kernel AT_PLATFORM name to canonical name if needed.
3125 Error if platform name is unknown.
3127 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3130 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3132 2018-02-01 Jeff Law <law@redhat.com>
3135 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3136 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3137 the scratch if RELEASE_VIA_POP is false.
3138 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3139 If we have to save a temporary register, decrement SIZE appropriately.
3140 Pass new arguments to release_scratch_register_on_entry.
3141 (ix86_adjust_stack_and_probe): Likewise.
3142 (ix86_emit_probe_stack_range): Pass new arguments to
3143 release_scratch_register_on_entry.
3145 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3147 PR rtl-optimization/84157
3148 * combine.c (change_zero_ext): Use REG_P predicate in
3149 front of HARD_REGISTER_P predicate.
3151 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3153 * config/avr/avr.c (avr_option_override): Move disabling of
3154 -fdelete-null-pointer-checks to...
3155 * common/config/avr/avr-common.c (avr_option_optimization_table):
3158 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3160 PR tree-optimization/81635
3161 * tree-data-ref.c (split_constant_offset_1): For types that
3162 wrap on overflow, try to use range info to prove that wrapping
3165 2018-02-01 Renlin Li <renlin.li@arm.com>
3168 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3170 (aarch64_register_move_cost): Likewise.
3171 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3173 (REG_CLASS_NAMES): Likewise.
3174 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3175 TAILCALL_ADDR_REGS. Remove IP registers.
3176 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3178 2018-02-01 Richard Biener <rguenther@suse.de>
3180 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3181 for specifying RPO order and allow NULL for that.
3182 * domwalk.c (dom_walker::dom_walker): Likewise.
3183 (dom_walker::walk): Handle NULL RPO order.
3184 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3186 (rewrite_update_dom_walker): Likewise.
3187 (mark_def_dom_walker): Likewise.
3189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3191 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3192 (aarch64_maybe_expand_sve_subreg_move): Declare.
3193 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3194 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3196 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3197 that are semantically a reverse operation.
3198 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3199 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3200 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3202 (aarch64_can_change_mode_class): For big-endian, forbid changes
3203 between two SVE modes if they have different element sizes.
3205 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3207 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3208 the TImode handling for big-endian targets.
3210 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3212 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3213 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3215 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3216 Remove BSWAP handing for big-endian targets and use the form of
3217 LD1RQ appropariate for the mode.
3219 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3221 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3222 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3225 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3228 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3229 check for operands that need to go through aarch64_sve_reload_be.
3231 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3233 PR tree-optimization/81661
3234 PR tree-optimization/84117
3235 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3236 * tree-eh.c: Include gimplify.h.
3237 (find_trapping_overflow, replace_trapping_overflow,
3238 rewrite_to_non_trapping_overflow): New functions.
3239 * tree-vect-loop.c: Include tree-eh.h.
3240 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3241 * tree-data-ref.c: Include tree-eh.h.
3242 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3244 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3246 PR rtl-optimization/84123
3247 * combine.c (change_zero_ext): Check if hard register satisfies
3248 can_change_dest_mode before calling gen_lowpart_SUBREG.
3250 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3253 * ira.c (ira_init_register_move_cost): Remove assert.
3255 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3257 PR rtl-optimization/84071
3258 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3259 * doc/tm.texi: Regenerate.
3261 2018-01-31 Richard Biener <rguenther@suse.de>
3263 PR tree-optimization/84132
3264 * tree-data-ref.c (analyze_miv_subscript): Properly
3265 check whether evolution_function_is_affine_multivariate_p
3266 before calling gcd_of_steps_may_divide_p.
3268 2018-01-31 Julia Koval <julia.koval@intel.com>
3271 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3272 * config/i386/i386.md (rdpid_rex64) New.
3273 (rdpid): Make 32bit only.
3275 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3278 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3279 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3281 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3284 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3286 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3288 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3290 PR rtl-optimization/84071
3291 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3292 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3294 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3296 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3297 (arc_attribute_table): Add 'aux' attribute.
3298 (arc_in_small_data_p): Consider aux like variables.
3299 (arc_is_aux_reg_p): New function.
3300 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3301 (arc_get_aux_arg): New function.
3302 (prepare_move_operands): Handle aux-register access.
3303 (arc_handle_aux_attribute): New function.
3304 * doc/extend.texi (ARC Variable attributes): Add subsection.
3306 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3308 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3309 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3310 (arc_attribute_table): Add 'uncached' attribute.
3311 (arc_print_operand): Print '.di' flag for uncached memory
3313 (arc_in_small_data_p): Do not consider for small data the uncached
3315 (arc_is_uncached_mem_p): New function.
3316 * config/arc/predicates.md (compact_store_memory_operand): Check
3317 for uncached memory accesses.
3318 (nonvol_nonimm_operand): Likewise.
3319 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3321 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3324 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3325 falign-loops=): Add Optimization flag.
3327 2018-01-30 Jeff Law <law@redhat.com>
3330 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3331 INT_REGISTERS_SAVED. Check it prior to calling
3332 get_scratch_register_on_entry.
3333 (ix86_adjust_stack_and_probe): Similarly.
3334 (ix86_emit_probe_stack_range): Similarly.
3335 (ix86_expand_prologue): Corresponding changes.
3337 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3340 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3341 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3343 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3346 * lra-constraints.c (curr_insn_transform): Process AND in the
3349 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3351 PR rtl-optimization/83986
3352 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3353 dependence against last_pending_memory_flush in addition to
3356 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3358 PR tree-optimization/81611
3359 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3362 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3365 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3368 2018-01-30 Richard Biener <rguenther@suse.de>
3369 Jakub Jelinek <jakub@redhat.com>
3371 PR tree-optimization/84111
3372 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3373 inner loops added during recursion, as they don't have up-to-date
3376 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3379 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3380 (can_inline_edge_by_limits_p): ... here.
3381 (can_early_inline_edge_p, check_callers,
3382 update_caller_keys, update_callee_keys, recursive_inlining,
3383 add_new_edges_to_heap, speculation_useful_p,
3384 inline_small_functions,
3385 inline_small_functions, flatten_function,
3386 inline_to_all_callers_1): Update.
3388 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3390 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3393 2018-01-30 Richard Biener <rguenther@suse.de>
3395 PR tree-optimization/83008
3396 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3397 invariant and constant vector uses in stmts when they need
3400 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3403 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3404 * configure: Regenerate.
3406 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3408 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3410 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3411 Use gen_rtx_REG rather than gen_lowpart.
3413 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3415 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3416 rather than 0 when creating partial subregs.
3418 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3420 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3423 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3426 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3427 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3428 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3429 flags. This restores the settings used before the 2017-07-24.
3430 Turning off pre increment/decrement/modify allows IVOPTS to
3431 optimize DF/SF loops where the index is an int.
3433 2018-01-29 Richard Biener <rguenther@suse.de>
3434 Kelvin Nilsen <kelvin@gcc.gnu.org>
3437 * tree-vect-stmts.c (vectorizable_call): Don't call
3438 targetm.vectorize_builtin_md_vectorized_function if callee is
3441 2018-01-22 Carl Love <cel@us.ibm.com>
3443 * doc/extend.tex: Fix typo in second arg in
3444 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3446 2018-01-29 Richard Biener <rguenther@suse.de>
3448 PR tree-optimization/84086
3449 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3450 (flush_ssaname_freelist): When SSA names were released reset
3451 the SCEV hash table.
3453 2018-01-29 Richard Biener <rguenther@suse.de>
3455 PR tree-optimization/84057
3456 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3457 removed paths when removing edges.
3459 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3461 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3462 -mfunction-return=@var{choice}.
3464 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3467 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3469 (layout::print_source_line): Likewise.
3470 (test_get_line_width_without_trailing_whitespace): Add test cases.
3472 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3475 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3478 2018-01-26 Jim Wilson <jimw@sifive.com>
3480 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3482 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3485 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3487 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3488 and CMP + SUB-immediate -> SUBS.
3490 2018-01-26 Martin Sebor <msebor@redhat.com>
3492 PR tree-optimization/83896
3493 * tree-ssa-strlen.c (get_string_len): Rename...
3494 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3495 Avoid assuming length is constant.
3496 (handle_char_store): Use HOST_WIDE_INT for string length.
3498 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3501 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3502 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3504 2018-01-26 Richard Biener <rguenther@suse.de>
3506 PR rtl-optimization/84003
3507 * dse.c (record_store): Only record redundant stores when
3508 the earlier store aliases at least all accesses the later one does.
3510 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3512 PR rtl-optimization/83985
3513 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3514 REG_CFA_RESTORE insns.
3515 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3516 REG_CFA_RESTORE insns here.
3519 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3520 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3522 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3524 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3525 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3526 (arc_init): Likewise.
3527 (arc_override_options): Likewise.
3528 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3530 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3532 * config/arc/arc.h (TARGET_DBNZ): Define.
3533 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3534 properly set the tune attribute.
3535 (dbnz): Use TARGET_DBNZ guard.
3536 * config/arc/arc.opt (mtune): Add core3 option.
3538 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3540 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3541 recognize new pic like addresses.
3542 (arc_delegitimize_address): Clean up.
3544 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3546 * config/arc/arc-arches.def: Option mrf16 valid for all
3548 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3549 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3550 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3551 * config/arc/arc-tables.opt: Regenerate.
3552 * config/arc/arc.c (arc_conditional_register_usage): Handle
3553 reduced register file case.
3554 (arc_file_start): Set must have build attributes.
3555 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3557 * config/arc/arc.opt (mrf16): Add new option.
3558 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3559 * config/arc/genmultilib.awk: Handle new mrf16 option.
3560 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3561 * config/arc/t-multilib: Regenerate.
3562 * doc/invoke.texi (ARC Options): Document mrf16 option.
3564 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3566 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3567 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3568 (arc_attribute_table): Add 'secure_call' attribute.
3569 (arc_print_operand): Print secure call operand.
3570 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3572 (arc_is_secure_call_p): New function. * config/arc/arc.md
3573 (call_i): Add support for sjli instruction.
3574 (call_value_i): Likewise.
3575 * config/arc/constraints.md (Csc): New constraint.
3577 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3578 John Eric Martin <John.Martin@emmicro-us.com>
3580 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3581 * config/arc/arc.c (_arc_jli_section): New struct.
3582 (arc_jli_section): New type.
3583 (rc_jli_sections): New static variable.
3584 (arc_handle_jli_attribute): New function.
3585 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3586 (arc_file_end): New function.
3587 (TARGET_ASM_FILE_END): Define.
3588 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3589 (arc_add_jli_section): New function.
3590 (jli_call_scan): Likewise.
3591 (arc_reorg): Call jli_call_scan.
3592 (arc_output_addsi): Remove 'S' from printing asm operand.
3593 (arc_is_jli_call_p): New function.
3594 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3596 (movhi_insn): Likewise.
3597 (movsi_insn): Likewise.
3598 (movsi_set_cc_insn): Likewise.
3599 (loadqi_update): Likewise.
3600 (load_zeroextendqisi_update): Likewise.
3601 (load_signextendqisi_update): Likewise.
3602 (loadhi_update): Likewise.
3603 (load_zeroextendhisi_update): Likewise.
3604 (load_signextendhisi_update): Likewise.
3605 (loadsi_update): Likewise.
3606 (loadsf_update): Likewise.
3607 (movsicc_insn): Likewise.
3608 (bset_insn): Likewise.
3609 (bxor_insn): Likewise.
3610 (bclr_insn): Likewise.
3611 (bmsk_insn): Likewise.
3612 (bicsi3_insn): Likewise.
3613 (cmpsi_cc_c_insn): Likewise.
3614 (movsi_ne): Likewise.
3615 (movsi_cond_exec): Likewise.
3616 (clrsbsi2): Likewise.
3627 (call_i): Remove 'S' asm letter, add jli instruction.
3628 (call_value_i): Likewise.
3629 * config/arc/arc.op (mjli-always): New option.
3630 * config/arc/constraints.md (Cji): New constraint.
3631 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3633 (subsf3_fpx): Likewise.
3634 (mulsf3_fpx): Likewise.
3635 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3637 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3639 * doc/invoke.texi (ARC): Document mjli-always option.
3641 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3643 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3644 avoid addition with 0 and use incw and decw where possible.
3646 2018-01-26 Richard Biener <rguenther@suse.de>
3648 PR tree-optimization/81082
3649 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3650 association if it requires casting to unsigned.
3651 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3652 from fold_plusminus_mult_expr to catch important cases late when
3653 range info is available.
3655 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3657 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3658 * configure.ac (hidden_linkonce): New test.
3659 * configure: Regenerate.
3660 * config.in: Regenerate.
3662 2018-01-26 Julia Koval <julia.koval@intel.com>
3664 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3665 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3666 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3667 _mm_mask_bitshuffle_epi64_mask): Fix type.
3668 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3669 USI_FTYPE_V4DI_V4DI_USI): Remove.
3670 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3671 __builtin_ia32_vpshufbitqmb256_mask,
3672 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3673 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3674 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3676 2018-01-26 Alan Modra <amodra@gmail.com>
3679 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3680 UNSPEC_VBPERMQ. Sort other unspecs.
3682 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3684 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3686 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3689 * predict.c (drop_profile): Do not push/pop cfun; update also
3691 (handle_missing_profiles): Fix logic looking for zero profiles.
3693 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3696 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3697 on functions with #pragma omp declare simd or functions with simd
3699 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3700 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3701 Remove trailing \n from warning_at calls.
3703 2018-01-25 Tom de Vries <tom@codesourcery.com>
3706 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3707 for neutered workers.
3709 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3712 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3713 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3715 2018-01-24 Jeff Law <law@redhat.com>
3718 * i386.c (get_probe_interval): Move to earlier point.
3719 (ix86_compute_frame_layout): If -fstack-clash-protection and
3720 the frame is larger than the probe interval, then use pushes
3721 to save registers rather than reg->mem moves.
3722 (ix86_expand_prologue): Remove conditional for int_registers_saved
3725 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3728 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3729 min/max for never referenced object.
3731 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3734 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3736 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3737 attributes from DECL_ATTRIBUTES (decl) without affecting
3738 DECL_ATTRIBUTES (current_function_decl).
3739 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3740 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3742 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3744 PR tree-optimization/83979
3745 * fold-const.c (fold_comparison): Use constant_boolean_node
3746 instead of boolean_{true,false}_node.
3748 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3750 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3753 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3755 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3756 Simplify the clause that sets the length attribute.
3757 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3758 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3759 clause that sets the length attribute.
3760 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3762 2018-01-24 Tom de Vries <tom@codesourcery.com>
3765 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3766 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3767 Add strict parameter.
3768 (prevent_branch_around_nothing): Insert dummy insn between branch to
3769 label and label with no ptx insn inbetween.
3770 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3772 2018-01-24 Tom de Vries <tom@codesourcery.com>
3775 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3776 for neutered threads in warp.
3777 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3779 2018-01-24 Richard Biener <rguenther@suse.de>
3781 PR tree-optimization/83176
3782 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3785 2018-01-24 Richard Biener <rguenther@suse.de>
3787 PR tree-optimization/82819
3788 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3789 code generating pluses that are no-ops in the target precision.
3791 2018-01-24 Richard Biener <rguenther@suse.de>
3794 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3796 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3798 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3799 to merge probabilities.
3800 * predict.c (probably_never_executed): Also mark as cold functions
3801 with global 0 profile and guessed local profile.
3802 * profile-count.c (profile_probability::combine_with_count): New
3804 * profile-count.h (profile_probability::operator*,
3805 profile_probability::operator*=, profile_probability::operator/,
3806 profile_probability::operator/=): Reduce precision to adjusted
3807 and set value to guessed on contradictory divisions.
3808 (profile_probability::combine_with_freq): Remove.
3809 (profile_probability::combine_wiht_count): Declare.
3810 (profile_count::force_nonzero):: Set to adjusted.
3811 (profile_count::probability_in):: Set quality to adjusted.
3812 * tree-ssa-tail-merge.c (replace_block_by): Use
3815 2018-01-23 Andrew Waterman <andrew@sifive.com>
3816 Jim Wilson <jimw@sifive.com>
3818 * config/riscv/riscv.c (riscv_stack_boundary): New.
3819 (riscv_option_override): Set riscv_stack_boundary. Handle
3820 riscv_preferred_stack_boundary_arg.
3821 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3822 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3823 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3824 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3825 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3826 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3828 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3831 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3832 of struct ix86_frame.
3833 (ix86_expand_epilogue): Likewise. Add a local variable for
3834 the reg_save_offset field in struct ix86_frame.
3836 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3838 PR tree-optimization/82604
3839 * tree-loop-distribution.c (enum partition_kind): New enum item
3840 PKIND_PARTIAL_MEMSET.
3841 (partition_builtin_p): Support above new enum item.
3842 (generate_code_for_partition): Ditto.
3843 (compute_access_range): Differentiate cases that equality can be
3844 proven at all loops, the innermost loops or no loops.
3845 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3846 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3847 (finalize_partitions, distribute_loop): Don't fuse partition of
3848 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3849 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3852 2018-01-23 Martin Liska <mliska@suse.cz>
3854 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3855 order to ignore the predictor.
3856 (PRED_POLYMORPHIC_CALL): Likewise.
3857 (PRED_RECURSIVE_CALL): Likewise.
3859 2018-01-23 Martin Liska <mliska@suse.cz>
3861 * tree-profile.c (tree_profiling): Print function header to
3862 aware reader which function we are working on.
3863 * value-prof.c (gimple_find_values_to_profile): Do not print
3864 not interesting value histograms.
3866 2018-01-23 Martin Liska <mliska@suse.cz>
3868 * profile-count.h (enum profile_quality): Add
3869 profile_uninitialized as the first value. Do not number values
3870 as they are zero based.
3871 (profile_count::verify): Update sanity check.
3872 (profile_probability::verify): Likewise.
3874 2018-01-23 Nathan Sidwell <nathan@acm.org>
3876 * doc/invoke.texi (ffor-scope): Deprecate.
3878 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3880 PR tree-optimization/83510
3881 * domwalk.c (set_all_edges_as_executable): New function.
3882 (dom_walker::dom_walker): Convert bool param
3883 "skip_unreachable_blocks" to enum reachability. Move setup of
3884 edge flags to set_all_edges_as_executable and only do it when
3885 reachability is REACHABLE_BLOCKS.
3886 * domwalk.h (enum dom_walker::reachability): New enum.
3887 (dom_walker::dom_walker): Convert bool param
3888 "skip_unreachable_blocks" to enum reachability.
3889 (set_all_edges_as_executable): New decl.
3890 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3891 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3893 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3894 but converting true to REACHABLE_BLOCKS.
3895 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3897 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3898 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3899 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3901 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3902 if check_all_array_refs will be called.
3904 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3906 * tree.c (selftest::test_location_wrappers): Add more test
3909 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3911 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3912 (selftest::test_bit_in_range): Likewise.
3914 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3917 * doc/sourcebuild.texi (vect_float): Say that the selector
3918 only describes the situation when -funsafe-math-optimizations is on.
3919 (vect_float_strict): Document.
3921 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3923 PR tree-optimization/83965
3924 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3925 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3926 instead of checking only for a reduction.
3927 (vect_recog_widen_sum_pattern): Likewise.
3929 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3931 * predict.c (probably_never_executed): Only use precise profile info.
3932 (compute_function_frequency): Skip after inlining hack since we now
3933 have quality checking.
3935 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3937 * profile-count.h (profile_probability::very_unlikely,
3938 profile_probability::unlikely, profile_probability::even): Set
3939 precision to guessed.
3941 2018-01-23 Richard Biener <rguenther@suse.de>
3943 PR tree-optimization/83963
3944 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3945 Properly terminate dominator walk when crossing the exit edge not
3946 when visiting its source block.
3948 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3951 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3952 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3954 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3956 PR tree-optimization/83957
3957 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3958 semicolon after for body surrounded by braces.
3960 PR tree-optimization/83081
3961 * profile-count.h (profile_probability::split): New method.
3962 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3963 Use profile_probability::split.
3964 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3965 when splitting a single conditional jump into 2.
3967 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3969 PR tree-optimization/69452
3970 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3973 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3975 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3976 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3977 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3979 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3981 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3982 * config/rl78/rl78.md (movdi): New define_expand.
3983 * config/rl78/rl78.c (rl78_split_movdi): New function.
3985 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3988 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3990 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3991 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3992 128-bit to produce an UNSPEC move to get the double word with the
3993 signbit and then a shift directly to do signbit.
3994 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3995 implementation with a new version that just does either a direct
3996 move or a regular move. Move memory interface to separate insns.
3997 Move insns so they are next to the expander.
3998 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3999 with signbit move. Split big and little endian case.
4000 (signbit<mode>2_dm_mem_le): Likewise.
4001 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4002 (signbit<mode>2_dm2): Likewise.
4004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4006 * config/rl78/rl78.md (anddi3): New define_expand.
4008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4010 * config/rl78/rl78.md (umindi3): New define_expand.
4012 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4014 * config/rl78/rl78.md (smindi3): New define_expand.
4016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4018 * config/rl78/rl78.md (smaxdi3): New define_expand.
4020 2018-01-22 Carl Love <cel@us.ibm.com>
4022 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4023 LVX_V1TI): Add macro expansion.
4024 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4025 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4026 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4027 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4028 Change check to determine if the instruction is a byte reversing
4029 entry. Fix typo in comment.
4030 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4031 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4032 Add def_builtin calls for new builtins.
4033 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4034 Add define_insn expansion.
4036 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4038 * config/rl78/rl78.md (umaxdi3): New define_expand.
4040 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4042 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4043 for non-QImode registers.
4045 2018-01-22 Richard Biener <rguenther@suse.de>
4047 PR tree-optimization/83963
4048 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4049 including the loop exit block.
4050 (scop_detection::merge_sese): Likewise.
4051 (scop_detection::add_scop): Do it here instead.
4053 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4055 * doc/sourcebuild.texi (arm_softfloat): Document.
4057 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4060 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4061 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4062 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4064 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4065 David Edelsohn <dje.gcc@gmail.com>
4068 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4069 Change "crset eq" to "crset 2".
4070 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4071 (*call_indirect_aix<mode>_nospec): Likewise.
4072 (*call_value_indirect_aix<mode>_nospec): Likewise.
4073 (*call_indirect_elfv2<mode>_nospec): Likewise.
4074 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4075 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4076 change assembly output from . to $.
4077 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4078 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4079 (*tablejump<mode>_internal1_nospec): Likewise.
4081 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4084 * config/sh/sh_optimize_sett_clrt.cc:
4085 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4087 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4089 PR tree-optimization/83940
4090 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4091 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4092 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4093 use it to pass back the definition type.
4094 (vect_check_store_rhs): Likewise rhs_dt_out.
4095 (vect_build_gather_load_calls): Add a mask_dt argument and use
4096 it instead of a call to vect_is_simple_use.
4097 (vectorizable_store): Update calls to vect_check_load_store_mask
4098 and vect_check_store_rhs. Use the dt returned by the latter instead
4099 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4100 instead of calls to vect_is_simple_use. Pass the scalar rather
4101 than the vector operand to vect_is_simple_use when handling
4102 second and subsequent copies of an rhs value.
4103 (vectorizable_load): Update calls to vect_check_load_store_mask
4104 and vect_build_gather_load_calls. Use the cached mask_dt and
4105 gs_info.offset_dt instead of calls to vect_is_simple_use.
4107 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4110 * tree-emutls.c: Include gimplify.h.
4111 (lower_emutls_2): New function.
4112 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4113 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4114 it before further processing.
4117 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4118 UINTVAL (trueop1) instead of INTVAL (op1).
4120 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4124 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4125 INCOMING_FRAME_SP_OFFSET if not defined.
4126 (scan_trace): Add ENTRY argument. If true and
4127 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4128 emit a note to adjust the CFA offset.
4129 (create_cfi_notes): Adjust scan_trace callers.
4130 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4131 INCOMING_FRAME_SP_OFFSET in the CIE.
4132 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4133 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4135 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4136 * doc/tm.texi: Regenerated.
4138 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4140 PR rtl-optimization/83147
4141 * lra-constraints.c (remove_inheritance_pseudos): Use
4142 lra_substitute_pseudo_within_insn.
4144 2018-01-19 Tom de Vries <tom@codesourcery.com>
4145 Cesar Philippidis <cesar@codesourcery.com>
4148 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4150 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4153 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4154 spaces for function labels.
4156 2018-01-19 Martin Liska <mliska@suse.cz>
4158 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4159 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4160 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4161 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4162 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4163 (PRED_CONST_RETURN): Change from 69 to 65.
4164 (PRED_NULL_RETURN): Change from 91 to 71.
4165 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4166 (PRED_LOOP_GUARD): Change from 66 to 73.
4168 2018-01-19 Martin Liska <mliska@suse.cz>
4170 * predict.c (predict_insn_def): Add new assert.
4171 (struct branch_predictor): Change type to signed integer.
4172 (test_prediction_value_range): Amend test to cover
4174 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4175 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4176 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4177 (PRED_LOOP_IV_COMPARE): Likewise.
4178 * predict.h (PROB_UNINITIALIZED): Define new constant.
4180 2018-01-19 Martin Liska <mliska@suse.cz>
4182 * predict.c (dump_prediction): Add new format for
4183 analyze_brprob.py script which is enabled with -details
4185 * profile-count.h (precise_p): New function.
4187 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4189 PR tree-optimization/83922
4190 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4191 there are no statements that need masking.
4192 (vect_active_double_reduction_p): New function.
4193 (vect_analyze_loop_operations): Use it when handling phis that
4194 are not in the loop header.
4196 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4198 PR tree-optimization/83914
4199 * tree-vect-loop.c (vectorizable_induction): Don't convert
4200 init_expr or apply the peeling adjustment for inductions
4201 that are nested within the vectorized loop.
4203 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4205 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4208 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4212 * function.h (gimplify_parameters): Add gimple_seq * argument.
4213 * function.c: Include gimple.h and options.h.
4214 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4215 for the added local temporaries if needed.
4216 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4217 if there are any parameter cleanups, wrap whole body into a
4218 try/finally with the cleanups.
4220 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4223 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4224 Use GET_MODE_CLASS for scalar floating point.
4226 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4230 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4231 Fix call of call_cgraph_insertion_hooks.
4233 2018-01-18 Martin Sebor <msebor@redhat.com>
4235 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4237 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4240 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4243 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4246 * common.opt: (-ffile-prefix-map): New option.
4247 * opts.c (common_handle_option): Defer it.
4248 * opts-global.c (handle_common_deferred_options): Handle it.
4249 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4250 * file-prefix-map.h: New file.
4251 (remap_debug_filename, add_debug_prefix_map): ...here.
4252 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4253 * final.c (debug_prefix_map, add_debug_prefix_map
4254 remap_debug_filename): Move to...
4255 * file-prefix-map.c: New file.
4256 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4257 generalize, get rid of alloca(), use strrchr() instead of strchr().
4258 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4259 Implement in terms of add_prefix_map().
4260 (remap_macro_filename, remap_debug_filename): Implement in term of
4262 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4263 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4264 * dbxout.c: Include file-prefix-map.h.
4265 * varasm.c: Likewise.
4266 * vmsdbgout.c: Likewise.
4267 * xcoffout.c: Likewise.
4268 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4269 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4270 * doc/invoke.texi (-ffile-prefix-map): Document.
4271 (-fdebug-prefix-map): Update description.
4273 2018-01-18 Martin Liska <mliska@suse.cz>
4275 * config/i386/i386.c (indirect_thunk_name): Document that also
4277 (output_indirect_thunk): Document why both instructions
4278 (pause and lfence) are generated.
4280 2018-01-18 Richard Biener <rguenther@suse.de>
4282 PR tree-optimization/83887
4283 * graphite-scop-detection.c
4284 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4285 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4286 (scop_detection::merge_sese): Re-implement with a flood-fill
4287 algorithm that properly finds a SESE region if it exists.
4289 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4292 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4293 pointer_diff optimizations use view_convert instead of convert.
4295 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4297 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4298 Generate different code for -mno-speculate-indirect-jumps.
4299 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4300 (*call_indirect_aix<mode>): Disable for
4301 -mno-speculate-indirect-jumps.
4302 (*call_indirect_aix<mode>_nospec): New define_insn.
4303 (*call_value_indirect_aix<mode>): Disable for
4304 -mno-speculate-indirect-jumps.
4305 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4306 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4307 -mno-speculate-indirect-jumps.
4308 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4310 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4312 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4313 long double type, set the flags for noting the default long double
4314 type, even if we don't pass or return a long double type.
4316 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4319 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4322 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4324 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4325 support for merge[hl].
4326 (fold_mergehl_helper): New helper function.
4327 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4328 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4329 (altivec_vmrglw_direct): Add xxmrglw insn.
4331 2018-01-17 Andrew Waterman <andrew@sifive.com>
4333 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4334 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4336 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4339 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4340 call the lto_location_cache before reading the
4341 DECL_SOURCE_LOCATION of the types.
4343 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4344 Richard Sandiford <richard.sandiford@linaro.org>
4346 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4347 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4348 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4349 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4350 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4352 * config/aarch64/constraints.md (aarch64_movti_operand):
4354 * config/aarch64/predicates.md (Uti): Add new constraint.
4356 2018-01-17 Carl Love <cel@us.ibm.com>
4358 * config/rs6000/vsx.md (define_expand xl_len_r,
4359 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4360 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4362 (define_expand, define_insn): Move the shift left from the
4363 define_insn to the define_expand for lxvl and stxvl instructions.
4364 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4365 and XL_LEN_R definitions to PURE.
4367 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4369 * config/i386/i386.c (indirect_thunk_name): Declare regno
4370 as unsigned int. Compare regno with INVALID_REGNUM.
4371 (output_indirect_thunk): Ditto.
4372 (output_indirect_thunk_function): Ditto.
4373 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4374 in the call to output_indirect_thunk_function.
4376 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4379 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4380 rather than the size of inner_type to determine the stack slot size
4381 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4383 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4386 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4389 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4391 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4392 endian Linux systems to optionally enable multilibs for selecting
4393 the long double type if the user configured an explicit type.
4394 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4395 have no long double multilibs if not defined.
4396 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4397 warn if the user used -mabi={ieee,ibm}longdouble and we built
4398 multilibs for long double.
4399 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4400 appropriate multilib option.
4401 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4403 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4404 for building long double multilibs.
4405 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4407 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4409 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4412 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4414 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4417 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4420 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4423 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4425 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4426 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4428 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4430 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4431 different rtl trees depending on TARGET_64BIT.
4432 (rs6000_gen_lvx): Likewise.
4434 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4436 * config/visium/visium.md (nop): Tweak comment.
4437 (hazard_nop): Likewise.
4439 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4441 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4442 -mspeculate-indirect-jumps.
4443 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4444 for -mno-speculate-indirect-jumps.
4445 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4446 (*call_value_indirect_elfv2<mode>): Disable for
4447 -mno-speculate-indirect-jumps.
4448 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4449 (indirect_jump): Emit different RTL for
4450 -mno-speculate-indirect-jumps.
4451 (*indirect_jump<mode>): Disable for
4452 -mno-speculate-indirect-jumps.
4453 (*indirect_jump<mode>_nospec): New define_insn.
4454 (tablejump): Emit different RTL for
4455 -mno-speculate-indirect-jumps.
4456 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4457 (tablejumpsi_nospec): New define_expand.
4458 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4459 (tablejumpdi_nospec): New define_expand.
4460 (*tablejump<mode>_internal1): Disable for
4461 -mno-speculate-indirect-jumps.
4462 (*tablejump<mode>_internal1_nospec): New define_insn.
4463 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4466 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4468 * caller-save.c (insert_save): Drop unnecessary parameter. All
4471 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4472 Richard Biener <rguenth@suse.de>
4475 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4476 return early, inline manually is_gimple_sizepos. Make sure if we
4477 call gimplify_expr we don't end up with a gimple constant.
4478 * tree.c (variably_modified_type_p): Don't return true for
4479 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4480 * gimplify.h (is_gimple_sizepos): Remove.
4482 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4484 PR tree-optimization/83857
4485 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4486 vectorizable_live_operation for pure SLP statements.
4487 (vectorizable_live_operation): Handle PHIs.
4489 2018-01-16 Richard Biener <rguenther@suse.de>
4491 PR tree-optimization/83867
4492 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4493 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4495 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4498 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4499 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4500 If off is not INTEGER_CST, issue a may not be aligned warning
4501 rather than isn't aligned. Use isn%'t rather than isn't.
4502 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4504 <case MULT_EXPR>: Improve the case when bottom and one of the
4505 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4506 operand, in that case check if the other operand is multiple of
4507 bottom divided by the INTEGER_CST operand.
4509 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4512 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4513 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4514 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4515 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4516 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4517 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4518 (pa_function_arg_size): New function.
4520 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4522 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4523 in a separate statement.
4525 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4527 PR tree-optimization/83847
4528 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4529 group gathers and scatters.
4531 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4533 PR rtl-optimization/86620
4534 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4536 PR rtl-optimization/83213
4537 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4538 to last if both are JUMP_INSNs.
4540 PR tree-optimization/83843
4541 * gimple-ssa-store-merging.c
4542 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4543 store_immediate_info for bswap/nop orig_stores.
4545 2018-01-15 Andrew Waterman <andrew@sifive.com>
4547 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4549 <UDIV>: Increase cost if !TARGET_DIV.
4551 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4553 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4554 (define_attr "cr_logical_3op"): New.
4555 (cceq_ior_compare): Adjust.
4556 (cceq_ior_compare_complement): Adjust.
4557 (*cceq_rev_compare): Adjust.
4558 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4559 (is_cracked_insn): Adjust.
4560 (insn_must_be_first_in_group): Adjust.
4561 * config/rs6000/40x.md: Adjust.
4562 * config/rs6000/440.md: Adjust.
4563 * config/rs6000/476.md: Adjust.
4564 * config/rs6000/601.md: Adjust.
4565 * config/rs6000/603.md: Adjust.
4566 * config/rs6000/6xx.md: Adjust.
4567 * config/rs6000/7450.md: Adjust.
4568 * config/rs6000/7xx.md: Adjust.
4569 * config/rs6000/8540.md: Adjust.
4570 * config/rs6000/cell.md: Adjust.
4571 * config/rs6000/e300c2c3.md: Adjust.
4572 * config/rs6000/e500mc.md: Adjust.
4573 * config/rs6000/e500mc64.md: Adjust.
4574 * config/rs6000/e5500.md: Adjust.
4575 * config/rs6000/e6500.md: Adjust.
4576 * config/rs6000/mpc.md: Adjust.
4577 * config/rs6000/power4.md: Adjust.
4578 * config/rs6000/power5.md: Adjust.
4579 * config/rs6000/power6.md: Adjust.
4580 * config/rs6000/power7.md: Adjust.
4581 * config/rs6000/power8.md: Adjust.
4582 * config/rs6000/power9.md: Adjust.
4583 * config/rs6000/rs64.md: Adjust.
4584 * config/rs6000/titan.md: Adjust.
4586 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4588 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4589 ix86_indirect_branch_register logic.
4591 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4593 * config/i386/constraints.md (Bs): Update
4594 ix86_indirect_branch_register check. Don't check
4595 ix86_indirect_branch_register with GOT_memory_operand.
4597 * config/i386/predicates.md (GOT_memory_operand): Don't check
4598 ix86_indirect_branch_register here.
4599 (GOT32_symbol_operand): Likewise.
4601 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4603 * config/i386/predicates.md (constant_call_address_operand):
4604 Rewrite ix86_indirect_branch_register logic.
4605 (sibcall_insn_operand): Likewise.
4607 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4609 * config/i386/constraints.md (Bs): Replace
4610 ix86_indirect_branch_thunk_register with
4611 ix86_indirect_branch_register.
4613 * config/i386/i386.md (indirect_jump): Likewise.
4614 (tablejump): Likewise.
4615 (*sibcall_memory): Likewise.
4616 (*sibcall_value_memory): Likewise.
4617 Peepholes of indirect call and jump via memory: Likewise.
4618 * config/i386/i386.opt: Likewise.
4619 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4620 (GOT_memory_operand): Likewise.
4621 (call_insn_operand): Likewise.
4622 (sibcall_insn_operand): Likewise.
4623 (GOT32_symbol_operand): Likewise.
4625 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4628 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4629 type rather than type addr's type points to.
4630 (expand_omp_atomic_mutex): Likewise.
4631 (expand_omp_atomic): Likewise.
4633 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4636 * config/i386/i386.c (output_indirect_thunk_function): Use
4637 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4638 for __x86_return_thunk.
4640 2018-01-15 Richard Biener <rguenther@suse.de>
4643 * expmed.c (extract_bit_field_1): Fix typo.
4645 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4648 * config/arm/iterators.md (VF): New mode iterator.
4649 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4650 Remove integer-related logic from pattern.
4651 (neon_vabd<mode>_3): Likewise.
4653 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4656 * common.opt (fstrict-overflow): No longer an alias.
4657 (fwrapv-pointer): New option.
4658 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4659 also for pointer types based on flag_wrapv_pointer.
4660 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4661 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4662 opts->x_flag_wrapv got set.
4663 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4664 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4665 POINTER_TYPE_OVERFLOW_UNDEFINED.
4666 * match.pd: Likewise in address comparison pattern.
4667 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4669 2018-01-15 Richard Biener <rguenther@suse.de>
4672 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4673 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4674 Reset type names to their identifier if their TYPE_DECL doesn't
4675 have linkage (and thus is used for ODR and devirt).
4676 (save_debug_info_for_decl): Remove.
4677 (save_debug_info_for_type): Likewise.
4678 (add_tree_to_fld_list): Adjust.
4679 * tree-pretty-print.c (dump_generic_node): Make dumping of
4680 type names more robust.
4682 2018-01-15 Richard Biener <rguenther@suse.de>
4684 * BASE-VER: Bump to 8.0.1.
4686 2018-01-14 Martin Sebor <msebor@redhat.com>
4689 * builtins.c (check_access): Avoid warning when the no-warning bit
4692 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4694 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4695 * ira-color (allocno_hard_regs_compare): Likewise.
4697 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4700 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4701 Use .pushsection/.popsection.
4703 2018-01-14 Martin Sebor <msebor@redhat.com>
4706 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4708 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4710 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4711 entry from extra_headers.
4712 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4713 extra_headers, make the list bitwise identical to the i?86-*-* one.
4715 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4717 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4718 -mcmodel=large with -mindirect-branch=thunk,
4719 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4720 -mfunction-return=thunk-extern.
4721 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4722 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4723 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4725 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4727 * config/i386/i386.c (print_reg): Print the name of the full
4728 integer register without '%'.
4729 (ix86_print_operand): Handle 'V'.
4730 * doc/extend.texi: Document 'V' modifier.
4732 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4734 * config/i386/constraints.md (Bs): Disallow memory operand for
4735 -mindirect-branch-register.
4737 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4738 (GOT_memory_operand): Likewise.
4739 (call_insn_operand): Likewise.
4740 (sibcall_insn_operand): Likewise.
4741 (GOT32_symbol_operand): Likewise.
4742 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4743 for -mindirect-branch-register.
4744 (tablejump): Likewise.
4745 (*sibcall_memory): Likewise.
4746 (*sibcall_value_memory): Likewise.
4747 Disallow peepholes of indirect call and jump via memory for
4748 -mindirect-branch-register.
4749 (*call_pop): Replace m with Bw.
4750 (*call_value_pop): Likewise.
4751 (*sibcall_pop_memory): Replace m with Bs.
4752 * config/i386/i386.opt (mindirect-branch-register): New option.
4753 * doc/invoke.texi: Document -mindirect-branch-register option.
4755 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4757 * config/i386/i386-protos.h (ix86_output_function_return): New.
4758 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4759 set function_return_type.
4760 (indirect_thunk_name): Add ret_p to indicate thunk for function
4762 (output_indirect_thunk_function): Pass false to
4763 indirect_thunk_name.
4764 (ix86_output_indirect_branch_via_reg): Likewise.
4765 (ix86_output_indirect_branch_via_push): Likewise.
4766 (output_indirect_thunk_function): Create alias for function
4767 return thunk if regno < 0.
4768 (ix86_output_function_return): New function.
4769 (ix86_handle_fndecl_attribute): Handle function_return.
4770 (ix86_attribute_table): Add function_return.
4771 * config/i386/i386.h (machine_function): Add
4772 function_return_type.
4773 * config/i386/i386.md (simple_return_internal): Use
4774 ix86_output_function_return.
4775 (simple_return_internal_long): Likewise.
4776 * config/i386/i386.opt (mfunction-return=): New option.
4777 (indirect_branch): Mention -mfunction-return=.
4778 * doc/extend.texi: Document function_return function attribute.
4779 * doc/invoke.texi: Document -mfunction-return= option.
4781 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4783 * config/i386/i386-opts.h (indirect_branch): New.
4784 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4785 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4786 with local indirect jump when converting indirect call and jump.
4787 (ix86_set_indirect_branch_type): New.
4788 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4789 (indirectlabelno): New.
4790 (indirect_thunk_needed): Likewise.
4791 (indirect_thunk_bnd_needed): Likewise.
4792 (indirect_thunks_used): Likewise.
4793 (indirect_thunks_bnd_used): Likewise.
4794 (INDIRECT_LABEL): Likewise.
4795 (indirect_thunk_name): Likewise.
4796 (output_indirect_thunk): Likewise.
4797 (output_indirect_thunk_function): Likewise.
4798 (ix86_output_indirect_branch_via_reg): Likewise.
4799 (ix86_output_indirect_branch_via_push): Likewise.
4800 (ix86_output_indirect_branch): Likewise.
4801 (ix86_output_indirect_jmp): Likewise.
4802 (ix86_code_end): Call output_indirect_thunk_function if needed.
4803 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4805 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4806 (ix86_attribute_table): Add indirect_branch.
4807 * config/i386/i386.h (machine_function): Add indirect_branch_type
4808 and has_local_indirect_jump.
4809 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4811 (tablejump): Likewise.
4812 (*indirect_jump): Use ix86_output_indirect_jmp.
4813 (*tablejump_1): Likewise.
4814 (simple_return_indirect_internal): Likewise.
4815 * config/i386/i386.opt (mindirect-branch=): New option.
4816 (indirect_branch): New.
4819 (thunk-inline): Likewise.
4820 (thunk-extern): Likewise.
4821 * doc/extend.texi: Document indirect_branch function attribute.
4822 * doc/invoke.texi: Document -mindirect-branch= option.
4824 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4827 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4829 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4831 * ipa-inline.c (want_inline_small_function_p): Return false if
4832 inlining has already failed with CIF_FINAL_ERROR.
4833 (update_caller_keys): Call want_inline_small_function_p before
4835 (update_callee_keys): Likewise.
4837 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4839 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4841 (rs6000_quadword_masked_address_p): Likewise.
4842 (quad_aligned_load_p): Likewise.
4843 (quad_aligned_store_p): Likewise.
4844 (const_load_sequence_p): Add comment to describe the outer-most loop.
4845 (mimic_memory_attributes_and_flags): New function.
4846 (rs6000_gen_stvx): Likewise.
4847 (replace_swapped_aligned_store): Likewise.
4848 (rs6000_gen_lvx): Likewise.
4849 (replace_swapped_aligned_load): Likewise.
4850 (replace_swapped_load_constant): Capitalize argument name in
4851 comment describing this function.
4852 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4853 and stores that access quad-word aligned addresses and replace
4854 with stvx or lvx instructions when appropriate.
4855 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4856 New function prototype.
4857 (rs6000_quadword_masked_address_p): Likewise.
4858 (rs6000_gen_lvx): Likewise.
4859 (rs6000_gen_stvx): Likewise.
4860 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4861 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4862 when memory address is aligned.
4863 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4864 this split to select lvx instruction when memory address is aligned.
4865 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4866 instruction when memory address is aligned.
4867 (*vsx_le_perm_load_v16qi): Likewise.
4868 (four unnamed splitters): Modify to select the stvx instruction
4869 when memory is aligned.
4871 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4873 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4874 which appears in the queue multiple times.
4876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4877 Alan Hayward <alan.hayward@arm.com>
4878 David Sherwood <david.sherwood@arm.com>
4880 * tree-vectorizer.h (vec_lower_bound): New structure.
4881 (_loop_vec_info): Add check_nonzero and lower_bounds.
4882 (LOOP_VINFO_CHECK_NONZERO): New macro.
4883 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4884 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4885 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4886 fields. Make seg_len the distance travelled, not including the
4888 (dr_direction_indicator): Declare.
4889 (dr_zero_step_indicator): Likewise.
4890 (dr_known_forward_stride_p): Likewise.
4891 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4893 (runtime_alias_check_p): Allow runtime alias checks with
4895 (operator ==): Compare access_size and align.
4896 (prune_runtime_alias_test_list): Rework for new distinction between
4897 the access_size and seg_len.
4898 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4900 (get_segment_min_max): New function.
4901 (create_intersect_range_checks): Use it.
4902 (dr_step_indicator): New function.
4903 (dr_direction_indicator): Likewise.
4904 (dr_zero_step_indicator): Likewise.
4905 (dr_known_forward_stride_p): Likewise.
4906 * tree-loop-distribution.c (data_ref_segment_size): Return
4907 DR_STEP * (niters - 1).
4908 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4910 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4911 (vect_preserves_scalar_order_p): New function, split out from...
4912 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4913 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4914 (vect_vfa_access_size): New function.
4915 (vect_vfa_align): Likewise.
4916 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4917 (dump_lower_bound): New function.
4918 (vect_check_lower_bound): Likewise.
4919 (vect_small_gap_p): Likewise.
4920 (vectorizable_with_step_bound_p): Likewise.
4921 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4922 depencies if the vectorization factor is 1. Convert the checks
4923 for nonzero steps into checks on the bounds of DR_STEP. Try using
4924 a bunds check for variable steps if the minimum required step is
4925 relatively small. Update calls to the dr_with_seg_len
4926 constructor and to vect_compile_time_alias.
4927 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4929 (vect_loop_versioning): Call it.
4930 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4932 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4935 Alan Hayward <alan.hayward@arm.com>
4936 David Sherwood <david.sherwood@arm.com>
4938 * doc/sourcebuild.texi (vect_scatter_store): Document.
4939 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4941 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4943 * genopinit.c (main): Add supports_vec_scatter_store and
4944 supports_vec_scatter_store_cached to target_optabs.
4945 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4946 IFN_MASK_SCATTER_STORE.
4947 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4949 * internal-fn.h (internal_store_fn_p): Declare.
4950 (internal_fn_stored_value_index): Likewise.
4951 * internal-fn.c (scatter_store_direct): New macro.
4952 (expand_scatter_store_optab_fn): New function.
4953 (direct_scatter_store_optab_supported_p): New macro.
4954 (internal_store_fn_p): New function.
4955 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4956 IFN_MASK_SCATTER_STORE.
4957 (internal_fn_mask_index): Likewise.
4958 (internal_fn_stored_value_index): New function.
4959 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4961 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4962 * optabs-query.c (supports_vec_scatter_store_p): New function.
4963 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4964 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4965 true for scatter stores.
4966 (vect_gather_scatter_fn_p): Handle scatter stores too.
4967 (vect_check_gather_scatter): Consider using scatter stores if
4968 supports_vec_scatter_store_p.
4969 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4971 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4972 internal_fn_stored_value_index.
4973 (check_load_store_masking): Handle scatter stores too.
4974 (vect_get_store_rhs): Make public.
4975 (vectorizable_call): Use internal_store_fn_p.
4976 (vectorizable_store): Handle scatter store internal functions.
4977 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4978 when deciding whether the end of the group has been reached.
4979 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4980 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4981 (mask_scatter_store<mode>): New insns.
4983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4984 Alan Hayward <alan.hayward@arm.com>
4985 David Sherwood <david.sherwood@arm.com>
4987 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4988 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4989 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4991 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4992 Use vect_truncate_gather_scatter_offset if we can't treat the
4993 operation as a normal gather load or scatter store.
4994 (get_group_load_store_type): Take the gather_scatter_info
4995 as argument. Try using a gather load or scatter store for
4996 single-element groups.
4997 (get_load_store_type): Update calls to get_group_load_store_type
4998 and vect_use_strided_gather_scatters_p.
5000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5001 Alan Hayward <alan.hayward@arm.com>
5002 David Sherwood <david.sherwood@arm.com>
5004 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5005 optional tree argument.
5006 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5008 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5009 but continue to use the current value as a fallback.
5010 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5011 to compare the updates.
5012 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5013 (get_load_store_type): Use it when handling a strided access.
5014 (vect_get_strided_load_store_ops): New function.
5015 (vect_get_data_ptr_increment): Likewise.
5016 (vectorizable_load): Handle strided gather loads. Always pass
5017 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5020 Alan Hayward <alan.hayward@arm.com>
5021 David Sherwood <david.sherwood@arm.com>
5023 * doc/md.texi (gather_load@var{m}): Document.
5024 (mask_gather_load@var{m}): Likewise.
5025 * genopinit.c (main): Add supports_vec_gather_load and
5026 supports_vec_gather_load_cached to target_optabs.
5027 * optabs-tree.c (init_tree_optimization_optabs): Use
5028 ggc_cleared_alloc to allocate target_optabs.
5029 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5030 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5032 * internal-fn.h (internal_load_fn_p): Declare.
5033 (internal_gather_scatter_fn_p): Likewise.
5034 (internal_fn_mask_index): Likewise.
5035 (internal_gather_scatter_fn_supported_p): Likewise.
5036 * internal-fn.c (gather_load_direct): New macro.
5037 (expand_gather_load_optab_fn): New function.
5038 (direct_gather_load_optab_supported_p): New macro.
5039 (direct_internal_fn_optab): New function.
5040 (internal_load_fn_p): Likewise.
5041 (internal_gather_scatter_fn_p): Likewise.
5042 (internal_fn_mask_index): Likewise.
5043 (internal_gather_scatter_fn_supported_p): Likewise.
5044 * optabs-query.c (supports_at_least_one_mode_p): New function.
5045 (supports_vec_gather_load_p): Likewise.
5046 * optabs-query.h (supports_vec_gather_load_p): Declare.
5047 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5048 and memory_type field.
5049 (NUM_PATTERNS): Bump to 15.
5050 * tree-vect-data-refs.c: Include internal-fn.h.
5051 (vect_gather_scatter_fn_p): New function.
5052 (vect_describe_gather_scatter_call): Likewise.
5053 (vect_check_gather_scatter): Try using internal functions for
5054 gather loads. Recognize existing calls to a gather load function.
5055 (vect_analyze_data_refs): Consider using gather loads if
5056 supports_vec_gather_load_p.
5057 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5058 (vect_get_gather_scatter_offset_type): Likewise.
5059 (vect_convert_mask_for_vectype): Likewise.
5060 (vect_add_conversion_to_patterm): Likewise.
5061 (vect_try_gather_scatter_pattern): Likewise.
5062 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5063 (vect_vect_recog_func_ptrs): Add it.
5064 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5065 internal_fn_mask_index and internal_gather_scatter_fn_p.
5066 (check_load_store_masking): Take the gather_scatter_info as an
5067 argument and handle gather loads.
5068 (vect_get_gather_scatter_ops): New function.
5069 (vectorizable_call): Check internal_load_fn_p.
5070 (vectorizable_load): Likewise. Handle gather load internal
5072 (vectorizable_store): Update call to check_load_store_masking.
5073 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5074 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5075 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5076 (aarch64_gather_scale_operand_d): New predicates.
5077 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5078 (mask_gather_load<mode>): New insns.
5080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5081 Alan Hayward <alan.hayward@arm.com>
5082 David Sherwood <david.sherwood@arm.com>
5084 * optabs.def (fold_left_plus_optab): New optab.
5085 * doc/md.texi (fold_left_plus_@var{m}): Document.
5086 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5087 * internal-fn.c (fold_left_direct): Define.
5088 (expand_fold_left_optab_fn): Likewise.
5089 (direct_fold_left_optab_supported_p): Likewise.
5090 * fold-const-call.c (fold_const_fold_left): New function.
5091 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5092 * tree-parloops.c (valid_reduction_p): New function.
5093 (gather_scalar_reductions): Use it.
5094 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5095 (vect_finish_replace_stmt): Declare.
5096 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5097 (needs_fold_left_reduction_p): New function, split out from...
5098 (vect_is_simple_reduction): ...here. Accept reductions that
5099 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5100 (vect_force_simple_reduction): Also store the reduction type in
5101 the assignment's STMT_VINFO_REDUC_TYPE.
5102 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5103 (merge_with_identity): New function.
5104 (vect_expand_fold_left): Likewise.
5105 (vectorize_fold_left_reduction): Likewise.
5106 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5107 scalar phi in place for it. Check for target support and reject
5108 cases that would reassociate the operation. Defer the transform
5109 phase to vectorize_fold_left_reduction.
5110 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5111 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5112 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5116 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5117 call to ifc_temp_var.
5119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5120 Alan Hayward <alan.hayward@arm.com>
5121 David Sherwood <david.sherwood@arm.com>
5123 * target.def (legitimize_address_displacement): Take the original
5124 offset as a poly_int.
5125 * targhooks.h (default_legitimize_address_displacement): Update
5127 * targhooks.c (default_legitimize_address_displacement): Likewise.
5128 * doc/tm.texi: Regenerate.
5129 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5130 as an argument, moving assert of ad->disp == ad->disp_term to...
5131 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5132 Try calling targetm.legitimize_address_displacement before expanding
5133 the address rather than afterwards, and adjust for the new interface.
5134 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5135 Match the new hook interface. Handle SVE addresses.
5136 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5139 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5141 * Makefile.in (OBJS): Add early-remat.o.
5142 * target.def (select_early_remat_modes): New hook.
5143 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5144 * doc/tm.texi: Regenerate.
5145 * targhooks.h (default_select_early_remat_modes): Declare.
5146 * targhooks.c (default_select_early_remat_modes): New function.
5147 * timevar.def (TV_EARLY_REMAT): New timevar.
5148 * passes.def (pass_early_remat): New pass.
5149 * tree-pass.h (make_pass_early_remat): Declare.
5150 * early-remat.c: New file.
5151 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5153 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5156 Alan Hayward <alan.hayward@arm.com>
5157 David Sherwood <david.sherwood@arm.com>
5159 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5160 vfm1 with a bound_epilog parameter.
5161 (vect_do_peeling): Update calls accordingly, and move the prologue
5162 call earlier in the function. Treat the base bound_epilog as 0 for
5163 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5164 this base when peeling for gaps.
5165 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5166 with fully-masked loops.
5167 (vect_estimate_min_profitable_iters): Handle the single peeled
5168 iteration in that case.
5170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5171 Alan Hayward <alan.hayward@arm.com>
5172 David Sherwood <david.sherwood@arm.com>
5174 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5175 single-element interleaving even if the size is not a power of 2.
5176 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5177 accesses for single-element interleaving if the group size is
5180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5181 Alan Hayward <alan.hayward@arm.com>
5182 David Sherwood <david.sherwood@arm.com>
5184 * doc/md.texi (fold_extract_last_@var{m}): Document.
5185 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5186 * optabs.def (fold_extract_last_optab): New optab.
5187 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5188 * internal-fn.c (fold_extract_direct): New macro.
5189 (expand_fold_extract_optab_fn): Likewise.
5190 (direct_fold_extract_optab_supported_p): Likewise.
5191 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5192 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5193 EXTRACT_LAST_REDUCTION.
5194 (get_initial_def_for_reduction): Do not create an initial vector
5195 for EXTRACT_LAST_REDUCTION reductions.
5196 (vectorizable_reduction): Leave the scalar phi in place for
5197 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5198 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5199 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5200 transform phase to vectorizable_condition.
5201 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5203 (vect_finish_stmt_generation): ...here.
5204 (vect_finish_replace_stmt): New function.
5205 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5206 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5208 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5211 Alan Hayward <alan.hayward@arm.com>
5212 David Sherwood <david.sherwood@arm.com>
5214 * doc/md.texi (extract_last_@var{m}): Document.
5215 * optabs.def (extract_last_optab): New optab.
5216 * internal-fn.def (EXTRACT_LAST): New internal function.
5217 * internal-fn.c (cond_unary_direct): New macro.
5218 (expand_cond_unary_optab_fn): Likewise.
5219 (direct_cond_unary_optab_supported_p): Likewise.
5220 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5221 loops using EXTRACT_LAST.
5222 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5223 (extract_last_<mode>): ...this optab.
5224 (vec_extract<mode><Vel>): Update accordingly.
5226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5227 Alan Hayward <alan.hayward@arm.com>
5228 David Sherwood <david.sherwood@arm.com>
5230 * target.def (empty_mask_is_expensive): New hook.
5231 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5232 * doc/tm.texi: Regenerate.
5233 * targhooks.h (default_empty_mask_is_expensive): Declare.
5234 * targhooks.c (default_empty_mask_is_expensive): New function.
5235 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5236 if the target says that empty masks are expensive.
5237 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5239 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5242 Alan Hayward <alan.hayward@arm.com>
5243 David Sherwood <david.sherwood@arm.com>
5245 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5246 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5247 (vect_use_loop_mask_for_alignment_p): New function.
5248 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5249 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5250 niters_skip argument. Make sure that the first niters_skip elements
5251 of the first iteration are inactive.
5252 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5253 Update call to vect_set_loop_masks_directly.
5254 (get_misalign_in_elems): New function, split out from...
5255 (vect_gen_prolog_loop_niters): ...here.
5256 (vect_update_init_of_dr): Take a code argument that specifies whether
5257 the adjustment should be added or subtracted.
5258 (vect_update_init_of_drs): Likewise.
5259 (vect_prepare_for_masked_peels): New function.
5260 (vect_do_peeling): Skip prologue peeling if we're using a mask
5261 instead. Update call to vect_update_inits_of_drs.
5262 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5264 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5265 alignment. Do not include the number of peeled iterations in
5266 the minimum threshold in that case.
5267 (vectorizable_induction): Adjust the start value down by
5268 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5269 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5270 Take the number of skipped iterations into account when calculating
5272 * tree-vect-stmts.c (vect_gen_while_not): New function.
5274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5275 Alan Hayward <alan.hayward@arm.com>
5276 David Sherwood <david.sherwood@arm.com>
5278 * doc/sourcebuild.texi (vect_fully_masked): Document.
5279 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5281 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5283 (vect_analyze_loop_2): ...here. Don't check the vectorization
5284 factor against the number of loop iterations if the loop is
5287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5288 Alan Hayward <alan.hayward@arm.com>
5289 David Sherwood <david.sherwood@arm.com>
5291 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5292 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5293 (dump_groups): Update accordingly.
5294 (iv_use::mem_type): New member variable.
5295 (address_p): New function.
5296 (record_use): Add a mem_type argument and initialize the new
5298 (record_group_use): Add a mem_type argument. Use address_p.
5299 Remove obsolete null checks of base_object. Update call to record_use.
5300 (find_interesting_uses_op): Update call to record_group_use.
5301 (find_interesting_uses_cond): Likewise.
5302 (find_interesting_uses_address): Likewise.
5303 (get_mem_type_for_internal_fn): New function.
5304 (find_address_like_use): Likewise.
5305 (find_interesting_uses_stmt): Try find_address_like_use before
5306 calling find_interesting_uses_op.
5307 (addr_offset_valid_p): Use the iv mem_type field as the type
5308 of the addressed memory.
5309 (add_autoinc_candidates): Likewise.
5310 (get_address_cost): Likewise.
5311 (split_small_address_groups_p): Use address_p.
5312 (split_address_groups): Likewise.
5313 (add_iv_candidate_for_use): Likewise.
5314 (autoinc_possible_for_pair): Likewise.
5315 (rewrite_groups): Likewise.
5316 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5317 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5318 (get_alias_ptr_type_for_ptr_address): New function.
5319 (rewrite_use_address): Rewrite address uses in calls that were
5320 identified by find_address_like_use.
5322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5323 Alan Hayward <alan.hayward@arm.com>
5324 David Sherwood <david.sherwood@arm.com>
5326 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5328 * gimple-expr.h (is_gimple_addressable: Likewise.
5329 * gimple-expr.c (is_gimple_address): Likewise.
5330 * internal-fn.c (expand_call_mem_ref): New function.
5331 (expand_mask_load_optab_fn): Use it.
5332 (expand_mask_store_optab_fn): Likewise.
5334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5335 Alan Hayward <alan.hayward@arm.com>
5336 David Sherwood <david.sherwood@arm.com>
5338 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5339 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5340 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5341 (cond_umax@var{mode}): Document.
5342 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5343 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5344 (cond_umin_optab, cond_umax_optab): New optabs.
5345 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5346 (COND_IOR, COND_XOR): New internal functions.
5347 * internal-fn.h (get_conditional_internal_fn): Declare.
5348 * internal-fn.c (cond_binary_direct): New macro.
5349 (expand_cond_binary_optab_fn): Likewise.
5350 (direct_cond_binary_optab_supported_p): Likewise.
5351 (get_conditional_internal_fn): New function.
5352 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5353 Cope with reduction statements that are vectorized as calls rather
5355 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5356 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5357 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5358 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5359 (UNSPEC_COND_EOR): New unspecs.
5360 (optab): Add mappings for them.
5361 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5362 (sve_int_op, sve_fp_op): New int attributes.
5364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5365 Alan Hayward <alan.hayward@arm.com>
5366 David Sherwood <david.sherwood@arm.com>
5368 * optabs.def (while_ult_optab): New optab.
5369 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5370 * internal-fn.def (WHILE_ULT): New internal function.
5371 * internal-fn.h (direct_internal_fn_supported_p): New override
5372 that takes two types as argument.
5373 * internal-fn.c (while_direct): New macro.
5374 (expand_while_optab_fn): New function.
5375 (convert_optab_supported_p): Likewise.
5376 (direct_while_optab_supported_p): New macro.
5377 * wide-int.h (wi::udiv_ceil): New function.
5378 * tree-vectorizer.h (rgroup_masks): New structure.
5379 (vec_loop_masks): New typedef.
5380 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5382 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5383 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5384 (vect_max_vf): New function.
5385 (slpeel_make_loop_iterate_ntimes): Delete.
5386 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5387 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5388 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5389 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5390 internal-fn.h, stor-layout.h and optabs-query.h.
5391 (vect_set_loop_mask): New function.
5392 (add_preheader_seq): Likewise.
5393 (add_header_seq): Likewise.
5394 (interleave_supported_p): Likewise.
5395 (vect_maybe_permute_loop_masks): Likewise.
5396 (vect_set_loop_masks_directly): Likewise.
5397 (vect_set_loop_condition_masked): Likewise.
5398 (vect_set_loop_condition_unmasked): New function, split out from
5399 slpeel_make_loop_iterate_ntimes.
5400 (slpeel_make_loop_iterate_ntimes): Rename to..
5401 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5402 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5403 (vect_do_peeling): Update call accordingly.
5404 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5406 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5407 mask_compare_type, can_fully_mask_p and fully_masked_p.
5408 (release_vec_loop_masks): New function.
5409 (_loop_vec_info): Use it to free the loop masks.
5410 (can_produce_all_loop_masks_p): New function.
5411 (vect_get_max_nscalars_per_iter): Likewise.
5412 (vect_verify_full_masking): Likewise.
5413 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5414 retries, and free the mask rgroups before retrying. Check loop-wide
5415 reasons for disallowing fully-masked loops. Make the final decision
5416 about whether use a fully-masked loop or not.
5417 (vect_estimate_min_profitable_iters): Do not assume that peeling
5418 for the number of iterations will be needed for fully-masked loops.
5419 (vectorizable_reduction): Disable fully-masked loops.
5420 (vectorizable_live_operation): Likewise.
5421 (vect_halve_mask_nunits): New function.
5422 (vect_double_mask_nunits): Likewise.
5423 (vect_record_loop_mask): Likewise.
5424 (vect_get_loop_mask): Likewise.
5425 (vect_transform_loop): Handle the case in which the final loop
5426 iteration might handle a partial vector. Call vect_set_loop_condition
5427 instead of slpeel_make_loop_iterate_ntimes.
5428 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5429 (check_load_store_masking): New function.
5430 (prepare_load_store_mask): Likewise.
5431 (vectorizable_store): Handle fully-masked loops.
5432 (vectorizable_load): Likewise.
5433 (supportable_widening_operation): Use vect_halve_mask_nunits for
5435 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5436 (vect_gen_while): New function.
5437 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5438 (aarch64_uqdec<mode>): New insn.
5440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5441 Alan Hayward <alan.hayward@arm.com>
5442 David Sherwood <david.sherwood@arm.com>
5444 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5445 (reduc_xor_scal_optab): New optabs.
5446 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5447 (reduc_xor_scal_@var{m}): Document.
5448 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5449 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5451 * fold-const-call.c (fold_const_call): Handle them.
5452 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5453 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5454 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5455 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5456 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5457 (UNSPEC_XORV): New unspecs.
5458 (optab): Add entries for them.
5459 (BITWISEV): New int iterator.
5460 (bit_reduc_op): New int attributes.
5462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5463 Alan Hayward <alan.hayward@arm.com>
5464 David Sherwood <david.sherwood@arm.com>
5466 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5467 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5468 * optabs.def (vec_shl_insert_optab): New optab.
5469 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5470 (duplicate_and_interleave): Likewise.
5471 * tree-vect-loop.c: Include internal-fn.h.
5472 (neutral_op_for_slp_reduction): New function, split out from
5473 get_initial_defs_for_reduction.
5474 (get_initial_def_for_reduction): Handle option 2 for variable-length
5475 vectors by loading the neutral value into a vector and then shifting
5476 the initial value into element 0.
5477 (get_initial_defs_for_reduction): Replace the code argument with
5478 the neutral value calculated by neutral_op_for_slp_reduction.
5479 Use gimple_build_vector for constant-length vectors.
5480 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5481 but the first group_size elements have a neutral value.
5482 Use duplicate_and_interleave otherwise.
5483 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5484 Update call to get_initial_defs_for_reduction. Handle SLP
5485 reductions for variable-length vectors by creating one vector
5486 result for each scalar result, with the elements associated
5487 with other scalar results stubbed out with the neutral value.
5488 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5489 Require IFN_VEC_SHL_INSERT for double reductions on
5490 variable-length vectors, or SLP reductions that have
5491 a neutral value. Require can_duplicate_and_interleave_p
5492 support for variable-length unchained SLP reductions if there
5493 is no neutral value, such as for MIN/MAX reductions. Also require
5494 the number of vector elements to be a multiple of the number of
5495 SLP statements when doing variable-length unchained SLP reductions.
5496 Update call to vect_create_epilog_for_reduction.
5497 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5498 and remove initial values.
5499 (duplicate_and_interleave): Make public.
5500 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5501 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5504 Alan Hayward <alan.hayward@arm.com>
5505 David Sherwood <david.sherwood@arm.com>
5507 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5508 (can_duplicate_and_interleave_p): New function.
5509 (vect_get_and_check_slp_defs): Take the vector of statements
5510 rather than just the current one. Remove excess parentheses.
5511 Restriction rejectinon of vect_constant_def and vect_external_def
5512 for variable-length vectors to boolean types, or types for which
5513 can_duplicate_and_interleave_p is false.
5514 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5515 (duplicate_and_interleave): New function.
5516 (vect_get_constant_vectors): Use gimple_build_vector for
5517 constant-length vectors and suitable variable-length constant
5518 vectors. Use duplicate_and_interleave for other variable-length
5519 vectors. Don't defer the update when inserting new statements.
5521 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5522 Alan Hayward <alan.hayward@arm.com>
5523 David Sherwood <david.sherwood@arm.com>
5525 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5526 min_profitable_iters doesn't go negative.
5528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5529 Alan Hayward <alan.hayward@arm.com>
5530 David Sherwood <david.sherwood@arm.com>
5532 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5533 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5534 * optabs.def (vec_mask_load_lanes_optab): New optab.
5535 (vec_mask_store_lanes_optab): Likewise.
5536 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5537 (MASK_STORE_LANES): Likewise.
5538 * internal-fn.c (mask_load_lanes_direct): New macro.
5539 (mask_store_lanes_direct): Likewise.
5540 (expand_mask_load_optab_fn): Handle masked operations.
5541 (expand_mask_load_lanes_optab_fn): New macro.
5542 (expand_mask_store_optab_fn): Handle masked operations.
5543 (expand_mask_store_lanes_optab_fn): New macro.
5544 (direct_mask_load_lanes_optab_supported_p): Likewise.
5545 (direct_mask_store_lanes_optab_supported_p): Likewise.
5546 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5548 (vect_load_lanes_supported): Likewise.
5549 * tree-vect-data-refs.c (strip_conversion): New function.
5550 (can_group_stmts_p): Likewise.
5551 (vect_analyze_data_ref_accesses): Use it instead of checking
5552 for a pair of assignments.
5553 (vect_store_lanes_supported): Take a masked_p parameter.
5554 (vect_load_lanes_supported): Likewise.
5555 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5556 vect_store_lanes_supported and vect_load_lanes_supported.
5557 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5558 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5559 parameter. Don't allow gaps for masked accesses.
5560 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5561 and vect_load_lanes_supported.
5562 (get_load_store_type): Take a masked_p parameter and update
5563 call to get_group_load_store_type.
5564 (vectorizable_store): Update call to get_load_store_type.
5565 Handle IFN_MASK_STORE_LANES.
5566 (vectorizable_load): Update call to get_load_store_type.
5567 Handle IFN_MASK_LOAD_LANES.
5569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5570 Alan Hayward <alan.hayward@arm.com>
5571 David Sherwood <david.sherwood@arm.com>
5573 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5575 * config/aarch64/aarch64-protos.h
5576 (aarch64_sve_struct_memory_operand_p): Declare.
5577 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5578 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5579 (VPRED, vpred): Handle SVE structure modes.
5580 * config/aarch64/constraints.md (Utx): New constraint.
5581 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5582 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5583 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5584 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5585 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5586 structure modes. Split into pieces after RA.
5587 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5588 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5590 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5591 SVE structure modes.
5592 (aarch64_classify_address): Likewise.
5593 (sizetochar): Move earlier in file.
5594 (aarch64_print_operand): Handle SVE register lists.
5595 (aarch64_array_mode): New function.
5596 (aarch64_sve_struct_memory_operand_p): Likewise.
5597 (TARGET_ARRAY_MODE): Redefine.
5599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5600 Alan Hayward <alan.hayward@arm.com>
5601 David Sherwood <david.sherwood@arm.com>
5603 * target.def (array_mode): New target hook.
5604 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5605 * doc/tm.texi: Regenerate.
5606 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5607 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5608 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5610 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5614 Alan Hayward <alan.hayward@arm.com>
5615 David Sherwood <david.sherwood@arm.com>
5617 * fold-const.c (fold_binary_loc): Check the argument types
5618 rather than the result type when testing for a vector operation.
5620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5622 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5623 * doc/tm.texi: Regenerate.
5625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5626 Alan Hayward <alan.hayward@arm.com>
5627 David Sherwood <david.sherwood@arm.com>
5629 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5630 (sve): Document new AArch64 extension.
5631 * doc/md.texi (w): Extend the description of the AArch64
5632 constraint to include SVE vectors.
5633 (Upl, Upa): Document new AArch64 predicate constraints.
5634 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5636 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5637 (msve-vector-bits=): New option.
5638 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5639 SVE when these are disabled.
5640 (sve): New extension.
5641 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5642 modes. Adjust their number of units based on aarch64_sve_vg.
5643 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5644 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5645 aarch64_addr_query_type.
5646 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5647 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5648 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5649 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5650 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5651 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5652 (aarch64_simd_imm_zero_p): Delete.
5653 (aarch64_check_zero_based_sve_index_immediate): Declare.
5654 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5655 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5656 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5657 (aarch64_sve_float_mul_immediate_p): Likewise.
5658 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5660 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5661 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5662 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5663 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5664 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5665 (aarch64_regmode_natural_size): Likewise.
5666 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5667 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5669 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5670 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5671 for VG and the SVE predicate registers.
5672 (V_ALIASES): Add a "z"-prefixed alias.
5673 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5674 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5675 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5676 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5677 (REG_CLASS_NAMES): Add entries for them.
5678 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5679 and the predicate registers.
5680 (aarch64_sve_vg): Declare.
5681 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5682 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5683 (REGMODE_NATURAL_SIZE): Define.
5684 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5686 * config/aarch64/aarch64.c: Include cfgrtl.h.
5687 (simd_immediate_info): Add a constructor for series vectors,
5688 and an associated step field.
5689 (aarch64_sve_vg): New variable.
5690 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5691 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5692 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5693 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5694 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5695 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5696 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5697 (aarch64_get_mask_mode): New functions.
5698 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5699 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5700 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5701 predicate modes and predicate registers. Explicitly restrict
5702 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5703 to store a vector mode if it is recognized by
5704 aarch64_classify_vector_mode.
5705 (aarch64_regmode_natural_size): New function.
5706 (aarch64_hard_regno_caller_save_mode): Return the original mode
5708 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5709 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5710 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5711 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5713 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5714 does not overlap dest if the function is frame-related. Handle
5716 (aarch64_split_add_offset): New function.
5717 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5718 them aarch64_add_offset.
5719 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5720 and update call to aarch64_sub_sp.
5721 (aarch64_add_cfa_expression): New function.
5722 (aarch64_expand_prologue): Pass extra temporary registers to the
5723 functions above. Handle the case in which we need to emit new
5724 DW_CFA_expressions for registers that were originally saved
5725 relative to the stack pointer, but now have to be expressed
5726 relative to the frame pointer.
5727 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5729 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5730 IP0 and IP1 values for SVE frames.
5731 (aarch64_expand_vec_series): New function.
5732 (aarch64_expand_sve_widened_duplicate): Likewise.
5733 (aarch64_expand_sve_const_vector): Likewise.
5734 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5735 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5736 into the register, rather than emitting a SET directly.
5737 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5738 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5739 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5740 (offset_9bit_signed_scaled_p): New functions.
5741 (aarch64_replicate_bitmask_imm): New function.
5742 (aarch64_bitmask_imm): Use it.
5743 (aarch64_cannot_force_const_mem): Reject expressions involving
5744 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5745 (aarch64_classify_index): Handle SVE indices, by requiring
5746 a plain register index with a scale that matches the element size.
5747 (aarch64_classify_address): Handle SVE addresses. Assert that
5748 the mode of the address is VOIDmode or an integer mode.
5749 Update call to aarch64_classify_symbol.
5750 (aarch64_classify_symbolic_expression): Update call to
5751 aarch64_classify_symbol.
5752 (aarch64_const_vec_all_in_range_p): New function.
5753 (aarch64_print_vector_float_operand): Likewise.
5754 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5755 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5756 and the FP immediates 1.0 and 0.5.
5757 (aarch64_print_address_internal): Handle SVE addresses.
5758 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5759 (aarch64_regno_regclass): Handle predicate registers.
5760 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5762 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5763 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5764 (aarch64_convert_sve_vector_bits): New function.
5765 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5766 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5768 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5769 Handle SVE vector and predicate modes. Accept VL-based constants
5770 that need only one temporary register, and VL offsets that require
5771 no temporary registers.
5772 (aarch64_conditional_register_usage): Mark the predicate registers
5773 as fixed if SVE isn't available.
5774 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5775 Return true for SVE vector and predicate modes.
5776 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5777 rather than an unsigned int. Handle SVE modes.
5778 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5780 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5782 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5783 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5784 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5785 (aarch64_sve_float_mul_immediate_p): New functions.
5786 (aarch64_sve_valid_immediate): New function.
5787 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5788 Explicitly reject structure modes. Check for INDEX constants.
5789 Handle PTRUE and PFALSE constants.
5790 (aarch64_check_zero_based_sve_index_immediate): New function.
5791 (aarch64_simd_imm_zero_p): Delete.
5792 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5793 vector modes. Accept constants in the range of CNT[BHWD].
5794 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5795 ask for an Advanced SIMD mode.
5796 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5797 (aarch64_simd_vector_alignment): Handle SVE predicates.
5798 (aarch64_vectorize_preferred_vector_alignment): New function.
5799 (aarch64_simd_vector_alignment_reachable): Use it instead of
5801 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5802 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5804 (MAX_VECT_LEN): Delete.
5805 (expand_vec_perm_d): Add a vec_flags field.
5806 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5807 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5808 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5810 (aarch64_evpc_rev): Rename to...
5811 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5812 (aarch64_evpc_rev_global): New function.
5813 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5814 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5816 (aarch64_evpc_sve_tbl): New function.
5817 (aarch64_expand_vec_perm_const_1): Update after rename of
5818 aarch64_evpc_rev. Handle SVE permutes too, trying
5819 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5820 than aarch64_evpc_tbl.
5821 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5822 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5823 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5824 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5825 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5826 (aarch64_expand_sve_vcond): New functions.
5827 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5828 of aarch64_vector_mode_p.
5829 (aarch64_dwarf_poly_indeterminate_value): New function.
5830 (aarch64_compute_pressure_classes): Likewise.
5831 (aarch64_can_change_mode_class): Likewise.
5832 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5833 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5834 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5835 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5836 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5837 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5838 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5839 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5841 (Dn, Dl, Dr): Accept const as well as const_vector.
5842 (Dz): Likewise. Compare against CONST0_RTX.
5843 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5844 of "vector" where appropriate.
5845 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5846 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5847 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5848 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5849 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5850 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5851 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5852 (v_int_equiv): Extend to SVE modes.
5853 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5855 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5856 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5857 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5858 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5859 (SVE_COND_FP_CMP): New int iterators.
5860 (perm_hilo): Handle the new unpack unspecs.
5861 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5863 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5864 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5865 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5866 (aarch64_equality_operator, aarch64_constant_vector_operand)
5867 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5868 (aarch64_sve_nonimmediate_operand): Likewise.
5869 (aarch64_sve_general_operand): Likewise.
5870 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5871 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5872 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5873 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5874 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5875 (aarch64_sve_float_arith_immediate): Likewise.
5876 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5877 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5878 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5879 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5880 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5881 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5882 (aarch64_sve_float_arith_operand): Likewise.
5883 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5884 (aarch64_sve_float_mul_operand): Likewise.
5885 (aarch64_sve_vec_perm_operand): Likewise.
5886 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5887 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5888 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5889 as well as const_vector.
5890 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5891 in file. Use CONST0_RTX and CONSTM1_RTX.
5892 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5893 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5894 Use aarch64_simd_imm_zero.
5895 * config/aarch64/aarch64-sve.md: New file.
5896 * config/aarch64/aarch64.md: Include it.
5897 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5898 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5899 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5900 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5901 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5902 (sve): New attribute.
5903 (enabled): Disable instructions with the sve attribute unless
5905 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5906 aarch64_expand_mov_immediate.
5907 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5908 CNT[BHSD] immediates.
5909 (movti): Split CONST_POLY_INT moves into two halves.
5910 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5911 Split additions that need a temporary here if the destination
5912 is the stack pointer.
5913 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5914 (*add<mode>3_poly_1): New instruction.
5915 (set_clobber_cc): New expander.
5917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5919 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5920 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5921 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5922 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5923 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5924 Change innermode from fixed_mode_size to machine_mode.
5925 (simplify_subreg): Update call accordingly. Handle a constant-sized
5926 subreg of a variable-length CONST_VECTOR.
5928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5929 Alan Hayward <alan.hayward@arm.com>
5930 David Sherwood <david.sherwood@arm.com>
5932 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5933 (add_offset_to_base): New function, split out from...
5934 (create_mem_ref): ...here. When handling a scale other than 1,
5935 check first whether the address is valid without the offset.
5936 Add it into the base if so, leaving the index and scale as-is.
5938 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5941 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5942 fold_for_warn before checking if arg2 is INTEGER_CST.
5944 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5946 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5947 (store_multiple_operation): Delete.
5948 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5949 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5950 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5951 guarded by TARGET_STRING.
5952 (rs6000_output_load_multiple): Delete.
5953 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5954 OPTION_MASK_STRING / TARGET_STRING handling.
5955 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5956 (const rs6000_opt_masks) <"string">: Change mask to 0.
5957 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5958 (MASK_STRING): Delete.
5959 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5961 (load_multiple): Delete.
5968 (store_multiple): Delete.
5975 (movmemsi_8reg): Delete.
5976 (corresponding unnamed define_insn): Delete.
5977 (movmemsi_6reg): Delete.
5978 (corresponding unnamed define_insn): Delete.
5979 (movmemsi_4reg): Delete.
5980 (corresponding unnamed define_insn): Delete.
5981 (movmemsi_2reg): Delete.
5982 (corresponding unnamed define_insn): Delete.
5983 (movmemsi_1reg): Delete.
5984 (corresponding unnamed define_insn): Delete.
5985 * config/rs6000/rs6000.opt (mno-string): New.
5986 (mstring): Replace by deprecation warning stub.
5987 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5989 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5991 * regrename.c (regrename_do_replace): If replacing the same
5992 reg multiple times, try to reuse last created gen_raw_REG.
5995 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5996 main to workaround a bug in GDB.
5998 2018-01-12 Tom de Vries <tom@codesourcery.com>
6001 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6003 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6005 PR rtl-optimization/80481
6006 * ira-color.c (get_cap_member): New function.
6007 (allocnos_conflict_by_live_ranges_p): Use it.
6008 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6009 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6011 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6014 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6015 (*saddl_se_1): Ditto.
6017 (*ssubl_se_1): Ditto.
6019 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6021 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6022 rather than wi::to_widest for DR_INITs.
6023 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6024 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6025 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6027 (vect_analyze_group_access_1): Note that here.
6029 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6031 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6032 polynomial type sizes.
6034 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6036 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6037 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6038 (gimple_add_tmp_var): Likewise.
6040 2018-01-12 Martin Liska <mliska@suse.cz>
6042 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6043 (gimple_alloc_sizes): Likewise.
6044 (dump_gimple_statistics): Use PRIu64 in printf format.
6045 * gimple.h: Change uint64_t to int.
6047 2018-01-12 Martin Liska <mliska@suse.cz>
6049 * tree-core.h: Use uint64_t instead of int.
6050 * tree.c (tree_node_counts): Likewise.
6051 (tree_node_sizes): Likewise.
6052 (dump_tree_statistics): Use PRIu64 in printf format.
6054 2018-01-12 Martin Liska <mliska@suse.cz>
6056 * Makefile.in: As qsort_chk is implemented in vec.c, add
6057 vec.o to linkage of gencfn-macros.
6058 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6059 passing the info to record_node_allocation_statistics.
6060 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6062 * ggc-common.c (struct ggc_usage): Add operator== and use
6063 it in operator< and compare function.
6064 * mem-stats.h (struct mem_usage): Likewise.
6065 * vec.c (struct vec_usage): Remove operator< and compare
6066 function. Can be simply inherited.
6068 2018-01-12 Martin Jambor <mjambor@suse.cz>
6071 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6072 * tree-ssa-math-opts.c: Include domwalk.h.
6073 (convert_mult_to_fma_1): New function.
6074 (fma_transformation_info): New type.
6075 (fma_deferring_state): Likewise.
6076 (cancel_fma_deferring): New function.
6077 (result_of_phi): Likewise.
6078 (last_fma_candidate_feeds_initial_phi): Likewise.
6079 (convert_mult_to_fma): Added deferring logic, split actual
6080 transformation to convert_mult_to_fma_1.
6081 (math_opts_dom_walker): New type.
6082 (math_opts_dom_walker::after_dom_children): New method, body moved
6083 here from pass_optimize_widening_mul::execute, added deferring logic
6085 (pass_optimize_widening_mul::execute): Moved most of code to
6086 math_opts_dom_walker::after_dom_children.
6087 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6088 * config/i386/i386.c (ix86_option_override_internal): Added
6089 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6091 2018-01-12 Richard Biener <rguenther@suse.de>
6094 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6095 inline instance vars.
6097 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6100 * config/rx/rx.c (rx_is_restricted_memory_address):
6103 2018-01-12 Richard Biener <rguenther@suse.de>
6105 PR tree-optimization/80846
6106 * target.def (split_reduction): New target hook.
6107 * targhooks.c (default_split_reduction): New function.
6108 * targhooks.h (default_split_reduction): Declare.
6109 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6110 target requests first reduce vectors by combining low and high
6112 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6113 (get_vectype_for_scalar_type_and_size): Export.
6114 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6115 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6116 * doc/tm.texi: Regenerate.
6117 * config/i386/i386.c (ix86_split_reduction): Implement
6118 TARGET_VECTORIZE_SPLIT_REDUCTION.
6120 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6123 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6124 in PIC mode except for TARGET_VXWORKS_RTP.
6125 * config/sparc/sparc.c: Include cfgrtl.h.
6126 (TARGET_INIT_PIC_REG): Define.
6127 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6128 (sparc_pic_register_p): New predicate.
6129 (sparc_legitimate_address_p): Use it.
6130 (sparc_legitimize_pic_address): Likewise.
6131 (sparc_delegitimize_address): Likewise.
6132 (sparc_mode_dependent_address_p): Likewise.
6133 (gen_load_pcrel_sym): Remove 4th parameter.
6134 (load_got_register): Adjust call to above. Remove obsolete stuff.
6135 (sparc_expand_prologue): Do not call load_got_register here.
6136 (sparc_flat_expand_prologue): Likewise.
6137 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6138 (sparc_use_pseudo_pic_reg): New function.
6139 (sparc_init_pic_reg): Likewise.
6140 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6141 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6143 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6145 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6146 Add item for branch_cost.
6148 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6150 PR rtl-optimization/83565
6151 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6152 not extend the result to a larger mode for rotate operations.
6153 (num_sign_bit_copies1): Likewise.
6155 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6158 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6160 Use values-Xc.o for -pedantic.
6161 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6163 2018-01-12 Martin Liska <mliska@suse.cz>
6166 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6168 (possible_polymorphic_call_targets): Use it.
6169 (ipa_devirt): Likewise.
6171 2018-01-12 Martin Liska <mliska@suse.cz>
6173 * profile-count.h (enum profile_quality): Use 0 as invalid
6174 enum value of profile_quality.
6176 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6178 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6179 -mext-string options.
6181 2018-01-12 Richard Biener <rguenther@suse.de>
6183 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6184 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6185 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6187 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6189 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6191 * configure.ac (--with-long-double-format): Add support for the
6192 configuration option to change the default long double format on
6194 * config.gcc (powerpc*-linux*-*): Likewise.
6195 * configure: Regenerate.
6196 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6197 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6198 used without modification.
6200 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6202 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6203 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6204 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6205 MISC_BUILTIN_SPEC_BARRIER.
6206 (rs6000_init_builtins): Likewise.
6207 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6209 (speculation_barrier): New define_insn.
6210 * doc/extend.texi: Document __builtin_speculation_barrier.
6212 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6215 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6216 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6217 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6219 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6220 integral modes instead of "ss" and "sd".
6221 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6222 vectors with 32-bit and 64-bit elements.
6223 (vecdupssescalarmodesuffix): New mode attribute.
6224 (vec_dup<mode>): Use it.
6226 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6229 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6230 frame if argument is passed on stack.
6232 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6235 * ree.c (combine_reaching_defs): Optimize also
6236 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6237 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6239 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6242 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6244 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6247 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6248 after they are computed.
6250 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6252 PR tree-optimization/83695
6253 * gimple-loop-linterchange.cc
6254 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6255 reset cached scev information after interchange.
6256 (pass_linterchange::execute): Remove call to scev_reset_htab.
6258 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6260 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6261 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6262 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6263 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6264 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6265 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6266 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6267 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6268 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6269 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6270 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6271 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6272 (V_lane_reg): Likewise.
6273 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6275 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6276 (vfmal_lane_low<mode>_intrinsic,
6277 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6278 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6279 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6280 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6281 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6282 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6284 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6286 * config/arm/arm-cpus.in (fp16fml): New feature.
6287 (ALL_SIMD): Add fp16fml.
6288 (armv8.2-a): Add fp16fml as an option.
6289 (armv8.3-a): Likewise.
6290 (armv8.4-a): Add fp16fml as part of fp16.
6291 * config/arm/arm.h (TARGET_FP16FML): Define.
6292 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6294 * config/arm/arm-modes.def (V2HF): Define.
6295 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6296 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6297 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6298 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6299 vfmsl_low, vfmsl_high): New set of builtins.
6300 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6301 (vfml_op): New code attribute.
6302 (VFMLHALVES): New int iterator.
6303 (VFML, VFMLSEL): New mode attributes.
6304 (V_reg): Define mapping for V2HF.
6305 (V_hi, V_lo): New mode attributes.
6306 (VF_constraint): Likewise.
6307 (vfml_half, vfml_half_selector): New int attributes.
6308 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6310 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6311 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6313 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6314 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6315 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6316 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6318 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6319 Document new effective target and option set.
6321 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6323 * config/arm/arm-cpus.in (armv8_4): New feature.
6324 (ARMv8_4a): New fgroup.
6325 (armv8.4-a): New arch.
6326 * config/arm/arm-tables.opt: Regenerate.
6327 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6328 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6329 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6330 Add matching rules for -march=armv8.4-a and extensions.
6331 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6333 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6336 * config/rx/rx.md (BW): New mode attribute.
6337 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6339 2018-01-11 Richard Biener <rguenther@suse.de>
6341 PR tree-optimization/83435
6342 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6343 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6344 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6346 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6347 Alan Hayward <alan.hayward@arm.com>
6348 David Sherwood <david.sherwood@arm.com>
6350 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6352 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6353 (aarch64_print_address_internal): Use it to check for a zero offset.
6355 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6356 Alan Hayward <alan.hayward@arm.com>
6357 David Sherwood <david.sherwood@arm.com>
6359 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6360 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6361 Return a poly_int64 rather than a HOST_WIDE_INT.
6362 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6363 rather than a HOST_WIDE_INT.
6364 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6365 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6366 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6367 final_offset from HOST_WIDE_INT to poly_int64.
6368 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6369 to_constant when getting the number of units in an Advanced SIMD
6371 (aarch64_builtin_vectorized_function): Check for a constant number
6373 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6375 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6376 attribute instead of GET_MODE_NUNITS.
6377 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6378 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6379 GET_MODE_SIZE for fixed-size registers.
6380 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6381 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6382 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6383 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6384 (aarch64_print_operand, aarch64_print_address_internal)
6385 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6386 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6387 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6388 Handle polynomial GET_MODE_SIZE.
6389 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6390 wider than SImode without modification.
6391 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6392 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6393 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6394 passing and returning SVE modes.
6395 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6396 rather than GEN_INT.
6397 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6398 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6399 (aarch64_allocate_and_probe_stack_space): Likewise.
6400 (aarch64_layout_frame): Cope with polynomial offsets.
6401 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6402 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6404 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6405 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6406 poly_int64 rather than a HOST_WIDE_INT.
6407 (aarch64_get_separate_components, aarch64_process_components)
6408 (aarch64_expand_prologue, aarch64_expand_epilogue)
6409 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6410 (aarch64_anchor_offset): New function, split out from...
6411 (aarch64_legitimize_address): ...here.
6412 (aarch64_builtin_vectorization_cost): Handle polynomial
6413 TYPE_VECTOR_SUBPARTS.
6414 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6416 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6417 number of elements from the PARALLEL rather than the mode.
6418 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6419 rather than GET_MODE_BITSIZE.
6420 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6421 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6422 (aarch64_expand_vec_perm_const_1): Handle polynomial
6423 d->perm.length () and d->perm elements.
6424 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6425 Apply to_constant to d->perm elements.
6426 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6427 polynomial CONST_VECTOR_NUNITS.
6428 (aarch64_move_pointer): Take amount as a poly_int64 rather
6430 (aarch64_progress_pointer): Avoid temporary variable.
6431 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6432 the mode attribute instead of GET_MODE.
6434 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6435 Alan Hayward <alan.hayward@arm.com>
6436 David Sherwood <david.sherwood@arm.com>
6438 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6439 x exists before using it.
6440 (aarch64_add_constant_internal): Rename to...
6441 (aarch64_add_offset_1): ...this. Replace regnum with separate
6442 src and dest rtxes. Handle the case in which they're different,
6443 including when the offset is zero. Replace scratchreg with an rtx.
6444 Use 2 additions if there is no spare register into which we can
6445 move a 16-bit constant.
6446 (aarch64_add_constant): Delete.
6447 (aarch64_add_offset): Replace reg with separate src and dest
6448 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6449 Use aarch64_add_offset_1.
6450 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6451 an rtx rather than an int. Take the delta as a poly_int64
6452 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6453 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6454 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6455 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6456 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6458 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6459 aarch64_add_constant.
6461 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6463 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6464 Use scalar_float_mode.
6466 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6468 * config/aarch64/aarch64-simd.md
6469 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6470 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6471 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6472 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6473 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6474 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6475 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6476 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6477 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6478 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6480 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6483 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6484 targ_options->x_arm_arch_string is non NULL.
6486 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6488 * config/aarch64/aarch64.h
6489 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6491 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6494 * expmed.c (emit_store_flag_force): Swap if const op0
6495 and change VOIDmode to mode of op0.
6497 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6499 PR rtl-optimization/83761
6500 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6501 than bytes to mode_for_size.
6503 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6506 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6507 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6510 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6513 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6514 when in layout mode.
6515 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6516 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6519 2018-01-10 Michael Collison <michael.collison@arm.com>
6521 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6522 * config/aarch64/aarch64-option-extension.def: Add
6523 AARCH64_OPT_EXTENSION of 'fp16fml'.
6524 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6525 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6526 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6527 * config/aarch64/constraints.md (Ui7): New constraint.
6528 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6529 (VFMLA_SEL_W): Ditto.
6532 (VFMLA16_LOW): New int iterator.
6533 (VFMLA16_HIGH): Ditto.
6534 (UNSPEC_FMLAL): New unspec.
6535 (UNSPEC_FMLSL): Ditto.
6536 (UNSPEC_FMLAL2): Ditto.
6537 (UNSPEC_FMLSL2): Ditto.
6538 (f16mac): New code attribute.
6539 * config/aarch64/aarch64-simd-builtins.def
6540 (aarch64_fmlal_lowv2sf): Ditto.
6541 (aarch64_fmlsl_lowv2sf): Ditto.
6542 (aarch64_fmlalq_lowv4sf): Ditto.
6543 (aarch64_fmlslq_lowv4sf): Ditto.
6544 (aarch64_fmlal_highv2sf): Ditto.
6545 (aarch64_fmlsl_highv2sf): Ditto.
6546 (aarch64_fmlalq_highv4sf): Ditto.
6547 (aarch64_fmlslq_highv4sf): Ditto.
6548 (aarch64_fmlal_lane_lowv2sf): Ditto.
6549 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6550 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6551 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6552 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6553 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6554 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6555 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6556 (aarch64_fmlal_lane_highv2sf): Ditto.
6557 (aarch64_fmlsl_lane_highv2sf): Ditto.
6558 (aarch64_fmlal_laneq_highv2sf): Ditto.
6559 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6560 (aarch64_fmlalq_lane_highv4sf): Ditto.
6561 (aarch64_fmlsl_lane_highv4sf): Ditto.
6562 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6563 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6564 * config/aarch64/aarch64-simd.md:
6565 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6566 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6567 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6568 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6569 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6570 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6571 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6572 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6573 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6574 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6575 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6576 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6577 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6578 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6579 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6580 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6581 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6582 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6583 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6584 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6585 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6586 (vfmlsl_low_u32): Ditto.
6587 (vfmlalq_low_u32): Ditto.
6588 (vfmlslq_low_u32): Ditto.
6589 (vfmlal_high_u32): Ditto.
6590 (vfmlsl_high_u32): Ditto.
6591 (vfmlalq_high_u32): Ditto.
6592 (vfmlslq_high_u32): Ditto.
6593 (vfmlal_lane_low_u32): Ditto.
6594 (vfmlsl_lane_low_u32): Ditto.
6595 (vfmlal_laneq_low_u32): Ditto.
6596 (vfmlsl_laneq_low_u32): Ditto.
6597 (vfmlalq_lane_low_u32): Ditto.
6598 (vfmlslq_lane_low_u32): Ditto.
6599 (vfmlalq_laneq_low_u32): Ditto.
6600 (vfmlslq_laneq_low_u32): Ditto.
6601 (vfmlal_lane_high_u32): Ditto.
6602 (vfmlsl_lane_high_u32): Ditto.
6603 (vfmlal_laneq_high_u32): Ditto.
6604 (vfmlsl_laneq_high_u32): Ditto.
6605 (vfmlalq_lane_high_u32): Ditto.
6606 (vfmlslq_lane_high_u32): Ditto.
6607 (vfmlalq_laneq_high_u32): Ditto.
6608 (vfmlslq_laneq_high_u32): Ditto.
6609 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6610 (AARCH64_FL_FOR_ARCH8_4): New.
6611 (AARCH64_ISA_F16FML): New ISA flag.
6612 (TARGET_F16FML): New feature flag for fp16fml.
6613 (doc/invoke.texi): Document new fp16fml option.
6615 2018-01-10 Michael Collison <michael.collison@arm.com>
6617 * config/aarch64/aarch64-builtins.c:
6618 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6619 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6620 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6621 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6622 (AARCH64_ISA_SHA3): New ISA flag.
6623 (TARGET_SHA3): New feature flag for sha3.
6624 * config/aarch64/iterators.md (sha512_op): New int attribute.
6625 (CRYPTO_SHA512): New int iterator.
6626 (UNSPEC_SHA512H): New unspec.
6627 (UNSPEC_SHA512H2): Ditto.
6628 (UNSPEC_SHA512SU0): Ditto.
6629 (UNSPEC_SHA512SU1): Ditto.
6630 * config/aarch64/aarch64-simd-builtins.def
6631 (aarch64_crypto_sha512hqv2di): New builtin.
6632 (aarch64_crypto_sha512h2qv2di): Ditto.
6633 (aarch64_crypto_sha512su0qv2di): Ditto.
6634 (aarch64_crypto_sha512su1qv2di): Ditto.
6635 (aarch64_eor3qv8hi): Ditto.
6636 (aarch64_rax1qv2di): Ditto.
6637 (aarch64_xarqv2di): Ditto.
6638 (aarch64_bcaxqv8hi): Ditto.
6639 * config/aarch64/aarch64-simd.md:
6640 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6641 (aarch64_crypto_sha512su0qv2di): Ditto.
6642 (aarch64_crypto_sha512su1qv2di): Ditto.
6643 (aarch64_eor3qv8hi): Ditto.
6644 (aarch64_rax1qv2di): Ditto.
6645 (aarch64_xarqv2di): Ditto.
6646 (aarch64_bcaxqv8hi): Ditto.
6647 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6648 (vsha512h2q_u64): Ditto.
6649 (vsha512su0q_u64): Ditto.
6650 (vsha512su1q_u64): Ditto.
6651 (veor3q_u16): Ditto.
6652 (vrax1q_u64): Ditto.
6654 (vbcaxq_u16): Ditto.
6655 * config/arm/types.md (crypto_sha512): New type attribute.
6656 (crypto_sha3): Ditto.
6657 (doc/invoke.texi): Document new sha3 option.
6659 2018-01-10 Michael Collison <michael.collison@arm.com>
6661 * config/aarch64/aarch64-builtins.c:
6662 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6663 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6664 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6665 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6666 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6667 (AARCH64_ISA_SM4): New ISA flag.
6668 (TARGET_SM4): New feature flag for sm4.
6669 * config/aarch64/aarch64-simd-builtins.def
6670 (aarch64_sm3ss1qv4si): Ditto.
6671 (aarch64_sm3tt1aq4si): Ditto.
6672 (aarch64_sm3tt1bq4si): Ditto.
6673 (aarch64_sm3tt2aq4si): Ditto.
6674 (aarch64_sm3tt2bq4si): Ditto.
6675 (aarch64_sm3partw1qv4si): Ditto.
6676 (aarch64_sm3partw2qv4si): Ditto.
6677 (aarch64_sm4eqv4si): Ditto.
6678 (aarch64_sm4ekeyqv4si): Ditto.
6679 * config/aarch64/aarch64-simd.md:
6680 (aarch64_sm3ss1qv4si): Ditto.
6681 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6682 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6683 (aarch64_sm4eqv4si): Ditto.
6684 (aarch64_sm4ekeyqv4si): Ditto.
6685 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6686 (sm3part_op): Ditto.
6687 (CRYPTO_SM3TT): Ditto.
6688 (CRYPTO_SM3PART): Ditto.
6689 (UNSPEC_SM3SS1): New unspec.
6690 (UNSPEC_SM3TT1A): Ditto.
6691 (UNSPEC_SM3TT1B): Ditto.
6692 (UNSPEC_SM3TT2A): Ditto.
6693 (UNSPEC_SM3TT2B): Ditto.
6694 (UNSPEC_SM3PARTW1): Ditto.
6695 (UNSPEC_SM3PARTW2): Ditto.
6696 (UNSPEC_SM4E): Ditto.
6697 (UNSPEC_SM4EKEY): Ditto.
6698 * config/aarch64/constraints.md (Ui2): New constraint.
6699 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6700 * config/arm/types.md (crypto_sm3): New type attribute.
6701 (crypto_sm4): Ditto.
6702 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6703 (vsm3tt1aq_u32): Ditto.
6704 (vsm3tt1bq_u32): Ditto.
6705 (vsm3tt2aq_u32): Ditto.
6706 (vsm3tt2bq_u32): Ditto.
6707 (vsm3partw1q_u32): Ditto.
6708 (vsm3partw2q_u32): Ditto.
6709 (vsm4eq_u32): Ditto.
6710 (vsm4ekeyq_u32): Ditto.
6711 (doc/invoke.texi): Document new sm4 option.
6713 2018-01-10 Michael Collison <michael.collison@arm.com>
6715 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6716 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6717 (AARCH64_FL_FOR_ARCH8_4): New.
6718 (AARCH64_FL_V8_4): New flag.
6719 (doc/invoke.texi): Document new armv8.4-a option.
6721 2018-01-10 Michael Collison <michael.collison@arm.com>
6723 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6724 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6725 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6726 * config/aarch64/aarch64-option-extension.def: Add
6727 AARCH64_OPT_EXTENSION of 'sha2'.
6728 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6729 (crypto): Disable sha2 and aes if crypto disabled.
6730 (crypto): Enable aes and sha2 if enabled.
6731 (simd): Disable sha2 and aes if simd disabled.
6732 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6734 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6735 (TARGET_SHA2): New feature flag for sha2.
6736 (TARGET_AES): New feature flag for aes.
6737 * config/aarch64/aarch64-simd.md:
6738 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6739 conditional on TARGET_AES.
6740 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6741 (aarch64_crypto_sha1hsi): Make pattern conditional
6743 (aarch64_crypto_sha1hv4si): Ditto.
6744 (aarch64_be_crypto_sha1hv4si): Ditto.
6745 (aarch64_crypto_sha1su1v4si): Ditto.
6746 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6747 (aarch64_crypto_sha1su0v4si): Ditto.
6748 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6749 (aarch64_crypto_sha256su0v4si): Ditto.
6750 (aarch64_crypto_sha256su1v4si): Ditto.
6751 (doc/invoke.texi): Document new aes and sha2 options.
6753 2018-01-10 Martin Sebor <msebor@redhat.com>
6755 PR tree-optimization/83781
6756 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6759 2018-01-11 Martin Sebor <msebor@gmail.com>
6760 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6762 PR tree-optimization/83501
6763 PR tree-optimization/81703
6765 * tree-ssa-strlen.c (get_string_cst): Rename...
6766 (get_string_len): ...to this. Handle global constants.
6767 (handle_char_store): Adjust.
6769 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6770 Jim Wilson <jimw@sifive.com>
6772 * config/riscv/riscv-protos.h (riscv_output_return): New.
6773 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6774 (riscv_attribute_table, riscv_output_return),
6775 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6776 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6777 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6779 (riscv_expand_prologue): Add early return for naked function.
6780 (riscv_expand_epilogue): Likewise.
6781 (riscv_function_ok_for_sibcall): Return false for naked function.
6782 (riscv_set_current_function): New.
6783 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6784 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6785 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6786 * doc/extend.texi (RISC-V Function Attributes): New.
6788 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6790 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6791 check for 128-bit long double before checking TCmode.
6792 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6793 128-bit long doubles before checking TFmode or TCmode.
6794 (FLOAT128_IBM_P): Likewise.
6796 2018-01-10 Martin Sebor <msebor@redhat.com>
6798 PR tree-optimization/83671
6799 * builtins.c (c_strlen): Unconditionally return zero for the empty
6801 Use -Warray-bounds for warnings.
6802 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6803 for non-constant array indices with COMPONENT_REF, arrays of
6804 arrays, and pointers to arrays.
6805 (gimple_fold_builtin_strlen): Determine and set length range for
6806 non-constant character arrays.
6808 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6811 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6814 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6816 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6818 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6821 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6822 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6823 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6824 indexed_or_indirect_operand predicate.
6825 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6826 (*vsx_le_perm_load_v8hi): Likewise.
6827 (*vsx_le_perm_load_v16qi): Likewise.
6828 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6829 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6830 (*vsx_le_perm_store_v8hi): Likewise.
6831 (*vsx_le_perm_store_v16qi): Likewise.
6832 (eight unnamed splitters): Likewise.
6834 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6836 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6837 * config/rs6000/emmintrin.h: Likewise.
6838 * config/rs6000/mmintrin.h: Likewise.
6839 * config/rs6000/xmmintrin.h: Likewise.
6841 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6844 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6846 * tree.c (tree_nop_conversion): Return true for location wrapper
6848 (maybe_wrap_with_location): New function.
6849 (selftest::check_strip_nops): New function.
6850 (selftest::test_location_wrappers): New function.
6851 (selftest::tree_c_tests): Call it.
6852 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6853 (maybe_wrap_with_location): New decl.
6854 (EXPR_LOCATION_WRAPPER_P): New macro.
6855 (location_wrapper_p): New inline function.
6856 (tree_strip_any_location_wrapper): New inline function.
6858 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6861 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6862 stack_realign_offset for the largest alignment of stack slot
6864 (ix86_find_max_used_stack_alignment): New function.
6865 (ix86_finalize_stack_frame_flags): Use it. Set
6866 max_used_stack_alignment if we don't realign stack.
6867 * config/i386/i386.h (machine_function): Add
6868 max_used_stack_alignment.
6870 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6872 * config/arm/arm.opt (-mbranch-cost): New option.
6873 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6876 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6879 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6880 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6882 2018-01-10 Richard Biener <rguenther@suse.de>
6885 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6886 early out so it also covers the case where we have a non-NULL
6889 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6891 PR tree-optimization/83753
6892 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6893 for non-strided grouped accesses if the number of elements is 1.
6895 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6898 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6899 * i386.h (TARGET_USE_GATHER): Define.
6900 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6902 2018-01-10 Martin Liska <mliska@suse.cz>
6905 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6906 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6908 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6909 CLEANUP_NO_PARTITIONING is not set.
6911 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6913 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6914 for vectors, as a partial revert of r254296.
6915 * rtl.h (const_vec_p): Delete.
6916 (const_vec_duplicate_p): Don't test for vector CONSTs.
6917 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6918 * expmed.c (make_tree): Likewise.
6921 * common.md (E, F): Use CONSTANT_P instead of checking for
6923 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6924 checking for CONST_VECTOR.
6926 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6929 * predict.c (force_edge_cold): Handle in more sane way edges
6932 2018-01-09 Carl Love <cel@us.ibm.com>
6934 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6936 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6937 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6938 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6939 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6940 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6941 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6942 * config/rs6000/rs6000-protos.h: Add extern defition for
6943 rs6000_generate_float2_double_code.
6944 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6946 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6947 (float2_v2df): Add define_expand.
6949 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6952 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6953 op_mode in the force_to_mode call.
6955 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6957 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6958 instead of checking each element individually.
6959 (aarch64_evpc_uzp): Likewise.
6960 (aarch64_evpc_zip): Likewise.
6961 (aarch64_evpc_ext): Likewise.
6962 (aarch64_evpc_rev): Likewise.
6963 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6964 instead of checking each element individually. Return true without
6966 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6967 whether all selected elements come from the same input, instead of
6968 checking each element individually. Remove calls to gen_rtx_REG,
6969 start_sequence and end_sequence and instead assert that no rtl is
6972 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6974 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6975 order of HIGH and CONST checks.
6977 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6979 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6980 if the destination isn't an SSA_NAME.
6982 2018-01-09 Richard Biener <rguenther@suse.de>
6984 PR tree-optimization/83668
6985 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6987 (canonicalize_loop_form): ... here, renamed from ...
6988 (canonicalize_loop_closed_ssa_form): ... this and amended to
6989 swap successor edges for loop exit blocks to make us use
6990 the RPO order we need for initial schedule generation.
6992 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6994 PR tree-optimization/64811
6995 * match.pd: When optimizing comparisons with Inf, avoid
6996 introducing or losing exceptions from comparisons with NaN.
6998 2018-01-09 Martin Liska <mliska@suse.cz>
7001 * asan.c (shadow_mem_size): Add gcc_assert.
7003 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7005 Don't save registers in main().
7008 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7009 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7010 * config/avr/avr.c (avr_set_current_function): Don't error if
7011 naked, OS_task or OS_main are specified at the same time.
7012 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7014 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7016 * common/config/avr/avr-common.c (avr_option_optimization_table):
7017 Switch on -mmain-is-OS_task for optimizing compilations.
7019 2018-01-09 Richard Biener <rguenther@suse.de>
7021 PR tree-optimization/83572
7022 * graphite.c: Include cfganal.h.
7023 (graphite_transform_loops): Connect infinite loops to exit
7024 and remove fake edges at the end.
7026 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7028 * ipa-inline.c (edge_badness): Revert accidental checkin.
7030 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7033 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7034 symbols; not inline clones.
7036 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7039 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7040 hard registers. Formatting fixes.
7042 PR preprocessor/83722
7043 * gcc.c (try_generate_repro): Pass
7044 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7045 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7048 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7049 Kito Cheng <kito.cheng@gmail.com>
7051 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7052 (riscv_leaf_function_p): Delete.
7053 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7055 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7057 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7059 (do_ifelse): New function.
7060 (do_isel): New function.
7061 (do_sub3): New function.
7062 (do_add3): New function.
7063 (do_load_mask_compare): New function.
7064 (do_overlap_load_compare): New function.
7065 (expand_compare_loop): New function.
7066 (expand_block_compare): Call expand_compare_loop() when appropriate.
7067 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7069 (-mblock-compare-inline-loop-limit): New option.
7071 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7074 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7075 Reverse order of second and third operands in first alternative.
7076 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7077 of first and second elements in UNSPEC_VPERMR vector.
7078 (altivec_expand_vec_perm_le): Likewise.
7080 2018-01-08 Jeff Law <law@redhat.com>
7082 PR rtl-optimizatin/81308
7083 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7084 (process_switch): If group_case_labels makes a change, then set
7086 (pass_convert_switch::execute): If a switch is converted, then
7087 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7089 PR rtl-optimization/81308
7090 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7093 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7095 PR target/83663 - Revert r255946
7096 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7097 generation for cases where splatting a value is not useful.
7098 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7099 across a vec_duplicate and a paradoxical subreg forming a vector
7100 mode to a vec_concat.
7102 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7104 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7105 -march=armv8.3-a variants.
7106 * config/arm/t-multilib: Likewise.
7107 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7109 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7111 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7113 (cceq_ior_compare_complement): Give it a name so I can use it, and
7114 change boolean_or_operator predicate to boolean_operator so it can
7115 be used to generate a crand.
7116 (eqne): New code iterator.
7117 (bd/bd_neg): New code_attrs.
7118 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7119 a single define_insn.
7120 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7121 decrement (bdnzt/bdnzf/bdzt/bdzf).
7122 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7123 with the new names of the branch decrement patterns, and added the
7124 names of the branch decrement conditional patterns.
7126 2018-01-08 Richard Biener <rguenther@suse.de>
7128 PR tree-optimization/83563
7129 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7132 2018-01-08 Richard Biener <rguenther@suse.de>
7135 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7137 2018-01-08 Richard Biener <rguenther@suse.de>
7139 PR tree-optimization/83685
7140 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7141 references to abnormals.
7143 2018-01-08 Richard Biener <rguenther@suse.de>
7146 * dwarf2out.c (output_indirect_strings): Handle empty
7147 skeleton_debug_str_hash.
7148 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7150 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7152 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7153 (emit_store_direct): Likewise.
7154 (arc_trampoline_adjust_address): Likewise.
7155 (arc_asm_trampoline_template): New function.
7156 (arc_initialize_trampoline): Use asm_trampoline_template.
7157 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7158 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7159 * config/arc/arc.md (flush_icache): Delete pattern.
7161 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7163 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7164 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7167 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7170 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7171 by not USED_FOR_TARGET.
7172 (make_pass_resolve_sw_modes): Likewise.
7174 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7176 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7179 2018-01-08 Richard Biener <rguenther@suse.de>
7182 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7184 2018-01-08 Richard Biener <rguenther@suse.de>
7187 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7189 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7192 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7193 basic blocks with a small number of successors.
7194 (convert_control_dep_chain_into_preds): Improve handling of
7196 (dump_predicates): Split apart into...
7197 (dump_pred_chain): ...here...
7198 (dump_pred_info): ...and here.
7199 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7200 (can_chain_union_be_invalidated_p): Improve check for invalidation
7202 (uninit_uses_cannot_happen): Avoid unnecessary if
7203 convert_control_dep_chain_into_preds yielded nothing.
7205 2018-01-06 Martin Sebor <msebor@redhat.com>
7207 PR tree-optimization/83640
7208 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7209 subtracting negative offset from size.
7210 (builtin_access::overlap): Adjust offset bounds of the access to fall
7211 within the size of the object if possible.
7213 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7215 PR rtl-optimization/83699
7216 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7217 extract_bit_field_as_subreg to cases in which the extracted
7218 value is also a vector.
7220 * lra-constraints.c (process_alt_operands): Test for the equivalence
7221 substitutions when detecting a possible reload cycle.
7223 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7226 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7227 by default if flag_selective_schedling{,2}. Formatting fixes.
7229 PR rtl-optimization/83682
7230 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7231 if it has non-VECTOR_MODE element mode.
7232 (vec_duplicate_p): Likewise.
7235 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7236 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7238 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7241 * config/i386/i386-builtin.def
7242 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7243 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7244 Require also OPTION_MASK_ISA_AVX512F in addition to
7245 OPTION_MASK_ISA_GFNI.
7246 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7247 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7248 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7249 to OPTION_MASK_ISA_GFNI.
7250 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7251 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7252 OPTION_MASK_ISA_AVX512BW.
7253 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7254 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7255 addition to OPTION_MASK_ISA_GFNI.
7256 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7257 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7258 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7259 to OPTION_MASK_ISA_GFNI.
7260 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7261 a requirement for all ISAs rather than any of them with a few
7263 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7265 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7266 bitmasks to be enabled with 3 exceptions, instead of requiring any
7267 enabled ISA with lots of exceptions.
7268 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7269 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7270 Change avx512bw in isa attribute to avx512f.
7271 * config/i386/sgxintrin.h: Add license boilerplate.
7272 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7273 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7274 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7275 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7277 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7278 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7279 temporarily sse2 rather than sse if not enabled already.
7282 * config/i386/sse.md (VI248_VLBW): Rename to ...
7283 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7284 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7285 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7286 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7287 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7288 mode iterator instead of VI248_VLBW.
7290 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7292 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7293 (record_modified): Skip clobbers; add debug output.
7294 (param_change_prob): Use sreal frequencies.
7296 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7298 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7299 punt for user-aligned variables.
7301 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7303 * tree-chrec.c (chrec_contains_symbols): Return true for
7306 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7309 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7310 of (x|y) == x for BICS pattern.
7312 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7314 PR tree-optimization/83605
7315 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7316 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7319 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7321 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7322 * config/epiphany/rtems.h: New file.
7324 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7325 Uros Bizjak <ubizjak@gmail.com>
7328 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7329 QIreg_operand instead of register_operand predicate.
7330 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7331 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7332 comments instead of -fmitigate[-_]rop.
7334 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7337 * cgraphunit.c (symbol_table::compile): Switch to text_section
7338 before calling assembly_start debug hook.
7339 * run-rtl-passes.c (run_rtl_passes): Likewise.
7342 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7344 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7345 range_int_cst_p rather than !symbolic_range_p before calling
7346 extract_range_from_multiplicative_op_1.
7348 2018-01-04 Jeff Law <law@redhat.com>
7350 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7351 redundant test in assertion.
7353 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7355 * doc/rtl.texi: Document machine_mode wrapper classes.
7357 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7359 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7362 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7364 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7365 the VEC_PERM_EXPR fold to fail.
7367 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7370 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7371 to switched_sections.
7373 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7376 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7379 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7382 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7383 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7385 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7388 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7389 is BLKmode and bitpos not zero or mode change is needed.
7391 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7394 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7397 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7400 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7401 instead of MULT rtx. Update all corresponding splitters.
7403 (*ssub<modesuffix>): Ditto.
7405 (*cmp_sadd_di): Update split patterns.
7406 (*cmp_sadd_si): Ditto.
7407 (*cmp_sadd_sidi): Ditto.
7408 (*cmp_ssub_di): Ditto.
7409 (*cmp_ssub_si): Ditto.
7410 (*cmp_ssub_sidi): Ditto.
7411 * config/alpha/predicates.md (const23_operand): New predicate.
7412 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7413 Look for ASHIFT, not MULT inner operand.
7414 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7416 2018-01-04 Martin Liska <mliska@suse.cz>
7418 PR gcov-profile/83669
7419 * gcov.c (output_intermediate_file): Add version to intermediate
7421 * doc/gcov.texi: Document new field 'version' in intermediate
7422 file format. Fix location of '-k' option of gcov command.
7424 2018-01-04 Martin Liska <mliska@suse.cz>
7427 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7429 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7431 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7433 2018-01-03 Martin Sebor <msebor@redhat.com>
7435 PR tree-optimization/83655
7436 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7437 checking calls with invalid arguments.
7439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7441 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7442 (vectorizable_mask_load_store): Delete.
7443 (vectorizable_call): Return false for masked loads and stores.
7444 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7445 instead of gimple_assign_rhs1.
7446 (vectorizable_load): Handle IFN_MASK_LOAD.
7447 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7451 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7453 (vectorizable_mask_load_store): ...here.
7454 (vectorizable_load): ...and here.
7456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7458 * tree-vect-stmts.c (vect_build_all_ones_mask)
7459 (vect_build_zero_merge_argument): New functions, split out from...
7460 (vectorizable_load): ...here.
7462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7464 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7466 (vectorizable_mask_load_store): ...here.
7467 (vectorizable_store): ...and here.
7469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7471 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7473 (vectorizable_mask_load_store): ...here.
7475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7477 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7478 (vect_model_store_cost): Take a vec_load_store_type instead of a
7480 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7481 (vect_model_store_cost): Take a vec_load_store_type instead of a
7483 (vectorizable_mask_load_store): Update accordingly.
7484 (vectorizable_store): Likewise.
7485 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7489 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7490 IFN_MASK_LOAD calls here rather than...
7491 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 Alan Hayward <alan.hayward@arm.com>
7495 David Sherwood <david.sherwood@arm.com>
7497 * expmed.c (extract_bit_field_1): For vector extracts,
7498 fall back to extract_bit_field_as_subreg if vec_extract
7501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7502 Alan Hayward <alan.hayward@arm.com>
7503 David Sherwood <david.sherwood@arm.com>
7505 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7506 they are variable or constant sized.
7507 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7508 slots for constant-sized data.
7510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7511 Alan Hayward <alan.hayward@arm.com>
7512 David Sherwood <david.sherwood@arm.com>
7514 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7515 handling COND_EXPRs with boolean comparisons, try to find a better
7516 basis for the mask type than the boolean itself.
7518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7520 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7521 is calculated and how it can be overridden.
7522 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7523 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7525 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7529 Alan Hayward <alan.hayward@arm.com>
7530 David Sherwood <david.sherwood@arm.com>
7532 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7533 Remove the mode argument.
7534 (aarch64_simd_valid_immediate): Remove the mode and inverse
7536 * config/aarch64/iterators.md (bitsize): New iterator.
7537 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7538 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7539 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7540 aarch64_simd_valid_immediate.
7541 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7542 (aarch64_reg_or_bic_imm): Likewise.
7543 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7544 with an insn_type enum and msl with a modifier_type enum.
7545 Replace element_width with a scalar_mode. Change the shift
7546 to unsigned int. Add constructors for scalar_float_mode and
7547 scalar_int_mode elements.
7548 (aarch64_vect_float_const_representable_p): Delete.
7549 (aarch64_can_const_movi_rtx_p)
7550 (aarch64_simd_scalar_immediate_valid_for_move)
7551 (aarch64_simd_make_constant): Update call to
7552 aarch64_simd_valid_immediate.
7553 (aarch64_advsimd_valid_immediate_hs): New function.
7554 (aarch64_advsimd_valid_immediate): Likewise.
7555 (aarch64_simd_valid_immediate): Remove mode and inverse
7556 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7557 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7558 and aarch64_float_const_representable_p on the result.
7559 (aarch64_output_simd_mov_immediate): Remove mode argument.
7560 Update call to aarch64_simd_valid_immediate and use of
7561 simd_immediate_info.
7562 (aarch64_output_scalar_simd_mov_immediate): Update call
7565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7566 Alan Hayward <alan.hayward@arm.com>
7567 David Sherwood <david.sherwood@arm.com>
7569 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7570 (mode_nunits): Likewise CONST_MODE_NUNITS.
7571 * machmode.def (ADJUST_NUNITS): Document.
7572 * genmodes.c (mode_data::need_nunits_adj): New field.
7573 (blank_mode): Update accordingly.
7574 (adj_nunits): New variable.
7575 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7577 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7578 listed in adj_nunits.
7579 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7580 listed in adj_nunits. Don't emit case statements for such modes.
7581 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7582 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7583 nothing if adj_nunits is nonnull.
7584 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7585 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7586 (emit_mode_fbit): Update use of print_maybe_const_decl.
7587 (emit_move_size): Likewise. Treat the array as non-const
7589 (emit_mode_adjustments): Handle adj_nunits.
7591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7593 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7594 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7595 (VECTOR_MODES): Use it.
7596 (make_vector_modes): Take the prefix as an argument.
7598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7599 Alan Hayward <alan.hayward@arm.com>
7600 David Sherwood <david.sherwood@arm.com>
7602 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7603 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7604 for MODE_VECTOR_BOOL.
7605 * machmode.def (VECTOR_BOOL_MODE): Document.
7606 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7607 (make_vector_bool_mode): New function.
7608 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7610 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7611 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7613 * stor-layout.c (int_mode_for_mode): Likewise.
7614 * tree.c (build_vector_type_for_mode): Likewise.
7615 * varasm.c (output_constant_pool_2): Likewise.
7616 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7617 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7618 for MODE_VECTOR_BOOL.
7619 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7620 of mode class checks.
7621 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7622 instead of a list of mode class checks.
7623 (expand_vector_scalar_condition): Likewise.
7624 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7627 Alan Hayward <alan.hayward@arm.com>
7628 David Sherwood <david.sherwood@arm.com>
7630 * machmode.h (mode_size): Change from unsigned short to
7632 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7633 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7634 or if measurement_type is not polynomial.
7635 (fixed_size_mode::includes_p): Check for constant-sized modes.
7636 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7637 return a poly_uint16 rather than an unsigned short.
7638 (emit_mode_size): Change the type of mode_size from unsigned short
7639 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7640 (emit_mode_adjustments): Cope with polynomial vector sizes.
7641 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7643 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7645 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7646 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7647 * caller-save.c (setup_save_areas): Likewise.
7648 (replace_reg_with_saved_mem): Likewise.
7649 * calls.c (emit_library_call_value_1): Likewise.
7650 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7651 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7652 (gen_lowpart_for_combine): Likewise.
7653 * convert.c (convert_to_integer_1): Likewise.
7654 * cse.c (equiv_constant, cse_insn): Likewise.
7655 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7656 (cselib_subst_to_values): Likewise.
7657 * dce.c (word_dce_process_block): Likewise.
7658 * df-problems.c (df_word_lr_mark_ref): Likewise.
7659 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7660 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7661 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7662 (rtl_for_decl_location): Likewise.
7663 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7664 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7665 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7666 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7667 (expand_expr_real_1): Likewise.
7668 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7669 (pad_below): Likewise.
7670 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7671 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7672 * ira.c (get_subreg_tracking_sizes): Likewise.
7673 * ira-build.c (ira_create_allocno_objects): Likewise.
7674 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7675 (ira_sort_regnos_for_alter_reg): Likewise.
7676 * ira-costs.c (record_operand_costs): Likewise.
7677 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7678 (resolve_simple_move): Likewise.
7679 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7680 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7681 (lra_constraints): Likewise.
7682 (CONST_POOL_OK_P): Reject variable-sized modes.
7683 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7684 (add_pseudo_to_slot, lra_spill): Likewise.
7685 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7686 * optabs-query.c (get_best_extraction_insn): Likewise.
7687 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7688 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7689 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7690 * recog.c (offsettable_address_addr_space_p): Likewise.
7691 * regcprop.c (maybe_mode_change): Likewise.
7692 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7693 * regrename.c (build_def_use): Likewise.
7694 * regstat.c (dump_reg_info): Likewise.
7695 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7696 (find_reloads, find_reloads_subreg_address): Likewise.
7697 * reload1.c (eliminate_regs_1): Likewise.
7698 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7699 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7700 (simplify_binary_operation_1, simplify_subreg): Likewise.
7701 * targhooks.c (default_function_arg_padding): Likewise.
7702 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7703 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7704 (verify_gimple_assign_ternary): Likewise.
7705 * tree-inline.c (estimate_move_cost): Likewise.
7706 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7707 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7708 (get_address_cost_ainc): Likewise.
7709 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7710 (vect_supportable_dr_alignment): Likewise.
7711 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7712 (vectorizable_reduction): Likewise.
7713 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7714 (vectorizable_operation, vectorizable_load): Likewise.
7715 * tree.c (build_same_sized_truth_vector_type): Likewise.
7716 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7717 * var-tracking.c (emit_note_insn_var_location): Likewise.
7718 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7719 (ADDR_VEC_ALIGN): Likewise.
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7727 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7728 or if measurement_type is polynomial.
7729 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7730 * combine.c (make_extraction): Likewise.
7731 * dse.c (find_shift_sequence): Likewise.
7732 * dwarf2out.c (mem_loc_descriptor): Likewise.
7733 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7734 (extract_bit_field, extract_low_bits): Likewise.
7735 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7736 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7737 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7738 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7739 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7740 * reload.c (find_reloads): Likewise.
7741 * reload1.c (alter_reg): Likewise.
7742 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7743 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7744 * tree-if-conv.c (predicate_mem_writes): Likewise.
7745 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7746 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7747 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7748 * valtrack.c (dead_debug_insert_temp): Likewise.
7749 * varasm.c (mergeable_constant_section): Likewise.
7750 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7753 Alan Hayward <alan.hayward@arm.com>
7754 David Sherwood <david.sherwood@arm.com>
7756 * expr.c (expand_assignment): Cope with polynomial mode sizes
7757 when assigning to a CONCAT.
7759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7760 Alan Hayward <alan.hayward@arm.com>
7761 David Sherwood <david.sherwood@arm.com>
7763 * machmode.h (mode_precision): Change from unsigned short to
7765 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7767 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7768 or if measurement_type is not polynomial.
7769 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7770 in which the mode is already known to be a scalar_int_mode.
7771 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7772 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7774 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7775 for GET_MODE_PRECISION.
7776 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7777 for GET_MODE_PRECISION.
7778 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7780 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7781 (expand_field_assignment, make_extraction): Likewise.
7782 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7783 (get_last_value): Likewise.
7784 * convert.c (convert_to_integer_1): Likewise.
7785 * cse.c (cse_insn): Likewise.
7786 * expr.c (expand_expr_real_1): Likewise.
7787 * lra-constraints.c (simplify_operand_subreg): Likewise.
7788 * optabs-query.c (can_atomic_load_p): Likewise.
7789 * optabs.c (expand_atomic_load): Likewise.
7790 (expand_atomic_store): Likewise.
7791 * ree.c (combine_reaching_defs): Likewise.
7792 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7793 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7794 * tree.h (type_has_mode_precision_p): Likewise.
7795 * ubsan.c (instrument_si_overflow): Likewise.
7797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7798 Alan Hayward <alan.hayward@arm.com>
7799 David Sherwood <david.sherwood@arm.com>
7801 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7802 polynomial numbers of units.
7803 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7804 (valid_vector_subparts_p): New function.
7805 (build_vector_type): Remove temporary shim and take the number
7806 of units as a poly_uint64 rather than an int.
7807 (build_opaque_vector_type): Take the number of units as a
7808 poly_uint64 rather than an int.
7809 * tree.c (build_vector_from_ctor): Handle polynomial
7810 TYPE_VECTOR_SUBPARTS.
7811 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7812 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7813 (build_vector_from_val): If the number of units is variable,
7814 use build_vec_duplicate_cst for constant operands and
7815 VEC_DUPLICATE_EXPR otherwise.
7816 (make_vector_type): Remove temporary is_constant ().
7817 (build_vector_type, build_opaque_vector_type): Take the number of
7818 units as a poly_uint64 rather than an int.
7819 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7821 * cfgexpand.c (expand_debug_expr): Likewise.
7822 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7823 (store_constructor, expand_expr_real_1): Likewise.
7824 (const_scalar_mask_from_tree): Likewise.
7825 * fold-const-call.c (fold_const_reduction): Likewise.
7826 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7827 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7828 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7829 (fold_relational_const): Likewise.
7830 (native_interpret_vector): Likewise. Change the size from an
7831 int to an unsigned int.
7832 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7833 TYPE_VECTOR_SUBPARTS.
7834 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7835 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7836 duplicating a non-constant operand into a variable-length vector.
7837 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7838 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7839 * ipa-icf.c (sem_variable::equals): Likewise.
7840 * match.pd: Likewise.
7841 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7842 * print-tree.c (print_node): Likewise.
7843 * stor-layout.c (layout_type): Likewise.
7844 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7845 * tree-cfg.c (verify_gimple_comparison): Likewise.
7846 (verify_gimple_assign_binary): Likewise.
7847 (verify_gimple_assign_ternary): Likewise.
7848 (verify_gimple_assign_single): Likewise.
7849 * tree-pretty-print.c (dump_generic_node): Likewise.
7850 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7851 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7852 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7853 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7854 (vect_shift_permute_load_chain): Likewise.
7855 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7856 (expand_vector_condition, optimize_vector_constructor): Likewise.
7857 (lower_vec_perm, get_compute_type): Likewise.
7858 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7859 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7860 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7861 (vect_recog_mask_conversion_pattern): Likewise.
7862 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7863 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7864 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7865 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7866 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7867 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7868 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7869 (supportable_widening_operation): Likewise.
7870 (supportable_narrowing_operation): Likewise.
7871 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7873 * varasm.c (output_constant): Likewise.
7875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7876 Alan Hayward <alan.hayward@arm.com>
7877 David Sherwood <david.sherwood@arm.com>
7879 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7880 so that both the length == 3 and length != 3 cases set up their
7881 own permute vectors. Add comments explaining why we know the
7882 number of elements is constant.
7883 (vect_permute_load_chain): Likewise.
7885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7886 Alan Hayward <alan.hayward@arm.com>
7887 David Sherwood <david.sherwood@arm.com>
7889 * machmode.h (mode_nunits): Change from unsigned char to
7891 (ONLY_FIXED_SIZE_MODES): New macro.
7892 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7893 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7894 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7896 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7897 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7898 or if measurement_type is not polynomial.
7899 * genmodes.c (ZERO_COEFFS): New macro.
7900 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7902 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7903 Use ZERO_COEFFS when emitting initializers.
7904 * data-streamer.h (bp_pack_poly_value): New function.
7905 (bp_unpack_poly_value): Likewise.
7906 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7907 for GET_MODE_NUNITS.
7908 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7909 for GET_MODE_NUNITS.
7910 * tree.c (make_vector_type): Remove temporary shim and make
7911 the real function take the number of units as a poly_uint64
7913 (build_vector_type_for_mode): Handle polynomial nunits.
7914 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7915 * emit-rtl.c (const_vec_series_p_1): Likewise.
7916 (gen_rtx_CONST_VECTOR): Likewise.
7917 * fold-const.c (test_vec_duplicate_folding): Likewise.
7918 * genrecog.c (validate_pattern): Likewise.
7919 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7920 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7921 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7922 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7923 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7924 * rtlanal.c (subreg_get_info): Likewise.
7925 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7926 (vect_grouped_load_supported): Likewise.
7927 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7928 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7929 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7930 (simplify_const_unary_operation, simplify_binary_operation_1)
7931 (simplify_const_binary_operation, simplify_ternary_operation)
7932 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7933 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7934 instead of CONST_VECTOR_NUNITS.
7935 * varasm.c (output_constant_pool_2): Likewise.
7936 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7937 explicit-encoded elements in the XVEC for variable-length vectors.
7939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7941 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7944 Alan Hayward <alan.hayward@arm.com>
7945 David Sherwood <david.sherwood@arm.com>
7947 * coretypes.h (fixed_size_mode): Declare.
7948 (fixed_size_mode_pod): New typedef.
7949 * builtins.h (target_builtins::x_apply_args_mode)
7950 (target_builtins::x_apply_result_mode): Change type to
7951 fixed_size_mode_pod.
7952 * builtins.c (apply_args_size, apply_result_size, result_vector)
7953 (expand_builtin_apply_args_1, expand_builtin_apply)
7954 (expand_builtin_return): Update accordingly.
7956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7958 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7959 * cselib.c (cselib_hash_rtx): Likewise.
7960 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7961 CONST_VECTOR encoding.
7963 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7964 Jeff Law <law@redhat.com>
7967 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7968 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7969 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7970 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7973 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7974 explicitly probe *sp in a noreturn function if there were any callee
7975 register saves or frame pointer is needed.
7977 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7980 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7981 BLKmode for ternary, binary or unary expressions.
7984 * var-tracking.c (delete_vta_debug_insn): New inline function.
7985 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7986 insns from get_insns () to NULL instead of each bb separately.
7987 Use delete_vta_debug_insn. No longer static.
7988 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7989 delete_vta_debug_insns callers.
7990 * rtl.h (delete_vta_debug_insns): Declare.
7991 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7992 instead of variable_tracking_main.
7994 2018-01-03 Martin Sebor <msebor@redhat.com>
7996 PR tree-optimization/83603
7997 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7998 arguments past the endof the argument list in functions declared
7999 without a prototype.
8000 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8001 Avoid checking when arguments are null.
8003 2018-01-03 Martin Sebor <msebor@redhat.com>
8006 * doc/extend.texi (attribute const): Fix a typo.
8007 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8008 issuing -Wsuggest-attribute for void functions.
8010 2018-01-03 Martin Sebor <msebor@redhat.com>
8012 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8013 offset_int::from instead of wide_int::to_shwi.
8014 (maybe_diag_overlap): Remove assertion.
8015 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8016 * gimple-ssa-sprintf.c (format_directive): Same.
8017 (parse_directive): Same.
8018 (sprintf_dom_walker::compute_format_length): Same.
8019 (try_substitute_return_value): Same.
8021 2018-01-03 Jeff Law <law@redhat.com>
8024 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8025 non-constant residual for zero at runtime and avoid probing in
8026 that case. Reorganize code for trailing problem to mirror handling
8029 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8031 PR tree-optimization/83501
8032 * tree-ssa-strlen.c (get_string_cst): New.
8033 (handle_char_store): Call get_string_cst.
8035 2018-01-03 Martin Liska <mliska@suse.cz>
8037 PR tree-optimization/83593
8038 * tree-ssa-strlen.c: Include tree-cfg.h.
8039 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8040 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8041 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8043 (strlen_dom_walker::before_dom_children): Call
8044 gimple_purge_dead_eh_edges. Dump tranformation with details
8046 (strlen_dom_walker::before_dom_children): Update call by adding
8047 new argument cleanup_eh.
8048 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8050 2018-01-03 Martin Liska <mliska@suse.cz>
8053 * cif-code.def (VARIADIC_THUNK): New enum value.
8054 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8057 2018-01-03 Jan Beulich <jbeulich@suse.com>
8059 * sse.md (mov<mode>_internal): Tighten condition for when to use
8060 vmovdqu<ssescalarsize> for TI and OI modes.
8062 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8064 Update copyright years.
8066 2018-01-03 Martin Liska <mliska@suse.cz>
8069 * ipa-visibility.c (function_and_variable_visibility): Skip
8070 functions with noipa attribure.
8072 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8074 * gcc.c (process_command): Update copyright notice dates.
8075 * gcov-dump.c (print_version): Ditto.
8076 * gcov.c (print_version): Ditto.
8077 * gcov-tool.c (print_version): Ditto.
8078 * gengtype.c (create_file): Ditto.
8079 * doc/cpp.texi: Bump @copying's copyright year.
8080 * doc/cppinternals.texi: Ditto.
8081 * doc/gcc.texi: Ditto.
8082 * doc/gccint.texi: Ditto.
8083 * doc/gcov.texi: Ditto.
8084 * doc/install.texi: Ditto.
8085 * doc/invoke.texi: Ditto.
8087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8089 * vector-builder.h (vector_builder::m_full_nelts): Change from
8090 unsigned int to poly_uint64.
8091 (vector_builder::full_nelts): Update prototype accordingly.
8092 (vector_builder::new_vector): Likewise.
8093 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8094 (vector_builder::operator ==): Likewise.
8095 (vector_builder::finalize): Likewise.
8096 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8097 Take the number of elements as a poly_uint64 rather than an
8099 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8100 from unsigned int to poly_uint64.
8101 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8102 (vec_perm_indices::new_vector): Likewise.
8103 (vec_perm_indices::length): Likewise.
8104 (vec_perm_indices::nelts_per_input): Likewise.
8105 (vec_perm_indices::input_nelts): Likewise.
8106 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8107 number of elements per input as a poly_uint64 rather than an
8108 unsigned int. Use the original encoding for variable-length
8109 vectors, rather than clamping each individual element.
8110 For the second and subsequent elements in each pattern,
8111 clamp the step and base before clamping their sum.
8112 (vec_perm_indices::series_p): Handle polynomial element counts.
8113 (vec_perm_indices::all_in_range_p): Likewise.
8114 (vec_perm_indices_to_tree): Likewise.
8115 (vec_perm_indices_to_rtx): Likewise.
8116 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8117 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8118 (tree_vector_builder::new_binary_operation): Handle polynomial
8119 element counts. Return false if we need to know the number
8120 of elements at compile time.
8121 * fold-const.c (fold_vec_perm): Punt if the number of elements
8122 isn't known at compile time.
8124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8126 * vec-perm-indices.h (vec_perm_builder): Change element type
8127 from HOST_WIDE_INT to poly_int64.
8128 (vec_perm_indices::element_type): Update accordingly.
8129 (vec_perm_indices::clamp): Handle polynomial element_types.
8130 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8131 (vec_perm_indices::all_in_range_p): Likewise.
8132 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8134 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8135 polynomial vec_perm_indices element types.
8136 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8137 * fold-const.c (fold_vec_perm): Likewise.
8138 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8139 * tree-vect-generic.c (lower_vec_perm): Likewise.
8140 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8141 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8142 element type to HOST_WIDE_INT.
8144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8145 Alan Hayward <alan.hayward@arm.com>
8146 David Sherwood <david.sherwood@arm.com>
8148 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8149 rather than an int. Use plus_constant.
8150 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8151 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8154 Alan Hayward <alan.hayward@arm.com>
8155 David Sherwood <david.sherwood@arm.com>
8157 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8158 a HOST_WIDE_INT to a poly_int64.
8160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8161 Alan Hayward <alan.hayward@arm.com>
8162 David Sherwood <david.sherwood@arm.com>
8164 * calls.c (load_register_parameters): Cope with polynomial
8165 mode sizes. Require a constant size for BLKmode parameters
8166 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8167 forces a parameter to be padded at the lsb end in order to
8168 fill a complete number of words, require the parameter size
8169 to be ordered wrt UNITS_PER_WORD.
8171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8172 Alan Hayward <alan.hayward@arm.com>
8173 David Sherwood <david.sherwood@arm.com>
8175 * reload1.c (spill_stack_slot_width): Change element type
8176 from unsigned int to poly_uint64_pod.
8177 (alter_reg): Treat mode sizes as polynomial.
8179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8180 Alan Hayward <alan.hayward@arm.com>
8181 David Sherwood <david.sherwood@arm.com>
8183 * reload.c (complex_word_subreg_p): New function.
8184 (reload_inner_reg_of_subreg, push_reload): Use it.
8186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8187 Alan Hayward <alan.hayward@arm.com>
8188 David Sherwood <david.sherwood@arm.com>
8190 * lra-constraints.c (process_alt_operands): Reject matched
8191 operands whose sizes aren't ordered.
8192 (match_reload): Refer to this check here.
8194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8195 Alan Hayward <alan.hayward@arm.com>
8196 David Sherwood <david.sherwood@arm.com>
8198 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8199 that the mode size is in the set {1, 2, 4, 8, 16}.
8201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8202 Alan Hayward <alan.hayward@arm.com>
8203 David Sherwood <david.sherwood@arm.com>
8205 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8206 Use plus_constant instead of gen_rtx_PLUS.
8208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8209 Alan Hayward <alan.hayward@arm.com>
8210 David Sherwood <david.sherwood@arm.com>
8212 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8213 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8214 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8215 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8216 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8217 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8218 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8219 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8220 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8221 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8223 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8224 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8225 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8226 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8227 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8228 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8229 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8230 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8231 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8232 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8234 * expr.c (emit_move_resolve_push): Treat the input and result
8235 of PUSH_ROUNDING as a poly_int64.
8236 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8237 (emit_push_insn): Likewise.
8238 * lra-eliminations.c (mark_not_eliminable): Likewise.
8239 * recog.c (push_operand): Likewise.
8240 * reload1.c (elimination_effects): Likewise.
8241 * rtlanal.c (nonzero_bits1): Likewise.
8242 * calls.c (store_one_arg): Likewise. Require the padding to be
8243 known at compile time.
8245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8246 Alan Hayward <alan.hayward@arm.com>
8247 David Sherwood <david.sherwood@arm.com>
8249 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8250 Use plus_constant instead of gen_rtx_PLUS.
8252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8253 Alan Hayward <alan.hayward@arm.com>
8254 David Sherwood <david.sherwood@arm.com>
8256 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8260 Alan Hayward <alan.hayward@arm.com>
8261 David Sherwood <david.sherwood@arm.com>
8263 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8264 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8265 via stack temporaries. Treat the mode size as polynomial too.
8267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8268 Alan Hayward <alan.hayward@arm.com>
8269 David Sherwood <david.sherwood@arm.com>
8271 * expr.c (expand_expr_real_2): When handling conversions involving
8272 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8273 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8274 as a poly_uint64 too.
8276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8277 Alan Hayward <alan.hayward@arm.com>
8278 David Sherwood <david.sherwood@arm.com>
8280 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8283 Alan Hayward <alan.hayward@arm.com>
8284 David Sherwood <david.sherwood@arm.com>
8286 * combine.c (can_change_dest_mode): Handle polynomial
8287 REGMODE_NATURAL_SIZE.
8288 * expmed.c (store_bit_field_1): Likewise.
8289 * expr.c (store_constructor): Likewise.
8290 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8291 and polynomial REGMODE_NATURAL_SIZE.
8292 (gen_lowpart_common): Likewise.
8293 * reginfo.c (record_subregs_of_mode): Likewise.
8294 * rtlanal.c (read_modify_subreg_p): Likewise.
8296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8297 Alan Hayward <alan.hayward@arm.com>
8298 David Sherwood <david.sherwood@arm.com>
8300 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8301 numbers of elements.
8303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8304 Alan Hayward <alan.hayward@arm.com>
8305 David Sherwood <david.sherwood@arm.com>
8307 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8314 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
8321 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8322 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8329 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8330 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8331 (expand_vector_condition, vector_element): Likewise.
8332 (subparts_gt): New function.
8333 (get_compute_type): Use subparts_gt.
8334 (count_type_subparts): Delete.
8335 (expand_vector_operations_1): Use subparts_gt instead of
8336 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8343 (vect_compile_time_alias): ...this new function. Do the calculation
8344 on poly_ints rather than trees.
8345 (vect_prune_runtime_alias_test_list): Update call accordingly.
8347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8348 Alan Hayward <alan.hayward@arm.com>
8349 David Sherwood <david.sherwood@arm.com>
8351 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8353 (vect_schedule_slp_instance): Likewise.
8355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8356 Alan Hayward <alan.hayward@arm.com>
8357 David Sherwood <david.sherwood@arm.com>
8359 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8360 constant and extern definitions for variable-length vectors.
8361 (vect_get_constant_vectors): Note that the number of units
8362 is known to be constant.
8364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8365 Alan Hayward <alan.hayward@arm.com>
8366 David Sherwood <david.sherwood@arm.com>
8368 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8369 of units as polynomial. Choose between WIDE and NARROW based
8372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8373 Alan Hayward <alan.hayward@arm.com>
8374 David Sherwood <david.sherwood@arm.com>
8376 * tree-vect-stmts.c (simd_clone_subparts): New function.
8377 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8380 Alan Hayward <alan.hayward@arm.com>
8381 David Sherwood <david.sherwood@arm.com>
8383 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8384 vectors as polynomial. Use build_index_vector for
8387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8388 Alan Hayward <alan.hayward@arm.com>
8389 David Sherwood <david.sherwood@arm.com>
8391 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8392 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8393 for variable-length vectors.
8394 (vectorizable_mask_load_store): Treat the number of units as
8395 polynomial, asserting that it is constant if the condition has
8396 already been enforced.
8397 (vectorizable_store, vectorizable_load): Likewise.
8399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8400 Alan Hayward <alan.hayward@arm.com>
8401 David Sherwood <david.sherwood@arm.com>
8403 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
8404 of units as polynomial. Punt if we can't tell at compile time
8405 which vector contains the final result.
8407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8408 Alan Hayward <alan.hayward@arm.com>
8409 David Sherwood <david.sherwood@arm.com>
8411 * tree-vect-loop.c (vectorizable_induction): Treat the number
8412 of units as polynomial. Punt on SLP inductions. Use an integer
8413 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8414 cast of such a series for variable-length floating-point
8417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8418 Alan Hayward <alan.hayward@arm.com>
8419 David Sherwood <david.sherwood@arm.com>
8421 * tree.h (build_index_vector): Declare.
8422 * tree.c (build_index_vector): New function.
8423 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8424 of units as polynomial, forcibly converting it to a constant if
8425 vectorizable_reduction has already enforced the condition.
8426 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8427 to create a {1,2,3,...} vector.
8428 (vectorizable_reduction): Treat the number of units as polynomial.
8429 Choose vectype_in based on the largest scalar element size rather
8430 than the smallest number of units. Enforce the restrictions
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 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8438 number of units as polynomial.
8440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8441 Alan Hayward <alan.hayward@arm.com>
8442 David Sherwood <david.sherwood@arm.com>
8444 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8445 * target.def (autovectorize_vector_sizes): Return the vector sizes
8446 by pointer, using vector_sizes rather than a bitmask.
8447 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8448 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8449 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8451 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8452 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8453 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8454 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8455 * omp-general.c (omp_max_vf): Likewise.
8456 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8457 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8458 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8459 * tree-vect-slp.c (vect_slp_bb): Likewise.
8460 * doc/tm.texi: Regenerate.
8461 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8463 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8464 the vector size as a poly_uint64 rather than an unsigned int.
8465 (current_vector_size): Change from an unsigned int to a poly_uint64.
8466 (get_vectype_for_scalar_type): Update accordingly.
8467 * tree.h (build_truth_vector_type): Take the size and number of
8468 units as a poly_uint64 rather than an unsigned int.
8469 (build_vector_type): Add a temporary overload that takes
8470 the number of units as a poly_uint64 rather than an unsigned int.
8471 * tree.c (make_vector_type): Likewise.
8472 (build_truth_vector_type): Take the number of units as a poly_uint64
8473 rather than an unsigned int.
8475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8476 Alan Hayward <alan.hayward@arm.com>
8477 David Sherwood <david.sherwood@arm.com>
8479 * target.def (get_mask_mode): Take the number of units and length
8480 as poly_uint64s rather than unsigned ints.
8481 * targhooks.h (default_get_mask_mode): Update accordingly.
8482 * targhooks.c (default_get_mask_mode): Likewise.
8483 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8484 * doc/tm.texi: Regenerate.
8486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8487 Alan Hayward <alan.hayward@arm.com>
8488 David Sherwood <david.sherwood@arm.com>
8490 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8491 * omp-general.c (omp_max_vf): Likewise.
8492 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8493 (expand_omp_simd): Handle polynomial safelen.
8494 * omp-low.c (omplow_simd_context): Add a default constructor.
8495 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8496 (lower_rec_simd_input_clauses): Update accordingly.
8497 (lower_rec_input_clauses): Likewise.
8499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8500 Alan Hayward <alan.hayward@arm.com>
8501 David Sherwood <david.sherwood@arm.com>
8503 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8504 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8505 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8506 (vect_analyze_slp_cost): Likewise.
8507 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8508 (vect_model_load_cost): Likewise.
8510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8511 Alan Hayward <alan.hayward@arm.com>
8512 David Sherwood <david.sherwood@arm.com>
8514 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8515 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8516 from an unsigned int * to a poly_uint64_pod *.
8517 (calculate_unrolling_factor): New function.
8518 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8521 Alan Hayward <alan.hayward@arm.com>
8522 David Sherwood <david.sherwood@arm.com>
8524 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8525 from an unsigned int to a poly_uint64.
8526 (_loop_vec_info::slp_unrolling_factor): Likewise.
8527 (_loop_vec_info::vectorization_factor): Change from an int
8529 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8530 (vect_get_num_vectors): New function.
8531 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8532 (vect_get_num_copies): Use vect_get_num_vectors.
8533 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8534 to an unsigned int *.
8535 (vect_analyze_data_refs): Change min_vf from an int * to a
8537 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8538 than an unsigned HOST_WIDE_INT.
8539 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8540 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8541 to an unsigned int *.
8542 (vect_analyze_data_ref_dependences): Likewise.
8543 (vect_compute_data_ref_alignment): Handle polynomial vf.
8544 (vect_enhance_data_refs_alignment): Likewise.
8545 (vect_prune_runtime_alias_test_list): Likewise.
8546 (vect_shift_permute_load_chain): Likewise.
8547 (vect_supportable_dr_alignment): Likewise.
8548 (dependence_distance_ge_vf): Take the vectorization factor as a
8549 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8550 (vect_analyze_data_refs): Change min_vf from an int * to a
8552 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8553 vfm1 as a poly_uint64 rather than an int. Make the same change
8554 for the returned bound_scalar.
8555 (vect_gen_vector_loop_niters): Handle polynomial vf.
8556 (vect_do_peeling): Likewise. Update call to
8557 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8558 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8560 * tree-vect-loop.c (vect_determine_vectorization_factor)
8561 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8562 (vect_get_known_peeling_cost): Likewise.
8563 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8564 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8565 (vect_transform_loop): Likewise. Use the lowest possible VF when
8566 updating the upper bounds of the loop.
8567 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8569 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8570 polynomial unroll factors.
8571 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8572 (vect_make_slp_decision): Likewise.
8573 (vect_supported_load_permutation_p): Likewise, and polynomial
8575 (vect_analyze_slp_cost): Handle polynomial vf.
8576 (vect_slp_analyze_node_operations): Likewise.
8577 (vect_slp_analyze_bb_1): Likewise.
8578 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8579 than an unsigned HOST_WIDE_INT.
8580 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8581 (vectorizable_load): Handle polynomial vf.
8582 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8584 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8587 Alan Hayward <alan.hayward@arm.com>
8588 David Sherwood <david.sherwood@arm.com>
8590 * match.pd: Handle bit operations involving three constants
8591 and try to fold one pair.
8593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8595 * tree-vect-loop-manip.c: Include gimple-fold.h.
8596 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8597 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8598 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8599 Add a path that uses a step of VF instead of 1, but disable it
8601 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8602 and niters_no_overflow parameters. Update calls to
8603 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8604 Create a new SSA name if the latter choses to use a ste other
8605 than zero, and return it via niters_vector_mult_vf_var.
8606 * tree-vect-loop.c (vect_transform_loop): Update calls to
8607 vect_do_peeling, vect_gen_vector_loop_niters and
8608 slpeel_make_loop_iterate_ntimes.
8609 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8610 (vect_gen_vector_loop_niters): Update declarations after above changes.
8612 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8614 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8615 128-bit round to integer instructions.
8616 (ceil<mode>2): Likewise.
8617 (btrunc<mode>2): Likewise.
8618 (round<mode>2): Likewise.
8620 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8622 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8623 unaligned VSX load/store on P8/P9.
8624 (expand_block_clear): Allow the use of unaligned VSX
8625 load/store on P8/P9.
8627 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8629 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8631 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8632 swap associated with both a load and a store.
8634 2018-01-02 Andrew Waterman <andrew@sifive.com>
8636 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8637 * config/riscv/riscv.md (clear_cache): Use it.
8639 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8641 * web.c: Remove out-of-date comment.
8643 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8645 * expr.c (fixup_args_size_notes): Check that any existing
8646 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8647 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8648 (emit_single_push_insn): ...here.
8650 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8652 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8653 (const_vector_encoded_nelts): New function.
8654 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8655 (const_vector_int_elt, const_vector_elt): Declare.
8656 * emit-rtl.c (const_vector_int_elt_1): New function.
8657 (const_vector_elt): Likewise.
8658 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8659 of CONST_VECTOR_ELT.
8661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8663 * expr.c: Include rtx-vector-builder.h.
8664 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8665 directly on the tree encoding.
8666 (const_vector_from_tree): Likewise.
8667 * optabs.c: Include rtx-vector-builder.h.
8668 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8669 sequence of "u" values.
8670 * vec-perm-indices.c: Include rtx-vector-builder.h.
8671 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8672 directly on the vec_perm_indices encoding.
8674 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8676 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8677 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8678 * rtx-vector-builder.h: New file.
8679 * rtx-vector-builder.c: Likewise.
8680 * rtl.h (rtx_def::u2): Add a const_vector field.
8681 (CONST_VECTOR_NPATTERNS): New macro.
8682 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8683 (CONST_VECTOR_DUPLICATE_P): Likewise.
8684 (CONST_VECTOR_STEPPED_P): Likewise.
8685 (CONST_VECTOR_ENCODED_ELT): Likewise.
8686 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8687 (unwrap_const_vec_duplicate): Likewise.
8688 (const_vec_series_p): Check for a non-duplicated vector encoding.
8689 Say that the function only returns true for integer vectors.
8690 * emit-rtl.c: Include rtx-vector-builder.h.
8691 (gen_const_vec_duplicate_1): Delete.
8692 (gen_const_vector): Call gen_const_vec_duplicate instead of
8693 gen_const_vec_duplicate_1.
8694 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8695 (gen_const_vec_duplicate): Use rtx_vector_builder.
8696 (gen_const_vec_series): Likewise.
8697 (gen_rtx_CONST_VECTOR): Likewise.
8698 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8699 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8700 Build a new vector rather than modifying a CONST_VECTOR in-place.
8701 (handle_special_swappables): Update call accordingly.
8702 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8703 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8704 Build a new vector rather than modifying a CONST_VECTOR in-place.
8705 (handle_special_swappables): Update call accordingly.
8707 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8709 * simplify-rtx.c (simplify_const_binary_operation): Use
8710 CONST_VECTOR_ELT instead of XVECEXP.
8712 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8714 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8715 the selector elements to be different from the data elements
8716 if the selector is a VECTOR_CST.
8717 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8718 ssizetype for the selector.
8720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8722 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8723 before testing each element individually.
8724 * tree-vect-generic.c (lower_vec_perm): Likewise.
8726 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8728 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8729 * selftest-run-tests.c (selftest::run_tests): Call it.
8730 * vector-builder.h (vector_builder::operator ==): New function.
8731 (vector_builder::operator !=): Likewise.
8732 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8733 (vec_perm_indices::all_from_input_p): New function.
8734 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8735 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8736 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8737 instead of reading the VECTOR_CST directly. Detect whether both
8738 vector inputs are the same before constructing the vec_perm_indices,
8739 and update the number of inputs argument accordingly. Use the
8740 utility functions added above. Only construct sel2 if we need to.
8742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8744 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8745 the broadcast of the low byte.
8746 (expand_mult_highpart): Use an explicit encoding for the permutes.
8747 * optabs-query.c (can_mult_highpart_p): Likewise.
8748 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8749 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8750 (vectorizable_bswap): Likewise.
8751 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8752 explicit encoding for the power-of-2 permutes.
8753 (vect_permute_store_chain): Likewise.
8754 (vect_grouped_load_supported): Likewise.
8755 (vect_permute_load_chain): Likewise.
8757 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8759 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8760 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8761 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8762 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8763 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8764 (vect_gen_perm_mask_any): Likewise.
8766 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8768 * int-vector-builder.h: New file.
8769 * vec-perm-indices.h: Include int-vector-builder.h.
8770 (vec_perm_indices): Redefine as an int_vector_builder.
8771 (auto_vec_perm_indices): Delete.
8772 (vec_perm_builder): Redefine as a stand-alone class.
8773 (vec_perm_indices::vec_perm_indices): New function.
8774 (vec_perm_indices::clamp): Likewise.
8775 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8776 (vec_perm_indices::new_vector): New function.
8777 (vec_perm_indices::new_expanded_vector): Update for new
8778 vec_perm_indices class.
8779 (vec_perm_indices::rotate_inputs): New function.
8780 (vec_perm_indices::all_in_range_p): Operate directly on the
8781 encoded form, without computing elided elements.
8782 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8783 encoding. Update for new vec_perm_indices class.
8784 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8785 the given vec_perm_builder.
8786 (expand_vec_perm_var): Update vec_perm_builder constructor.
8787 (expand_mult_highpart): Use vec_perm_builder instead of
8788 auto_vec_perm_indices.
8789 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8790 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8791 or double series encoding as appropriate.
8792 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8793 vec_perm_indices instead of auto_vec_perm_indices.
8794 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8795 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8796 (vect_permute_store_chain): Likewise.
8797 (vect_grouped_load_supported): Likewise.
8798 (vect_permute_load_chain): Likewise.
8799 (vect_shift_permute_load_chain): Likewise.
8800 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8801 (vect_transform_slp_perm_load): Likewise.
8802 (vect_schedule_slp_instance): Likewise.
8803 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8804 (vectorizable_mask_load_store): Likewise.
8805 (vectorizable_bswap): Likewise.
8806 (vectorizable_store): Likewise.
8807 (vectorizable_load): Likewise.
8808 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8809 vec_perm_indices instead of auto_vec_perm_indices. Use
8810 tree_to_vec_perm_builder to read the vector from a tree.
8811 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8812 vec_perm_builder instead of a vec_perm_indices.
8813 (have_whole_vector_shift): Use vec_perm_builder and
8814 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8815 truncation to calc_vec_perm_mask_for_shift.
8816 (vect_create_epilog_for_reduction): Likewise.
8817 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8818 from auto_vec_perm_indices to vec_perm_indices.
8819 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8820 instead of changing individual elements.
8821 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8822 the vector in d.perm.
8823 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8824 from auto_vec_perm_indices to vec_perm_indices.
8825 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8826 instead of changing individual elements.
8827 (arm_vectorize_vec_perm_const): Use new_vector to install
8828 the vector in d.perm.
8829 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8830 Update vec_perm_builder constructor.
8831 (rs6000_expand_interleave): Likewise.
8832 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8833 (rs6000_expand_interleave): Likewise.
8835 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8837 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8838 to qimode could truncate the indices.
8839 * optabs.c (expand_vec_perm_var): Likewise.
8841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8843 * Makefile.in (OBJS): Add vec-perm-indices.o.
8844 * vec-perm-indices.h: New file.
8845 * vec-perm-indices.c: Likewise.
8846 * target.h (vec_perm_indices): Replace with a forward class
8848 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8849 * optabs.h: Include vec-perm-indices.h.
8850 (expand_vec_perm): Delete.
8851 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8852 (expand_vec_perm_const): Declare.
8853 * target.def (vec_perm_const_ok): Replace with...
8854 (vec_perm_const): ...this new hook.
8855 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8856 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8857 * doc/tm.texi: Regenerate.
8858 * optabs.def (vec_perm_const): Delete.
8859 * doc/md.texi (vec_perm_const): Likewise.
8860 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8861 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8862 expand_vec_perm for constant permutation vectors. Assert that
8863 the mode of variable permutation vectors is the integer equivalent
8864 of the mode that is being permuted.
8865 * optabs-query.h (selector_fits_mode_p): Declare.
8866 * optabs-query.c: Include vec-perm-indices.h.
8867 (selector_fits_mode_p): New function.
8868 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8869 is defined, instead of checking whether the vec_perm_const_optab
8870 exists. Use targetm.vectorize.vec_perm_const instead of
8871 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8872 fit in the vector mode before using a variable permute.
8873 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8874 vec_perm_indices instead of an rtx.
8875 (expand_vec_perm): Replace with...
8876 (expand_vec_perm_const): ...this new function. Take the selector
8877 as a vec_perm_indices rather than an rtx. Also take the mode of
8878 the selector. Update call to shift_amt_for_vec_perm_mask.
8879 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8880 Use vec_perm_indices::new_expanded_vector to expand the original
8881 selector into bytes. Check whether the indices fit in the vector
8882 mode before using a variable permute.
8883 (expand_vec_perm_var): Make global.
8884 (expand_mult_highpart): Use expand_vec_perm_const.
8885 * fold-const.c: Includes vec-perm-indices.h.
8886 * tree-ssa-forwprop.c: Likewise.
8887 * tree-vect-data-refs.c: Likewise.
8888 * tree-vect-generic.c: Likewise.
8889 * tree-vect-loop.c: Likewise.
8890 * tree-vect-slp.c: Likewise.
8891 * tree-vect-stmts.c: Likewise.
8892 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8894 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8895 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8896 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8897 (aarch64_vectorize_vec_perm_const): ...this new function.
8898 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8899 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8900 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8901 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8902 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8903 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8904 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8906 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8907 check for NEON modes.
8908 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8909 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8910 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8911 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8913 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8914 the old VEC_PERM_CONST conditions.
8915 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8916 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8917 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8918 (ia64_vectorize_vec_perm_const_ok): Merge into...
8919 (ia64_vectorize_vec_perm_const): ...this new function.
8920 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8921 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8922 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8923 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8924 * config/mips/mips.c (mips_expand_vec_perm_const)
8925 (mips_vectorize_vec_perm_const_ok): Merge into...
8926 (mips_vectorize_vec_perm_const): ...this new function.
8927 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8928 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8929 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8930 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8931 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8932 (rs6000_expand_vec_perm_const): Delete.
8933 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8935 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8936 (altivec_expand_vec_perm_const_le): Take each operand individually.
8937 Operate on constant selectors rather than rtxes.
8938 (altivec_expand_vec_perm_const): Likewise. Update call to
8939 altivec_expand_vec_perm_const_le.
8940 (rs6000_expand_vec_perm_const): Delete.
8941 (rs6000_vectorize_vec_perm_const_ok): Delete.
8942 (rs6000_vectorize_vec_perm_const): New function.
8943 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8944 an element count and rtx array.
8945 (rs6000_expand_extract_even): Update call accordingly.
8946 (rs6000_expand_interleave): Likewise.
8947 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8948 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8949 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8950 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8951 (rs6000_expand_vec_perm_const): Delete.
8952 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8953 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8954 (altivec_expand_vec_perm_const_le): Take each operand individually.
8955 Operate on constant selectors rather than rtxes.
8956 (altivec_expand_vec_perm_const): Likewise. Update call to
8957 altivec_expand_vec_perm_const_le.
8958 (rs6000_expand_vec_perm_const): Delete.
8959 (rs6000_vectorize_vec_perm_const_ok): Delete.
8960 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8961 reference to the SPE evmerge intructions.
8962 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8963 an element count and rtx array.
8964 (rs6000_expand_extract_even): Update call accordingly.
8965 (rs6000_expand_interleave): Likewise.
8966 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8967 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8969 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8973 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8974 vector mode and that that mode matches the mode of the data
8976 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8977 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8978 directly using expand_vec_perm_1 when forcing selectors into
8980 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8982 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8984 * optabs-query.h (can_vec_perm_p): Delete.
8985 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8986 * optabs-query.c (can_vec_perm_p): Split into...
8987 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8988 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8989 particular selector is valid.
8990 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8991 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8992 (vect_grouped_load_supported): Likewise.
8993 (vect_shift_permute_load_chain): Likewise.
8994 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8995 (vect_transform_slp_perm_load): Likewise.
8996 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8997 (vectorizable_bswap): Likewise.
8998 (vect_gen_perm_mask_checked): Likewise.
8999 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9000 implementations of variable permutation vectors into account
9001 when deciding which selector to use.
9002 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9003 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9004 with a false third argument.
9005 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9006 to test whether the constant selector is valid and can_vec_perm_var_p
9007 to test whether a variable selector is valid.
9009 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9011 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9012 * optabs-query.c (can_vec_perm_p): Likewise.
9013 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9014 instead of vec_perm_indices.
9015 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9016 (vect_gen_perm_mask_checked): Likewise,
9017 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9018 (vect_gen_perm_mask_checked): Likewise,
9020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9022 * optabs-query.h (qimode_for_vec_perm): Declare.
9023 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9024 (qimode_for_vec_perm): ...this new function.
9025 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9027 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9029 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9030 does not have a conditional at the top.
9032 2018-01-02 Richard Biener <rguenther@suse.de>
9034 * ipa-inline.c (big_speedup_p): Fix expression.
9036 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9039 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9042 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9046 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9047 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9048 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9049 cond_taken_branch_cost 3->4.
9051 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9053 PR tree-optimization/83581
9054 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9055 TODO_cleanup_cfg if any changes have been made.
9058 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9059 convert_modes if target mode has the right side, but different mode
9063 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9064 last argument when extracting from CONCAT. If either from_real or
9065 from_imag is NULL, use expansion through memory. If result is not
9066 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9067 the parts directly to inner mode, if even that fails, use expansion
9071 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9072 check for bswap in mode rather than HImode and use that in expand_unop
9075 Copyright (C) 2018 Free Software Foundation, Inc.
9077 Copying and distribution of this file, with or without modification,
9078 are permitted in any medium without royalty provided the copyright
9079 notice and this notice are preserved.