1 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6 Kito Cheng <kito.cheng@gmail.com>
8 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10 (nds32_adjust_reg_alloc_order): New function.
11 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
13 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16 nds32_print_operand, nds32_print_operand_address): Use
17 HOST_WIDE_INT_PRINT_DEC instead.
19 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
21 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23 2018-03-17 Jakub Jelinek <jakub@redhat.com>
26 * config/i386/i386.c (initial_ix86_tune_features,
27 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
29 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
30 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
31 rather than 1u << ix86_tune. Formatting fix.
32 (ix86_option_override_internal): Change ix86_arch_mask from
33 unsigned int to unsigned HOST_WIDE_INT, initialize to
34 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
35 (ix86_function_specific_restore): Likewise.
37 2018-03-16 Jakub Jelinek <jakub@redhat.com>
40 * postreload.c (reload_combine_recognize_pattern): Perform
41 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
42 truncate_int_for_mode the result for the destination's mode.
45 * hsa-gen.c (mem_type_for_type): Fix comment typo.
46 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
48 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
51 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
54 * lra-assigns.c (lra_split_hard_reg_for): Don't use
55 regno_allocno_class_array and sorted_pseudos.
56 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
57 insns where regno is used.
59 2018-03-16 Martin Liska <mliska@suse.cz>
62 * multiple_target.c (create_dispatcher_calls): Redirect
63 reference in the symbol table.
65 2018-03-16 Martin Liska <mliska@suse.cz>
68 * multiple_target.c (create_dispatcher_calls): Redirect also
71 2018-03-16 Jakub Jelinek <jakub@redhat.com>
75 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
76 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
77 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
79 2018-03-16 Julia Koval <julia.koval@intel.com>
81 * doc/invoke.texi (Skylake Server): Add CLWB.
82 Cannonlake): Remove CLWB.
84 2018-03-16 Jakub Jelinek <jakub@redhat.com>
86 PR tree-optimization/84841
87 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
89 (FLOAT_ONE_CONST_TYPE): Define.
90 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
91 (sort_by_operand_rank): Put entries with higher constant_type last
92 rather than first to match comments.
94 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
96 * config/nios2/nios2.md (movsi_internal): Fix thinko in
99 2018-03-15 Jakub Jelinek <jakub@redhat.com>
102 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
103 check and use address of target always.
105 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
108 * config/i386/i386.c (indirect_thunk_needed): Update comments.
109 (indirect_thunk_bnd_needed): Likewise.
110 (indirect_thunks_used): Likewise.
111 (indirect_thunks_bnd_used): Likewise.
112 (indirect_return_needed): New.
113 (indirect_return_bnd_needed): Likewise.
114 (output_indirect_thunk_function): Add a bool argument for
116 (output_indirect_thunk_function): Don't generate alias for
117 function return thunk.
118 (ix86_code_end): Call output_indirect_thunk_function to generate
119 function return thunks.
120 (ix86_output_function_return): Set indirect_return_bnd_needed
121 and indirect_return_needed instead of indirect_thunk_bnd_needed
122 and indirect_thunk_needed.
124 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
126 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
127 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
128 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
130 2018-03-15 David Malcolm <dmalcolm@redhat.com>
131 Paul Hua <paul.hua.gm@gmail.com>
134 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
136 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
138 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
139 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
142 2018-03-15 Tamar Christina <tamar.christina@arm.com>
145 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
146 instead of GET_MODE_SIZE when comparing Units.
148 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
151 * varasm.c (hash_section): Return an unchangeble hash value
152 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
153 Return !aarch64_can_use_per_function_literal_pools_p ().
155 2018-03-15 Jakub Jelinek <jakub@redhat.com>
158 * optabs.c (emit_conditional_move): Pass address of cmode's copy
159 rather than address of cmode as last argument to prepare_cmp_insn.
161 2018-03-15 Julia Koval <julia.koval@intel.com>
163 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
164 F_AVX512VNNI, F_AVX512BITALG): New.
166 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
169 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
170 insn for floating-point loads and stores.
172 2018-03-14 Carl Love <cel@us.ibm.com>
174 * config/rs6000/rs6000-c.c: Add macro definitions for
175 ALTIVEC_BUILTIN_VEC_PERMXOR.
176 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
177 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
178 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
179 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
181 * config/doc/extend.texi: Add prototypes for vec_permxor.
183 2018-03-14 David Malcolm <dmalcolm@redhat.com>
186 * diagnostic-show-locus.c (class layout_point): Convert m_line
187 from int to linenum_type.
188 (line_span::comparator): Use linenum "compare" function when
189 comparing line numbers.
190 (test_line_span): New function.
191 (layout_range::contains_point): Convert param "row" from int to
193 (layout_range::intersects_line_p): Likewise.
194 (layout::will_show_line_p): Likewise.
195 (layout::print_source_line): Likewise.
196 (layout::should_print_annotation_line_p): Likewise.
197 (layout::print_annotation_line): Likewise.
198 (layout::print_leading_fixits): Likewise.
199 (layout::annotation_line_showed_range_p): Likewise.
200 (struct line_corrections): Likewise for field m_row.
201 (line_corrections::line_corrections): Likewise for param "row".
202 (layout::print_trailing_fixits): Likewise.
203 (layout::get_state_at_point): Likewise.
204 (layout::get_x_bound_for_row): Likewise.
205 (layout::print_line): Likewise.
206 (diagnostic_show_locus): Likewise for locals "last_line" and
208 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
209 * input.c (selftest::test_linenum_comparisons): New function.
210 (selftest::input_c_tests): Call it.
211 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
212 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
213 * selftest.h (ASSERT_GT): New macro.
214 (ASSERT_GT_AT): New macro.
215 (ASSERT_LT): New macro.
216 (ASSERT_LT_AT): New macro.
218 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
220 PR rtl-optimization/84780
221 * combine.c (distribute_links): Don't make a link based on pc_rtx.
223 2018-03-14 Martin Liska <mliska@suse.cz>
225 * tree.c (record_node_allocation_statistics): Use
227 (get_stats_node_kind): New function extracted from
228 record_node_allocation_statistics.
229 (free_node): Use get_stats_node_kind.
231 2018-03-14 Richard Biener <rguenther@suse.de>
233 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
234 that the value-set of ANTIC_IN doesn't grow.
237 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
239 (BB_VISITED_WITH_VISITED_SUCCS): New define.
240 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
242 2018-03-14 Julia Koval <julia.koval@intel.com>
244 * config.gcc (icelake-client, icelake-server): New.
246 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
247 (initial_ix86_arch_features): Ditto.
248 (PTA_SKYLAKE): Add SGX.
249 (PTA_ICELAKE): Remove.
250 (PTA_ICELAKE_CLIENT): New.
251 (PTA_ICELAKE_SERVER): New.
252 (ix86_option_override_internal): Split up icelake on icelake client and
254 (get_builtin_code_for_version): Ditto.
255 (fold_builtin_cpu): Ditto.
256 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
257 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
258 * config/i386/i386.h (processor_type): Ditto.
259 * doc/invoke.texi: Ditto.
261 2018-03-14 Jakub Jelinek <jakub@redhat.com>
264 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
265 INTEGER_CST offset, add it together with bitpos / 8 and
266 sign extend based on POINTER_SIZE.
270 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
273 * config/i386/constraints.md (Yc): New register constraint.
274 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
275 Use Yc constraint for alternative 2 of operand 0. Remove
276 preferred_for_speed attribute.
278 2018-03-14 Richard Biener <rguenther@suse.de>
280 PR tree-optimization/84830
281 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
282 with the old one to avoid oscillations.
284 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
287 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
289 (assign_by_spills): Return a flag of reload assignment failure.
290 Do not process the reload assignment failures. Do not spill other
291 reload pseudos if they has the same reg class. Update n if
293 (lra_assign): Add a return arg. Set up from the result of
294 assign_by_spills call.
295 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
296 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
297 usage_insns if it is not NULL.
298 (spill_hard_reg_in_range): New function.
299 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
300 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
302 (lra_assign): Change prototype.
303 * lra.c (lra): Add code to deal with fails by splitting hard reg
306 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
308 * config/riscv/riscv.opt (mrelax): New option.
309 * config/riscv/riscv.c (riscv_file_start): Emit ".option
310 "norelax" when riscv_mrelax is disabled.
311 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
313 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
316 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
317 reassociation for int modes.
319 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
321 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
322 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
324 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
325 * config/aarch64/aarch64-sve.md
326 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
327 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
328 (*extend<mode><Vwide>2): Rename to...
329 (aarch64_sve_extend<mode><Vwide>2): ...this.
330 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
331 renaming the old pattern to...
332 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
334 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
335 define_expand, renaming the old pattern to...
336 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
337 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
338 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
339 account when deciding which SVE instruction the optab should use.
340 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
342 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
344 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
345 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
346 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
347 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
348 (tlsdesc_small_<mode>): Turn a define_expand and use
349 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
350 (tlsdesc_small_advsimd_<mode>): ...this.
351 (tlsdesc_small_sve_<mode>): New pattern.
353 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
355 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
356 (UNSPEC_UMUL_HIGHPART): New constants.
357 (MUL_HIGHPART): New int iteraor.
358 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
359 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
361 (*<su>mul<mode>3_highpart): New define_insn.
363 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
366 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
369 2018-03-13 Martin Liska <mliska@suse.cz>
372 * (sem_item_optimizer::sem_item_optimizer): Initialize new
374 (sem_item_optimizer::~sem_item_optimizer): Release it.
375 (sem_item_optimizer::merge_classes): Register variable aliases.
376 (sem_item_optimizer::fixup_pt_set): New function.
377 (sem_item_optimizer::fixup_points_to_sets): Likewise.
378 * ipa-icf.h: Declare new variables and functions.
380 2018-03-13 Jakub Jelinek <jakub@redhat.com>
383 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
384 integer_pow2p@2 and test integer_pow2p in condition.
385 (A < 0 ? C : 0): Similarly for @1.
388 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
389 characters starting at p contain '\0' character, don't look beyond
393 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
394 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
397 * reg-stack.c (change_stack): Change update_end var from int to
398 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
399 also call set_block_for_insn on the newly added insns and rescan.
402 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
406 * tree.c (stabilize_reference_1): Return save_expr (e) for
407 STATEMENT_LIST even if it doesn't have side-effects.
409 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
411 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
413 2018-03-12 Renlin Li <renlin.li@arm.com>
415 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
416 aarch64_output_scalar_simd_mov_immediate.
418 2018-03-12 Martin Sebor <msebor@redhat.com>
420 PR tree-optimization/83456
421 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
422 for perfectly overlapping calls to memcpy.
423 (gimple_fold_builtin_memory_chk): Same.
424 (gimple_fold_builtin_strcpy): Handle no-warning.
425 (gimple_fold_builtin_stxcpy_chk): Same.
426 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
428 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
430 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
431 parameter. Use it for SFmode.
432 (rs6000_function_arg_advance_1): Adjust.
433 (rs6000_function_arg): Adjust.
434 (rs6000_gimplify_va_arg): Pass false for that new parameter.
436 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
438 PR rtl-optimization/84169
439 PR rtl-optimization/84780
440 * combine.c (can_combine_p): Check for a 2-insn combination whether
441 the destination register is used between the two insns, too.
443 2018-03-12 Richard Biener <rguenther@suse.de>
445 PR tree-optimization/84803
446 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
447 for refs DR analysis didn't process.
449 2018-03-12 Richard Biener <rguenther@suse.de>
451 PR tree-optimization/84777
452 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
453 force-vectorize loops ignore whether we are optimizing for size.
455 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
457 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
458 (TARGET_MD_ASM_ADJUST): Define.
460 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
461 Kito Cheng <kito.cheng@gmail.com>
462 Chung-Ju Wu <jasonwucj@gmail.com>
464 * config/nds32/nds32.c (nds32_compute_stack_frame,
465 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
466 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
467 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
468 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
469 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
470 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
471 * config/nds32/nds32.md (prologue, epilogue): Use macro
472 NDS32_V3PUSH_AVAILABLE_P to do checking.
474 2018-03-11 Jakub Jelinek <jakub@redhat.com>
477 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
478 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
479 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
480 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
481 addition of most attributes on !orig_type_die or the attribute not
482 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
484 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
485 Chung-Ju Wu <jasonwucj@gmail.com>
487 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
489 * config/nds32/nds32.opt (mvh): New option.
491 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
492 Chung-Ju Wu <jasonwucj@gmail.com>
494 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
496 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
497 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
500 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
501 Chung-Ju Wu <jasonwucj@gmail.com>
503 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
505 * config/nds32/nds32-multiple.md (strlensi): New pattern.
506 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
508 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
509 Kito Cheng <kito.cheng@gmail.com>
510 Chung-Ju Wu <jasonwucj@gmail.com>
512 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
513 UNSPEC_FFMISM and UNSPEC_FLMISM.
514 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
515 for ffb, ffmism and flmism.
516 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
517 (unspec_ffmism): Ditto.
518 (unspec_flmism): Ditto.
519 (nds32_expand_builtin_impl): Check if string extension is available.
520 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
521 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
523 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
526 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
529 * lra-assigns.c (assign_by_spills): Return a flag of reload
530 assignment failure. Do not process the reload assignment
531 failures. Do not spill other reload pseudos if they has the same
533 (lra_assign): Add a return arg. Set up from the result of
534 assign_by_spills call.
535 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
536 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
537 usage_insns if it is not NULL.
538 (spill_hard_reg_in_range): New function.
539 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
540 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
542 (lra_assign): Change prototype.
543 * lra.c (lra): Add code to deal with fails by splitting hard reg
546 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
549 * config/i386/i386.opt: Replace Enforcment with Enforcement.
551 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
554 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
555 (dw_val_node): Add val_symbolic_view.
556 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
557 (symview_upper_bound): New.
558 (new_line_info_table): Initialize symviews_since_reset.
559 (dwarf2out_source_line): Count symviews_since_reset and set
561 (dw_val_equal_p): Handle symview.
562 (add_AT_symview): New.
563 (print_dw_val): Handle symview.
564 (attr_checksum, attr_checksum_ordered): Likewise.
565 (same_dw_val_p, size_of_die): Likewise.
566 (value_format, output_die): Likewise.
567 (add_high_low_attributes): Use add_AT_symview for entry_view.
568 (dwarf2out_finish): Reset symview_upper_bound, clear
571 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
574 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
575 Add strict argument and use it.
576 (rs6000_split_multireg_move): Update for new strict argument.
577 (mem_operand_gpr): Disallow all non-offsettable addresses.
578 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
580 2018-03-09 Jakub Jelinek <jakub@redhat.com>
583 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
584 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
585 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
588 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
589 decl, use remap_type if we want to use the type.
591 2018-03-09 Martin Sebor <msebor@redhat.com>
593 PR tree-optimization/84526
594 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
596 (builtin_access::generic_overlap): Be prepared to handle non-array
599 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
601 PR rtl-optimization/84682
602 * lra-constraints.c (process_address_1): Check is_address flag
603 for address constraints.
604 (process_alt_operands): Likewise.
605 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
606 preprocess_constraints.
607 * recog.h (preprocess_constraints): Add oploc parameter.
609 * recog.c (preprocess_constraints): Test address_operand for
610 CT_ADDRESS constraints.
612 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
615 * lra-assigns.c (assign_by_spills): Return a flag of reload
616 assignment failure. Do not process the reload assignment
617 failures. Do not spill other reload pseudos if they has the same
619 (lra_assign): Add a return arg. Set up from the result of
620 assign_by_spills call.
621 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
622 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
623 usage_insns if it is not NULL.
624 (spill_hard_reg_in_range): New function.
625 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
626 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
628 (lra_assign): Change prototype.
629 * lra.c (lra): Add code to deal with fails by splitting hard reg
632 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
635 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
636 Accept complain bool parameter. Only emit errors if it is true.
637 (arm_parse_cpu_option_name): Likewise.
638 (arm_target_thumb_only): Adjust callers of the above.
639 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
640 prototype to take a default true bool parameter.
641 (arm_parse_arch_option_name): Likewise.
643 2018-03-09 David Malcolm <dmalcolm@redhat.com>
644 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
648 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
649 * configure: Regenerate.
650 * configure.ac ("linker --version-script option"): New.
651 ("linker soname option"): New.
653 2018-03-09 Richard Biener <rguenther@suse.de>
655 PR tree-optimization/84775
656 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
657 immediate uses of predicate stmts and mark them modified.
660 PR tree-optimization/84178
661 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
663 (version_loop_for_if_conversion): Delay update_ssa call.
664 (tree_if_conversion): Delay update_ssa until after predicate
667 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
670 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
671 when the function accesses prior frames.
673 2018-03-08 Jakub Jelinek <jakub@redhat.com>
676 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
677 gen_llsym, otherwise call maybe_gen_llsym.
680 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
681 has ',' character inside of it.
683 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
686 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
687 as clobbering CC_REGNUM.
689 2018-03-08 Richard Biener <rguenther@suse.de>
692 * tree-scalar-evolution.c: Include tree-into-ssa.h.
693 (follow_copies_to_constant): Do not follow SSA names registered
696 2018-03-08 Richard Biener <rguenther@suse.de>
698 PR tree-optimization/84178
699 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
701 (version_loop_for_if_conversion): Delay update_ssa call.
702 (tree_if_conversion): Delay update_ssa until after predicate
705 2018-03-08 David Malcolm <dmalcolm@redhat.com>
707 PR tree-optimization/84178
708 * tree-if-conv.c (release_bb_predicate): Remove the
709 the assertion that the stmts have NULL use_ops.
710 Discard the statements, asserting that they haven't
711 yet been added to a BB.
713 2018-03-08 Richard Biener <rguenther@suse.de>
715 PR tree-optimization/84746
716 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
717 (phi_translate): Pass in destination ANTIC_OUT set.
718 (phi_translate_1): Likewise. For a simplified result lookup
719 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
720 (phi_translate_set): Adjust.
721 (do_pre_regular_insertion): Likewise.
722 (do_pre_partial_partial_insertion): Likewise.
724 2018-03-08 Martin Liska <mliska@suse.cz>
726 PR gcov-profile/84735
727 * doc/gcov.texi: Document usage of profile files.
728 * gcov-io.h: Document changes in the format.
730 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
734 * dwarf2out.c (struct dw_line_info_table): Update comments for
736 (FORCE_RESET_NEXT_VIEW): New.
737 (FORCE_RESETTING_VIEW_P): New.
738 (RESETTING_VIEW_P): Check for -1 too.
739 (ZERO_VIEW_P): Likewise.
740 (new_line_info_table): Force-reset next view.
741 (dwarf2out_begin_function): Likewise.
742 (dwarf2out_source_line): Simplify zero_view_p initialization.
743 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
744 view directly. Omit view when omitting .loc at line 0.
746 2018-03-08 Jakub Jelinek <jakub@redhat.com>
748 PR tree-optimization/84740
749 * tree-switch-conversion.c (process_switch): Call build_constructors
750 only if info.phi_count is non-zero.
752 PR tree-optimization/84739
753 * tree-tailcall.c (find_tail_calls): Check call arguments against
754 DECL_ARGUMENTS (current_function_decl) rather than
755 DECL_ARGUMENTS (func) when checking for tail recursion.
757 2018-03-07 Jakub Jelinek <jakub@redhat.com>
759 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
760 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
761 Volker Reichelt's entry and add entries for people that perform
762 GCC fuzzy testing and report numerous bugs.
764 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
767 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
768 readonly data in sdata, if that is disabled.
769 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
770 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
771 -mreadonly-in-sdata option.
773 2018-03-07 Martin Sebor <msebor@redhat.com>
775 PR tree-optimization/84468
776 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
777 basic block when looking for nul assignment.
779 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
782 * except.h (output_function_exception_table): Adjust prototype.
783 * except.c (output_function_exception_table): Remove FNNAME parameter
784 and add SECTION parameter. Ouput one part of the table at a time.
785 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
786 the first part of the exception table and emit unwind directives.
787 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
788 (i386_pe_seh_cold_init): Likewise.
789 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
790 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
791 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
792 (ix86_output_call_insn): Emit a nop in one more case for SEH.
793 * config/i386/winnt.c: Include except.h.
794 (struct seh_frame_state): Add reg_offset, after_prologue and
795 in_cold_section fields.
796 (i386_pe_seh_end_prologue): Set seh->after_prologue.
797 (i386_pe_seh_cold_init): New function.
798 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
799 to seh->in_cold_section.
800 (seh_emit_push): Record the offset of the push.
801 (seh_emit_save): Record the offet of the save.
802 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
803 Test seh->after_prologue to disregard the epilogue.
804 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
805 (i386_pe_end_cold_function): New function.
807 2018-03-07 Jakub Jelinek <jakub@redhat.com>
810 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
811 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
814 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
816 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
817 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
820 * multiple_target.c: Include tree-inline.h and intl.h.
821 (expand_target_clones): Diagnose and fail if node->definition and
822 !tree_versionable_function_p (node->decl).
824 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
826 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
828 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
829 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
830 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
832 2018-03-06 Jakub Jelinek <jakub@redhat.com>
835 * combine.c (try_combine): Use reg_or_subregno instead of handling
836 just paradoxical SUBREGs and REGs.
838 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
840 * config/arc/arc.c (arc_finalize_pic): Remove function.
841 (arc_must_save_register): We use single base PIC register, remove
842 checks to save/restore the PIC register.
843 (arc_expand_prologue): Likewise.
844 * config/arc/arc-protos.h (arc_set_default_type_attributes):
846 (arc_verify_short): Likewise.
847 (arc_attr_type): Likewise.
848 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
849 (walk_stores): Likewise.
850 (arc_address_cost): Make it static.
851 (arc_verify_short): Likewise.
852 (branch_dest): Likewise.
853 (arc_attr_type): Likewise.
854 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
855 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
856 (arc_final_prescan_insn): Remove inserting the nops due to
857 hardware hazards. It is done in reorg step.
858 (insn_length_variant_t): Remove.
859 (insn_length_parameters_t): Likewise.
860 (arc_insn_length_parameters): Likewise.
861 (arc_get_insn_variants): Likewise.
862 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
864 2018-03-06 Jakub Jelinek <jakub@redhat.com>
867 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
870 PR tree-optimization/84687
871 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
873 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
875 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
877 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
878 Rename to ppc_speculation_barrier.
879 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
880 __builtin_ppc_speculation_barrier.
882 2018-03-05 Jakub Jelinek <jakub@redhat.com>
885 * combine.c (combine_simplify_rtx): Don't try to simplify if
886 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
889 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
891 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
892 to 32 bytes when compiling for POWER9.
894 2018-03-05 Jakub Jelinek <jakub@redhat.com>
897 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
898 regparm >= 3 with no arg reg available also for calls with
899 flag_force_indirect_call. Pass decl to ix86_function_regparm.
902 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
904 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
906 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
909 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
911 2018-03-05 Richard Biener <rguenther@suse.de>
913 PR tree-optimization/84486
914 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
915 When inserting a __builtin_assume_aligned call set the LHS
916 SSA name alignment info accordingly.
918 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
920 PR tree-optimization/84114
921 * config/aarch64/aarch64.c (aarch64_reassociation_width)
922 Avoid reassociation of FLOAT_MODE addition.
924 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
926 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
927 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
928 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
929 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
930 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
931 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
932 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
934 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
935 __builtin_ia32_wbinvd): New builtins.
936 (SPECIAL_ARGS2): New.
937 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
938 (SPECIAL_ARGS2): New.
939 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
940 (ix86_valid_target_attribute_inner_p): Ditto.
941 (ix86_init_mmx_sse_builtins): Add special_args2.
942 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
943 TARGET_WBNOINVD_P): New.
944 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
945 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
946 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
947 * config/i386/immintrin.h (_wbinvd): New intrinsic.
948 * config/i386/pconfigintrin.h: New file.
949 * config/i386/wbnoinvdintrin.h: Ditto.
950 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
951 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
953 2018-03-05 Richard Biener <rguenther@suse.de>
955 PR tree-optimization/84670
956 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
958 (BB_VISITED_WITH_VISITED_SUCCS): New define.
959 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
960 (compute_antic_aux): Only assert the number of values in ANTIC_IN
961 doesn't grow if all successors (recursively) were visited at least
964 2018-03-05 Richard Biener <rguenther@suse.de>
966 PR tree-optimization/84650
967 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
968 if executed in the loop pipeline.
970 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
972 * doc/configfiles.texi (Configuration Files): Move info about
973 conditionalizing $target-protos.h to...
974 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
975 differs from $target-protos.h.
977 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
978 Chung-Ju Wu <jasonwucj@gmail.com>
980 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
981 * config/nds32/nds32-multiple.md (setmemsi): Define.
982 * config/nds32/nds32-memory-manipulation.c
983 (nds32_gen_dup_4_byte_to_word_value): New.
984 (emit_setmem_word_loop): New.
985 (emit_setmem_byte_loop): New.
986 (nds32_expand_setmem_loop): New.
987 (nds32_expand_setmem_loop_v3m): New.
988 (nds32_expand_setmem_unroll): New.
989 (nds32_expand_setmem): New.
991 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
992 Chung-Ju Wu <jasonwucj@gmail.com>
994 * config/nds32/nds32-memory-manipulation.c
995 (nds32_emit_load_store): New.
996 (nds32_emit_post_inc_load_store): New.
997 (nds32_emit_mem_move): New.
998 (nds32_emit_mem_move_block): New.
999 (nds32_expand_movmemsi_loop_unknown_size): New.
1000 (nds32_expand_movmemsi_loop_known_size): New.
1001 (nds32_expand_movmemsi_loop): New.
1002 (nds32_expand_movmemsi_unroll): New.
1003 (nds32_expand_movmemqi): Rename ...
1004 (nds32_expand_movmemsi): ... to this.
1005 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1006 (movmemsi): ... to this.
1007 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1008 (nds32_expand_movmemsi): ... to this.
1010 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1011 Monk Chiang <sh.chiang04@gmail.com>
1012 Chung-Ju Wu <jasonwucj@gmail.com>
1014 * config/nds32/nds32-protos.h
1015 (nds32_expand_load_multiple): New arguments.
1016 (nds32_expand_store_multiple): Ditto.
1017 (nds32_valid_multiple_load_store): Rename ...
1018 (nds32_valid_multiple_load_store_p): ... to this.
1019 * config/nds32/nds32-memory-manipulation.c
1020 (nds32_expand_load_multiple): Refine implementation.
1021 (nds32_expand_store_multiple): Ditto.
1022 * config/nds32/nds32-multiple.md
1023 (load_multiple): Update nds32_expand_load_multiple interface.
1024 (store_multiple): Update nds32_expand_store_multiple interface.
1025 * config/nds32/nds32-predicates.c
1026 (nds32_valid_multiple_load_store): Rename ...
1027 (nds32_valid_multiple_load_store_p): ... to this and refine
1029 * config/nds32/predicates.md
1030 (nds32_load_multiple_and_update_address_operation): New predicate.
1031 (nds32_store_multiple_and_update_address_operation): New predicate.
1033 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1034 Chung-Ju Wu <jasonwucj@gmail.com>
1036 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1037 (combo): New attribute.
1038 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1040 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1042 * config/nds32/nds32.opt: Change -mcmodel= default value.
1044 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1045 Monk Chiang <sh.chiang04@gmail.com>
1046 Chung-Ju Wu <jasonwucj@gmail.com>
1048 * config/nds32/constants.md (unspec_element): New enum.
1049 * config/nds32/constraints.md (Umw): New constraint.
1050 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1051 * config/nds32/nds32-intrinsic.md: Likewise.
1052 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1053 (nds32_valid_smw_lwm_base_p): New.
1054 (nds32_output_smw_single_word): New.
1055 (nds32_output_lmw_single_word): New.
1056 (nds32_expand_unaligned_load): New.
1057 (nds32_expand_unaligned_store): New.
1058 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1059 (nds32_output_smw_single_word): Declare.
1060 (nds32_output_lmw_single_word): Declare.
1061 (nds32_expand_unaligned_load): Declare.
1062 (nds32_expand_unaligned_store): Declare.
1063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1064 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1065 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1066 NDS32_BUILTIN_UASTORE_DW.
1067 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1070 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1071 Kito Cheng <kito.cheng@gmail.com>
1072 Chung-Ju Wu <jasonwucj@gmail.com>
1074 * config/nds32/nds32-intrinsic.c
1075 (nds32_expand_builtin_null_ftype_reg): Delete.
1076 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1077 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1078 (nds32_read_argument): New.
1079 (nds32_legitimize_target): Ditto.
1080 (nds32_legitimize_argument): Ditto.
1081 (nds32_check_constant_argument): Ditto.
1082 (nds32_expand_unop_builtin): Ditto.
1083 (nds32_expand_unopimm_builtin): Ditto.
1084 (nds32_expand_binop_builtin): Ditto.
1085 (nds32_builtin_decl_impl): Ditto.
1086 (builtin_description): Ditto.
1087 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1088 (nds32_init_builtins_impl): Ditto.
1089 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1090 (nds32_builtin_decl): New.
1091 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1092 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1094 2018-03-02 Jeff Law <law@redhat.com>
1096 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1097 (redundant_insn, fill_simple_delay_slots): Likewise.
1098 (fill_slots_from_thread): Likewise.
1099 * resource.c (mark_referenced_resources): Likewise.
1100 (mark_set_resources, find_dead_or_set_registers): Likewise.
1102 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1104 * substring-locations.h (format_warning_va): Formatting fix for
1106 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1108 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1109 * substring-locations.c: Include intl.h.
1110 (format_warning_va): Turned into small wrapper around
1111 format_warning_n_va, renamed to ...
1112 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1113 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1115 (format_warning_at_substring_n): New function.
1116 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1117 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1118 format_warning_at_substring with just a shorter name instead of
1119 const function pointer.
1120 (fmtwarn_n): New function.
1121 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1122 appropriate, get rid of all the fmtstr temporaries, move conditionals
1123 with G_() wrapped string literals directly into fmtwarn arguments,
1124 cast dir.len to (int), formatting fixes.
1126 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1128 * doc/invoke.texi: Remove "Cilk Plus" references.
1130 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1131 Richard Biener <rguenther@suse.de>
1134 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1135 for error or warning attributes if CALL_FROM_THUNK_P is set.
1138 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1141 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1142 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1145 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1146 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1148 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1149 instead of -1U in last predictors element's probability member.
1151 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1154 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1155 arguments if they are comparable.
1157 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1159 PR tree-optimization/84634
1160 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1161 masks and masked_loop_p with a single loop_masks, making sure it's
1162 null for bb vectorization.
1164 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1166 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1167 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1168 loop->force_vectorize to check whether there is no alias.
1170 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1173 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1175 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1177 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1178 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1179 instead of a loop around prev_real_insn.
1180 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1184 * config/i386/i386.c (ix86_print_operand): Use conditional
1185 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1188 2018-03-02 Richard Biener <rguenther@suse.de>
1190 PR tree-optimization/84427
1191 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1192 (bitmap_set_subtract_values): Rewrite to handle multiple
1195 (prune_clobbered_mems): Likewise.
1196 (phi_translate): Take edge instead of pred/phiblock.
1197 (phi_translate_1): Likewise.
1198 (phi_translate_set): Likewise. Insert all translated
1199 exprs for a value into the set, keeping possibly multiple
1200 expressions per value.
1201 (compute_antic_aux): Adjust for phi_translate changes.
1202 When intersecting union the expressions and prune those
1203 not in the final value set, keeping possibly multiple
1204 expressions per value. Do not use value-insertion
1205 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1206 all expressions. Add verification that the value-sets
1207 only shrink during iteration.
1208 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1209 (do_pre_regular_insertion): Likewise.
1210 (do_pre_partial_partial_insertion): Likewise.
1212 2018-03-02 Richard Biener <rguenther@suse.de>
1215 * config/darwin.c (saved_debug_info_level): New static global.
1216 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1217 (darwin_asm_lto_end): Restore debug info generation settings.
1219 2018-03-01 Martin Liska <mliska@suse.cz>
1222 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1225 2018-03-01 Richard Biener <rguenther@suse.de>
1228 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1229 type annotation with LTO when debug was disabled at compile-time.
1231 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1233 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1235 (mips_final_postscan_insn): Likewise.
1237 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1239 PR rtl-optimization/84528
1240 * alias.c (init_alias_target): Add commentary.
1241 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1242 a unique base value if the frame pointer is not eliminated
1243 to the stack pointer.
1245 2018-03-01 Tom de Vries <tom@codesourcery.com>
1247 PR rtl-optimization/83327
1248 * lra-int.h (hard_regs_spilled_into): Declare.
1249 * lra.c (hard_regs_spilled_into): Define.
1250 (init_reg_info): Init hard_regs_spilled_into.
1251 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1252 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1253 (process_bb_lives): Handle hard_regs_spilled_into.
1254 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1255 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1257 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1259 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1260 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1261 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1262 * config/rs6000/aix72.h: New file.
1264 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1266 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1267 instead of warning_at with conditional singular and plural messages
1271 * stor-layout.c (update_alignment_for_field): For
1272 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1273 && !DECL_PACKED (field), do the alignment update, just use
1274 only desired_align instead of MAX (type_align, desired_align)
1276 (place_field): Don't do known_align < desired_align handling
1277 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1278 is non-NULL, instead do it after rli->prev_field handling and
1279 only if not within a bitfield word. For DECL_PACKED (field)
1280 use type_align of BITS_PER_UNIT.
1282 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1284 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1285 superfluous parentheses and trailing spaces.
1287 2018-02-28 Richard Biener <rguenther@suse.de>
1289 PR tree-optimization/84584
1290 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1291 SCoPs with fake exit edge.
1293 2018-02-28 Martin Liska <mliska@suse.cz>
1296 * timevar.c (timer::print): Fix format to properly print 100%
1299 2018-02-28 Richard Biener <rguenther@suse.de>
1302 * genmatch.c (capture_info::walk_match): Do not mark
1303 captured expressions without operands as expr_p given
1304 they act more like predicates and should be subject to
1305 "lost tail" side-effect preserving.
1307 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1309 PR rtl-optimization/81611
1310 * auto-inc-dec.c (attempt_change): Move dead note from
1311 mem_insn if it's the next use of regno
1312 (find_address): Take address use of reg holding
1313 non-incremented value. Add parm to limit search to the named
1315 (merge_in_block): Attempt to use a mem insn that is the next
1316 use of the original regno.
1318 2018-02-27 Martin Sebor <msebor@redhat.com>
1321 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1322 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1324 2018-02-27 Martin Sebor <msebor@redhat.com>
1326 PR translation/84207
1327 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1328 n argument to unsigned HOST_WIDE_INT.
1329 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1330 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1331 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1332 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1334 2018-02-27 Richard Biener <rguenther@suse.de>
1336 PR tree-optimization/84512
1337 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1338 Do not use the estimate returned from record_stmt_cost for
1339 the scalar iteration cost but sum properly using add_stmt_cost.
1341 2018-02-27 Richard Biener <rguenther@suse.de>
1343 PR tree-optimization/84466
1344 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1345 Adjust last change to less strictly validate use operands.
1347 2018-02-27 Martin Liska <mliska@suse.cz>
1349 PR gcov-profile/84548
1350 * gcov.c (process_file): Allow partial overlap and consider it
1351 also as group functions.
1352 (output_lines): Properly calculate range of lines for a group.
1354 2018-02-27 Martin Liska <mliska@suse.cz>
1356 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
1357 'ggc' suffixes. Change first column width.
1358 (timer::print): Fix formatting of the column.
1360 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
1362 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
1363 preserve inline entry blocks for the sake of debug inline
1364 entry point markers alone.
1365 (remove_unused_locals): Suggest in comments a better place to
1366 force the preservation of inline entry blocks that are
1367 otherwise unused, but do not preserve them.
1369 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1371 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
1373 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1376 * config/i386/constraints.md (Bs): Replace
1377 ix86_indirect_branch_register with
1378 TARGET_INDIRECT_BRANCH_REGISTER.
1380 * config/i386/i386.md (indirect_jump): Likewise.
1381 (tablejump): Likewise.
1382 (*sibcall_memory): Likewise.
1383 (*sibcall_value_memory): Likewise.
1384 Peepholes of indirect call and jump via memory: Likewise.
1385 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
1386 (*sibcall_value_GOT_32): Likewise.
1387 * config/i386/predicates.md (indirect_branch_operand): Likewise.
1388 (GOT_memory_operand): Likewise.
1389 (call_insn_operand): Likewise.
1390 (sibcall_insn_operand): Likewise.
1391 (GOT32_symbol_operand): Likewise.
1392 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
1394 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1396 PR rtl-optimization/83496
1397 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
1398 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
1399 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
1400 redundant insn, if any.
1401 (relax_delay_slots): Likewise.
1402 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
1404 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
1406 PR tree-optimization/83965
1407 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
1408 that grouped statements are part of a reduction chain. Return
1409 true if the statement is not marked as a reduction itself but
1411 (vect_recog_dot_prod_pattern): Don't check whether the statement
1412 is part of a group here.
1413 (vect_recog_sad_pattern): Likewise.
1414 (vect_recog_widen_sum_pattern): Likewise.
1416 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
1419 * final.c (rest_of_clean_state): Also look for calls inside sequences.
1421 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
1424 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
1426 (ix86_output_indirect_function_return): New prototype.
1427 (ix86_split_simple_return_pop_internal): Likewise.
1428 * config/i386/i386.c (indirect_return_via_cx): New.
1429 (indirect_return_via_cx_bnd): Likewise.
1430 (indirect_thunk_name): Handle return va CX_REG.
1431 (output_indirect_thunk_function): Create alias for
1432 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
1433 (ix86_output_indirect_jmp): Remove the bool argument.
1434 (ix86_output_indirect_function_return): New function.
1435 (ix86_split_simple_return_pop_internal): Likewise.
1436 * config/i386/i386.md (*indirect_jump): Don't pass false
1437 to ix86_output_indirect_jmp.
1438 (*tablejump_1): Likewise.
1439 (simple_return_pop_internal): Change it to define_insn_and_split.
1440 Call ix86_split_simple_return_pop_internal to split it for
1442 (simple_return_indirect_internal): Call
1443 ix86_output_indirect_function_return instead of
1444 ix86_output_indirect_jmp.
1446 2018-02-26 Jakub Jelinek <jakub@redhat.com>
1449 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
1450 memset and value initialization afterwards.
1452 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
1454 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
1456 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1459 * common/config/aarch64/aarch64-common.c
1460 (aarch_option_optimization_table[]): Switch
1461 off fomit-frame-pointer
1463 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1464 Chung-Ju Wu <jasonwucj@gmail.com>
1466 * config/nds32/nds32-multiple.md (load_multiple): Disallow
1468 (store_multiple): Ditto.
1470 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
1472 * config.gcc: Add --with-cpu support for nds32 target.
1473 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
1474 * config/nds32/nds32.opt: Add -mcpu= option.
1476 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
1478 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
1479 isel=yes): Warn for these deprecated options.
1481 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
1483 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
1484 ISA_2_5_MASKS_EMBEDDED.
1486 2018-02-23 Jakub Jelinek <jakub@redhat.com>
1488 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
1489 p->max as pointers rather than using iterative_hash_expr.
1491 2018-02-23 Carl Love <cel@us.ibm.com>
1493 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
1494 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
1496 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
1497 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
1498 P8V_BUILTIN_VEC_VUNSIGNED2.
1500 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
1503 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
1504 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
1506 * lra-constraints.c (curr_insn_transform): Set up
1507 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
1509 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
1510 * lra-eliminations.c (spill_pseudos): Ditto.
1511 (process_insn_for_elimination): Ditto.
1512 * lra-lives.c (reg_early_clobber_p): Use the new macros.
1513 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
1514 LRA_NON_CLOBBERED_ALT.
1516 2018-02-22 Martin Sebor <msebor@redhat.com>
1518 PR tree-optimization/84480
1519 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
1520 to maybe_diag_stxncpy_trunc. Call it.
1521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
1522 from gimple_fold_builtin_strcpy. Print inlining stack.
1523 (handle_builtin_stxncpy): Print inlining stack.
1524 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
1526 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
1529 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
1530 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
1531 and -fcheck-pointer-bounds are used together.
1532 (indirect_thunk_prefix): New enum.
1533 (indirect_thunk_need_prefix): New function.
1534 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
1535 "_nt" instead of "_bnd" for NOTRACK prefix.
1536 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
1537 (output_indirect_thunk_function): Likewise.
1539 (ix86_code_end): Update output_indirect_thunk_function calls.
1540 (ix86_output_indirect_branch_via_reg): Replace
1541 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
1542 (ix86_output_indirect_branch_via_push): Likewise.
1543 (ix86_output_function_return): Likewise.
1544 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
1545 incompatible with -fcf-protection=branch and
1546 -fcheck-pointer-bounds.
1548 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1551 * config/aarch64/aarch64.c (aarch64_print_address_internal):
1552 Change gcc_assert call to output_operand_lossage.
1554 2018-02-22 Steve Ellcey <sellcey@cavium.com>
1556 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
1558 2018-02-22 DJ Delorie <dj@redhat.com>
1559 Sebastian Perta <sebastian.perta@renesas.com>
1560 Oleg Endo <olegendo@gcc.gnu.org>
1562 * config/rx/rx.c (rx_rtx_costs): New function.
1563 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
1565 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
1567 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
1569 2018-02-22 Martin Liska <mliska@suse.cz>
1572 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1573 Add "native" as a possible value.
1575 2018-02-22 Martin Liska <mliska@suse.cz>
1578 * config/i386/i386.c (ix86_option_override_internal):
1579 Add "native" as a possible value for -march and -mtune.
1581 2018-02-22 Jakub Jelinek <jakub@redhat.com>
1584 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
1585 to all type variants.
1587 PR tree-optimization/84503
1588 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
1589 width as info->bitpos + info->bitsize - start.
1590 (merged_store_group::merge_overlapping): Simplify width computation.
1591 (check_no_overlap): New function.
1592 (imm_store_chain_info::try_coalesce_bswap): Compute expected
1593 start + width and last_order of the group, fail if check_no_overlap
1595 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
1596 to group if check_no_overlap fails.
1598 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1600 * config/rs6000/altivec.md: Delete contraint arguments to
1601 define_expand, define_split, and define_peephole2, and in
1602 define_insn_and_split if always unused.
1603 * config/rs6000/darwin.md: Ditto.
1604 * config/rs6000/dfp.md: Ditto.
1605 * config/rs6000/rs6000.md: Ditto.
1606 * config/rs6000/sync.md: Ditto.
1607 * config/rs6000/vector.md: Ditto.
1608 * config/rs6000/vsx.md: Ditto.
1610 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
1612 * config/rs6000/altivec.md: Write output control strings as braced
1613 blocks instead of double-quoted strings.
1614 * config/rs6000/darwin.md: Ditto.
1615 * config/rs6000/rs6000.md: Ditto.
1616 * config/rs6000/vector.md: Ditto.
1617 * config/rs6000/vsx.md: Ditto.
1619 2018-02-21 Jason Merrill <jason@redhat.com>
1621 PR c++/84314 - ICE with templates and fastcall attribute.
1622 * attribs.c (build_type_attribute_qual_variant): Remove assert.
1624 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1626 * ipa-cp.c (determine_versionability): Fix comment typos.
1628 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
1631 * ipa-cp.c (determine_versionability): Do not version functions caling
1634 2018-02-21 Martin Liska <mliska@suse.cz>
1637 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
1638 Add "native" as a possible value.
1639 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
1640 the macro when native cpu detection is available.
1642 2018-02-21 Martin Liska <mliska@suse.cz>
1645 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
1646 Add "native" as a possible value.
1647 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
1648 when native cpu detection is available.
1650 2018-02-21 Jakub Jelinek <jakub@redhat.com>
1651 Martin Sebor <msebor@redhat.com>
1653 PR tree-optimization/84478
1654 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
1656 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
1657 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
1658 support which is conservatively correct, for 2 only stay conservative
1659 for maxlen. Formatting and comment capitalization fixes. Add STRICT
1660 argument to the 2 argument get_range_strlen, adjust 6 arg
1661 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
1663 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
1664 (gimple_fold_builtin_strlen): Pass true as last argument to
1667 2018-02-20 Martin Sebor <msebor@redhat.com>
1670 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1671 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1672 (builtin_memref::builtin_memref): Factor out parts into
1673 set_base_and_offset and call it.
1675 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1678 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1679 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1680 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1681 search at the associated MODE_INT.
1683 2018-02-20 Jeff Law <law@redhat.com>
1686 PR tree-optimization/81592
1688 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1689 for range data rather than using global data.
1690 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1691 range data rather than using global data.
1692 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1693 pass it to children as needed.
1694 (struct directive::fmtresult): Similarly.
1695 (struct directive::set_width): Similarly.
1696 (struct directive::set_precision): Similarly.
1697 (format_integer, format_directive, parse_directive): Similarly.
1698 (format_none): Accept unnamed vr_values parameter.
1699 (format_percent, format_floating, format_character): Similarly.
1700 (format_string, format_plain): Similarly.
1701 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1702 the EVRP range analyzer for range data rather than using global data.
1703 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1704 gimple-ssa-evrp-analyze.h
1705 (class sprintf_dom_walker): Add after_dom_children member function.
1706 Add evrp_range_analyzer member.
1707 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1708 range analyzer as needed.
1709 (sprintf_dom_walker::after_dom_children): New member function.
1710 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1712 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1713 (evrp_range_analyzer::pop_to_marker): Likewise.
1715 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1717 PR tree-optimization/84419
1718 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1719 with the required type if its current type is compatible but
1722 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1725 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1726 after vectorization.
1728 2018-02-20 Martin Liska <mliska@suse.cz>
1731 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1732 possible values if we don't have a hint.
1734 2018-02-20 Martin Liska <mliska@suse.cz>
1738 * final.c (shorten_branches): Build align_tab array with one
1740 * opts.c (finish_options): Add alignment option limit check.
1741 (MAX_CODE_ALIGN): Likewise.
1742 (MAX_CODE_ALIGN_VALUE): Likewise.
1743 * doc/invoke.texi: Document maximum allowed option value for
1744 all -falign-* options.
1746 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1749 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1750 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1751 * var-tracking.c (emit_note_insn_var_location): Remove all references
1752 to NOTE_INSN_CALL_ARG_LOCATION.
1753 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1754 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1755 Use copy_rtx_if_shared.
1756 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1757 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1758 (dwarf2out_var_location): Remove handling of
1759 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1761 * final.c (final_scan_insn): Remove all references to
1762 NOTE_INSN_CALL_ARG_LOCATION.
1763 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1764 before dumping final insns.
1765 * except.c (emit_note_eh_region_end): Remove all references to
1766 NOTE_INSN_CALL_ARG_LOCATION.
1767 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1768 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1769 * config/arc/arc.c (hwloop_optimize): Likewise.
1770 * config/arm/arm.c (create_fix_barrier): Likewise.
1771 * config/s390/s390.c (s390_chunkify_start): Likewise.
1772 * config/sh/sh.c (find_barrier): Likewise.
1773 * config/i386/i386.c (rest_of_insert_endbranch,
1774 ix86_seh_fixup_eh_fallthru): Likewise.
1775 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1776 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1777 * config/frv/frv.c (frv_function_prologue): Likewise.
1778 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1780 (note_outside_basic_block_p): Remove all references to
1781 NOTE_INSN_CALL_ARG_LOCATION.
1782 * gengtype.c (adjust_field_rtx_def): Likewise.
1783 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1785 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1786 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1789 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1792 PR tree-optimization/84452
1793 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1794 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1797 2018-02-19 Martin Liska <mliska@suse.cz>
1800 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1802 2018-02-19 Martin Liska <mliska@suse.cz>
1803 Richard Sandiford <richard.sandiford@linaro.org>
1805 PR tree-optimization/82491
1806 * gimple-fold.c (get_base_constructor): Make earlier bail out
1809 2018-02-19 Carl Love <cel@us.ibm.com>
1811 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1812 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1814 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1815 P8V_BUILTIN_VEC_NEG.
1817 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1819 * config/rl78/rl78.md (movdf): New define expand.
1821 2018-02-19 Martin Liska <mliska@suse.cz>
1824 * doc/invoke.texi: Fix typo.
1825 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1827 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1829 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1830 handle rs6000_single_float and rs6000_double_float specially for
1833 2018-02-16 Jeff Law <law@redhat.com>
1835 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1838 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1841 * ipa-inline.c (inline_small_functions): Fix a typo.
1843 2018-02-16 Nathan Sidwell <nathan@acm.org>
1845 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1847 2018-02-16 Carl Love <cel@us.ibm.com>
1849 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1850 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1851 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1852 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1853 expansion to P8V_BUILTIN_VEC_FLOAT2.
1855 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1857 PR rtl-optimization/70023
1858 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1859 src_regno into account.
1861 2018-02-16 Carl Love <cel@us.ibm.com>
1863 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1864 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1865 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1866 * config/rs6000/rs6000.c: Remove case statements for
1867 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1868 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1869 and P9V_BUILTIN_VEC_VINSERT4B.
1870 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1871 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1872 * config/rs6000/vsx.md:
1873 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1876 2018-02-16 Carl Love <cel@us.ibm.com>
1878 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1880 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1882 * config/rs6000/rs6000-c.c: Add the definitions for
1883 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1884 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1885 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1886 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1887 definition for insert4b and define insn *insert3b_internal.
1888 * doc/extend.texi: Add documentation for vec_extract4b.
1890 2018-02-16 Nathan Sidwell <nathan@acm.org>
1892 * doc/extend.texi (Backwards Compatibility): Mention friend
1893 injection. Note for-scope is deprecated.
1894 * doc/invoke.texi (-ffriend-injection): Deprecate.
1896 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1898 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1899 that moved to I2, also allow destinations that are a paradoxical
1900 subreg (instead of a normal reg).
1902 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1905 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1908 2018-02-16 Richard Biener <rguenther@suse.de>
1910 PR tree-optimization/84037
1911 PR tree-optimization/84016
1913 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1914 Adjust vec_construct for the fact we need additional higher latency
1915 128bit inserts for AVX256 and AVX512 vector builds.
1916 (ix86_add_stmt_cost): Scale vector construction cost for
1919 2018-02-16 Richard Biener <rguenther@suse.de>
1921 PR tree-optimization/84417
1922 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1923 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1924 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1926 2018-02-16 Martin Liska <mliska@suse.cz>
1929 * internal-fn.def (ASAN_CHECK): Set proper flags.
1930 (ASAN_MARK): Likewise.
1932 2018-02-16 Julia Koval <julia.koval@intel.com>
1934 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1935 from PTA_CANNONLAKE.
1937 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1940 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1941 Use ++iter rather than iter++ for std::list iterators.
1942 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1943 defer deleting them until all nodes in the forest are processed. Do
1944 free even leaf nodes. Change to_process into auto_vec.
1947 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1948 * vec.h (vec_default_construct): Use memset instead of placement new
1949 if BROKEN_VALUE_INITIALIZATION is defined.
1950 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1951 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1954 PR rtl-optimization/83723
1955 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1956 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1957 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1959 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1961 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1963 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1965 PR rtl-optimization/81443
1966 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1967 from inner REGs to paradoxical SUBREGs.
1969 2018-02-16 Richard Biener <rguenther@suse.de>
1971 PR tree-optimization/84399
1972 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1973 For operands we can analyze at their definition make sure we can
1974 analyze them at each use as well.
1976 2018-02-16 Richard Biener <rguenther@suse.de>
1978 PR tree-optimization/84190
1979 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1980 volatile accesses if the decl isn't volatile.
1982 2018-02-15 Jason Merrill <jason@redhat.com>
1984 PR c++/84314 - ICE with templates and fastcall attribute.
1985 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1986 TYPE_CANONICAL on an existing type.
1988 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1990 PR tree-optimization/84383
1991 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1992 dstoff nor call operand_equal_p if dstbase is NULL.
1994 PR tree-optimization/84334
1995 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1996 also a CONSTANT_CLASS_P, punt.
1998 2018-02-14 Jim Wilson <jimw@sifive.com>
2000 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2001 first SMALL_OPERAND check. New local min_second_step. Move assert
2002 to where locals are set. Add TARGET_RVC support.
2003 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2005 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2007 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2009 2018-02-14 Martin Sebor <msebor@redhat.com>
2011 PR tree-optimization/83698
2012 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2013 arrays constrain the offset range to their bounds.
2014 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2015 (builtin_access::overlap): Avoid setting the size of overlap if it's
2017 (maybe_diag_overlap): Also consider arrays when deciding what values
2018 of offsets to include in diagnostics.
2020 2018-02-14 Martin Sebor <msebor@redhat.com>
2023 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2024 that correspond to the kind of a declaration.
2026 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2029 * config/pa/pa.md: Load address of PIC label using the linkage table
2030 if the label is nonlocal.
2032 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2034 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2035 warning message if user requests -maltivec=be.
2036 * doc/invoke.texi: Document deprecation of -maltivec=be.
2038 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2041 * config/rs6000/rs6000-c.c: Update definitions for
2042 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2043 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2045 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2048 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2049 add _get_ssp intrinsics. Remove argument from
2050 __builtin_ia32_rdssp[d|q].
2051 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2052 * config/i386/i386-builtin.def: Remove argument from
2053 __builtin_ia32_rdssp[d|q].
2054 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2055 ix86_expand_special_args_builtin for _rdssp[d|q].
2056 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2057 Clear register before usage.
2058 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2059 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2061 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2063 PR tree-optimization/84357
2064 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2065 operand 1 of an ARRAY_REF too.
2067 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2070 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2071 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2073 (set_of_reg): New struct.
2074 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2075 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2076 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2078 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2079 Split into bitclr, bitset, bitinvert patterns if appropriate.
2080 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2081 use rx_fuse_in_memory_bitop.
2082 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2083 to named insn, correct maximum insn length.
2085 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2088 * machmode.def: Define a complex mode for PARTIAL_INT.
2089 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2091 * doc/rtl.texi: Document CSPImode.
2092 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2094 (msp430_hard_regno_nregs_with_padding): Likewise.
2096 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2099 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2101 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2103 PR rtl-optimization/84169
2104 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2105 we generated a parallel as new i3 and we split that to new i2 and i3
2106 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2107 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2108 those to i2, not i1. Partially rewrite this scan code.
2110 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2113 * stor-layout.c (place_field): For variable length fields, adjust
2114 offset_align afterwards not just based on the field's alignment,
2115 but also on the size.
2118 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2119 of exps and logs in the use_exp2 case.
2121 2018-02-13 Jeff Law <law@redhat.com>
2123 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2126 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2129 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2133 * common.opt (gas-loc-support, gas-locview-support): New.
2134 (ginline-points, ginternal-reset-location-views): New.
2135 * doc/invoke.texi: Document them. Use @itemx where intended.
2136 (gvariable-location-views): Adjust.
2137 * target.def (reset_location_view): New.
2138 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2139 (TARGET_RESET_LOCATION_VIEW): New.
2140 * doc/tm.texi: Rebuilt.
2141 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2142 (dwarf2out_default_as_locview_support): New.
2143 (output_asm_line_debug_info): Use option variables.
2144 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2145 (output_loc_list): Likewise.
2146 (add_high_low_attributes): Check option variables.
2147 Don't output entry view attribute in strict mode.
2148 (gen_inlined_subroutine_die): Check option variables.
2149 (dwarf2out_inline_entry): Likewise.
2150 (init_sections_and_labels): Likewise.
2151 (dwarf2out_early_finish): Likewise.
2152 (maybe_reset_location_view): New, from...
2153 (dwarf2out_var_location): ... here. Call it.
2154 * debug.h (dwarf2out_default_as_loc_support): Declare.
2155 (dwarf2out_default_as_locview_support): Declare.
2156 * hooks.c (hook_int_rtx_insn_0): New.
2157 * hooks.h (hook_int_rtx_insn_0): Declare.
2158 * toplev.c (process_options): Take -gas-loc-support and
2159 -gas-locview-support from dwarf2out. Enable
2160 -gvariable-location-views by default only with locview
2161 assembler support. Enable -ginternal-reset-location-views by
2162 default only if the target defines the corresponding hook.
2163 Enable -ginline-points by default if location views are
2164 enabled; force it disabled if statement frontiers are
2166 * tree-inline.c (expand_call_inline): Check option variables.
2167 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2169 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2171 PR tree-optimization/84321
2172 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2173 handling. Also check whether the anti-range contains any values
2174 that satisfy the mask; switch to a VR_RANGE if not.
2176 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2179 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2181 2018-02-13 Martin Jambor <mjambor@suse.cz>
2184 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2185 of call statements, also set location of a load to a temporary.
2187 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2189 * config/rl78/rl78.c (add_vector_labels): New function.
2190 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2191 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2192 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2193 which checks that no arguments are passed.
2194 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2195 * doc/extend.texi: Documentation for the new attribute.
2197 2018-02-13 Andreas Schwab <schwab@suse.de>
2199 * config/riscv/linux.h (CPP_SPEC): Define.
2201 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2204 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2205 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2206 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2207 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2208 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2209 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2210 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2211 temporarily for AES and PCLMUL builtins.
2213 PR tree-optimization/84339
2214 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2215 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2219 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2220 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2221 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2223 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2225 * omp-simd-clone.h: New file.
2226 * omp-simd-clone.c: Include omp-simd-clone.h.
2227 (expand_simd_clones): No longer static.
2228 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2229 cgraph.h and omp-simd-clone.h.
2230 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2231 (vect_recog_widen_shift_pattern): Formatting fix.
2232 (vect_pattern_recog_1): Don't check optab for calls.
2235 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2236 operands[2] into a REG before using gen_lowpart on it.
2238 2018-02-12 Jeff Law <law@redhat.com>
2241 * config/sh/sh.c (find_barrier): Consider a sibling call
2244 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2245 successfully back substituting a reg.
2247 2018-02-12 Richard Biener <rguenther@suse.de>
2249 PR tree-optimization/84037
2250 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2251 parameter, move visited init to caller.
2252 (vect_slp_analyze_operations): Separate cost from validity
2253 check, initialize visited once for all instances.
2254 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2256 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2258 (vect_model_promotion_demotion_cost): Likewise.
2259 (vectorizable_bswap): Guard cost modeling with !slp_node
2260 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2262 (vectorizable_call): Likewise.
2263 (vectorizable_conversion): Likewise.
2264 (vectorizable_assignment): Likewise.
2265 (vectorizable_shift): Likewise.
2266 (vectorizable_operation): Likewise.
2267 (vectorizable_store): Likewise.
2268 (vectorizable_load): Likewise.
2269 (vectorizable_condition): Likewise.
2270 (vectorizable_comparison): Likewise.
2272 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2275 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2276 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2277 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2279 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2282 * params.def (inline-min-speedup): Increase from 8 to 15.
2283 (max-inline-insns-auto): Decrease from 40 to 30.
2284 * ipa-split.c (consider_split): Add some buffer for function to
2285 be considered inlining candidate.
2286 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2289 2018-02-12 Richard Biener <rguenther@suse.de>
2291 PR tree-optimization/84037
2292 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2293 matched stmts if we cannot swap the non-matched ones.
2295 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2297 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2298 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2299 _mm_maskz_scalef_round_ss): New intrinsics.
2300 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2301 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2302 __builtin_ia32_scalefss_round): Remove.
2303 (__builtin_ia32_scalefsd_mask_round,
2304 __builtin_ia32_scalefss_mask_round): New intrinsics.
2305 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2306 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2307 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2308 "<round_constraint>")): Changed to ...
2309 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2310 "<round_scalar_constraint>")): ... this.
2311 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2312 %0, %1, %2<round_op3>}"): Changed to ...
2313 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2314 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2315 %2<round_scalar_mask_op3>}"): ... this.
2316 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2318 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2320 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2321 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2322 (_mm_maskz_sqrt_round_ss): New intrinsics.
2323 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2324 (__builtin_ia32_sqrtsd_mask_round)
2325 (__builtin_ia32_sqrtss_mask_round): New builtins.
2326 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2327 (__builtin_ia32_sqrtss_round): Remove.
2328 (__builtin_ia32_sqrtsd_mask_round)
2329 (__builtin_ia32_sqrtss_mask_round): New builtins.
2330 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2331 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2332 ((match_operand:VF_128 1 "vector_operand"
2333 "xBm,<round_constraint>")): Changed to ...
2334 ((match_operand:VF_128 1 "vector_operand"
2335 "xBm,<round_scalar_constraint>")): ... this.
2336 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
2337 %0, %2, %<iptr>1<round_op3>}): Changed to ...
2338 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
2339 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
2340 %<iptr>1<round_scalar_mask_op3>}): ... this.
2341 ((set_attr "prefix" "<round_prefix>")): Changed to ...
2342 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
2344 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
2347 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
2348 Cast vec_cmpeq result to correct type.
2349 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
2350 Cast vec_cmpgt result to correct type.
2352 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
2354 * final.c (final_scan_insn_1): Renamed from...
2355 (final_scan_insn): ... this. New wrapper, to recover
2356 seen from the outermost call in recursive ones.
2357 * config/sparc/sparc.c (output_return): Drop seen from call.
2358 (output_sibcall): Likewise.
2359 * config/visium/visium.c (output_branch): Likewise.
2361 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
2363 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
2366 2018-02-10 Alan Modra <amodra@gmail.com>
2369 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
2370 Specify LR as an input.
2372 2018-02-10 Jakub Jelinek <jakub@redhat.com>
2375 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
2376 remove_member_access_dummy_vars): New functions.
2377 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
2378 lower_omp_1, execute_lower_omp): Use them.
2380 PR rtl-optimization/84308
2381 * shrink-wrap.c (spread_components): Release todo vector.
2383 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
2385 PR rtl-optimization/57193
2386 * ira-color.c (struct allocno_color_data): Add member
2387 conflict_allocno_hard_prefs.
2388 (update_conflict_allocno_hard_prefs): New.
2389 (bucket_allocno_compare_func): Add a preference based on
2390 conflict_allocno_hard_prefs.
2391 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
2392 (color_allocnos): Remove a dead code. Initiate
2393 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
2395 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2398 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
2399 constraint from =wa to wa. Avoid a subreg on the output operand,
2400 instead use a pseudo and subreg it in a move.
2401 (p9_xxbrd_<mode>): Changed to ...
2402 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
2403 (p9_xxbrd_v2df): New expander.
2404 (p9_xxbrw_<mode>): Changed to ...
2405 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
2406 (p9_xxbrw_v4sf): New expander.
2408 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2410 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
2412 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
2415 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
2416 multiply in 32-bit mode.
2417 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
2418 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
2421 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
2423 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
2424 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
2425 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
2426 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
2428 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
2431 * dwarf2out.c (is_trivial_indirect_ref): New function.
2432 (dwarf2out_late_global_decl): Do not generate a location
2433 attribute for variables that have a non-trivial DECL_VALUE_EXPR
2434 and that are not defined in the current unit.
2436 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2438 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
2439 instead of a libcall for UNORDERED.
2441 2018-02-09 Tamar Christina <tamar.christina@arm.com>
2444 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
2445 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
2447 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2450 * config/s390/s390.c (s390_set_current_function): Invoke
2451 s390_indirect_branch_settings also if fndecl didn't change.
2453 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2455 * config/rs6000/rs6000.md (blockage): Set length to zero.
2457 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
2459 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
2461 2018-02-09 Jakub Jelinek <jakub@redhat.com>
2464 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
2465 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
2469 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
2470 PARALLEL incoming that failed vt_get_decl_and_offset check.
2473 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
2474 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
2476 (get_variable_section): For decls in named .bss* sections pass true as
2477 second argument to bss_initializer_p.
2479 2018-02-09 Marek Polacek <polacek@redhat.com>
2480 Jakub Jelinek <jakub@redhat.com>
2483 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
2484 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
2485 Sync some changes from cxx_fold_indirect_ref.
2487 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
2489 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
2491 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
2492 (BLOCK_INLINE_ENTRY_LABEL): New.
2493 (dwarf2out_var_location): Disregard inline entry markers.
2494 (inline_entry_data): New struct.
2495 (inline_entry_data_hasher): New hashtable type.
2496 (inline_entry_data_hasher::hash): New.
2497 (inline_entry_data_hasher::equal): New.
2498 (inline_entry_data_table): New variable.
2499 (add_high_low_attributes): Add DW_AT_entry_pc and
2500 DW_AT_GNU_entry_view attributes if a pending entry is found
2501 in inline_entry_data_table. Add old entry_pc attribute only
2502 if debug nonbinding markers are disabled.
2503 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
2504 markers are enabled.
2505 (block_within_block_p, dwarf2out_inline_entry): New.
2506 (dwarf2out_finish): Check that no entries remained in
2507 inline_entry_data_table.
2508 * final.c (reemit_insn_block_notes): Handle inline entry notes.
2509 (final_scan_insn, notice_source_line): Likewise.
2510 (rest_of_clean_state): Skip inline entry markers.
2511 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
2513 * gimple.c (gimple_build_debug_inline_entry): New.
2514 * gimple.h (enum gimple_debug_subcode): Add
2515 GIMPLE_DEBUG_INLINE_ENTRY.
2516 (gimple_build_debug_inline_entry): Declare.
2517 (gimple_debug_inline_entry_p): New.
2518 (gimple_debug_nonbind_marker_p): Adjust.
2519 * insn-notes.def (INLINE_ENTRY): New.
2520 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
2521 inline entry marker notes.
2522 (print_insn): Likewise.
2523 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
2524 (INSN_DEBUG_MARKER_KIND): Likewise.
2525 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
2526 * tree-inline.c (expand_call_inline): Build and insert
2527 debug_inline_entry stmt.
2528 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
2529 inline entry blocks early, if nonbind markers are enabled.
2530 (dump_scope_block): Dump fragment info.
2531 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
2532 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
2533 (gimple_build_debug_inline_entry): New.
2534 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
2535 Enable/disable inline entry points too.
2536 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
2537 (DEBUG_INSN): Describe inline entry markers.
2539 * common.opt (gvariable-location-views): New.
2540 (gvariable-location-views=incompat5): New.
2541 * config.in: Rebuilt.
2542 * configure: Rebuilt.
2543 * configure.ac: Test assembler for view support.
2544 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
2545 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
2546 * dwarf2out.c (var_loc_view): New typedef.
2547 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
2548 (dwarf2out_locviews_in_attribute): New.
2549 (dwarf2out_locviews_in_loclist): New.
2550 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
2551 (enum dw_line_info_opcode): Add LI_adv_address.
2552 (struct dw_line_info_table): Add view.
2553 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
2554 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
2555 (zero_view_p): New variable.
2556 (ZERO_VIEW_P): New macro.
2557 (output_asm_line_debug_info): New.
2558 (struct var_loc_node): Add view.
2559 (add_AT_view_list, AT_loc_list): New.
2560 (add_var_loc_to_decl): Add view param. Test it against last.
2561 (new_loc_list): Add view params. Record them.
2562 (AT_loc_list_ptr): Handle loc and view lists.
2563 (view_list_to_loc_list_val_node): New.
2564 (print_dw_val): Handle dw_val_class_view_list.
2565 (size_of_die): Likewise.
2566 (value_format): Likewise.
2567 (loc_list_has_views): New.
2568 (gen_llsym): Set vl_symbol too.
2569 (maybe_gen_llsym, skip_loc_list_entry): New.
2570 (dwarf2out_maybe_output_loclist_view_pair): New.
2571 (output_loc_list): Output view list or entries too.
2572 (output_view_list_offset): New.
2573 (output_die): Handle dw_val_class_view_list.
2574 (output_dwarf_version): New.
2575 (output_compilation_unit_header): Use it.
2576 (output_skeleton_debug_sections): Likewise.
2577 (output_rnglists, output_line_info): Likewise.
2578 (output_pubnames, output_aranges): Update version comments.
2579 (output_one_line_info_table): Output view numbers in asm comments.
2580 (dw_loc_list): Determine current endview, pass it to new_loc_list.
2581 Call maybe_gen_llsym.
2582 (loc_list_from_tree_1): Adjust.
2583 (add_AT_location_description): Create view list attribute if
2584 needed, check it's absent otherwise.
2585 (convert_cfa_to_fb_loc_list): Adjust.
2586 (maybe_emit_file): Call output_asm_line_debug_info for test.
2587 (dwarf2out_var_location): Reset views as needed. Precompute
2588 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
2589 attribute. Set view.
2590 (new_line_info_table): Reset next view.
2591 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
2592 (dwarf2out_source_line): Likewise. Output view resets and labels to
2593 the assembler, or select appropriate line info opcodes.
2594 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
2595 (optimize_string_length): Catch it. Adjust.
2596 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
2597 dw_val_class_view_list, and remove it if no longer needed.
2598 (hash_loc_list): Hash view numbers.
2599 (loc_list_hasher::equal): Compare them.
2600 (optimize_location_lists): Check whether a view list symbol is
2601 needed, and whether the locview attribute is present, and
2602 whether they match. Remove the locview attribute if no longer
2604 (index_location_lists): Call skip_loc_list_entry for test.
2605 (dwarf2out_finish): Call output_asm_line_debug_info for test.
2606 Use output_dwarf_version.
2607 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
2608 (struct dw_val_node): Add val_view_list.
2609 * final.c (SEEN_NEXT_VIEW): New.
2610 (set_next_view_needed): New.
2611 (clear_next_view_needed): New.
2612 (maybe_output_next_view): New.
2613 (final_start_function): Rename to...
2614 (final_start_function_1): ... this. Take pointer to FIRST,
2615 add SEEN parameter. Emit param bindings in the initial view.
2616 (final_start_function): Reintroduce SEEN-less interface.
2617 (final): Rename to...
2618 (final_1): ... this. Take SEEN parameter. Output final pending
2619 next view at the end.
2620 (final): Reintroduce seen-less interface.
2621 (final_scan_insn): Output pending next view before switching
2622 sections or ending a block. Mark the next view as needed when
2623 outputting variable locations. Notify debug backend of section
2624 changes, and of location view changes.
2625 (rest_of_handle_final): Adjust.
2626 * toplev.c (process_options): Autodetect value for debug variable
2627 location views option. Warn on incompat5 without -gdwarf-5.
2628 * doc/invoke.texi (gvariable-location-views): New.
2629 (gvariable-location-views=incompat5): New.
2630 (gno-variable-location-views): New.
2632 2018-02-08 David Malcolm <dmalcolm@redhat.com>
2634 PR tree-optimization/84136
2635 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
2636 that the result of find_edge is non-NULL.
2638 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2641 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
2642 storing integer register in SImode. Fix cost of 256 and 512
2643 byte aligned SSE register store.
2645 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
2647 * config/i386/i386.c (ix86_multiplication_cost): Fix
2648 multiplication cost for TARGET_AVX512DQ.
2650 2018-02-08 Marek Polacek <polacek@redhat.com>
2652 PR tree-optimization/84238
2653 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
2656 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2658 PR tree-optimization/84265
2659 * tree-vect-stmts.c (vectorizable_store): Don't treat
2660 VMAT_CONTIGUOUS accesses as grouped.
2661 (vectorizable_load): Likewise.
2663 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2665 PR tree-optimization/81635
2666 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2667 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2668 (test_round_for_mask): New functions.
2669 (wide_int_cc_tests): Call test_round_for_mask.
2670 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2671 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2672 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2673 range returned by get_range_info.
2675 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2678 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2679 * symtab.c: Include builtins.h
2680 (symtab_node::output_to_lto_symbol_table_p): Move here
2681 from lto-streamer-out.c:output_symbol_p.
2682 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2683 (output_symbol_p): Move all logic to symtab.c
2684 (produce_symtab): Update.
2686 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2688 * config/s390/s390-opts.h (enum indirect_branch): Define.
2689 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2690 (s390_indirect_branch_via_thunk)
2691 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2692 (enum s390_indirect_branch_type): Define.
2693 * config/s390/s390.c (struct s390_frame_layout, struct
2694 machine_function): Remove.
2695 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2696 (indirect_branch_table_label_no, indirect_branch_table_name):
2698 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2699 (enum s390_indirect_branch_option): Define.
2700 (s390_return_addr_from_memory): New function.
2701 (s390_handle_string_attribute): New function.
2702 (s390_attribute_table): Add new attribute handler.
2703 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2704 (s390_indirect_branch_via_thunk): New function.
2705 (s390_indirect_branch_via_inline_thunk): New function.
2706 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2707 sibling call optimization for non z10 compiles.
2708 (s390_emit_call): Force indirect branch target to be a single
2709 register. Add r1 clobber for non-z10 compiles.
2710 (s390_emit_epilogue): Emit return jump via return_use expander.
2711 (s390_reorg): Handle JUMP_INSNs as execute targets.
2712 (s390_option_override_internal): Perform validity checks for the
2713 new command line options.
2714 (s390_indirect_branch_attrvalue): New function.
2715 (s390_indirect_branch_settings): New function.
2716 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2717 (s390_output_indirect_thunk_function): New function.
2718 (s390_code_end): Implement target hook.
2719 (s390_case_values_threshold): Implement target hook.
2720 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2722 * config/s390/s390.h (struct s390_frame_layout)
2723 (struct machine_function): Move here from s390.c.
2724 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2725 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2726 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2727 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2728 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2729 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2730 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2731 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2732 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2733 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2734 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2735 (mnemonic attribute): Add values which aren't recognized
2737 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2738 pattern for branch conversion. Fix mnemonic attribute.
2739 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2740 indirect branch via thunk if requested.
2741 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2742 ("*indirect_jump"): Disable for branch conversion using out of
2744 ("indirect_jump_via_thunk<mode>_z10")
2745 ("indirect_jump_via_thunk<mode>")
2746 ("indirect_jump_via_inlinethunk<mode>_z10")
2747 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2748 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2749 ("casesi_jump_via_inlinethunk<mode>_z10")
2750 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2751 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2752 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2753 ("*indirect2_jump"): Disable for branch conversion.
2754 ("casesi_jump"): Turn into expander and expand patterns for branch
2756 ("return_use"): New expander.
2757 ("*return"): Emit return via thunk and rename it to ...
2758 ("*return<mode>"): ... this one.
2759 * config/s390/s390.opt: Add new options and and enum for the
2762 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2764 * lra-constraints.c (match_reload): Unconditionally use
2765 gen_lowpart_SUBREG, rather than selecting between that
2766 and equivalent gen_rtx_SUBREG code.
2768 2018-02-08 Richard Biener <rguenther@suse.de>
2770 PR tree-optimization/84233
2771 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2772 changed flag instead of boguously re-using phi_inserted.
2774 2018-02-08 Martin Jambor <mjambor@suse.cz>
2776 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2777 static local variables.
2779 2018-02-08 Richard Biener <rguenther@suse.de>
2781 PR tree-optimization/84278
2782 * tree-vect-stmts.c (vectorizable_store): When looking for
2783 smaller vector types to perform grouped strided loads/stores
2784 make sure the mode is supported by the target.
2785 (vectorizable_load): Likewise.
2787 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2789 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2790 Increase LDP/STP opportunities by adding adjacent callee-saves.
2792 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2794 PR rtl-optimization/84068
2795 PR rtl-optimization/83459
2796 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2798 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2800 PR tree-optimization/84224
2801 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2802 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2805 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2808 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2809 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2810 count, remove one USE.
2812 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2814 * doc/install.texi (Configuration): Document the
2815 --with-long-double-format={ibm,ieee} PowerPC configuration
2819 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2820 Convert from define_expand to be define_insn_and_split. Rework
2821 float/double/_Float128 conversions to QI/HI/SImode to work with
2822 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2823 conversions to QI/HImode types did a store and then a load to
2824 truncate the value. For conversions to VSX registers, don't split
2825 the insn, instead emit the code directly. Use the code iterator
2826 any_fix to combine signed and unsigned conversions.
2827 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2828 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2829 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2830 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2831 (fix_<mode>di2_hw): Likewise.
2832 (fixuns_<mode>di2_hw): Likewise.
2833 (fix_<mode>si2_hw): Likewise.
2834 (fixuns_<mode>si2_hw): Likewise.
2835 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2836 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2837 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2838 fix<uns>_trunc<SFDF:mode>si2_p8.
2839 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2841 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2842 (fix<uns>_<mode>_mem): Likewise.
2843 (fctiw<u>z_<mode>_mem): Likewise.
2844 (fix<uns>_<mode>_mem): Likewise.
2845 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2846 the register allocator from doing a direct move to the GPRs to do
2847 a store, and instead use the ISA 3.0 store byte/half-word from
2848 vector register instruction. For IEEE 128-bit floating point,
2849 also optimize stores of 32-bit ints.
2850 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2852 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2854 * genextract.c (push_pathstr_operand): New function to support
2856 (walk_rtx): Call push_pathstr_operand.
2857 (print_path): Support [a-zA-Z].
2859 2018-02-07 Richard Biener <rguenther@suse.de>
2861 PR tree-optimization/84037
2862 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2863 (cse_and_gimplify_to_preheader): Declare.
2864 (vect_get_place_in_interleaving_chain): Likewise.
2865 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2867 (_loop_vec_info::~_loop_vec_info): Delete it.
2868 (cse_and_gimplify_to_preheader): New function.
2869 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2870 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2871 (vectorizable_load): Likewise. For grouped stores always base
2872 the IV on the first element.
2873 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2874 condition before gimplifying.
2876 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2878 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2879 *DIV_EXPR and *MOD_EXPR.
2881 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2884 * config/i386/i386.c (ix86_option_override_internal): Mask out
2885 the CF_SET bit when checking -fcf-protection.
2887 2018-02-07 Tom de Vries <tom@codesourcery.com>
2890 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2893 2018-02-07 Richard Biener <rguenther@suse.de>
2895 PR tree-optimization/84204
2896 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2899 PR tree-optimization/84205
2900 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2901 special-case isl_ast_op_zdiv_r.
2903 PR tree-optimization/84223
2904 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2905 Only add conditions from within the region.
2906 (gather_bbs::after_dom_children): Adjust.
2908 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2911 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2912 * config/avr/avr.md: Only post-reload split REG-REG moves if
2913 either register is GENERAL_REG_P.
2915 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2917 PR tree-optimization/84235
2918 * tree-ssa-scopedtables.c
2919 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2920 if the subtraction is performed in floating point type where NaNs are
2921 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2922 build 1. Formatting fix.
2924 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2927 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2928 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2929 and skip it regardless of bb boundaries. Use CALL_P macro,
2930 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2931 unnecessarily, formatting fix.
2933 2018-02-06 Michael Collison <michael.collison@arm.com>
2935 * config/arm/thumb2.md:
2936 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2937 (*thumb_mov_notscc): Ditto.
2939 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2942 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2943 unsigned_fix, not "s".
2945 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2947 * configure.ac (gcc_fn_eh_frame_ro): New function.
2948 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2949 correct .eh_frame permissions.
2950 * configure: Regenerate.
2952 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2954 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2957 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2959 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2960 Display warning message for -mno-speculate-indirect-jumps.
2962 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2964 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2966 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2968 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2970 PR tree-optimization/84225
2971 * tree-eh.c (find_trapping_overflow): Only call
2972 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2974 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2977 * config/i386/i386.c: Reimplement the check of possible options
2978 -mibt/-mshstk conbination. Change error messages.
2979 * doc/invoke.texi: Fix a typo: remove extra '='.
2981 2018-02-06 Marek Polacek <polacek@redhat.com>
2983 PR tree-optimization/84228
2984 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2986 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2989 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2990 emitted arch directives.
2991 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2992 __ARM_FEATURE_COPROC before changing architectures.
2994 2018-02-06 Richard Biener <rguenther@suse.de>
2996 * config/i386/i386.c (print_reg): Fix typo.
2997 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2999 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3001 * configure: Regenerate.
3003 2018-02-05 Martin Sebor <msebor@redhat.com>
3005 PR tree-optimization/83369
3006 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3009 2018-02-05 Martin Liska <mliska@suse.cz>
3011 * doc/invoke.texi: Cherry-pick upstream r323995.
3013 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3015 * ira.c (ira_init_register_move_cost): Adjust comment.
3017 2018-02-05 Martin Liska <mliska@suse.cz>
3019 PR gcov-profile/84137
3020 * doc/gcov.texi: Fix typo in documentation.
3022 2018-02-05 Martin Liska <mliska@suse.cz>
3024 PR gcov-profile/83879
3025 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3026 using dlopen functionality.
3028 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3030 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3031 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3032 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3033 _mm_maskz_range_round_ss): New intrinsics.
3034 (__builtin_ia32_rangesd128_round)
3035 (__builtin_ia32_rangess128_round): Remove.
3036 (__builtin_ia32_rangesd128_mask_round,
3037 __builtin_ia32_rangess128_mask_round): New builtins.
3038 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3039 __builtin_ia32_rangess128_round): Remove.
3040 (__builtin_ia32_rangesd128_mask_round,
3041 __builtin_ia32_rangess128_mask_round): New builtins.
3042 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3043 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3044 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3045 "<round_saeonly_constraint>")): Changed to ...
3046 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3047 "<round_saeonly_scalar_constraint>")): ... this.
3048 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3049 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3050 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3051 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3052 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3054 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3056 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3058 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3059 Remove all values except native, 8540 and 8548.
3061 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3063 * config/i386/i386.c (ix86_output_function_return): Pass
3064 INVALID_REGNUM, instead of -1, as invalid register number to
3065 indirect_thunk_name and output_indirect_thunk.
3067 2018-02-02 Julia Koval <julia.koval@intel.com>
3069 * config.gcc: Add -march=icelake.
3070 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3071 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3072 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3073 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3074 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3075 (processor_target_table): Add icelake.
3076 (ix86_option_override_internal): Handle new PTAs.
3077 (get_builtin_code_for_version): Handle icelake.
3078 (M_INTEL_COREI7_ICELAKE): New.
3079 (fold_builtin_cpu): Handle icelake.
3080 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3081 * doc/invoke.texi: Add -march=icelake.
3083 2018-02-02 Julia Koval <julia.koval@intel.com>
3085 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3086 to wide_int_bitmask.
3087 * wide-int-bitmask.h: New.
3089 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3092 * config/i386/i386.md: Replace Pmode with word_mode in
3093 builtin_setjmp_setup and builtin_longjmp to support x32.
3095 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3099 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3101 (rs6000_supported_cpu_names): New static variable.
3102 (linux_cpu_translation_table): Likewise.
3103 (elf_platform) <cpu>: Define new static variable and use it.
3104 Translate kernel AT_PLATFORM name to canonical name if needed.
3105 Error if platform name is unknown.
3107 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3110 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3112 2018-02-01 Jeff Law <law@redhat.com>
3115 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3116 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3117 the scratch if RELEASE_VIA_POP is false.
3118 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3119 If we have to save a temporary register, decrement SIZE appropriately.
3120 Pass new arguments to release_scratch_register_on_entry.
3121 (ix86_adjust_stack_and_probe): Likewise.
3122 (ix86_emit_probe_stack_range): Pass new arguments to
3123 release_scratch_register_on_entry.
3125 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3127 PR rtl-optimization/84157
3128 * combine.c (change_zero_ext): Use REG_P predicate in
3129 front of HARD_REGISTER_P predicate.
3131 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3133 * config/avr/avr.c (avr_option_override): Move disabling of
3134 -fdelete-null-pointer-checks to...
3135 * common/config/avr/avr-common.c (avr_option_optimization_table):
3138 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3140 PR tree-optimization/81635
3141 * tree-data-ref.c (split_constant_offset_1): For types that
3142 wrap on overflow, try to use range info to prove that wrapping
3145 2018-02-01 Renlin Li <renlin.li@arm.com>
3148 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3150 (aarch64_register_move_cost): Likewise.
3151 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3153 (REG_CLASS_NAMES): Likewise.
3154 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3155 TAILCALL_ADDR_REGS. Remove IP registers.
3156 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3158 2018-02-01 Richard Biener <rguenther@suse.de>
3160 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3161 for specifying RPO order and allow NULL for that.
3162 * domwalk.c (dom_walker::dom_walker): Likewise.
3163 (dom_walker::walk): Handle NULL RPO order.
3164 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3166 (rewrite_update_dom_walker): Likewise.
3167 (mark_def_dom_walker): Likewise.
3169 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3171 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3172 (aarch64_maybe_expand_sve_subreg_move): Declare.
3173 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3174 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3176 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3177 that are semantically a reverse operation.
3178 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3179 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3180 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3182 (aarch64_can_change_mode_class): For big-endian, forbid changes
3183 between two SVE modes if they have different element sizes.
3185 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3187 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3188 the TImode handling for big-endian targets.
3190 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3192 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3193 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3195 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3196 Remove BSWAP handing for big-endian targets and use the form of
3197 LD1RQ appropariate for the mode.
3199 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3201 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3202 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3205 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3208 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3209 check for operands that need to go through aarch64_sve_reload_be.
3211 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3213 PR tree-optimization/81661
3214 PR tree-optimization/84117
3215 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3216 * tree-eh.c: Include gimplify.h.
3217 (find_trapping_overflow, replace_trapping_overflow,
3218 rewrite_to_non_trapping_overflow): New functions.
3219 * tree-vect-loop.c: Include tree-eh.h.
3220 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3221 * tree-data-ref.c: Include tree-eh.h.
3222 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3224 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3226 PR rtl-optimization/84123
3227 * combine.c (change_zero_ext): Check if hard register satisfies
3228 can_change_dest_mode before calling gen_lowpart_SUBREG.
3230 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3233 * ira.c (ira_init_register_move_cost): Remove assert.
3235 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3237 PR rtl-optimization/84071
3238 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3239 * doc/tm.texi: Regenerate.
3241 2018-01-31 Richard Biener <rguenther@suse.de>
3243 PR tree-optimization/84132
3244 * tree-data-ref.c (analyze_miv_subscript): Properly
3245 check whether evolution_function_is_affine_multivariate_p
3246 before calling gcd_of_steps_may_divide_p.
3248 2018-01-31 Julia Koval <julia.koval@intel.com>
3251 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3252 * config/i386/i386.md (rdpid_rex64) New.
3253 (rdpid): Make 32bit only.
3255 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3258 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3259 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3261 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3264 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3266 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3268 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3270 PR rtl-optimization/84071
3271 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3272 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3274 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3276 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3277 (arc_attribute_table): Add 'aux' attribute.
3278 (arc_in_small_data_p): Consider aux like variables.
3279 (arc_is_aux_reg_p): New function.
3280 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3281 (arc_get_aux_arg): New function.
3282 (prepare_move_operands): Handle aux-register access.
3283 (arc_handle_aux_attribute): New function.
3284 * doc/extend.texi (ARC Variable attributes): Add subsection.
3286 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3288 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3289 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3290 (arc_attribute_table): Add 'uncached' attribute.
3291 (arc_print_operand): Print '.di' flag for uncached memory
3293 (arc_in_small_data_p): Do not consider for small data the uncached
3295 (arc_is_uncached_mem_p): New function.
3296 * config/arc/predicates.md (compact_store_memory_operand): Check
3297 for uncached memory accesses.
3298 (nonvol_nonimm_operand): Likewise.
3299 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3301 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3304 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3305 falign-loops=): Add Optimization flag.
3307 2018-01-30 Jeff Law <law@redhat.com>
3310 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3311 INT_REGISTERS_SAVED. Check it prior to calling
3312 get_scratch_register_on_entry.
3313 (ix86_adjust_stack_and_probe): Similarly.
3314 (ix86_emit_probe_stack_range): Similarly.
3315 (ix86_expand_prologue): Corresponding changes.
3317 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3320 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3321 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3323 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3326 * lra-constraints.c (curr_insn_transform): Process AND in the
3329 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3331 PR rtl-optimization/83986
3332 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3333 dependence against last_pending_memory_flush in addition to
3336 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
3338 PR tree-optimization/81611
3339 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
3342 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3345 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
3348 2018-01-30 Richard Biener <rguenther@suse.de>
3349 Jakub Jelinek <jakub@redhat.com>
3351 PR tree-optimization/84111
3352 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
3353 inner loops added during recursion, as they don't have up-to-date
3356 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3359 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
3360 (can_inline_edge_by_limits_p): ... here.
3361 (can_early_inline_edge_p, check_callers,
3362 update_caller_keys, update_callee_keys, recursive_inlining,
3363 add_new_edges_to_heap, speculation_useful_p,
3364 inline_small_functions,
3365 inline_small_functions, flatten_function,
3366 inline_to_all_callers_1): Update.
3368 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
3370 * profile-count.c (profile_count::combine_with_ipa_count): Handle
3373 2018-01-30 Richard Biener <rguenther@suse.de>
3375 PR tree-optimization/83008
3376 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
3377 invariant and constant vector uses in stmts when they need
3380 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3383 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
3384 * configure: Regenerate.
3386 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3388 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
3390 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
3391 Use gen_rtx_REG rather than gen_lowpart.
3393 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3395 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
3396 rather than 0 when creating partial subregs.
3398 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
3400 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
3403 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3406 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
3407 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
3408 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
3409 flags. This restores the settings used before the 2017-07-24.
3410 Turning off pre increment/decrement/modify allows IVOPTS to
3411 optimize DF/SF loops where the index is an int.
3413 2018-01-29 Richard Biener <rguenther@suse.de>
3414 Kelvin Nilsen <kelvin@gcc.gnu.org>
3417 * tree-vect-stmts.c (vectorizable_call): Don't call
3418 targetm.vectorize_builtin_md_vectorized_function if callee is
3421 2018-01-22 Carl Love <cel@us.ibm.com>
3423 * doc/extend.tex: Fix typo in second arg in
3424 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
3426 2018-01-29 Richard Biener <rguenther@suse.de>
3428 PR tree-optimization/84086
3429 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
3430 (flush_ssaname_freelist): When SSA names were released reset
3431 the SCEV hash table.
3433 2018-01-29 Richard Biener <rguenther@suse.de>
3435 PR tree-optimization/84057
3436 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
3437 removed paths when removing edges.
3439 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
3441 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
3442 -mfunction-return=@var{choice}.
3444 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
3447 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
3449 (layout::print_source_line): Likewise.
3450 (test_get_line_width_without_trailing_whitespace): Add test cases.
3452 2018-01-27 Jakub Jelinek <jakub@redhat.com>
3455 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
3458 2018-01-26 Jim Wilson <jimw@sifive.com>
3460 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
3462 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
3465 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3467 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
3468 and CMP + SUB-immediate -> SUBS.
3470 2018-01-26 Martin Sebor <msebor@redhat.com>
3472 PR tree-optimization/83896
3473 * tree-ssa-strlen.c (get_string_len): Rename...
3474 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
3475 Avoid assuming length is constant.
3476 (handle_char_store): Use HOST_WIDE_INT for string length.
3478 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
3481 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
3482 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
3484 2018-01-26 Richard Biener <rguenther@suse.de>
3486 PR rtl-optimization/84003
3487 * dse.c (record_store): Only record redundant stores when
3488 the earlier store aliases at least all accesses the later one does.
3490 2018-01-26 Jakub Jelinek <jakub@redhat.com>
3492 PR rtl-optimization/83985
3493 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
3494 REG_CFA_RESTORE insns.
3495 (delete_unmarked_insns): Don't ignore separate shrink wrapping
3496 REG_CFA_RESTORE insns here.
3499 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
3500 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
3502 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3504 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
3505 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
3506 (arc_init): Likewise.
3507 (arc_override_options): Likewise.
3508 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
3510 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
3512 * config/arc/arc.h (TARGET_DBNZ): Define.
3513 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
3514 properly set the tune attribute.
3515 (dbnz): Use TARGET_DBNZ guard.
3516 * config/arc/arc.opt (mtune): Add core3 option.
3518 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3520 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
3521 recognize new pic like addresses.
3522 (arc_delegitimize_address): Clean up.
3524 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3526 * config/arc/arc-arches.def: Option mrf16 valid for all
3528 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
3529 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
3530 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
3531 * config/arc/arc-tables.opt: Regenerate.
3532 * config/arc/arc.c (arc_conditional_register_usage): Handle
3533 reduced register file case.
3534 (arc_file_start): Set must have build attributes.
3535 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
3537 * config/arc/arc.opt (mrf16): Add new option.
3538 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
3539 * config/arc/genmultilib.awk: Handle new mrf16 option.
3540 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
3541 * config/arc/t-multilib: Regenerate.
3542 * doc/invoke.texi (ARC Options): Document mrf16 option.
3544 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3546 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
3547 * config/arc/arc.c (arc_handle_secure_attribute): New function.
3548 (arc_attribute_table): Add 'secure_call' attribute.
3549 (arc_print_operand): Print secure call operand.
3550 (arc_function_ok_for_sibcall): Don't optimize tail calls when
3552 (arc_is_secure_call_p): New function. * config/arc/arc.md
3553 (call_i): Add support for sjli instruction.
3554 (call_value_i): Likewise.
3555 * config/arc/constraints.md (Csc): New constraint.
3557 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
3558 John Eric Martin <John.Martin@emmicro-us.com>
3560 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
3561 * config/arc/arc.c (_arc_jli_section): New struct.
3562 (arc_jli_section): New type.
3563 (rc_jli_sections): New static variable.
3564 (arc_handle_jli_attribute): New function.
3565 (arc_attribute_table): Add jli_always and jli_fixed attribute.
3566 (arc_file_end): New function.
3567 (TARGET_ASM_FILE_END): Define.
3568 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
3569 (arc_add_jli_section): New function.
3570 (jli_call_scan): Likewise.
3571 (arc_reorg): Call jli_call_scan.
3572 (arc_output_addsi): Remove 'S' from printing asm operand.
3573 (arc_is_jli_call_p): New function.
3574 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
3576 (movhi_insn): Likewise.
3577 (movsi_insn): Likewise.
3578 (movsi_set_cc_insn): Likewise.
3579 (loadqi_update): Likewise.
3580 (load_zeroextendqisi_update): Likewise.
3581 (load_signextendqisi_update): Likewise.
3582 (loadhi_update): Likewise.
3583 (load_zeroextendhisi_update): Likewise.
3584 (load_signextendhisi_update): Likewise.
3585 (loadsi_update): Likewise.
3586 (loadsf_update): Likewise.
3587 (movsicc_insn): Likewise.
3588 (bset_insn): Likewise.
3589 (bxor_insn): Likewise.
3590 (bclr_insn): Likewise.
3591 (bmsk_insn): Likewise.
3592 (bicsi3_insn): Likewise.
3593 (cmpsi_cc_c_insn): Likewise.
3594 (movsi_ne): Likewise.
3595 (movsi_cond_exec): Likewise.
3596 (clrsbsi2): Likewise.
3607 (call_i): Remove 'S' asm letter, add jli instruction.
3608 (call_value_i): Likewise.
3609 * config/arc/arc.op (mjli-always): New option.
3610 * config/arc/constraints.md (Cji): New constraint.
3611 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
3613 (subsf3_fpx): Likewise.
3614 (mulsf3_fpx): Likewise.
3615 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
3617 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
3619 * doc/invoke.texi (ARC): Document mjli-always option.
3621 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
3623 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
3624 avoid addition with 0 and use incw and decw where possible.
3626 2018-01-26 Richard Biener <rguenther@suse.de>
3628 PR tree-optimization/81082
3629 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
3630 association if it requires casting to unsigned.
3631 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
3632 from fold_plusminus_mult_expr to catch important cases late when
3633 range info is available.
3635 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3637 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
3638 * configure.ac (hidden_linkonce): New test.
3639 * configure: Regenerate.
3640 * config.in: Regenerate.
3642 2018-01-26 Julia Koval <julia.koval@intel.com>
3644 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
3645 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
3646 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
3647 _mm_mask_bitshuffle_epi64_mask): Fix type.
3648 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
3649 USI_FTYPE_V4DI_V4DI_USI): Remove.
3650 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
3651 __builtin_ia32_vpshufbitqmb256_mask,
3652 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
3653 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
3654 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
3656 2018-01-26 Alan Modra <amodra@gmail.com>
3659 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
3660 UNSPEC_VBPERMQ. Sort other unspecs.
3662 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
3664 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3666 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3669 * predict.c (drop_profile): Do not push/pop cfun; update also
3671 (handle_missing_profiles): Fix logic looking for zero profiles.
3673 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3676 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3677 on functions with #pragma omp declare simd or functions with simd
3679 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3680 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3681 Remove trailing \n from warning_at calls.
3683 2018-01-25 Tom de Vries <tom@codesourcery.com>
3686 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3687 for neutered workers.
3689 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3692 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3693 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3695 2018-01-24 Jeff Law <law@redhat.com>
3698 * i386.c (get_probe_interval): Move to earlier point.
3699 (ix86_compute_frame_layout): If -fstack-clash-protection and
3700 the frame is larger than the probe interval, then use pushes
3701 to save registers rather than reg->mem moves.
3702 (ix86_expand_prologue): Remove conditional for int_registers_saved
3705 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3708 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3709 min/max for never referenced object.
3711 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3714 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3716 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3717 attributes from DECL_ATTRIBUTES (decl) without affecting
3718 DECL_ATTRIBUTES (current_function_decl).
3719 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3720 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3722 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3724 PR tree-optimization/83979
3725 * fold-const.c (fold_comparison): Use constant_boolean_node
3726 instead of boolean_{true,false}_node.
3728 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3730 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3733 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3735 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3736 Simplify the clause that sets the length attribute.
3737 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3738 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3739 clause that sets the length attribute.
3740 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3742 2018-01-24 Tom de Vries <tom@codesourcery.com>
3745 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3746 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3747 Add strict parameter.
3748 (prevent_branch_around_nothing): Insert dummy insn between branch to
3749 label and label with no ptx insn inbetween.
3750 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3752 2018-01-24 Tom de Vries <tom@codesourcery.com>
3755 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3756 for neutered threads in warp.
3757 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3759 2018-01-24 Richard Biener <rguenther@suse.de>
3761 PR tree-optimization/83176
3762 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3765 2018-01-24 Richard Biener <rguenther@suse.de>
3767 PR tree-optimization/82819
3768 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3769 code generating pluses that are no-ops in the target precision.
3771 2018-01-24 Richard Biener <rguenther@suse.de>
3774 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3776 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3778 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3779 to merge probabilities.
3780 * predict.c (probably_never_executed): Also mark as cold functions
3781 with global 0 profile and guessed local profile.
3782 * profile-count.c (profile_probability::combine_with_count): New
3784 * profile-count.h (profile_probability::operator*,
3785 profile_probability::operator*=, profile_probability::operator/,
3786 profile_probability::operator/=): Reduce precision to adjusted
3787 and set value to guessed on contradictory divisions.
3788 (profile_probability::combine_with_freq): Remove.
3789 (profile_probability::combine_wiht_count): Declare.
3790 (profile_count::force_nonzero):: Set to adjusted.
3791 (profile_count::probability_in):: Set quality to adjusted.
3792 * tree-ssa-tail-merge.c (replace_block_by): Use
3795 2018-01-23 Andrew Waterman <andrew@sifive.com>
3796 Jim Wilson <jimw@sifive.com>
3798 * config/riscv/riscv.c (riscv_stack_boundary): New.
3799 (riscv_option_override): Set riscv_stack_boundary. Handle
3800 riscv_preferred_stack_boundary_arg.
3801 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3802 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3803 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3804 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3805 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3806 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3808 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3811 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3812 of struct ix86_frame.
3813 (ix86_expand_epilogue): Likewise. Add a local variable for
3814 the reg_save_offset field in struct ix86_frame.
3816 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3818 PR tree-optimization/82604
3819 * tree-loop-distribution.c (enum partition_kind): New enum item
3820 PKIND_PARTIAL_MEMSET.
3821 (partition_builtin_p): Support above new enum item.
3822 (generate_code_for_partition): Ditto.
3823 (compute_access_range): Differentiate cases that equality can be
3824 proven at all loops, the innermost loops or no loops.
3825 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3826 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3827 (finalize_partitions, distribute_loop): Don't fuse partition of
3828 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3829 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3832 2018-01-23 Martin Liska <mliska@suse.cz>
3834 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3835 order to ignore the predictor.
3836 (PRED_POLYMORPHIC_CALL): Likewise.
3837 (PRED_RECURSIVE_CALL): Likewise.
3839 2018-01-23 Martin Liska <mliska@suse.cz>
3841 * tree-profile.c (tree_profiling): Print function header to
3842 aware reader which function we are working on.
3843 * value-prof.c (gimple_find_values_to_profile): Do not print
3844 not interesting value histograms.
3846 2018-01-23 Martin Liska <mliska@suse.cz>
3848 * profile-count.h (enum profile_quality): Add
3849 profile_uninitialized as the first value. Do not number values
3850 as they are zero based.
3851 (profile_count::verify): Update sanity check.
3852 (profile_probability::verify): Likewise.
3854 2018-01-23 Nathan Sidwell <nathan@acm.org>
3856 * doc/invoke.texi (ffor-scope): Deprecate.
3858 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3860 PR tree-optimization/83510
3861 * domwalk.c (set_all_edges_as_executable): New function.
3862 (dom_walker::dom_walker): Convert bool param
3863 "skip_unreachable_blocks" to enum reachability. Move setup of
3864 edge flags to set_all_edges_as_executable and only do it when
3865 reachability is REACHABLE_BLOCKS.
3866 * domwalk.h (enum dom_walker::reachability): New enum.
3867 (dom_walker::dom_walker): Convert bool param
3868 "skip_unreachable_blocks" to enum reachability.
3869 (set_all_edges_as_executable): New decl.
3870 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3871 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3873 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3874 but converting true to REACHABLE_BLOCKS.
3875 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3877 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3878 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3879 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3881 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3882 if check_all_array_refs will be called.
3884 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3886 * tree.c (selftest::test_location_wrappers): Add more test
3889 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3891 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3892 (selftest::test_bit_in_range): Likewise.
3894 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3897 * doc/sourcebuild.texi (vect_float): Say that the selector
3898 only describes the situation when -funsafe-math-optimizations is on.
3899 (vect_float_strict): Document.
3901 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3903 PR tree-optimization/83965
3904 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3905 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3906 instead of checking only for a reduction.
3907 (vect_recog_widen_sum_pattern): Likewise.
3909 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3911 * predict.c (probably_never_executed): Only use precise profile info.
3912 (compute_function_frequency): Skip after inlining hack since we now
3913 have quality checking.
3915 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3917 * profile-count.h (profile_probability::very_unlikely,
3918 profile_probability::unlikely, profile_probability::even): Set
3919 precision to guessed.
3921 2018-01-23 Richard Biener <rguenther@suse.de>
3923 PR tree-optimization/83963
3924 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3925 Properly terminate dominator walk when crossing the exit edge not
3926 when visiting its source block.
3928 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3931 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3932 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3934 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3936 PR tree-optimization/83957
3937 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3938 semicolon after for body surrounded by braces.
3940 PR tree-optimization/83081
3941 * profile-count.h (profile_probability::split): New method.
3942 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3943 Use profile_probability::split.
3944 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3945 when splitting a single conditional jump into 2.
3947 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3949 PR tree-optimization/69452
3950 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3953 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3955 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3956 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3957 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3959 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3961 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3962 * config/rl78/rl78.md (movdi): New define_expand.
3963 * config/rl78/rl78.c (rl78_split_movdi): New function.
3965 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3968 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3970 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3971 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3972 128-bit to produce an UNSPEC move to get the double word with the
3973 signbit and then a shift directly to do signbit.
3974 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3975 implementation with a new version that just does either a direct
3976 move or a regular move. Move memory interface to separate insns.
3977 Move insns so they are next to the expander.
3978 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3979 with signbit move. Split big and little endian case.
3980 (signbit<mode>2_dm_mem_le): Likewise.
3981 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3982 (signbit<mode>2_dm2): Likewise.
3984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3986 * config/rl78/rl78.md (anddi3): New define_expand.
3988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3990 * config/rl78/rl78.md (umindi3): New define_expand.
3992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3994 * config/rl78/rl78.md (smindi3): New define_expand.
3996 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3998 * config/rl78/rl78.md (smaxdi3): New define_expand.
4000 2018-01-22 Carl Love <cel@us.ibm.com>
4002 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4003 LVX_V1TI): Add macro expansion.
4004 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4005 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4006 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4007 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4008 Change check to determine if the instruction is a byte reversing
4009 entry. Fix typo in comment.
4010 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4011 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4012 Add def_builtin calls for new builtins.
4013 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4014 Add define_insn expansion.
4016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4018 * config/rl78/rl78.md (umaxdi3): New define_expand.
4020 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4022 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4023 for non-QImode registers.
4025 2018-01-22 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/83963
4028 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4029 including the loop exit block.
4030 (scop_detection::merge_sese): Likewise.
4031 (scop_detection::add_scop): Do it here instead.
4033 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4035 * doc/sourcebuild.texi (arm_softfloat): Document.
4037 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4040 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4041 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4042 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4044 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4045 David Edelsohn <dje.gcc@gmail.com>
4048 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4049 Change "crset eq" to "crset 2".
4050 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4051 (*call_indirect_aix<mode>_nospec): Likewise.
4052 (*call_value_indirect_aix<mode>_nospec): Likewise.
4053 (*call_indirect_elfv2<mode>_nospec): Likewise.
4054 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4055 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4056 change assembly output from . to $.
4057 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4058 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4059 (*tablejump<mode>_internal1_nospec): Likewise.
4061 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4064 * config/sh/sh_optimize_sett_clrt.cc:
4065 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4067 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4069 PR tree-optimization/83940
4070 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4071 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4072 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4073 use it to pass back the definition type.
4074 (vect_check_store_rhs): Likewise rhs_dt_out.
4075 (vect_build_gather_load_calls): Add a mask_dt argument and use
4076 it instead of a call to vect_is_simple_use.
4077 (vectorizable_store): Update calls to vect_check_load_store_mask
4078 and vect_check_store_rhs. Use the dt returned by the latter instead
4079 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4080 instead of calls to vect_is_simple_use. Pass the scalar rather
4081 than the vector operand to vect_is_simple_use when handling
4082 second and subsequent copies of an rhs value.
4083 (vectorizable_load): Update calls to vect_check_load_store_mask
4084 and vect_build_gather_load_calls. Use the cached mask_dt and
4085 gs_info.offset_dt instead of calls to vect_is_simple_use.
4087 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4090 * tree-emutls.c: Include gimplify.h.
4091 (lower_emutls_2): New function.
4092 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4093 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4094 it before further processing.
4097 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4098 UINTVAL (trueop1) instead of INTVAL (op1).
4100 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4104 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4105 INCOMING_FRAME_SP_OFFSET if not defined.
4106 (scan_trace): Add ENTRY argument. If true and
4107 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4108 emit a note to adjust the CFA offset.
4109 (create_cfi_notes): Adjust scan_trace callers.
4110 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4111 INCOMING_FRAME_SP_OFFSET in the CIE.
4112 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4113 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4115 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4116 * doc/tm.texi: Regenerated.
4118 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4120 PR rtl-optimization/83147
4121 * lra-constraints.c (remove_inheritance_pseudos): Use
4122 lra_substitute_pseudo_within_insn.
4124 2018-01-19 Tom de Vries <tom@codesourcery.com>
4125 Cesar Philippidis <cesar@codesourcery.com>
4128 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4130 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4133 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4134 spaces for function labels.
4136 2018-01-19 Martin Liska <mliska@suse.cz>
4138 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4139 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4140 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4141 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4142 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4143 (PRED_CONST_RETURN): Change from 69 to 65.
4144 (PRED_NULL_RETURN): Change from 91 to 71.
4145 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4146 (PRED_LOOP_GUARD): Change from 66 to 73.
4148 2018-01-19 Martin Liska <mliska@suse.cz>
4150 * predict.c (predict_insn_def): Add new assert.
4151 (struct branch_predictor): Change type to signed integer.
4152 (test_prediction_value_range): Amend test to cover
4154 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4155 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4156 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4157 (PRED_LOOP_IV_COMPARE): Likewise.
4158 * predict.h (PROB_UNINITIALIZED): Define new constant.
4160 2018-01-19 Martin Liska <mliska@suse.cz>
4162 * predict.c (dump_prediction): Add new format for
4163 analyze_brprob.py script which is enabled with -details
4165 * profile-count.h (precise_p): New function.
4167 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4169 PR tree-optimization/83922
4170 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4171 there are no statements that need masking.
4172 (vect_active_double_reduction_p): New function.
4173 (vect_analyze_loop_operations): Use it when handling phis that
4174 are not in the loop header.
4176 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4178 PR tree-optimization/83914
4179 * tree-vect-loop.c (vectorizable_induction): Don't convert
4180 init_expr or apply the peeling adjustment for inductions
4181 that are nested within the vectorized loop.
4183 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4185 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4188 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4192 * function.h (gimplify_parameters): Add gimple_seq * argument.
4193 * function.c: Include gimple.h and options.h.
4194 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4195 for the added local temporaries if needed.
4196 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4197 if there are any parameter cleanups, wrap whole body into a
4198 try/finally with the cleanups.
4200 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4203 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4204 Use GET_MODE_CLASS for scalar floating point.
4206 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4210 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4211 Fix call of call_cgraph_insertion_hooks.
4213 2018-01-18 Martin Sebor <msebor@redhat.com>
4215 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4217 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4220 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4223 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4226 * common.opt: (-ffile-prefix-map): New option.
4227 * opts.c (common_handle_option): Defer it.
4228 * opts-global.c (handle_common_deferred_options): Handle it.
4229 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4230 * file-prefix-map.h: New file.
4231 (remap_debug_filename, add_debug_prefix_map): ...here.
4232 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4233 * final.c (debug_prefix_map, add_debug_prefix_map
4234 remap_debug_filename): Move to...
4235 * file-prefix-map.c: New file.
4236 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4237 generalize, get rid of alloca(), use strrchr() instead of strchr().
4238 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4239 Implement in terms of add_prefix_map().
4240 (remap_macro_filename, remap_debug_filename): Implement in term of
4242 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4243 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4244 * dbxout.c: Include file-prefix-map.h.
4245 * varasm.c: Likewise.
4246 * vmsdbgout.c: Likewise.
4247 * xcoffout.c: Likewise.
4248 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4249 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4250 * doc/invoke.texi (-ffile-prefix-map): Document.
4251 (-fdebug-prefix-map): Update description.
4253 2018-01-18 Martin Liska <mliska@suse.cz>
4255 * config/i386/i386.c (indirect_thunk_name): Document that also
4257 (output_indirect_thunk): Document why both instructions
4258 (pause and lfence) are generated.
4260 2018-01-18 Richard Biener <rguenther@suse.de>
4262 PR tree-optimization/83887
4263 * graphite-scop-detection.c
4264 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4265 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4266 (scop_detection::merge_sese): Re-implement with a flood-fill
4267 algorithm that properly finds a SESE region if it exists.
4269 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4272 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4273 pointer_diff optimizations use view_convert instead of convert.
4275 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4277 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4278 Generate different code for -mno-speculate-indirect-jumps.
4279 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4280 (*call_indirect_aix<mode>): Disable for
4281 -mno-speculate-indirect-jumps.
4282 (*call_indirect_aix<mode>_nospec): New define_insn.
4283 (*call_value_indirect_aix<mode>): Disable for
4284 -mno-speculate-indirect-jumps.
4285 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4286 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4287 -mno-speculate-indirect-jumps.
4288 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4290 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4292 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4293 long double type, set the flags for noting the default long double
4294 type, even if we don't pass or return a long double type.
4296 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4299 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4302 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4304 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4305 support for merge[hl].
4306 (fold_mergehl_helper): New helper function.
4307 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4308 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4309 (altivec_vmrglw_direct): Add xxmrglw insn.
4311 2018-01-17 Andrew Waterman <andrew@sifive.com>
4313 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4314 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4316 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4319 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4320 call the lto_location_cache before reading the
4321 DECL_SOURCE_LOCATION of the types.
4323 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4324 Richard Sandiford <richard.sandiford@linaro.org>
4326 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4327 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4328 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4329 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4330 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4332 * config/aarch64/constraints.md (aarch64_movti_operand):
4334 * config/aarch64/predicates.md (Uti): Add new constraint.
4336 2018-01-17 Carl Love <cel@us.ibm.com>
4338 * config/rs6000/vsx.md (define_expand xl_len_r,
4339 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
4340 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
4342 (define_expand, define_insn): Move the shift left from the
4343 define_insn to the define_expand for lxvl and stxvl instructions.
4344 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
4345 and XL_LEN_R definitions to PURE.
4347 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
4349 * config/i386/i386.c (indirect_thunk_name): Declare regno
4350 as unsigned int. Compare regno with INVALID_REGNUM.
4351 (output_indirect_thunk): Ditto.
4352 (output_indirect_thunk_function): Ditto.
4353 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
4354 in the call to output_indirect_thunk_function.
4356 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
4359 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
4360 rather than the size of inner_type to determine the stack slot size
4361 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
4363 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
4366 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
4369 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4371 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
4372 endian Linux systems to optionally enable multilibs for selecting
4373 the long double type if the user configured an explicit type.
4374 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
4375 have no long double multilibs if not defined.
4376 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
4377 warn if the user used -mabi={ieee,ibm}longdouble and we built
4378 multilibs for long double.
4379 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
4380 appropriate multilib option.
4381 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
4383 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
4384 for building long double multilibs.
4385 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
4387 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
4389 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
4392 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
4394 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
4397 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
4400 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
4403 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4405 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
4406 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
4408 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4410 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
4411 different rtl trees depending on TARGET_64BIT.
4412 (rs6000_gen_lvx): Likewise.
4414 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
4416 * config/visium/visium.md (nop): Tweak comment.
4417 (hazard_nop): Likewise.
4419 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4421 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
4422 -mspeculate-indirect-jumps.
4423 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
4424 for -mno-speculate-indirect-jumps.
4425 (*call_indirect_elfv2<mode>_nospec): New define_insn.
4426 (*call_value_indirect_elfv2<mode>): Disable for
4427 -mno-speculate-indirect-jumps.
4428 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
4429 (indirect_jump): Emit different RTL for
4430 -mno-speculate-indirect-jumps.
4431 (*indirect_jump<mode>): Disable for
4432 -mno-speculate-indirect-jumps.
4433 (*indirect_jump<mode>_nospec): New define_insn.
4434 (tablejump): Emit different RTL for
4435 -mno-speculate-indirect-jumps.
4436 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
4437 (tablejumpsi_nospec): New define_expand.
4438 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
4439 (tablejumpdi_nospec): New define_expand.
4440 (*tablejump<mode>_internal1): Disable for
4441 -mno-speculate-indirect-jumps.
4442 (*tablejump<mode>_internal1_nospec): New define_insn.
4443 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
4446 2018-01-16 Artyom Skrobov tyomitch@gmail.com
4448 * caller-save.c (insert_save): Drop unnecessary parameter. All
4451 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4452 Richard Biener <rguenth@suse.de>
4455 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
4456 return early, inline manually is_gimple_sizepos. Make sure if we
4457 call gimplify_expr we don't end up with a gimple constant.
4458 * tree.c (variably_modified_type_p): Don't return true for
4459 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
4460 * gimplify.h (is_gimple_sizepos): Remove.
4462 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4464 PR tree-optimization/83857
4465 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
4466 vectorizable_live_operation for pure SLP statements.
4467 (vectorizable_live_operation): Handle PHIs.
4469 2018-01-16 Richard Biener <rguenther@suse.de>
4471 PR tree-optimization/83867
4472 * tree-vect-stmts.c (vect_transform_stmt): Precompute
4473 nested_in_vect_loop_p since the scalar stmt may get invalidated.
4475 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4478 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
4479 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
4480 If off is not INTEGER_CST, issue a may not be aligned warning
4481 rather than isn't aligned. Use isn%'t rather than isn't.
4482 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
4484 <case MULT_EXPR>: Improve the case when bottom and one of the
4485 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
4486 operand, in that case check if the other operand is multiple of
4487 bottom divided by the INTEGER_CST operand.
4489 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4492 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
4493 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
4494 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
4495 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
4496 * config/pa/pa.c (pa_function_arg_advance): Likewise.
4497 (pa_function_arg, pa_arg_partial_bytes): Likewise.
4498 (pa_function_arg_size): New function.
4500 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4502 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
4503 in a separate statement.
4505 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
4507 PR tree-optimization/83847
4508 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
4509 group gathers and scatters.
4511 2018-01-16 Jakub Jelinek <jakub@redhat.com>
4513 PR rtl-optimization/86620
4514 * params.def (max-sched-ready-insns): Bump minimum value to 1.
4516 PR rtl-optimization/83213
4517 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
4518 to last if both are JUMP_INSNs.
4520 PR tree-optimization/83843
4521 * gimple-ssa-store-merging.c
4522 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
4523 store_immediate_info for bswap/nop orig_stores.
4525 2018-01-15 Andrew Waterman <andrew@sifive.com>
4527 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
4529 <UDIV>: Increase cost if !TARGET_DIV.
4531 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
4533 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
4534 (define_attr "cr_logical_3op"): New.
4535 (cceq_ior_compare): Adjust.
4536 (cceq_ior_compare_complement): Adjust.
4537 (*cceq_rev_compare): Adjust.
4538 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
4539 (is_cracked_insn): Adjust.
4540 (insn_must_be_first_in_group): Adjust.
4541 * config/rs6000/40x.md: Adjust.
4542 * config/rs6000/440.md: Adjust.
4543 * config/rs6000/476.md: Adjust.
4544 * config/rs6000/601.md: Adjust.
4545 * config/rs6000/603.md: Adjust.
4546 * config/rs6000/6xx.md: Adjust.
4547 * config/rs6000/7450.md: Adjust.
4548 * config/rs6000/7xx.md: Adjust.
4549 * config/rs6000/8540.md: Adjust.
4550 * config/rs6000/cell.md: Adjust.
4551 * config/rs6000/e300c2c3.md: Adjust.
4552 * config/rs6000/e500mc.md: Adjust.
4553 * config/rs6000/e500mc64.md: Adjust.
4554 * config/rs6000/e5500.md: Adjust.
4555 * config/rs6000/e6500.md: Adjust.
4556 * config/rs6000/mpc.md: Adjust.
4557 * config/rs6000/power4.md: Adjust.
4558 * config/rs6000/power5.md: Adjust.
4559 * config/rs6000/power6.md: Adjust.
4560 * config/rs6000/power7.md: Adjust.
4561 * config/rs6000/power8.md: Adjust.
4562 * config/rs6000/power9.md: Adjust.
4563 * config/rs6000/rs64.md: Adjust.
4564 * config/rs6000/titan.md: Adjust.
4566 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4568 * config/i386/predicates.md (indirect_branch_operand): Rewrite
4569 ix86_indirect_branch_register logic.
4571 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4573 * config/i386/constraints.md (Bs): Update
4574 ix86_indirect_branch_register check. Don't check
4575 ix86_indirect_branch_register with GOT_memory_operand.
4577 * config/i386/predicates.md (GOT_memory_operand): Don't check
4578 ix86_indirect_branch_register here.
4579 (GOT32_symbol_operand): Likewise.
4581 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4583 * config/i386/predicates.md (constant_call_address_operand):
4584 Rewrite ix86_indirect_branch_register logic.
4585 (sibcall_insn_operand): Likewise.
4587 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4589 * config/i386/constraints.md (Bs): Replace
4590 ix86_indirect_branch_thunk_register with
4591 ix86_indirect_branch_register.
4593 * config/i386/i386.md (indirect_jump): Likewise.
4594 (tablejump): Likewise.
4595 (*sibcall_memory): Likewise.
4596 (*sibcall_value_memory): Likewise.
4597 Peepholes of indirect call and jump via memory: Likewise.
4598 * config/i386/i386.opt: Likewise.
4599 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4600 (GOT_memory_operand): Likewise.
4601 (call_insn_operand): Likewise.
4602 (sibcall_insn_operand): Likewise.
4603 (GOT32_symbol_operand): Likewise.
4605 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4608 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
4609 type rather than type addr's type points to.
4610 (expand_omp_atomic_mutex): Likewise.
4611 (expand_omp_atomic): Likewise.
4613 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
4616 * config/i386/i386.c (output_indirect_thunk_function): Use
4617 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
4618 for __x86_return_thunk.
4620 2018-01-15 Richard Biener <rguenther@suse.de>
4623 * expmed.c (extract_bit_field_1): Fix typo.
4625 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4628 * config/arm/iterators.md (VF): New mode iterator.
4629 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
4630 Remove integer-related logic from pattern.
4631 (neon_vabd<mode>_3): Likewise.
4633 2018-01-15 Jakub Jelinek <jakub@redhat.com>
4636 * common.opt (fstrict-overflow): No longer an alias.
4637 (fwrapv-pointer): New option.
4638 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
4639 also for pointer types based on flag_wrapv_pointer.
4640 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
4641 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
4642 opts->x_flag_wrapv got set.
4643 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
4644 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
4645 POINTER_TYPE_OVERFLOW_UNDEFINED.
4646 * match.pd: Likewise in address comparison pattern.
4647 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
4649 2018-01-15 Richard Biener <rguenther@suse.de>
4652 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
4653 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
4654 Reset type names to their identifier if their TYPE_DECL doesn't
4655 have linkage (and thus is used for ODR and devirt).
4656 (save_debug_info_for_decl): Remove.
4657 (save_debug_info_for_type): Likewise.
4658 (add_tree_to_fld_list): Adjust.
4659 * tree-pretty-print.c (dump_generic_node): Make dumping of
4660 type names more robust.
4662 2018-01-15 Richard Biener <rguenther@suse.de>
4664 * BASE-VER: Bump to 8.0.1.
4666 2018-01-14 Martin Sebor <msebor@redhat.com>
4669 * builtins.c (check_access): Avoid warning when the no-warning bit
4672 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4674 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4675 * ira-color (allocno_hard_regs_compare): Likewise.
4677 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4680 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4681 Use .pushsection/.popsection.
4683 2018-01-14 Martin Sebor <msebor@redhat.com>
4686 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4688 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4690 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4691 entry from extra_headers.
4692 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4693 extra_headers, make the list bitwise identical to the i?86-*-* one.
4695 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4697 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4698 -mcmodel=large with -mindirect-branch=thunk,
4699 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4700 -mfunction-return=thunk-extern.
4701 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4702 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4703 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4705 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4707 * config/i386/i386.c (print_reg): Print the name of the full
4708 integer register without '%'.
4709 (ix86_print_operand): Handle 'V'.
4710 * doc/extend.texi: Document 'V' modifier.
4712 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4714 * config/i386/constraints.md (Bs): Disallow memory operand for
4715 -mindirect-branch-register.
4717 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4718 (GOT_memory_operand): Likewise.
4719 (call_insn_operand): Likewise.
4720 (sibcall_insn_operand): Likewise.
4721 (GOT32_symbol_operand): Likewise.
4722 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4723 for -mindirect-branch-register.
4724 (tablejump): Likewise.
4725 (*sibcall_memory): Likewise.
4726 (*sibcall_value_memory): Likewise.
4727 Disallow peepholes of indirect call and jump via memory for
4728 -mindirect-branch-register.
4729 (*call_pop): Replace m with Bw.
4730 (*call_value_pop): Likewise.
4731 (*sibcall_pop_memory): Replace m with Bs.
4732 * config/i386/i386.opt (mindirect-branch-register): New option.
4733 * doc/invoke.texi: Document -mindirect-branch-register option.
4735 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4737 * config/i386/i386-protos.h (ix86_output_function_return): New.
4738 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4739 set function_return_type.
4740 (indirect_thunk_name): Add ret_p to indicate thunk for function
4742 (output_indirect_thunk_function): Pass false to
4743 indirect_thunk_name.
4744 (ix86_output_indirect_branch_via_reg): Likewise.
4745 (ix86_output_indirect_branch_via_push): Likewise.
4746 (output_indirect_thunk_function): Create alias for function
4747 return thunk if regno < 0.
4748 (ix86_output_function_return): New function.
4749 (ix86_handle_fndecl_attribute): Handle function_return.
4750 (ix86_attribute_table): Add function_return.
4751 * config/i386/i386.h (machine_function): Add
4752 function_return_type.
4753 * config/i386/i386.md (simple_return_internal): Use
4754 ix86_output_function_return.
4755 (simple_return_internal_long): Likewise.
4756 * config/i386/i386.opt (mfunction-return=): New option.
4757 (indirect_branch): Mention -mfunction-return=.
4758 * doc/extend.texi: Document function_return function attribute.
4759 * doc/invoke.texi: Document -mfunction-return= option.
4761 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4763 * config/i386/i386-opts.h (indirect_branch): New.
4764 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4765 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4766 with local indirect jump when converting indirect call and jump.
4767 (ix86_set_indirect_branch_type): New.
4768 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4769 (indirectlabelno): New.
4770 (indirect_thunk_needed): Likewise.
4771 (indirect_thunk_bnd_needed): Likewise.
4772 (indirect_thunks_used): Likewise.
4773 (indirect_thunks_bnd_used): Likewise.
4774 (INDIRECT_LABEL): Likewise.
4775 (indirect_thunk_name): Likewise.
4776 (output_indirect_thunk): Likewise.
4777 (output_indirect_thunk_function): Likewise.
4778 (ix86_output_indirect_branch_via_reg): Likewise.
4779 (ix86_output_indirect_branch_via_push): Likewise.
4780 (ix86_output_indirect_branch): Likewise.
4781 (ix86_output_indirect_jmp): Likewise.
4782 (ix86_code_end): Call output_indirect_thunk_function if needed.
4783 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4785 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4786 (ix86_attribute_table): Add indirect_branch.
4787 * config/i386/i386.h (machine_function): Add indirect_branch_type
4788 and has_local_indirect_jump.
4789 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4791 (tablejump): Likewise.
4792 (*indirect_jump): Use ix86_output_indirect_jmp.
4793 (*tablejump_1): Likewise.
4794 (simple_return_indirect_internal): Likewise.
4795 * config/i386/i386.opt (mindirect-branch=): New option.
4796 (indirect_branch): New.
4799 (thunk-inline): Likewise.
4800 (thunk-extern): Likewise.
4801 * doc/extend.texi: Document indirect_branch function attribute.
4802 * doc/invoke.texi: Document -mindirect-branch= option.
4804 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4807 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4809 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4811 * ipa-inline.c (want_inline_small_function_p): Return false if
4812 inlining has already failed with CIF_FINAL_ERROR.
4813 (update_caller_keys): Call want_inline_small_function_p before
4815 (update_callee_keys): Likewise.
4817 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4819 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4821 (rs6000_quadword_masked_address_p): Likewise.
4822 (quad_aligned_load_p): Likewise.
4823 (quad_aligned_store_p): Likewise.
4824 (const_load_sequence_p): Add comment to describe the outer-most loop.
4825 (mimic_memory_attributes_and_flags): New function.
4826 (rs6000_gen_stvx): Likewise.
4827 (replace_swapped_aligned_store): Likewise.
4828 (rs6000_gen_lvx): Likewise.
4829 (replace_swapped_aligned_load): Likewise.
4830 (replace_swapped_load_constant): Capitalize argument name in
4831 comment describing this function.
4832 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4833 and stores that access quad-word aligned addresses and replace
4834 with stvx or lvx instructions when appropriate.
4835 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4836 New function prototype.
4837 (rs6000_quadword_masked_address_p): Likewise.
4838 (rs6000_gen_lvx): Likewise.
4839 (rs6000_gen_stvx): Likewise.
4840 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4841 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4842 when memory address is aligned.
4843 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4844 this split to select lvx instruction when memory address is aligned.
4845 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4846 instruction when memory address is aligned.
4847 (*vsx_le_perm_load_v16qi): Likewise.
4848 (four unnamed splitters): Modify to select the stvx instruction
4849 when memory is aligned.
4851 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4853 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4854 which appears in the queue multiple times.
4856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4857 Alan Hayward <alan.hayward@arm.com>
4858 David Sherwood <david.sherwood@arm.com>
4860 * tree-vectorizer.h (vec_lower_bound): New structure.
4861 (_loop_vec_info): Add check_nonzero and lower_bounds.
4862 (LOOP_VINFO_CHECK_NONZERO): New macro.
4863 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4864 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4865 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4866 fields. Make seg_len the distance travelled, not including the
4868 (dr_direction_indicator): Declare.
4869 (dr_zero_step_indicator): Likewise.
4870 (dr_known_forward_stride_p): Likewise.
4871 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4873 (runtime_alias_check_p): Allow runtime alias checks with
4875 (operator ==): Compare access_size and align.
4876 (prune_runtime_alias_test_list): Rework for new distinction between
4877 the access_size and seg_len.
4878 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4880 (get_segment_min_max): New function.
4881 (create_intersect_range_checks): Use it.
4882 (dr_step_indicator): New function.
4883 (dr_direction_indicator): Likewise.
4884 (dr_zero_step_indicator): Likewise.
4885 (dr_known_forward_stride_p): Likewise.
4886 * tree-loop-distribution.c (data_ref_segment_size): Return
4887 DR_STEP * (niters - 1).
4888 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4890 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4891 (vect_preserves_scalar_order_p): New function, split out from...
4892 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4893 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4894 (vect_vfa_access_size): New function.
4895 (vect_vfa_align): Likewise.
4896 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4897 (dump_lower_bound): New function.
4898 (vect_check_lower_bound): Likewise.
4899 (vect_small_gap_p): Likewise.
4900 (vectorizable_with_step_bound_p): Likewise.
4901 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4902 depencies if the vectorization factor is 1. Convert the checks
4903 for nonzero steps into checks on the bounds of DR_STEP. Try using
4904 a bunds check for variable steps if the minimum required step is
4905 relatively small. Update calls to the dr_with_seg_len
4906 constructor and to vect_compile_time_alias.
4907 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4909 (vect_loop_versioning): Call it.
4910 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4912 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4915 Alan Hayward <alan.hayward@arm.com>
4916 David Sherwood <david.sherwood@arm.com>
4918 * doc/sourcebuild.texi (vect_scatter_store): Document.
4919 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4921 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4923 * genopinit.c (main): Add supports_vec_scatter_store and
4924 supports_vec_scatter_store_cached to target_optabs.
4925 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4926 IFN_MASK_SCATTER_STORE.
4927 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4929 * internal-fn.h (internal_store_fn_p): Declare.
4930 (internal_fn_stored_value_index): Likewise.
4931 * internal-fn.c (scatter_store_direct): New macro.
4932 (expand_scatter_store_optab_fn): New function.
4933 (direct_scatter_store_optab_supported_p): New macro.
4934 (internal_store_fn_p): New function.
4935 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4936 IFN_MASK_SCATTER_STORE.
4937 (internal_fn_mask_index): Likewise.
4938 (internal_fn_stored_value_index): New function.
4939 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4941 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4942 * optabs-query.c (supports_vec_scatter_store_p): New function.
4943 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4944 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4945 true for scatter stores.
4946 (vect_gather_scatter_fn_p): Handle scatter stores too.
4947 (vect_check_gather_scatter): Consider using scatter stores if
4948 supports_vec_scatter_store_p.
4949 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4951 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4952 internal_fn_stored_value_index.
4953 (check_load_store_masking): Handle scatter stores too.
4954 (vect_get_store_rhs): Make public.
4955 (vectorizable_call): Use internal_store_fn_p.
4956 (vectorizable_store): Handle scatter store internal functions.
4957 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4958 when deciding whether the end of the group has been reached.
4959 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4960 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4961 (mask_scatter_store<mode>): New insns.
4963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4964 Alan Hayward <alan.hayward@arm.com>
4965 David Sherwood <david.sherwood@arm.com>
4967 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4968 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4969 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4971 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4972 Use vect_truncate_gather_scatter_offset if we can't treat the
4973 operation as a normal gather load or scatter store.
4974 (get_group_load_store_type): Take the gather_scatter_info
4975 as argument. Try using a gather load or scatter store for
4976 single-element groups.
4977 (get_load_store_type): Update calls to get_group_load_store_type
4978 and vect_use_strided_gather_scatters_p.
4980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4981 Alan Hayward <alan.hayward@arm.com>
4982 David Sherwood <david.sherwood@arm.com>
4984 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4985 optional tree argument.
4986 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4988 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4989 but continue to use the current value as a fallback.
4990 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4991 to compare the updates.
4992 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4993 (get_load_store_type): Use it when handling a strided access.
4994 (vect_get_strided_load_store_ops): New function.
4995 (vect_get_data_ptr_increment): Likewise.
4996 (vectorizable_load): Handle strided gather loads. Always pass
4997 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5000 Alan Hayward <alan.hayward@arm.com>
5001 David Sherwood <david.sherwood@arm.com>
5003 * doc/md.texi (gather_load@var{m}): Document.
5004 (mask_gather_load@var{m}): Likewise.
5005 * genopinit.c (main): Add supports_vec_gather_load and
5006 supports_vec_gather_load_cached to target_optabs.
5007 * optabs-tree.c (init_tree_optimization_optabs): Use
5008 ggc_cleared_alloc to allocate target_optabs.
5009 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5010 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5012 * internal-fn.h (internal_load_fn_p): Declare.
5013 (internal_gather_scatter_fn_p): Likewise.
5014 (internal_fn_mask_index): Likewise.
5015 (internal_gather_scatter_fn_supported_p): Likewise.
5016 * internal-fn.c (gather_load_direct): New macro.
5017 (expand_gather_load_optab_fn): New function.
5018 (direct_gather_load_optab_supported_p): New macro.
5019 (direct_internal_fn_optab): New function.
5020 (internal_load_fn_p): Likewise.
5021 (internal_gather_scatter_fn_p): Likewise.
5022 (internal_fn_mask_index): Likewise.
5023 (internal_gather_scatter_fn_supported_p): Likewise.
5024 * optabs-query.c (supports_at_least_one_mode_p): New function.
5025 (supports_vec_gather_load_p): Likewise.
5026 * optabs-query.h (supports_vec_gather_load_p): Declare.
5027 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5028 and memory_type field.
5029 (NUM_PATTERNS): Bump to 15.
5030 * tree-vect-data-refs.c: Include internal-fn.h.
5031 (vect_gather_scatter_fn_p): New function.
5032 (vect_describe_gather_scatter_call): Likewise.
5033 (vect_check_gather_scatter): Try using internal functions for
5034 gather loads. Recognize existing calls to a gather load function.
5035 (vect_analyze_data_refs): Consider using gather loads if
5036 supports_vec_gather_load_p.
5037 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5038 (vect_get_gather_scatter_offset_type): Likewise.
5039 (vect_convert_mask_for_vectype): Likewise.
5040 (vect_add_conversion_to_patterm): Likewise.
5041 (vect_try_gather_scatter_pattern): Likewise.
5042 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5043 (vect_vect_recog_func_ptrs): Add it.
5044 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5045 internal_fn_mask_index and internal_gather_scatter_fn_p.
5046 (check_load_store_masking): Take the gather_scatter_info as an
5047 argument and handle gather loads.
5048 (vect_get_gather_scatter_ops): New function.
5049 (vectorizable_call): Check internal_load_fn_p.
5050 (vectorizable_load): Likewise. Handle gather load internal
5052 (vectorizable_store): Update call to check_load_store_masking.
5053 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5054 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5055 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5056 (aarch64_gather_scale_operand_d): New predicates.
5057 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5058 (mask_gather_load<mode>): New insns.
5060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5061 Alan Hayward <alan.hayward@arm.com>
5062 David Sherwood <david.sherwood@arm.com>
5064 * optabs.def (fold_left_plus_optab): New optab.
5065 * doc/md.texi (fold_left_plus_@var{m}): Document.
5066 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5067 * internal-fn.c (fold_left_direct): Define.
5068 (expand_fold_left_optab_fn): Likewise.
5069 (direct_fold_left_optab_supported_p): Likewise.
5070 * fold-const-call.c (fold_const_fold_left): New function.
5071 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5072 * tree-parloops.c (valid_reduction_p): New function.
5073 (gather_scalar_reductions): Use it.
5074 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5075 (vect_finish_replace_stmt): Declare.
5076 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5077 (needs_fold_left_reduction_p): New function, split out from...
5078 (vect_is_simple_reduction): ...here. Accept reductions that
5079 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5080 (vect_force_simple_reduction): Also store the reduction type in
5081 the assignment's STMT_VINFO_REDUC_TYPE.
5082 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5083 (merge_with_identity): New function.
5084 (vect_expand_fold_left): Likewise.
5085 (vectorize_fold_left_reduction): Likewise.
5086 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5087 scalar phi in place for it. Check for target support and reject
5088 cases that would reassociate the operation. Defer the transform
5089 phase to vectorize_fold_left_reduction.
5090 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5091 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5092 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5096 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5097 call to ifc_temp_var.
5099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5100 Alan Hayward <alan.hayward@arm.com>
5101 David Sherwood <david.sherwood@arm.com>
5103 * target.def (legitimize_address_displacement): Take the original
5104 offset as a poly_int.
5105 * targhooks.h (default_legitimize_address_displacement): Update
5107 * targhooks.c (default_legitimize_address_displacement): Likewise.
5108 * doc/tm.texi: Regenerate.
5109 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5110 as an argument, moving assert of ad->disp == ad->disp_term to...
5111 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5112 Try calling targetm.legitimize_address_displacement before expanding
5113 the address rather than afterwards, and adjust for the new interface.
5114 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5115 Match the new hook interface. Handle SVE addresses.
5116 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5121 * Makefile.in (OBJS): Add early-remat.o.
5122 * target.def (select_early_remat_modes): New hook.
5123 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5124 * doc/tm.texi: Regenerate.
5125 * targhooks.h (default_select_early_remat_modes): Declare.
5126 * targhooks.c (default_select_early_remat_modes): New function.
5127 * timevar.def (TV_EARLY_REMAT): New timevar.
5128 * passes.def (pass_early_remat): New pass.
5129 * tree-pass.h (make_pass_early_remat): Declare.
5130 * early-remat.c: New file.
5131 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5133 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5136 Alan Hayward <alan.hayward@arm.com>
5137 David Sherwood <david.sherwood@arm.com>
5139 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5140 vfm1 with a bound_epilog parameter.
5141 (vect_do_peeling): Update calls accordingly, and move the prologue
5142 call earlier in the function. Treat the base bound_epilog as 0 for
5143 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5144 this base when peeling for gaps.
5145 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5146 with fully-masked loops.
5147 (vect_estimate_min_profitable_iters): Handle the single peeled
5148 iteration in that case.
5150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5151 Alan Hayward <alan.hayward@arm.com>
5152 David Sherwood <david.sherwood@arm.com>
5154 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5155 single-element interleaving even if the size is not a power of 2.
5156 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5157 accesses for single-element interleaving if the group size is
5160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5161 Alan Hayward <alan.hayward@arm.com>
5162 David Sherwood <david.sherwood@arm.com>
5164 * doc/md.texi (fold_extract_last_@var{m}): Document.
5165 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5166 * optabs.def (fold_extract_last_optab): New optab.
5167 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5168 * internal-fn.c (fold_extract_direct): New macro.
5169 (expand_fold_extract_optab_fn): Likewise.
5170 (direct_fold_extract_optab_supported_p): Likewise.
5171 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5172 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5173 EXTRACT_LAST_REDUCTION.
5174 (get_initial_def_for_reduction): Do not create an initial vector
5175 for EXTRACT_LAST_REDUCTION reductions.
5176 (vectorizable_reduction): Leave the scalar phi in place for
5177 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5178 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5179 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5180 transform phase to vectorizable_condition.
5181 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5183 (vect_finish_stmt_generation): ...here.
5184 (vect_finish_replace_stmt): New function.
5185 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5186 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5188 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5191 Alan Hayward <alan.hayward@arm.com>
5192 David Sherwood <david.sherwood@arm.com>
5194 * doc/md.texi (extract_last_@var{m}): Document.
5195 * optabs.def (extract_last_optab): New optab.
5196 * internal-fn.def (EXTRACT_LAST): New internal function.
5197 * internal-fn.c (cond_unary_direct): New macro.
5198 (expand_cond_unary_optab_fn): Likewise.
5199 (direct_cond_unary_optab_supported_p): Likewise.
5200 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5201 loops using EXTRACT_LAST.
5202 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5203 (extract_last_<mode>): ...this optab.
5204 (vec_extract<mode><Vel>): Update accordingly.
5206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5207 Alan Hayward <alan.hayward@arm.com>
5208 David Sherwood <david.sherwood@arm.com>
5210 * target.def (empty_mask_is_expensive): New hook.
5211 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5212 * doc/tm.texi: Regenerate.
5213 * targhooks.h (default_empty_mask_is_expensive): Declare.
5214 * targhooks.c (default_empty_mask_is_expensive): New function.
5215 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5216 if the target says that empty masks are expensive.
5217 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5219 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5222 Alan Hayward <alan.hayward@arm.com>
5223 David Sherwood <david.sherwood@arm.com>
5225 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5226 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5227 (vect_use_loop_mask_for_alignment_p): New function.
5228 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5229 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5230 niters_skip argument. Make sure that the first niters_skip elements
5231 of the first iteration are inactive.
5232 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5233 Update call to vect_set_loop_masks_directly.
5234 (get_misalign_in_elems): New function, split out from...
5235 (vect_gen_prolog_loop_niters): ...here.
5236 (vect_update_init_of_dr): Take a code argument that specifies whether
5237 the adjustment should be added or subtracted.
5238 (vect_update_init_of_drs): Likewise.
5239 (vect_prepare_for_masked_peels): New function.
5240 (vect_do_peeling): Skip prologue peeling if we're using a mask
5241 instead. Update call to vect_update_inits_of_drs.
5242 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5244 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5245 alignment. Do not include the number of peeled iterations in
5246 the minimum threshold in that case.
5247 (vectorizable_induction): Adjust the start value down by
5248 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5249 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5250 Take the number of skipped iterations into account when calculating
5252 * tree-vect-stmts.c (vect_gen_while_not): New function.
5254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5255 Alan Hayward <alan.hayward@arm.com>
5256 David Sherwood <david.sherwood@arm.com>
5258 * doc/sourcebuild.texi (vect_fully_masked): Document.
5259 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5261 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5263 (vect_analyze_loop_2): ...here. Don't check the vectorization
5264 factor against the number of loop iterations if the loop is
5267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5268 Alan Hayward <alan.hayward@arm.com>
5269 David Sherwood <david.sherwood@arm.com>
5271 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5272 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5273 (dump_groups): Update accordingly.
5274 (iv_use::mem_type): New member variable.
5275 (address_p): New function.
5276 (record_use): Add a mem_type argument and initialize the new
5278 (record_group_use): Add a mem_type argument. Use address_p.
5279 Remove obsolete null checks of base_object. Update call to record_use.
5280 (find_interesting_uses_op): Update call to record_group_use.
5281 (find_interesting_uses_cond): Likewise.
5282 (find_interesting_uses_address): Likewise.
5283 (get_mem_type_for_internal_fn): New function.
5284 (find_address_like_use): Likewise.
5285 (find_interesting_uses_stmt): Try find_address_like_use before
5286 calling find_interesting_uses_op.
5287 (addr_offset_valid_p): Use the iv mem_type field as the type
5288 of the addressed memory.
5289 (add_autoinc_candidates): Likewise.
5290 (get_address_cost): Likewise.
5291 (split_small_address_groups_p): Use address_p.
5292 (split_address_groups): Likewise.
5293 (add_iv_candidate_for_use): Likewise.
5294 (autoinc_possible_for_pair): Likewise.
5295 (rewrite_groups): Likewise.
5296 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5297 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5298 (get_alias_ptr_type_for_ptr_address): New function.
5299 (rewrite_use_address): Rewrite address uses in calls that were
5300 identified by find_address_like_use.
5302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5303 Alan Hayward <alan.hayward@arm.com>
5304 David Sherwood <david.sherwood@arm.com>
5306 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5308 * gimple-expr.h (is_gimple_addressable: Likewise.
5309 * gimple-expr.c (is_gimple_address): Likewise.
5310 * internal-fn.c (expand_call_mem_ref): New function.
5311 (expand_mask_load_optab_fn): Use it.
5312 (expand_mask_store_optab_fn): Likewise.
5314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5315 Alan Hayward <alan.hayward@arm.com>
5316 David Sherwood <david.sherwood@arm.com>
5318 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5319 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5320 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5321 (cond_umax@var{mode}): Document.
5322 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5323 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5324 (cond_umin_optab, cond_umax_optab): New optabs.
5325 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5326 (COND_IOR, COND_XOR): New internal functions.
5327 * internal-fn.h (get_conditional_internal_fn): Declare.
5328 * internal-fn.c (cond_binary_direct): New macro.
5329 (expand_cond_binary_optab_fn): Likewise.
5330 (direct_cond_binary_optab_supported_p): Likewise.
5331 (get_conditional_internal_fn): New function.
5332 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5333 Cope with reduction statements that are vectorized as calls rather
5335 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5336 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
5337 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
5338 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5339 (UNSPEC_COND_EOR): New unspecs.
5340 (optab): Add mappings for them.
5341 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
5342 (sve_int_op, sve_fp_op): New int attributes.
5344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5345 Alan Hayward <alan.hayward@arm.com>
5346 David Sherwood <david.sherwood@arm.com>
5348 * optabs.def (while_ult_optab): New optab.
5349 * doc/md.texi (while_ult@var{m}@var{n}): Document.
5350 * internal-fn.def (WHILE_ULT): New internal function.
5351 * internal-fn.h (direct_internal_fn_supported_p): New override
5352 that takes two types as argument.
5353 * internal-fn.c (while_direct): New macro.
5354 (expand_while_optab_fn): New function.
5355 (convert_optab_supported_p): Likewise.
5356 (direct_while_optab_supported_p): New macro.
5357 * wide-int.h (wi::udiv_ceil): New function.
5358 * tree-vectorizer.h (rgroup_masks): New structure.
5359 (vec_loop_masks): New typedef.
5360 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
5362 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
5363 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
5364 (vect_max_vf): New function.
5365 (slpeel_make_loop_iterate_ntimes): Delete.
5366 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
5367 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
5368 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
5369 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
5370 internal-fn.h, stor-layout.h and optabs-query.h.
5371 (vect_set_loop_mask): New function.
5372 (add_preheader_seq): Likewise.
5373 (add_header_seq): Likewise.
5374 (interleave_supported_p): Likewise.
5375 (vect_maybe_permute_loop_masks): Likewise.
5376 (vect_set_loop_masks_directly): Likewise.
5377 (vect_set_loop_condition_masked): Likewise.
5378 (vect_set_loop_condition_unmasked): New function, split out from
5379 slpeel_make_loop_iterate_ntimes.
5380 (slpeel_make_loop_iterate_ntimes): Rename to..
5381 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
5382 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
5383 (vect_do_peeling): Update call accordingly.
5384 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
5386 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5387 mask_compare_type, can_fully_mask_p and fully_masked_p.
5388 (release_vec_loop_masks): New function.
5389 (_loop_vec_info): Use it to free the loop masks.
5390 (can_produce_all_loop_masks_p): New function.
5391 (vect_get_max_nscalars_per_iter): Likewise.
5392 (vect_verify_full_masking): Likewise.
5393 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
5394 retries, and free the mask rgroups before retrying. Check loop-wide
5395 reasons for disallowing fully-masked loops. Make the final decision
5396 about whether use a fully-masked loop or not.
5397 (vect_estimate_min_profitable_iters): Do not assume that peeling
5398 for the number of iterations will be needed for fully-masked loops.
5399 (vectorizable_reduction): Disable fully-masked loops.
5400 (vectorizable_live_operation): Likewise.
5401 (vect_halve_mask_nunits): New function.
5402 (vect_double_mask_nunits): Likewise.
5403 (vect_record_loop_mask): Likewise.
5404 (vect_get_loop_mask): Likewise.
5405 (vect_transform_loop): Handle the case in which the final loop
5406 iteration might handle a partial vector. Call vect_set_loop_condition
5407 instead of slpeel_make_loop_iterate_ntimes.
5408 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
5409 (check_load_store_masking): New function.
5410 (prepare_load_store_mask): Likewise.
5411 (vectorizable_store): Handle fully-masked loops.
5412 (vectorizable_load): Likewise.
5413 (supportable_widening_operation): Use vect_halve_mask_nunits for
5415 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
5416 (vect_gen_while): New function.
5417 * config/aarch64/aarch64.md (umax<mode>3): New expander.
5418 (aarch64_uqdec<mode>): New insn.
5420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5421 Alan Hayward <alan.hayward@arm.com>
5422 David Sherwood <david.sherwood@arm.com>
5424 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
5425 (reduc_xor_scal_optab): New optabs.
5426 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
5427 (reduc_xor_scal_@var{m}): Document.
5428 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
5429 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
5431 * fold-const-call.c (fold_const_call): Handle them.
5432 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
5433 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
5434 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
5435 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
5436 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
5437 (UNSPEC_XORV): New unspecs.
5438 (optab): Add entries for them.
5439 (BITWISEV): New int iterator.
5440 (bit_reduc_op): New int attributes.
5442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5443 Alan Hayward <alan.hayward@arm.com>
5444 David Sherwood <david.sherwood@arm.com>
5446 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
5447 * internal-fn.def (VEC_SHL_INSERT): New internal function.
5448 * optabs.def (vec_shl_insert_optab): New optab.
5449 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
5450 (duplicate_and_interleave): Likewise.
5451 * tree-vect-loop.c: Include internal-fn.h.
5452 (neutral_op_for_slp_reduction): New function, split out from
5453 get_initial_defs_for_reduction.
5454 (get_initial_def_for_reduction): Handle option 2 for variable-length
5455 vectors by loading the neutral value into a vector and then shifting
5456 the initial value into element 0.
5457 (get_initial_defs_for_reduction): Replace the code argument with
5458 the neutral value calculated by neutral_op_for_slp_reduction.
5459 Use gimple_build_vector for constant-length vectors.
5460 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
5461 but the first group_size elements have a neutral value.
5462 Use duplicate_and_interleave otherwise.
5463 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
5464 Update call to get_initial_defs_for_reduction. Handle SLP
5465 reductions for variable-length vectors by creating one vector
5466 result for each scalar result, with the elements associated
5467 with other scalar results stubbed out with the neutral value.
5468 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
5469 Require IFN_VEC_SHL_INSERT for double reductions on
5470 variable-length vectors, or SLP reductions that have
5471 a neutral value. Require can_duplicate_and_interleave_p
5472 support for variable-length unchained SLP reductions if there
5473 is no neutral value, such as for MIN/MAX reductions. Also require
5474 the number of vector elements to be a multiple of the number of
5475 SLP statements when doing variable-length unchained SLP reductions.
5476 Update call to vect_create_epilog_for_reduction.
5477 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
5478 and remove initial values.
5479 (duplicate_and_interleave): Make public.
5480 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
5481 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
5483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5484 Alan Hayward <alan.hayward@arm.com>
5485 David Sherwood <david.sherwood@arm.com>
5487 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
5488 (can_duplicate_and_interleave_p): New function.
5489 (vect_get_and_check_slp_defs): Take the vector of statements
5490 rather than just the current one. Remove excess parentheses.
5491 Restriction rejectinon of vect_constant_def and vect_external_def
5492 for variable-length vectors to boolean types, or types for which
5493 can_duplicate_and_interleave_p is false.
5494 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
5495 (duplicate_and_interleave): New function.
5496 (vect_get_constant_vectors): Use gimple_build_vector for
5497 constant-length vectors and suitable variable-length constant
5498 vectors. Use duplicate_and_interleave for other variable-length
5499 vectors. Don't defer the update when inserting new statements.
5501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5502 Alan Hayward <alan.hayward@arm.com>
5503 David Sherwood <david.sherwood@arm.com>
5505 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
5506 min_profitable_iters doesn't go negative.
5508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5509 Alan Hayward <alan.hayward@arm.com>
5510 David Sherwood <david.sherwood@arm.com>
5512 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
5513 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
5514 * optabs.def (vec_mask_load_lanes_optab): New optab.
5515 (vec_mask_store_lanes_optab): Likewise.
5516 * internal-fn.def (MASK_LOAD_LANES): New internal function.
5517 (MASK_STORE_LANES): Likewise.
5518 * internal-fn.c (mask_load_lanes_direct): New macro.
5519 (mask_store_lanes_direct): Likewise.
5520 (expand_mask_load_optab_fn): Handle masked operations.
5521 (expand_mask_load_lanes_optab_fn): New macro.
5522 (expand_mask_store_optab_fn): Handle masked operations.
5523 (expand_mask_store_lanes_optab_fn): New macro.
5524 (direct_mask_load_lanes_optab_supported_p): Likewise.
5525 (direct_mask_store_lanes_optab_supported_p): Likewise.
5526 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
5528 (vect_load_lanes_supported): Likewise.
5529 * tree-vect-data-refs.c (strip_conversion): New function.
5530 (can_group_stmts_p): Likewise.
5531 (vect_analyze_data_ref_accesses): Use it instead of checking
5532 for a pair of assignments.
5533 (vect_store_lanes_supported): Take a masked_p parameter.
5534 (vect_load_lanes_supported): Likewise.
5535 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
5536 vect_store_lanes_supported and vect_load_lanes_supported.
5537 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5538 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
5539 parameter. Don't allow gaps for masked accesses.
5540 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
5541 and vect_load_lanes_supported.
5542 (get_load_store_type): Take a masked_p parameter and update
5543 call to get_group_load_store_type.
5544 (vectorizable_store): Update call to get_load_store_type.
5545 Handle IFN_MASK_STORE_LANES.
5546 (vectorizable_load): Update call to get_load_store_type.
5547 Handle IFN_MASK_LOAD_LANES.
5549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5550 Alan Hayward <alan.hayward@arm.com>
5551 David Sherwood <david.sherwood@arm.com>
5553 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
5555 * config/aarch64/aarch64-protos.h
5556 (aarch64_sve_struct_memory_operand_p): Declare.
5557 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
5558 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
5559 (VPRED, vpred): Handle SVE structure modes.
5560 * config/aarch64/constraints.md (Utx): New constraint.
5561 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
5562 (aarch64_sve_struct_nonimmediate_operand): New predicates.
5563 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
5564 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
5565 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
5566 structure modes. Split into pieces after RA.
5567 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
5568 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
5570 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
5571 SVE structure modes.
5572 (aarch64_classify_address): Likewise.
5573 (sizetochar): Move earlier in file.
5574 (aarch64_print_operand): Handle SVE register lists.
5575 (aarch64_array_mode): New function.
5576 (aarch64_sve_struct_memory_operand_p): Likewise.
5577 (TARGET_ARRAY_MODE): Redefine.
5579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5580 Alan Hayward <alan.hayward@arm.com>
5581 David Sherwood <david.sherwood@arm.com>
5583 * target.def (array_mode): New target hook.
5584 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
5585 * doc/tm.texi: Regenerate.
5586 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
5587 * hooks.c (hook_optmode_mode_uhwi_none): New function.
5588 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
5590 * stor-layout.c (mode_for_array): Likewise. Support polynomial
5593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5594 Alan Hayward <alan.hayward@arm.com>
5595 David Sherwood <david.sherwood@arm.com>
5597 * fold-const.c (fold_binary_loc): Check the argument types
5598 rather than the result type when testing for a vector operation.
5600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5602 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
5603 * doc/tm.texi: Regenerate.
5605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5606 Alan Hayward <alan.hayward@arm.com>
5607 David Sherwood <david.sherwood@arm.com>
5609 * doc/invoke.texi (-msve-vector-bits=): Document new option.
5610 (sve): Document new AArch64 extension.
5611 * doc/md.texi (w): Extend the description of the AArch64
5612 constraint to include SVE vectors.
5613 (Upl, Upa): Document new AArch64 predicate constraints.
5614 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
5616 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
5617 (msve-vector-bits=): New option.
5618 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
5619 SVE when these are disabled.
5620 (sve): New extension.
5621 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
5622 modes. Adjust their number of units based on aarch64_sve_vg.
5623 (MAX_BITSIZE_MODE_ANY_MODE): Define.
5624 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
5625 aarch64_addr_query_type.
5626 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
5627 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
5628 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
5629 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
5630 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
5631 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
5632 (aarch64_simd_imm_zero_p): Delete.
5633 (aarch64_check_zero_based_sve_index_immediate): Declare.
5634 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5635 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5636 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5637 (aarch64_sve_float_mul_immediate_p): Likewise.
5638 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5640 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
5641 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
5642 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
5643 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
5644 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
5645 (aarch64_regmode_natural_size): Likewise.
5646 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
5647 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
5649 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
5650 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
5651 for VG and the SVE predicate registers.
5652 (V_ALIASES): Add a "z"-prefixed alias.
5653 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
5654 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
5655 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
5656 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
5657 (REG_CLASS_NAMES): Add entries for them.
5658 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
5659 and the predicate registers.
5660 (aarch64_sve_vg): Declare.
5661 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
5662 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
5663 (REGMODE_NATURAL_SIZE): Define.
5664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5666 * config/aarch64/aarch64.c: Include cfgrtl.h.
5667 (simd_immediate_info): Add a constructor for series vectors,
5668 and an associated step field.
5669 (aarch64_sve_vg): New variable.
5670 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5671 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5672 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5673 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5674 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5675 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5676 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5677 (aarch64_get_mask_mode): New functions.
5678 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5679 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5680 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5681 predicate modes and predicate registers. Explicitly restrict
5682 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5683 to store a vector mode if it is recognized by
5684 aarch64_classify_vector_mode.
5685 (aarch64_regmode_natural_size): New function.
5686 (aarch64_hard_regno_caller_save_mode): Return the original mode
5688 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5689 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5690 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5691 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5693 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5694 does not overlap dest if the function is frame-related. Handle
5696 (aarch64_split_add_offset): New function.
5697 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5698 them aarch64_add_offset.
5699 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5700 and update call to aarch64_sub_sp.
5701 (aarch64_add_cfa_expression): New function.
5702 (aarch64_expand_prologue): Pass extra temporary registers to the
5703 functions above. Handle the case in which we need to emit new
5704 DW_CFA_expressions for registers that were originally saved
5705 relative to the stack pointer, but now have to be expressed
5706 relative to the frame pointer.
5707 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5709 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5710 IP0 and IP1 values for SVE frames.
5711 (aarch64_expand_vec_series): New function.
5712 (aarch64_expand_sve_widened_duplicate): Likewise.
5713 (aarch64_expand_sve_const_vector): Likewise.
5714 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5715 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5716 into the register, rather than emitting a SET directly.
5717 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5718 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5719 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5720 (offset_9bit_signed_scaled_p): New functions.
5721 (aarch64_replicate_bitmask_imm): New function.
5722 (aarch64_bitmask_imm): Use it.
5723 (aarch64_cannot_force_const_mem): Reject expressions involving
5724 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5725 (aarch64_classify_index): Handle SVE indices, by requiring
5726 a plain register index with a scale that matches the element size.
5727 (aarch64_classify_address): Handle SVE addresses. Assert that
5728 the mode of the address is VOIDmode or an integer mode.
5729 Update call to aarch64_classify_symbol.
5730 (aarch64_classify_symbolic_expression): Update call to
5731 aarch64_classify_symbol.
5732 (aarch64_const_vec_all_in_range_p): New function.
5733 (aarch64_print_vector_float_operand): Likewise.
5734 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5735 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5736 and the FP immediates 1.0 and 0.5.
5737 (aarch64_print_address_internal): Handle SVE addresses.
5738 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5739 (aarch64_regno_regclass): Handle predicate registers.
5740 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5742 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5743 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5744 (aarch64_convert_sve_vector_bits): New function.
5745 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5746 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5748 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5749 Handle SVE vector and predicate modes. Accept VL-based constants
5750 that need only one temporary register, and VL offsets that require
5751 no temporary registers.
5752 (aarch64_conditional_register_usage): Mark the predicate registers
5753 as fixed if SVE isn't available.
5754 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5755 Return true for SVE vector and predicate modes.
5756 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5757 rather than an unsigned int. Handle SVE modes.
5758 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5760 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5762 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5763 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5764 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5765 (aarch64_sve_float_mul_immediate_p): New functions.
5766 (aarch64_sve_valid_immediate): New function.
5767 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5768 Explicitly reject structure modes. Check for INDEX constants.
5769 Handle PTRUE and PFALSE constants.
5770 (aarch64_check_zero_based_sve_index_immediate): New function.
5771 (aarch64_simd_imm_zero_p): Delete.
5772 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5773 vector modes. Accept constants in the range of CNT[BHWD].
5774 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5775 ask for an Advanced SIMD mode.
5776 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5777 (aarch64_simd_vector_alignment): Handle SVE predicates.
5778 (aarch64_vectorize_preferred_vector_alignment): New function.
5779 (aarch64_simd_vector_alignment_reachable): Use it instead of
5781 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5782 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5784 (MAX_VECT_LEN): Delete.
5785 (expand_vec_perm_d): Add a vec_flags field.
5786 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5787 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5788 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5790 (aarch64_evpc_rev): Rename to...
5791 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5792 (aarch64_evpc_rev_global): New function.
5793 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5794 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5796 (aarch64_evpc_sve_tbl): New function.
5797 (aarch64_expand_vec_perm_const_1): Update after rename of
5798 aarch64_evpc_rev. Handle SVE permutes too, trying
5799 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5800 than aarch64_evpc_tbl.
5801 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5802 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5803 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5804 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5805 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5806 (aarch64_expand_sve_vcond): New functions.
5807 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5808 of aarch64_vector_mode_p.
5809 (aarch64_dwarf_poly_indeterminate_value): New function.
5810 (aarch64_compute_pressure_classes): Likewise.
5811 (aarch64_can_change_mode_class): Likewise.
5812 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5813 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5814 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5815 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5816 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5817 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5818 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5819 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5821 (Dn, Dl, Dr): Accept const as well as const_vector.
5822 (Dz): Likewise. Compare against CONST0_RTX.
5823 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5824 of "vector" where appropriate.
5825 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5826 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5827 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5828 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5829 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5830 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5831 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5832 (v_int_equiv): Extend to SVE modes.
5833 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5835 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5836 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5837 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5838 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5839 (SVE_COND_FP_CMP): New int iterators.
5840 (perm_hilo): Handle the new unpack unspecs.
5841 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5843 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5844 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5845 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5846 (aarch64_equality_operator, aarch64_constant_vector_operand)
5847 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5848 (aarch64_sve_nonimmediate_operand): Likewise.
5849 (aarch64_sve_general_operand): Likewise.
5850 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5851 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5852 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5853 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5854 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5855 (aarch64_sve_float_arith_immediate): Likewise.
5856 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5857 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5858 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5859 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5860 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5861 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5862 (aarch64_sve_float_arith_operand): Likewise.
5863 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5864 (aarch64_sve_float_mul_operand): Likewise.
5865 (aarch64_sve_vec_perm_operand): Likewise.
5866 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5867 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5868 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5869 as well as const_vector.
5870 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5871 in file. Use CONST0_RTX and CONSTM1_RTX.
5872 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5873 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5874 Use aarch64_simd_imm_zero.
5875 * config/aarch64/aarch64-sve.md: New file.
5876 * config/aarch64/aarch64.md: Include it.
5877 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5878 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5879 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5880 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5881 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5882 (sve): New attribute.
5883 (enabled): Disable instructions with the sve attribute unless
5885 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5886 aarch64_expand_mov_immediate.
5887 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5888 CNT[BHSD] immediates.
5889 (movti): Split CONST_POLY_INT moves into two halves.
5890 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5891 Split additions that need a temporary here if the destination
5892 is the stack pointer.
5893 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5894 (*add<mode>3_poly_1): New instruction.
5895 (set_clobber_cc): New expander.
5897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5899 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5900 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5901 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5902 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5903 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5904 Change innermode from fixed_mode_size to machine_mode.
5905 (simplify_subreg): Update call accordingly. Handle a constant-sized
5906 subreg of a variable-length CONST_VECTOR.
5908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5909 Alan Hayward <alan.hayward@arm.com>
5910 David Sherwood <david.sherwood@arm.com>
5912 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5913 (add_offset_to_base): New function, split out from...
5914 (create_mem_ref): ...here. When handling a scale other than 1,
5915 check first whether the address is valid without the offset.
5916 Add it into the base if so, leaving the index and scale as-is.
5918 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5921 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5922 fold_for_warn before checking if arg2 is INTEGER_CST.
5924 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5926 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5927 (store_multiple_operation): Delete.
5928 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5929 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5930 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5931 guarded by TARGET_STRING.
5932 (rs6000_output_load_multiple): Delete.
5933 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5934 OPTION_MASK_STRING / TARGET_STRING handling.
5935 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5936 (const rs6000_opt_masks) <"string">: Change mask to 0.
5937 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5938 (MASK_STRING): Delete.
5939 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5941 (load_multiple): Delete.
5948 (store_multiple): Delete.
5955 (movmemsi_8reg): Delete.
5956 (corresponding unnamed define_insn): Delete.
5957 (movmemsi_6reg): Delete.
5958 (corresponding unnamed define_insn): Delete.
5959 (movmemsi_4reg): Delete.
5960 (corresponding unnamed define_insn): Delete.
5961 (movmemsi_2reg): Delete.
5962 (corresponding unnamed define_insn): Delete.
5963 (movmemsi_1reg): Delete.
5964 (corresponding unnamed define_insn): Delete.
5965 * config/rs6000/rs6000.opt (mno-string): New.
5966 (mstring): Replace by deprecation warning stub.
5967 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5969 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5971 * regrename.c (regrename_do_replace): If replacing the same
5972 reg multiple times, try to reuse last created gen_raw_REG.
5975 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5976 main to workaround a bug in GDB.
5978 2018-01-12 Tom de Vries <tom@codesourcery.com>
5981 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5983 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5985 PR rtl-optimization/80481
5986 * ira-color.c (get_cap_member): New function.
5987 (allocnos_conflict_by_live_ranges_p): Use it.
5988 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5989 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5991 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5994 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5995 (*saddl_se_1): Ditto.
5997 (*ssubl_se_1): Ditto.
5999 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6001 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6002 rather than wi::to_widest for DR_INITs.
6003 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6004 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6005 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6007 (vect_analyze_group_access_1): Note that here.
6009 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6011 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6012 polynomial type sizes.
6014 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6016 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6017 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6018 (gimple_add_tmp_var): Likewise.
6020 2018-01-12 Martin Liska <mliska@suse.cz>
6022 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6023 (gimple_alloc_sizes): Likewise.
6024 (dump_gimple_statistics): Use PRIu64 in printf format.
6025 * gimple.h: Change uint64_t to int.
6027 2018-01-12 Martin Liska <mliska@suse.cz>
6029 * tree-core.h: Use uint64_t instead of int.
6030 * tree.c (tree_node_counts): Likewise.
6031 (tree_node_sizes): Likewise.
6032 (dump_tree_statistics): Use PRIu64 in printf format.
6034 2018-01-12 Martin Liska <mliska@suse.cz>
6036 * Makefile.in: As qsort_chk is implemented in vec.c, add
6037 vec.o to linkage of gencfn-macros.
6038 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6039 passing the info to record_node_allocation_statistics.
6040 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6042 * ggc-common.c (struct ggc_usage): Add operator== and use
6043 it in operator< and compare function.
6044 * mem-stats.h (struct mem_usage): Likewise.
6045 * vec.c (struct vec_usage): Remove operator< and compare
6046 function. Can be simply inherited.
6048 2018-01-12 Martin Jambor <mjambor@suse.cz>
6051 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6052 * tree-ssa-math-opts.c: Include domwalk.h.
6053 (convert_mult_to_fma_1): New function.
6054 (fma_transformation_info): New type.
6055 (fma_deferring_state): Likewise.
6056 (cancel_fma_deferring): New function.
6057 (result_of_phi): Likewise.
6058 (last_fma_candidate_feeds_initial_phi): Likewise.
6059 (convert_mult_to_fma): Added deferring logic, split actual
6060 transformation to convert_mult_to_fma_1.
6061 (math_opts_dom_walker): New type.
6062 (math_opts_dom_walker::after_dom_children): New method, body moved
6063 here from pass_optimize_widening_mul::execute, added deferring logic
6065 (pass_optimize_widening_mul::execute): Moved most of code to
6066 math_opts_dom_walker::after_dom_children.
6067 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6068 * config/i386/i386.c (ix86_option_override_internal): Added
6069 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6071 2018-01-12 Richard Biener <rguenther@suse.de>
6074 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6075 inline instance vars.
6077 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6080 * config/rx/rx.c (rx_is_restricted_memory_address):
6083 2018-01-12 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/80846
6086 * target.def (split_reduction): New target hook.
6087 * targhooks.c (default_split_reduction): New function.
6088 * targhooks.h (default_split_reduction): Declare.
6089 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6090 target requests first reduce vectors by combining low and high
6092 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6093 (get_vectype_for_scalar_type_and_size): Export.
6094 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6095 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6096 * doc/tm.texi: Regenerate.
6097 * config/i386/i386.c (ix86_split_reduction): Implement
6098 TARGET_VECTORIZE_SPLIT_REDUCTION.
6100 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6103 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6104 in PIC mode except for TARGET_VXWORKS_RTP.
6105 * config/sparc/sparc.c: Include cfgrtl.h.
6106 (TARGET_INIT_PIC_REG): Define.
6107 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6108 (sparc_pic_register_p): New predicate.
6109 (sparc_legitimate_address_p): Use it.
6110 (sparc_legitimize_pic_address): Likewise.
6111 (sparc_delegitimize_address): Likewise.
6112 (sparc_mode_dependent_address_p): Likewise.
6113 (gen_load_pcrel_sym): Remove 4th parameter.
6114 (load_got_register): Adjust call to above. Remove obsolete stuff.
6115 (sparc_expand_prologue): Do not call load_got_register here.
6116 (sparc_flat_expand_prologue): Likewise.
6117 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6118 (sparc_use_pseudo_pic_reg): New function.
6119 (sparc_init_pic_reg): Likewise.
6120 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6121 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6123 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6125 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6126 Add item for branch_cost.
6128 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6130 PR rtl-optimization/83565
6131 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6132 not extend the result to a larger mode for rotate operations.
6133 (num_sign_bit_copies1): Likewise.
6135 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6138 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6140 Use values-Xc.o for -pedantic.
6141 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6143 2018-01-12 Martin Liska <mliska@suse.cz>
6146 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6148 (possible_polymorphic_call_targets): Use it.
6149 (ipa_devirt): Likewise.
6151 2018-01-12 Martin Liska <mliska@suse.cz>
6153 * profile-count.h (enum profile_quality): Use 0 as invalid
6154 enum value of profile_quality.
6156 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6158 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6159 -mext-string options.
6161 2018-01-12 Richard Biener <rguenther@suse.de>
6163 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6164 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6165 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6167 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6169 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6171 * configure.ac (--with-long-double-format): Add support for the
6172 configuration option to change the default long double format on
6174 * config.gcc (powerpc*-linux*-*): Likewise.
6175 * configure: Regenerate.
6176 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6177 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6178 used without modification.
6180 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6182 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6183 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6184 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6185 MISC_BUILTIN_SPEC_BARRIER.
6186 (rs6000_init_builtins): Likewise.
6187 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6189 (speculation_barrier): New define_insn.
6190 * doc/extend.texi: Document __builtin_speculation_barrier.
6192 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6195 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6196 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6197 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6199 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6200 integral modes instead of "ss" and "sd".
6201 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6202 vectors with 32-bit and 64-bit elements.
6203 (vecdupssescalarmodesuffix): New mode attribute.
6204 (vec_dup<mode>): Use it.
6206 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6209 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6210 frame if argument is passed on stack.
6212 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6215 * ree.c (combine_reaching_defs): Optimize also
6216 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6217 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6219 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6222 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6224 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6227 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6228 after they are computed.
6230 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6232 PR tree-optimization/83695
6233 * gimple-loop-linterchange.cc
6234 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6235 reset cached scev information after interchange.
6236 (pass_linterchange::execute): Remove call to scev_reset_htab.
6238 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6240 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6241 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6242 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6243 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6244 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6245 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6246 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6247 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6248 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6249 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6250 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6251 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6252 (V_lane_reg): Likewise.
6253 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6255 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6256 (vfmal_lane_low<mode>_intrinsic,
6257 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6258 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6259 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6260 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6261 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6262 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6264 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6266 * config/arm/arm-cpus.in (fp16fml): New feature.
6267 (ALL_SIMD): Add fp16fml.
6268 (armv8.2-a): Add fp16fml as an option.
6269 (armv8.3-a): Likewise.
6270 (armv8.4-a): Add fp16fml as part of fp16.
6271 * config/arm/arm.h (TARGET_FP16FML): Define.
6272 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6274 * config/arm/arm-modes.def (V2HF): Define.
6275 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6276 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6277 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6278 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6279 vfmsl_low, vfmsl_high): New set of builtins.
6280 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6281 (vfml_op): New code attribute.
6282 (VFMLHALVES): New int iterator.
6283 (VFML, VFMLSEL): New mode attributes.
6284 (V_reg): Define mapping for V2HF.
6285 (V_hi, V_lo): New mode attributes.
6286 (VF_constraint): Likewise.
6287 (vfml_half, vfml_half_selector): New int attributes.
6288 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6290 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6291 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6293 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6294 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6295 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6296 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6298 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6299 Document new effective target and option set.
6301 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6303 * config/arm/arm-cpus.in (armv8_4): New feature.
6304 (ARMv8_4a): New fgroup.
6305 (armv8.4-a): New arch.
6306 * config/arm/arm-tables.opt: Regenerate.
6307 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6308 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6309 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6310 Add matching rules for -march=armv8.4-a and extensions.
6311 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6313 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6316 * config/rx/rx.md (BW): New mode attribute.
6317 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6319 2018-01-11 Richard Biener <rguenther@suse.de>
6321 PR tree-optimization/83435
6322 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6323 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6324 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6326 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6327 Alan Hayward <alan.hayward@arm.com>
6328 David Sherwood <david.sherwood@arm.com>
6330 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6332 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6333 (aarch64_print_address_internal): Use it to check for a zero offset.
6335 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6336 Alan Hayward <alan.hayward@arm.com>
6337 David Sherwood <david.sherwood@arm.com>
6339 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
6340 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
6341 Return a poly_int64 rather than a HOST_WIDE_INT.
6342 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
6343 rather than a HOST_WIDE_INT.
6344 * config/aarch64/aarch64.h (aarch64_frame): Protect with
6345 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
6346 hard_fp_offset, frame_size, initial_adjust, callee_offset and
6347 final_offset from HOST_WIDE_INT to poly_int64.
6348 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
6349 to_constant when getting the number of units in an Advanced SIMD
6351 (aarch64_builtin_vectorized_function): Check for a constant number
6353 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
6355 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
6356 attribute instead of GET_MODE_NUNITS.
6357 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
6358 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
6359 GET_MODE_SIZE for fixed-size registers.
6360 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
6361 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
6362 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
6363 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
6364 (aarch64_print_operand, aarch64_print_address_internal)
6365 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
6366 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
6367 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
6368 Handle polynomial GET_MODE_SIZE.
6369 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
6370 wider than SImode without modification.
6371 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
6372 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
6373 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
6374 passing and returning SVE modes.
6375 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
6376 rather than GEN_INT.
6377 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
6378 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
6379 (aarch64_allocate_and_probe_stack_space): Likewise.
6380 (aarch64_layout_frame): Cope with polynomial offsets.
6381 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
6382 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
6384 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
6385 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
6386 poly_int64 rather than a HOST_WIDE_INT.
6387 (aarch64_get_separate_components, aarch64_process_components)
6388 (aarch64_expand_prologue, aarch64_expand_epilogue)
6389 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
6390 (aarch64_anchor_offset): New function, split out from...
6391 (aarch64_legitimize_address): ...here.
6392 (aarch64_builtin_vectorization_cost): Handle polynomial
6393 TYPE_VECTOR_SUBPARTS.
6394 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
6396 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
6397 number of elements from the PARALLEL rather than the mode.
6398 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
6399 rather than GET_MODE_BITSIZE.
6400 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
6401 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
6402 (aarch64_expand_vec_perm_const_1): Handle polynomial
6403 d->perm.length () and d->perm elements.
6404 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
6405 Apply to_constant to d->perm elements.
6406 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
6407 polynomial CONST_VECTOR_NUNITS.
6408 (aarch64_move_pointer): Take amount as a poly_int64 rather
6410 (aarch64_progress_pointer): Avoid temporary variable.
6411 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
6412 the mode attribute instead of GET_MODE.
6414 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6415 Alan Hayward <alan.hayward@arm.com>
6416 David Sherwood <david.sherwood@arm.com>
6418 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
6419 x exists before using it.
6420 (aarch64_add_constant_internal): Rename to...
6421 (aarch64_add_offset_1): ...this. Replace regnum with separate
6422 src and dest rtxes. Handle the case in which they're different,
6423 including when the offset is zero. Replace scratchreg with an rtx.
6424 Use 2 additions if there is no spare register into which we can
6425 move a 16-bit constant.
6426 (aarch64_add_constant): Delete.
6427 (aarch64_add_offset): Replace reg with separate src and dest
6428 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
6429 Use aarch64_add_offset_1.
6430 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
6431 an rtx rather than an int. Take the delta as a poly_int64
6432 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
6433 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
6434 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
6435 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
6436 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
6438 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
6439 aarch64_add_constant.
6441 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6443 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
6444 Use scalar_float_mode.
6446 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6448 * config/aarch64/aarch64-simd.md
6449 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
6450 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
6451 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
6452 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
6453 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
6454 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
6455 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
6456 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
6457 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
6458 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
6460 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6463 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
6464 targ_options->x_arm_arch_string is non NULL.
6466 2018-01-11 Tamar Christina <tamar.christina@arm.com>
6468 * config/aarch64/aarch64.h
6469 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
6471 2018-01-11 Sudakshina Das <sudi.das@arm.com>
6474 * expmed.c (emit_store_flag_force): Swap if const op0
6475 and change VOIDmode to mode of op0.
6477 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6479 PR rtl-optimization/83761
6480 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
6481 than bytes to mode_for_size.
6483 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6486 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
6487 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
6490 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6493 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
6494 when in layout mode.
6495 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
6496 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
6499 2018-01-10 Michael Collison <michael.collison@arm.com>
6501 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
6502 * config/aarch64/aarch64-option-extension.def: Add
6503 AARCH64_OPT_EXTENSION of 'fp16fml'.
6504 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6505 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
6506 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
6507 * config/aarch64/constraints.md (Ui7): New constraint.
6508 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
6509 (VFMLA_SEL_W): Ditto.
6512 (VFMLA16_LOW): New int iterator.
6513 (VFMLA16_HIGH): Ditto.
6514 (UNSPEC_FMLAL): New unspec.
6515 (UNSPEC_FMLSL): Ditto.
6516 (UNSPEC_FMLAL2): Ditto.
6517 (UNSPEC_FMLSL2): Ditto.
6518 (f16mac): New code attribute.
6519 * config/aarch64/aarch64-simd-builtins.def
6520 (aarch64_fmlal_lowv2sf): Ditto.
6521 (aarch64_fmlsl_lowv2sf): Ditto.
6522 (aarch64_fmlalq_lowv4sf): Ditto.
6523 (aarch64_fmlslq_lowv4sf): Ditto.
6524 (aarch64_fmlal_highv2sf): Ditto.
6525 (aarch64_fmlsl_highv2sf): Ditto.
6526 (aarch64_fmlalq_highv4sf): Ditto.
6527 (aarch64_fmlslq_highv4sf): Ditto.
6528 (aarch64_fmlal_lane_lowv2sf): Ditto.
6529 (aarch64_fmlsl_lane_lowv2sf): Ditto.
6530 (aarch64_fmlal_laneq_lowv2sf): Ditto.
6531 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
6532 (aarch64_fmlalq_lane_lowv4sf): Ditto.
6533 (aarch64_fmlsl_lane_lowv4sf): Ditto.
6534 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
6535 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
6536 (aarch64_fmlal_lane_highv2sf): Ditto.
6537 (aarch64_fmlsl_lane_highv2sf): Ditto.
6538 (aarch64_fmlal_laneq_highv2sf): Ditto.
6539 (aarch64_fmlsl_laneq_highv2sf): Ditto.
6540 (aarch64_fmlalq_lane_highv4sf): Ditto.
6541 (aarch64_fmlsl_lane_highv4sf): Ditto.
6542 (aarch64_fmlalq_laneq_highv4sf): Ditto.
6543 (aarch64_fmlsl_laneq_highv4sf): Ditto.
6544 * config/aarch64/aarch64-simd.md:
6545 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
6546 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6547 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
6548 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
6549 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
6550 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
6551 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
6552 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
6553 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
6554 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
6555 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
6556 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
6557 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
6558 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
6559 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
6560 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
6561 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
6562 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
6563 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
6564 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
6565 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
6566 (vfmlsl_low_u32): Ditto.
6567 (vfmlalq_low_u32): Ditto.
6568 (vfmlslq_low_u32): Ditto.
6569 (vfmlal_high_u32): Ditto.
6570 (vfmlsl_high_u32): Ditto.
6571 (vfmlalq_high_u32): Ditto.
6572 (vfmlslq_high_u32): Ditto.
6573 (vfmlal_lane_low_u32): Ditto.
6574 (vfmlsl_lane_low_u32): Ditto.
6575 (vfmlal_laneq_low_u32): Ditto.
6576 (vfmlsl_laneq_low_u32): Ditto.
6577 (vfmlalq_lane_low_u32): Ditto.
6578 (vfmlslq_lane_low_u32): Ditto.
6579 (vfmlalq_laneq_low_u32): Ditto.
6580 (vfmlslq_laneq_low_u32): Ditto.
6581 (vfmlal_lane_high_u32): Ditto.
6582 (vfmlsl_lane_high_u32): Ditto.
6583 (vfmlal_laneq_high_u32): Ditto.
6584 (vfmlsl_laneq_high_u32): Ditto.
6585 (vfmlalq_lane_high_u32): Ditto.
6586 (vfmlslq_lane_high_u32): Ditto.
6587 (vfmlalq_laneq_high_u32): Ditto.
6588 (vfmlslq_laneq_high_u32): Ditto.
6589 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
6590 (AARCH64_FL_FOR_ARCH8_4): New.
6591 (AARCH64_ISA_F16FML): New ISA flag.
6592 (TARGET_F16FML): New feature flag for fp16fml.
6593 (doc/invoke.texi): Document new fp16fml option.
6595 2018-01-10 Michael Collison <michael.collison@arm.com>
6597 * config/aarch64/aarch64-builtins.c:
6598 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
6599 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6600 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
6601 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
6602 (AARCH64_ISA_SHA3): New ISA flag.
6603 (TARGET_SHA3): New feature flag for sha3.
6604 * config/aarch64/iterators.md (sha512_op): New int attribute.
6605 (CRYPTO_SHA512): New int iterator.
6606 (UNSPEC_SHA512H): New unspec.
6607 (UNSPEC_SHA512H2): Ditto.
6608 (UNSPEC_SHA512SU0): Ditto.
6609 (UNSPEC_SHA512SU1): Ditto.
6610 * config/aarch64/aarch64-simd-builtins.def
6611 (aarch64_crypto_sha512hqv2di): New builtin.
6612 (aarch64_crypto_sha512h2qv2di): Ditto.
6613 (aarch64_crypto_sha512su0qv2di): Ditto.
6614 (aarch64_crypto_sha512su1qv2di): Ditto.
6615 (aarch64_eor3qv8hi): Ditto.
6616 (aarch64_rax1qv2di): Ditto.
6617 (aarch64_xarqv2di): Ditto.
6618 (aarch64_bcaxqv8hi): Ditto.
6619 * config/aarch64/aarch64-simd.md:
6620 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
6621 (aarch64_crypto_sha512su0qv2di): Ditto.
6622 (aarch64_crypto_sha512su1qv2di): Ditto.
6623 (aarch64_eor3qv8hi): Ditto.
6624 (aarch64_rax1qv2di): Ditto.
6625 (aarch64_xarqv2di): Ditto.
6626 (aarch64_bcaxqv8hi): Ditto.
6627 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
6628 (vsha512h2q_u64): Ditto.
6629 (vsha512su0q_u64): Ditto.
6630 (vsha512su1q_u64): Ditto.
6631 (veor3q_u16): Ditto.
6632 (vrax1q_u64): Ditto.
6634 (vbcaxq_u16): Ditto.
6635 * config/arm/types.md (crypto_sha512): New type attribute.
6636 (crypto_sha3): Ditto.
6637 (doc/invoke.texi): Document new sha3 option.
6639 2018-01-10 Michael Collison <michael.collison@arm.com>
6641 * config/aarch64/aarch64-builtins.c:
6642 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
6643 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6644 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
6645 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
6646 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
6647 (AARCH64_ISA_SM4): New ISA flag.
6648 (TARGET_SM4): New feature flag for sm4.
6649 * config/aarch64/aarch64-simd-builtins.def
6650 (aarch64_sm3ss1qv4si): Ditto.
6651 (aarch64_sm3tt1aq4si): Ditto.
6652 (aarch64_sm3tt1bq4si): Ditto.
6653 (aarch64_sm3tt2aq4si): Ditto.
6654 (aarch64_sm3tt2bq4si): Ditto.
6655 (aarch64_sm3partw1qv4si): Ditto.
6656 (aarch64_sm3partw2qv4si): Ditto.
6657 (aarch64_sm4eqv4si): Ditto.
6658 (aarch64_sm4ekeyqv4si): Ditto.
6659 * config/aarch64/aarch64-simd.md:
6660 (aarch64_sm3ss1qv4si): Ditto.
6661 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
6662 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
6663 (aarch64_sm4eqv4si): Ditto.
6664 (aarch64_sm4ekeyqv4si): Ditto.
6665 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6666 (sm3part_op): Ditto.
6667 (CRYPTO_SM3TT): Ditto.
6668 (CRYPTO_SM3PART): Ditto.
6669 (UNSPEC_SM3SS1): New unspec.
6670 (UNSPEC_SM3TT1A): Ditto.
6671 (UNSPEC_SM3TT1B): Ditto.
6672 (UNSPEC_SM3TT2A): Ditto.
6673 (UNSPEC_SM3TT2B): Ditto.
6674 (UNSPEC_SM3PARTW1): Ditto.
6675 (UNSPEC_SM3PARTW2): Ditto.
6676 (UNSPEC_SM4E): Ditto.
6677 (UNSPEC_SM4EKEY): Ditto.
6678 * config/aarch64/constraints.md (Ui2): New constraint.
6679 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6680 * config/arm/types.md (crypto_sm3): New type attribute.
6681 (crypto_sm4): Ditto.
6682 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6683 (vsm3tt1aq_u32): Ditto.
6684 (vsm3tt1bq_u32): Ditto.
6685 (vsm3tt2aq_u32): Ditto.
6686 (vsm3tt2bq_u32): Ditto.
6687 (vsm3partw1q_u32): Ditto.
6688 (vsm3partw2q_u32): Ditto.
6689 (vsm4eq_u32): Ditto.
6690 (vsm4ekeyq_u32): Ditto.
6691 (doc/invoke.texi): Document new sm4 option.
6693 2018-01-10 Michael Collison <michael.collison@arm.com>
6695 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6696 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6697 (AARCH64_FL_FOR_ARCH8_4): New.
6698 (AARCH64_FL_V8_4): New flag.
6699 (doc/invoke.texi): Document new armv8.4-a option.
6701 2018-01-10 Michael Collison <michael.collison@arm.com>
6703 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6704 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6705 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6706 * config/aarch64/aarch64-option-extension.def: Add
6707 AARCH64_OPT_EXTENSION of 'sha2'.
6708 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6709 (crypto): Disable sha2 and aes if crypto disabled.
6710 (crypto): Enable aes and sha2 if enabled.
6711 (simd): Disable sha2 and aes if simd disabled.
6712 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6714 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6715 (TARGET_SHA2): New feature flag for sha2.
6716 (TARGET_AES): New feature flag for aes.
6717 * config/aarch64/aarch64-simd.md:
6718 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6719 conditional on TARGET_AES.
6720 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6721 (aarch64_crypto_sha1hsi): Make pattern conditional
6723 (aarch64_crypto_sha1hv4si): Ditto.
6724 (aarch64_be_crypto_sha1hv4si): Ditto.
6725 (aarch64_crypto_sha1su1v4si): Ditto.
6726 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6727 (aarch64_crypto_sha1su0v4si): Ditto.
6728 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6729 (aarch64_crypto_sha256su0v4si): Ditto.
6730 (aarch64_crypto_sha256su1v4si): Ditto.
6731 (doc/invoke.texi): Document new aes and sha2 options.
6733 2018-01-10 Martin Sebor <msebor@redhat.com>
6735 PR tree-optimization/83781
6736 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6739 2018-01-11 Martin Sebor <msebor@gmail.com>
6740 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6742 PR tree-optimization/83501
6743 PR tree-optimization/81703
6745 * tree-ssa-strlen.c (get_string_cst): Rename...
6746 (get_string_len): ...to this. Handle global constants.
6747 (handle_char_store): Adjust.
6749 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6750 Jim Wilson <jimw@sifive.com>
6752 * config/riscv/riscv-protos.h (riscv_output_return): New.
6753 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6754 (riscv_attribute_table, riscv_output_return),
6755 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6756 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6757 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6759 (riscv_expand_prologue): Add early return for naked function.
6760 (riscv_expand_epilogue): Likewise.
6761 (riscv_function_ok_for_sibcall): Return false for naked function.
6762 (riscv_set_current_function): New.
6763 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6764 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6765 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6766 * doc/extend.texi (RISC-V Function Attributes): New.
6768 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6770 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6771 check for 128-bit long double before checking TCmode.
6772 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6773 128-bit long doubles before checking TFmode or TCmode.
6774 (FLOAT128_IBM_P): Likewise.
6776 2018-01-10 Martin Sebor <msebor@redhat.com>
6778 PR tree-optimization/83671
6779 * builtins.c (c_strlen): Unconditionally return zero for the empty
6781 Use -Warray-bounds for warnings.
6782 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6783 for non-constant array indices with COMPONENT_REF, arrays of
6784 arrays, and pointers to arrays.
6785 (gimple_fold_builtin_strlen): Determine and set length range for
6786 non-constant character arrays.
6788 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6791 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6794 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6796 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6798 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6801 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6802 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6803 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6804 indexed_or_indirect_operand predicate.
6805 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6806 (*vsx_le_perm_load_v8hi): Likewise.
6807 (*vsx_le_perm_load_v16qi): Likewise.
6808 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6809 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6810 (*vsx_le_perm_store_v8hi): Likewise.
6811 (*vsx_le_perm_store_v16qi): Likewise.
6812 (eight unnamed splitters): Likewise.
6814 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6816 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6817 * config/rs6000/emmintrin.h: Likewise.
6818 * config/rs6000/mmintrin.h: Likewise.
6819 * config/rs6000/xmmintrin.h: Likewise.
6821 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6824 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6826 * tree.c (tree_nop_conversion): Return true for location wrapper
6828 (maybe_wrap_with_location): New function.
6829 (selftest::check_strip_nops): New function.
6830 (selftest::test_location_wrappers): New function.
6831 (selftest::tree_c_tests): Call it.
6832 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6833 (maybe_wrap_with_location): New decl.
6834 (EXPR_LOCATION_WRAPPER_P): New macro.
6835 (location_wrapper_p): New inline function.
6836 (tree_strip_any_location_wrapper): New inline function.
6838 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6841 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6842 stack_realign_offset for the largest alignment of stack slot
6844 (ix86_find_max_used_stack_alignment): New function.
6845 (ix86_finalize_stack_frame_flags): Use it. Set
6846 max_used_stack_alignment if we don't realign stack.
6847 * config/i386/i386.h (machine_function): Add
6848 max_used_stack_alignment.
6850 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6852 * config/arm/arm.opt (-mbranch-cost): New option.
6853 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6856 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6859 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6860 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6862 2018-01-10 Richard Biener <rguenther@suse.de>
6865 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6866 early out so it also covers the case where we have a non-NULL
6869 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6871 PR tree-optimization/83753
6872 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6873 for non-strided grouped accesses if the number of elements is 1.
6875 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6878 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6879 * i386.h (TARGET_USE_GATHER): Define.
6880 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6882 2018-01-10 Martin Liska <mliska@suse.cz>
6885 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6886 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6888 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6889 CLEANUP_NO_PARTITIONING is not set.
6891 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6893 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6894 for vectors, as a partial revert of r254296.
6895 * rtl.h (const_vec_p): Delete.
6896 (const_vec_duplicate_p): Don't test for vector CONSTs.
6897 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6898 * expmed.c (make_tree): Likewise.
6901 * common.md (E, F): Use CONSTANT_P instead of checking for
6903 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6904 checking for CONST_VECTOR.
6906 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6909 * predict.c (force_edge_cold): Handle in more sane way edges
6912 2018-01-09 Carl Love <cel@us.ibm.com>
6914 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6916 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6917 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6918 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6919 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6920 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6921 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6922 * config/rs6000/rs6000-protos.h: Add extern defition for
6923 rs6000_generate_float2_double_code.
6924 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6926 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6927 (float2_v2df): Add define_expand.
6929 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6932 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6933 op_mode in the force_to_mode call.
6935 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6937 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6938 instead of checking each element individually.
6939 (aarch64_evpc_uzp): Likewise.
6940 (aarch64_evpc_zip): Likewise.
6941 (aarch64_evpc_ext): Likewise.
6942 (aarch64_evpc_rev): Likewise.
6943 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6944 instead of checking each element individually. Return true without
6946 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6947 whether all selected elements come from the same input, instead of
6948 checking each element individually. Remove calls to gen_rtx_REG,
6949 start_sequence and end_sequence and instead assert that no rtl is
6952 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6954 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6955 order of HIGH and CONST checks.
6957 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6959 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6960 if the destination isn't an SSA_NAME.
6962 2018-01-09 Richard Biener <rguenther@suse.de>
6964 PR tree-optimization/83668
6965 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6967 (canonicalize_loop_form): ... here, renamed from ...
6968 (canonicalize_loop_closed_ssa_form): ... this and amended to
6969 swap successor edges for loop exit blocks to make us use
6970 the RPO order we need for initial schedule generation.
6972 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6974 PR tree-optimization/64811
6975 * match.pd: When optimizing comparisons with Inf, avoid
6976 introducing or losing exceptions from comparisons with NaN.
6978 2018-01-09 Martin Liska <mliska@suse.cz>
6981 * asan.c (shadow_mem_size): Add gcc_assert.
6983 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6985 Don't save registers in main().
6988 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6989 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6990 * config/avr/avr.c (avr_set_current_function): Don't error if
6991 naked, OS_task or OS_main are specified at the same time.
6992 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6994 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6996 * common/config/avr/avr-common.c (avr_option_optimization_table):
6997 Switch on -mmain-is-OS_task for optimizing compilations.
6999 2018-01-09 Richard Biener <rguenther@suse.de>
7001 PR tree-optimization/83572
7002 * graphite.c: Include cfganal.h.
7003 (graphite_transform_loops): Connect infinite loops to exit
7004 and remove fake edges at the end.
7006 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7008 * ipa-inline.c (edge_badness): Revert accidental checkin.
7010 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7013 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7014 symbols; not inline clones.
7016 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7019 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7020 hard registers. Formatting fixes.
7022 PR preprocessor/83722
7023 * gcc.c (try_generate_repro): Pass
7024 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7025 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7028 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7029 Kito Cheng <kito.cheng@gmail.com>
7031 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7032 (riscv_leaf_function_p): Delete.
7033 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7035 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7037 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7039 (do_ifelse): New function.
7040 (do_isel): New function.
7041 (do_sub3): New function.
7042 (do_add3): New function.
7043 (do_load_mask_compare): New function.
7044 (do_overlap_load_compare): New function.
7045 (expand_compare_loop): New function.
7046 (expand_block_compare): Call expand_compare_loop() when appropriate.
7047 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7049 (-mblock-compare-inline-loop-limit): New option.
7051 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7054 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7055 Reverse order of second and third operands in first alternative.
7056 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7057 of first and second elements in UNSPEC_VPERMR vector.
7058 (altivec_expand_vec_perm_le): Likewise.
7060 2018-01-08 Jeff Law <law@redhat.com>
7062 PR rtl-optimizatin/81308
7063 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7064 (process_switch): If group_case_labels makes a change, then set
7066 (pass_convert_switch::execute): If a switch is converted, then
7067 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7069 PR rtl-optimization/81308
7070 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7073 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7075 PR target/83663 - Revert r255946
7076 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7077 generation for cases where splatting a value is not useful.
7078 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7079 across a vec_duplicate and a paradoxical subreg forming a vector
7080 mode to a vec_concat.
7082 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7084 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7085 -march=armv8.3-a variants.
7086 * config/arm/t-multilib: Likewise.
7087 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7089 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7091 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7093 (cceq_ior_compare_complement): Give it a name so I can use it, and
7094 change boolean_or_operator predicate to boolean_operator so it can
7095 be used to generate a crand.
7096 (eqne): New code iterator.
7097 (bd/bd_neg): New code_attrs.
7098 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7099 a single define_insn.
7100 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7101 decrement (bdnzt/bdnzf/bdzt/bdzf).
7102 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7103 with the new names of the branch decrement patterns, and added the
7104 names of the branch decrement conditional patterns.
7106 2018-01-08 Richard Biener <rguenther@suse.de>
7108 PR tree-optimization/83563
7109 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7112 2018-01-08 Richard Biener <rguenther@suse.de>
7115 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7117 2018-01-08 Richard Biener <rguenther@suse.de>
7119 PR tree-optimization/83685
7120 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7121 references to abnormals.
7123 2018-01-08 Richard Biener <rguenther@suse.de>
7126 * dwarf2out.c (output_indirect_strings): Handle empty
7127 skeleton_debug_str_hash.
7128 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7130 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7132 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7133 (emit_store_direct): Likewise.
7134 (arc_trampoline_adjust_address): Likewise.
7135 (arc_asm_trampoline_template): New function.
7136 (arc_initialize_trampoline): Use asm_trampoline_template.
7137 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7138 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7139 * config/arc/arc.md (flush_icache): Delete pattern.
7141 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7143 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7144 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7147 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7150 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7151 by not USED_FOR_TARGET.
7152 (make_pass_resolve_sw_modes): Likewise.
7154 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7156 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7159 2018-01-08 Richard Biener <rguenther@suse.de>
7162 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7164 2018-01-08 Richard Biener <rguenther@suse.de>
7167 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7169 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7172 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7173 basic blocks with a small number of successors.
7174 (convert_control_dep_chain_into_preds): Improve handling of
7176 (dump_predicates): Split apart into...
7177 (dump_pred_chain): ...here...
7178 (dump_pred_info): ...and here.
7179 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7180 (can_chain_union_be_invalidated_p): Improve check for invalidation
7182 (uninit_uses_cannot_happen): Avoid unnecessary if
7183 convert_control_dep_chain_into_preds yielded nothing.
7185 2018-01-06 Martin Sebor <msebor@redhat.com>
7187 PR tree-optimization/83640
7188 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7189 subtracting negative offset from size.
7190 (builtin_access::overlap): Adjust offset bounds of the access to fall
7191 within the size of the object if possible.
7193 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7195 PR rtl-optimization/83699
7196 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7197 extract_bit_field_as_subreg to cases in which the extracted
7198 value is also a vector.
7200 * lra-constraints.c (process_alt_operands): Test for the equivalence
7201 substitutions when detecting a possible reload cycle.
7203 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7206 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7207 by default if flag_selective_schedling{,2}. Formatting fixes.
7209 PR rtl-optimization/83682
7210 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7211 if it has non-VECTOR_MODE element mode.
7212 (vec_duplicate_p): Likewise.
7215 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7216 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7218 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7221 * config/i386/i386-builtin.def
7222 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7223 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7224 Require also OPTION_MASK_ISA_AVX512F in addition to
7225 OPTION_MASK_ISA_GFNI.
7226 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7227 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7228 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7229 to OPTION_MASK_ISA_GFNI.
7230 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7231 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7232 OPTION_MASK_ISA_AVX512BW.
7233 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7234 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7235 addition to OPTION_MASK_ISA_GFNI.
7236 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7237 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7238 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7239 to OPTION_MASK_ISA_GFNI.
7240 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7241 a requirement for all ISAs rather than any of them with a few
7243 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7245 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7246 bitmasks to be enabled with 3 exceptions, instead of requiring any
7247 enabled ISA with lots of exceptions.
7248 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7249 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7250 Change avx512bw in isa attribute to avx512f.
7251 * config/i386/sgxintrin.h: Add license boilerplate.
7252 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7253 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7254 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7255 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7257 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7258 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7259 temporarily sse2 rather than sse if not enabled already.
7262 * config/i386/sse.md (VI248_VLBW): Rename to ...
7263 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7264 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7265 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7266 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7267 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7268 mode iterator instead of VI248_VLBW.
7270 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7272 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7273 (record_modified): Skip clobbers; add debug output.
7274 (param_change_prob): Use sreal frequencies.
7276 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7278 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7279 punt for user-aligned variables.
7281 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7283 * tree-chrec.c (chrec_contains_symbols): Return true for
7286 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7289 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7290 of (x|y) == x for BICS pattern.
7292 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7294 PR tree-optimization/83605
7295 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7296 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7299 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7301 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7302 * config/epiphany/rtems.h: New file.
7304 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7305 Uros Bizjak <ubizjak@gmail.com>
7308 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7309 QIreg_operand instead of register_operand predicate.
7310 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7311 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7312 comments instead of -fmitigate[-_]rop.
7314 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7317 * cgraphunit.c (symbol_table::compile): Switch to text_section
7318 before calling assembly_start debug hook.
7319 * run-rtl-passes.c (run_rtl_passes): Likewise.
7322 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7324 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7325 range_int_cst_p rather than !symbolic_range_p before calling
7326 extract_range_from_multiplicative_op_1.
7328 2018-01-04 Jeff Law <law@redhat.com>
7330 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7331 redundant test in assertion.
7333 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7335 * doc/rtl.texi: Document machine_mode wrapper classes.
7337 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7339 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
7342 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7344 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
7345 the VEC_PERM_EXPR fold to fail.
7347 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7350 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
7351 to switched_sections.
7353 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7356 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
7359 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
7362 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
7363 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
7365 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7368 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
7369 is BLKmode and bitpos not zero or mode change is needed.
7371 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7374 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
7377 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
7380 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
7381 instead of MULT rtx. Update all corresponding splitters.
7383 (*ssub<modesuffix>): Ditto.
7385 (*cmp_sadd_di): Update split patterns.
7386 (*cmp_sadd_si): Ditto.
7387 (*cmp_sadd_sidi): Ditto.
7388 (*cmp_ssub_di): Ditto.
7389 (*cmp_ssub_si): Ditto.
7390 (*cmp_ssub_sidi): Ditto.
7391 * config/alpha/predicates.md (const23_operand): New predicate.
7392 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
7393 Look for ASHIFT, not MULT inner operand.
7394 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
7396 2018-01-04 Martin Liska <mliska@suse.cz>
7398 PR gcov-profile/83669
7399 * gcov.c (output_intermediate_file): Add version to intermediate
7401 * doc/gcov.texi: Document new field 'version' in intermediate
7402 file format. Fix location of '-k' option of gcov command.
7404 2018-01-04 Martin Liska <mliska@suse.cz>
7407 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
7409 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7411 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
7413 2018-01-03 Martin Sebor <msebor@redhat.com>
7415 PR tree-optimization/83655
7416 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
7417 checking calls with invalid arguments.
7419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7421 * tree-vect-stmts.c (vect_get_store_rhs): New function.
7422 (vectorizable_mask_load_store): Delete.
7423 (vectorizable_call): Return false for masked loads and stores.
7424 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
7425 instead of gimple_assign_rhs1.
7426 (vectorizable_load): Handle IFN_MASK_LOAD.
7427 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7431 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
7433 (vectorizable_mask_load_store): ...here.
7434 (vectorizable_load): ...and here.
7436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 * tree-vect-stmts.c (vect_build_all_ones_mask)
7439 (vect_build_zero_merge_argument): New functions, split out from...
7440 (vectorizable_load): ...here.
7442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7444 * tree-vect-stmts.c (vect_check_store_rhs): New function,
7446 (vectorizable_mask_load_store): ...here.
7447 (vectorizable_store): ...and here.
7449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7451 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
7453 (vectorizable_mask_load_store): ...here.
7455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7457 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
7458 (vect_model_store_cost): Take a vec_load_store_type instead of a
7460 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
7461 (vect_model_store_cost): Take a vec_load_store_type instead of a
7463 (vectorizable_mask_load_store): Update accordingly.
7464 (vectorizable_store): Likewise.
7465 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
7467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7469 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
7470 IFN_MASK_LOAD calls here rather than...
7471 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
7473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7474 Alan Hayward <alan.hayward@arm.com>
7475 David Sherwood <david.sherwood@arm.com>
7477 * expmed.c (extract_bit_field_1): For vector extracts,
7478 fall back to extract_bit_field_as_subreg if vec_extract
7481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7482 Alan Hayward <alan.hayward@arm.com>
7483 David Sherwood <david.sherwood@arm.com>
7485 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
7486 they are variable or constant sized.
7487 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
7488 slots for constant-sized data.
7490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7491 Alan Hayward <alan.hayward@arm.com>
7492 David Sherwood <david.sherwood@arm.com>
7494 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
7495 handling COND_EXPRs with boolean comparisons, try to find a better
7496 basis for the mask type than the boolean itself.
7498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7500 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
7501 is calculated and how it can be overridden.
7502 * genmodes.c (max_bitsize_mode_any_mode): New variable.
7503 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
7505 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
7508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7509 Alan Hayward <alan.hayward@arm.com>
7510 David Sherwood <david.sherwood@arm.com>
7512 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
7513 Remove the mode argument.
7514 (aarch64_simd_valid_immediate): Remove the mode and inverse
7516 * config/aarch64/iterators.md (bitsize): New iterator.
7517 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
7518 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
7519 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
7520 aarch64_simd_valid_immediate.
7521 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
7522 (aarch64_reg_or_bic_imm): Likewise.
7523 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
7524 with an insn_type enum and msl with a modifier_type enum.
7525 Replace element_width with a scalar_mode. Change the shift
7526 to unsigned int. Add constructors for scalar_float_mode and
7527 scalar_int_mode elements.
7528 (aarch64_vect_float_const_representable_p): Delete.
7529 (aarch64_can_const_movi_rtx_p)
7530 (aarch64_simd_scalar_immediate_valid_for_move)
7531 (aarch64_simd_make_constant): Update call to
7532 aarch64_simd_valid_immediate.
7533 (aarch64_advsimd_valid_immediate_hs): New function.
7534 (aarch64_advsimd_valid_immediate): Likewise.
7535 (aarch64_simd_valid_immediate): Remove mode and inverse
7536 arguments. Rewrite to use the above. Use const_vec_duplicate_p
7537 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
7538 and aarch64_float_const_representable_p on the result.
7539 (aarch64_output_simd_mov_immediate): Remove mode argument.
7540 Update call to aarch64_simd_valid_immediate and use of
7541 simd_immediate_info.
7542 (aarch64_output_scalar_simd_mov_immediate): Update call
7545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7546 Alan Hayward <alan.hayward@arm.com>
7547 David Sherwood <david.sherwood@arm.com>
7549 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
7550 (mode_nunits): Likewise CONST_MODE_NUNITS.
7551 * machmode.def (ADJUST_NUNITS): Document.
7552 * genmodes.c (mode_data::need_nunits_adj): New field.
7553 (blank_mode): Update accordingly.
7554 (adj_nunits): New variable.
7555 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
7557 (emit_mode_size_inline): Set need_bytesize_adj for all modes
7558 listed in adj_nunits.
7559 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
7560 listed in adj_nunits. Don't emit case statements for such modes.
7561 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
7562 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
7563 nothing if adj_nunits is nonnull.
7564 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
7565 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
7566 (emit_mode_fbit): Update use of print_maybe_const_decl.
7567 (emit_move_size): Likewise. Treat the array as non-const
7569 (emit_mode_adjustments): Handle adj_nunits.
7571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7573 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
7574 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
7575 (VECTOR_MODES): Use it.
7576 (make_vector_modes): Take the prefix as an argument.
7578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7579 Alan Hayward <alan.hayward@arm.com>
7580 David Sherwood <david.sherwood@arm.com>
7582 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
7583 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
7584 for MODE_VECTOR_BOOL.
7585 * machmode.def (VECTOR_BOOL_MODE): Document.
7586 * genmodes.c (VECTOR_BOOL_MODE): New macro.
7587 (make_vector_bool_mode): New function.
7588 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
7590 * lto-streamer-in.c (lto_input_mode_table): Likewise.
7591 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
7593 * stor-layout.c (int_mode_for_mode): Likewise.
7594 * tree.c (build_vector_type_for_mode): Likewise.
7595 * varasm.c (output_constant_pool_2): Likewise.
7596 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
7597 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
7598 for MODE_VECTOR_BOOL.
7599 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
7600 of mode class checks.
7601 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
7602 instead of a list of mode class checks.
7603 (expand_vector_scalar_condition): Likewise.
7604 (type_for_widest_vector_mode): Handle BImode as an inner mode.
7606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7607 Alan Hayward <alan.hayward@arm.com>
7608 David Sherwood <david.sherwood@arm.com>
7610 * machmode.h (mode_size): Change from unsigned short to
7612 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
7613 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7614 or if measurement_type is not polynomial.
7615 (fixed_size_mode::includes_p): Check for constant-sized modes.
7616 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
7617 return a poly_uint16 rather than an unsigned short.
7618 (emit_mode_size): Change the type of mode_size from unsigned short
7619 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
7620 (emit_mode_adjustments): Cope with polynomial vector sizes.
7621 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7623 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7625 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
7626 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
7627 * caller-save.c (setup_save_areas): Likewise.
7628 (replace_reg_with_saved_mem): Likewise.
7629 * calls.c (emit_library_call_value_1): Likewise.
7630 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
7631 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
7632 (gen_lowpart_for_combine): Likewise.
7633 * convert.c (convert_to_integer_1): Likewise.
7634 * cse.c (equiv_constant, cse_insn): Likewise.
7635 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
7636 (cselib_subst_to_values): Likewise.
7637 * dce.c (word_dce_process_block): Likewise.
7638 * df-problems.c (df_word_lr_mark_ref): Likewise.
7639 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
7640 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
7641 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
7642 (rtl_for_decl_location): Likewise.
7643 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
7644 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
7645 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
7646 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
7647 (expand_expr_real_1): Likewise.
7648 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
7649 (pad_below): Likewise.
7650 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7651 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
7652 * ira.c (get_subreg_tracking_sizes): Likewise.
7653 * ira-build.c (ira_create_allocno_objects): Likewise.
7654 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
7655 (ira_sort_regnos_for_alter_reg): Likewise.
7656 * ira-costs.c (record_operand_costs): Likewise.
7657 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
7658 (resolve_simple_move): Likewise.
7659 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
7660 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
7661 (lra_constraints): Likewise.
7662 (CONST_POOL_OK_P): Reject variable-sized modes.
7663 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
7664 (add_pseudo_to_slot, lra_spill): Likewise.
7665 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7666 * optabs-query.c (get_best_extraction_insn): Likewise.
7667 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7668 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7669 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7670 * recog.c (offsettable_address_addr_space_p): Likewise.
7671 * regcprop.c (maybe_mode_change): Likewise.
7672 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7673 * regrename.c (build_def_use): Likewise.
7674 * regstat.c (dump_reg_info): Likewise.
7675 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7676 (find_reloads, find_reloads_subreg_address): Likewise.
7677 * reload1.c (eliminate_regs_1): Likewise.
7678 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7679 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7680 (simplify_binary_operation_1, simplify_subreg): Likewise.
7681 * targhooks.c (default_function_arg_padding): Likewise.
7682 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7683 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7684 (verify_gimple_assign_ternary): Likewise.
7685 * tree-inline.c (estimate_move_cost): Likewise.
7686 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7687 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7688 (get_address_cost_ainc): Likewise.
7689 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7690 (vect_supportable_dr_alignment): Likewise.
7691 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7692 (vectorizable_reduction): Likewise.
7693 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7694 (vectorizable_operation, vectorizable_load): Likewise.
7695 * tree.c (build_same_sized_truth_vector_type): Likewise.
7696 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7697 * var-tracking.c (emit_note_insn_var_location): Likewise.
7698 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7699 (ADDR_VEC_ALIGN): Likewise.
7701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7702 Alan Hayward <alan.hayward@arm.com>
7703 David Sherwood <david.sherwood@arm.com>
7705 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7707 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7708 or if measurement_type is polynomial.
7709 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7710 * combine.c (make_extraction): Likewise.
7711 * dse.c (find_shift_sequence): Likewise.
7712 * dwarf2out.c (mem_loc_descriptor): Likewise.
7713 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7714 (extract_bit_field, extract_low_bits): Likewise.
7715 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7716 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7717 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7718 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7719 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7720 * reload.c (find_reloads): Likewise.
7721 * reload1.c (alter_reg): Likewise.
7722 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7723 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7724 * tree-if-conv.c (predicate_mem_writes): Likewise.
7725 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7726 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7727 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7728 * valtrack.c (dead_debug_insert_temp): Likewise.
7729 * varasm.c (mergeable_constant_section): Likewise.
7730 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7733 Alan Hayward <alan.hayward@arm.com>
7734 David Sherwood <david.sherwood@arm.com>
7736 * expr.c (expand_assignment): Cope with polynomial mode sizes
7737 when assigning to a CONCAT.
7739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7740 Alan Hayward <alan.hayward@arm.com>
7741 David Sherwood <david.sherwood@arm.com>
7743 * machmode.h (mode_precision): Change from unsigned short to
7745 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7747 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7748 or if measurement_type is not polynomial.
7749 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7750 in which the mode is already known to be a scalar_int_mode.
7751 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7752 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7754 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7755 for GET_MODE_PRECISION.
7756 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7757 for GET_MODE_PRECISION.
7758 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7760 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7761 (expand_field_assignment, make_extraction): Likewise.
7762 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7763 (get_last_value): Likewise.
7764 * convert.c (convert_to_integer_1): Likewise.
7765 * cse.c (cse_insn): Likewise.
7766 * expr.c (expand_expr_real_1): Likewise.
7767 * lra-constraints.c (simplify_operand_subreg): Likewise.
7768 * optabs-query.c (can_atomic_load_p): Likewise.
7769 * optabs.c (expand_atomic_load): Likewise.
7770 (expand_atomic_store): Likewise.
7771 * ree.c (combine_reaching_defs): Likewise.
7772 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7773 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7774 * tree.h (type_has_mode_precision_p): Likewise.
7775 * ubsan.c (instrument_si_overflow): Likewise.
7777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7778 Alan Hayward <alan.hayward@arm.com>
7779 David Sherwood <david.sherwood@arm.com>
7781 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7782 polynomial numbers of units.
7783 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7784 (valid_vector_subparts_p): New function.
7785 (build_vector_type): Remove temporary shim and take the number
7786 of units as a poly_uint64 rather than an int.
7787 (build_opaque_vector_type): Take the number of units as a
7788 poly_uint64 rather than an int.
7789 * tree.c (build_vector_from_ctor): Handle polynomial
7790 TYPE_VECTOR_SUBPARTS.
7791 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7792 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7793 (build_vector_from_val): If the number of units is variable,
7794 use build_vec_duplicate_cst for constant operands and
7795 VEC_DUPLICATE_EXPR otherwise.
7796 (make_vector_type): Remove temporary is_constant ().
7797 (build_vector_type, build_opaque_vector_type): Take the number of
7798 units as a poly_uint64 rather than an int.
7799 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7801 * cfgexpand.c (expand_debug_expr): Likewise.
7802 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7803 (store_constructor, expand_expr_real_1): Likewise.
7804 (const_scalar_mask_from_tree): Likewise.
7805 * fold-const-call.c (fold_const_reduction): Likewise.
7806 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7807 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7808 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7809 (fold_relational_const): Likewise.
7810 (native_interpret_vector): Likewise. Change the size from an
7811 int to an unsigned int.
7812 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7813 TYPE_VECTOR_SUBPARTS.
7814 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7815 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7816 duplicating a non-constant operand into a variable-length vector.
7817 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7818 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7819 * ipa-icf.c (sem_variable::equals): Likewise.
7820 * match.pd: Likewise.
7821 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7822 * print-tree.c (print_node): Likewise.
7823 * stor-layout.c (layout_type): Likewise.
7824 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7825 * tree-cfg.c (verify_gimple_comparison): Likewise.
7826 (verify_gimple_assign_binary): Likewise.
7827 (verify_gimple_assign_ternary): Likewise.
7828 (verify_gimple_assign_single): Likewise.
7829 * tree-pretty-print.c (dump_generic_node): Likewise.
7830 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7831 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7832 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7833 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7834 (vect_shift_permute_load_chain): Likewise.
7835 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7836 (expand_vector_condition, optimize_vector_constructor): Likewise.
7837 (lower_vec_perm, get_compute_type): Likewise.
7838 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7839 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7840 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7841 (vect_recog_mask_conversion_pattern): Likewise.
7842 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7843 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7844 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7845 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7846 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7847 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7848 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7849 (supportable_widening_operation): Likewise.
7850 (supportable_narrowing_operation): Likewise.
7851 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7853 * varasm.c (output_constant): Likewise.
7855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7856 Alan Hayward <alan.hayward@arm.com>
7857 David Sherwood <david.sherwood@arm.com>
7859 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7860 so that both the length == 3 and length != 3 cases set up their
7861 own permute vectors. Add comments explaining why we know the
7862 number of elements is constant.
7863 (vect_permute_load_chain): Likewise.
7865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7866 Alan Hayward <alan.hayward@arm.com>
7867 David Sherwood <david.sherwood@arm.com>
7869 * machmode.h (mode_nunits): Change from unsigned char to
7871 (ONLY_FIXED_SIZE_MODES): New macro.
7872 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7873 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7874 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7876 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7877 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7878 or if measurement_type is not polynomial.
7879 * genmodes.c (ZERO_COEFFS): New macro.
7880 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7882 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7883 Use ZERO_COEFFS when emitting initializers.
7884 * data-streamer.h (bp_pack_poly_value): New function.
7885 (bp_unpack_poly_value): Likewise.
7886 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7887 for GET_MODE_NUNITS.
7888 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7889 for GET_MODE_NUNITS.
7890 * tree.c (make_vector_type): Remove temporary shim and make
7891 the real function take the number of units as a poly_uint64
7893 (build_vector_type_for_mode): Handle polynomial nunits.
7894 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7895 * emit-rtl.c (const_vec_series_p_1): Likewise.
7896 (gen_rtx_CONST_VECTOR): Likewise.
7897 * fold-const.c (test_vec_duplicate_folding): Likewise.
7898 * genrecog.c (validate_pattern): Likewise.
7899 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7900 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7901 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7902 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7903 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7904 * rtlanal.c (subreg_get_info): Likewise.
7905 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7906 (vect_grouped_load_supported): Likewise.
7907 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7908 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7909 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7910 (simplify_const_unary_operation, simplify_binary_operation_1)
7911 (simplify_const_binary_operation, simplify_ternary_operation)
7912 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7913 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7914 instead of CONST_VECTOR_NUNITS.
7915 * varasm.c (output_constant_pool_2): Likewise.
7916 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7917 explicit-encoded elements in the XVEC for variable-length vectors.
7919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7921 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7924 Alan Hayward <alan.hayward@arm.com>
7925 David Sherwood <david.sherwood@arm.com>
7927 * coretypes.h (fixed_size_mode): Declare.
7928 (fixed_size_mode_pod): New typedef.
7929 * builtins.h (target_builtins::x_apply_args_mode)
7930 (target_builtins::x_apply_result_mode): Change type to
7931 fixed_size_mode_pod.
7932 * builtins.c (apply_args_size, apply_result_size, result_vector)
7933 (expand_builtin_apply_args_1, expand_builtin_apply)
7934 (expand_builtin_return): Update accordingly.
7936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7938 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7939 * cselib.c (cselib_hash_rtx): Likewise.
7940 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7941 CONST_VECTOR encoding.
7943 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7944 Jeff Law <law@redhat.com>
7947 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7948 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7949 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7950 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7953 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7954 explicitly probe *sp in a noreturn function if there were any callee
7955 register saves or frame pointer is needed.
7957 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7960 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7961 BLKmode for ternary, binary or unary expressions.
7964 * var-tracking.c (delete_vta_debug_insn): New inline function.
7965 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7966 insns from get_insns () to NULL instead of each bb separately.
7967 Use delete_vta_debug_insn. No longer static.
7968 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7969 delete_vta_debug_insns callers.
7970 * rtl.h (delete_vta_debug_insns): Declare.
7971 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7972 instead of variable_tracking_main.
7974 2018-01-03 Martin Sebor <msebor@redhat.com>
7976 PR tree-optimization/83603
7977 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7978 arguments past the endof the argument list in functions declared
7979 without a prototype.
7980 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7981 Avoid checking when arguments are null.
7983 2018-01-03 Martin Sebor <msebor@redhat.com>
7986 * doc/extend.texi (attribute const): Fix a typo.
7987 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7988 issuing -Wsuggest-attribute for void functions.
7990 2018-01-03 Martin Sebor <msebor@redhat.com>
7992 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7993 offset_int::from instead of wide_int::to_shwi.
7994 (maybe_diag_overlap): Remove assertion.
7995 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7996 * gimple-ssa-sprintf.c (format_directive): Same.
7997 (parse_directive): Same.
7998 (sprintf_dom_walker::compute_format_length): Same.
7999 (try_substitute_return_value): Same.
8001 2018-01-03 Jeff Law <law@redhat.com>
8004 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8005 non-constant residual for zero at runtime and avoid probing in
8006 that case. Reorganize code for trailing problem to mirror handling
8009 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8011 PR tree-optimization/83501
8012 * tree-ssa-strlen.c (get_string_cst): New.
8013 (handle_char_store): Call get_string_cst.
8015 2018-01-03 Martin Liska <mliska@suse.cz>
8017 PR tree-optimization/83593
8018 * tree-ssa-strlen.c: Include tree-cfg.h.
8019 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8020 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8021 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8023 (strlen_dom_walker::before_dom_children): Call
8024 gimple_purge_dead_eh_edges. Dump tranformation with details
8026 (strlen_dom_walker::before_dom_children): Update call by adding
8027 new argument cleanup_eh.
8028 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8030 2018-01-03 Martin Liska <mliska@suse.cz>
8033 * cif-code.def (VARIADIC_THUNK): New enum value.
8034 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8037 2018-01-03 Jan Beulich <jbeulich@suse.com>
8039 * sse.md (mov<mode>_internal): Tighten condition for when to use
8040 vmovdqu<ssescalarsize> for TI and OI modes.
8042 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8044 Update copyright years.
8046 2018-01-03 Martin Liska <mliska@suse.cz>
8049 * ipa-visibility.c (function_and_variable_visibility): Skip
8050 functions with noipa attribure.
8052 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8054 * gcc.c (process_command): Update copyright notice dates.
8055 * gcov-dump.c (print_version): Ditto.
8056 * gcov.c (print_version): Ditto.
8057 * gcov-tool.c (print_version): Ditto.
8058 * gengtype.c (create_file): Ditto.
8059 * doc/cpp.texi: Bump @copying's copyright year.
8060 * doc/cppinternals.texi: Ditto.
8061 * doc/gcc.texi: Ditto.
8062 * doc/gccint.texi: Ditto.
8063 * doc/gcov.texi: Ditto.
8064 * doc/install.texi: Ditto.
8065 * doc/invoke.texi: Ditto.
8067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8069 * vector-builder.h (vector_builder::m_full_nelts): Change from
8070 unsigned int to poly_uint64.
8071 (vector_builder::full_nelts): Update prototype accordingly.
8072 (vector_builder::new_vector): Likewise.
8073 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8074 (vector_builder::operator ==): Likewise.
8075 (vector_builder::finalize): Likewise.
8076 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8077 Take the number of elements as a poly_uint64 rather than an
8079 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8080 from unsigned int to poly_uint64.
8081 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8082 (vec_perm_indices::new_vector): Likewise.
8083 (vec_perm_indices::length): Likewise.
8084 (vec_perm_indices::nelts_per_input): Likewise.
8085 (vec_perm_indices::input_nelts): Likewise.
8086 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8087 number of elements per input as a poly_uint64 rather than an
8088 unsigned int. Use the original encoding for variable-length
8089 vectors, rather than clamping each individual element.
8090 For the second and subsequent elements in each pattern,
8091 clamp the step and base before clamping their sum.
8092 (vec_perm_indices::series_p): Handle polynomial element counts.
8093 (vec_perm_indices::all_in_range_p): Likewise.
8094 (vec_perm_indices_to_tree): Likewise.
8095 (vec_perm_indices_to_rtx): Likewise.
8096 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8097 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8098 (tree_vector_builder::new_binary_operation): Handle polynomial
8099 element counts. Return false if we need to know the number
8100 of elements at compile time.
8101 * fold-const.c (fold_vec_perm): Punt if the number of elements
8102 isn't known at compile time.
8104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8106 * vec-perm-indices.h (vec_perm_builder): Change element type
8107 from HOST_WIDE_INT to poly_int64.
8108 (vec_perm_indices::element_type): Update accordingly.
8109 (vec_perm_indices::clamp): Handle polynomial element_types.
8110 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8111 (vec_perm_indices::all_in_range_p): Likewise.
8112 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8114 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8115 polynomial vec_perm_indices element types.
8116 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8117 * fold-const.c (fold_vec_perm): Likewise.
8118 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8119 * tree-vect-generic.c (lower_vec_perm): Likewise.
8120 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8121 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8122 element type to HOST_WIDE_INT.
8124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8125 Alan Hayward <alan.hayward@arm.com>
8126 David Sherwood <david.sherwood@arm.com>
8128 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8129 rather than an int. Use plus_constant.
8130 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8131 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8134 Alan Hayward <alan.hayward@arm.com>
8135 David Sherwood <david.sherwood@arm.com>
8137 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8138 a HOST_WIDE_INT to a poly_int64.
8140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8141 Alan Hayward <alan.hayward@arm.com>
8142 David Sherwood <david.sherwood@arm.com>
8144 * calls.c (load_register_parameters): Cope with polynomial
8145 mode sizes. Require a constant size for BLKmode parameters
8146 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8147 forces a parameter to be padded at the lsb end in order to
8148 fill a complete number of words, require the parameter size
8149 to be ordered wrt UNITS_PER_WORD.
8151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8152 Alan Hayward <alan.hayward@arm.com>
8153 David Sherwood <david.sherwood@arm.com>
8155 * reload1.c (spill_stack_slot_width): Change element type
8156 from unsigned int to poly_uint64_pod.
8157 (alter_reg): Treat mode sizes as polynomial.
8159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8160 Alan Hayward <alan.hayward@arm.com>
8161 David Sherwood <david.sherwood@arm.com>
8163 * reload.c (complex_word_subreg_p): New function.
8164 (reload_inner_reg_of_subreg, push_reload): Use it.
8166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8167 Alan Hayward <alan.hayward@arm.com>
8168 David Sherwood <david.sherwood@arm.com>
8170 * lra-constraints.c (process_alt_operands): Reject matched
8171 operands whose sizes aren't ordered.
8172 (match_reload): Refer to this check here.
8174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8175 Alan Hayward <alan.hayward@arm.com>
8176 David Sherwood <david.sherwood@arm.com>
8178 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8179 that the mode size is in the set {1, 2, 4, 8, 16}.
8181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8182 Alan Hayward <alan.hayward@arm.com>
8183 David Sherwood <david.sherwood@arm.com>
8185 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8186 Use plus_constant instead of gen_rtx_PLUS.
8188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8189 Alan Hayward <alan.hayward@arm.com>
8190 David Sherwood <david.sherwood@arm.com>
8192 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8193 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8194 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8195 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8196 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8197 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8198 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8199 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8200 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8201 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8203 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8204 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8205 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8206 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8207 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8208 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8209 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8210 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8211 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8212 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8214 * expr.c (emit_move_resolve_push): Treat the input and result
8215 of PUSH_ROUNDING as a poly_int64.
8216 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8217 (emit_push_insn): Likewise.
8218 * lra-eliminations.c (mark_not_eliminable): Likewise.
8219 * recog.c (push_operand): Likewise.
8220 * reload1.c (elimination_effects): Likewise.
8221 * rtlanal.c (nonzero_bits1): Likewise.
8222 * calls.c (store_one_arg): Likewise. Require the padding to be
8223 known at compile time.
8225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8226 Alan Hayward <alan.hayward@arm.com>
8227 David Sherwood <david.sherwood@arm.com>
8229 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8230 Use plus_constant instead of gen_rtx_PLUS.
8232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8233 Alan Hayward <alan.hayward@arm.com>
8234 David Sherwood <david.sherwood@arm.com>
8236 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8240 Alan Hayward <alan.hayward@arm.com>
8241 David Sherwood <david.sherwood@arm.com>
8243 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8244 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8245 via stack temporaries. Treat the mode size as polynomial too.
8247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8248 Alan Hayward <alan.hayward@arm.com>
8249 David Sherwood <david.sherwood@arm.com>
8251 * expr.c (expand_expr_real_2): When handling conversions involving
8252 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8253 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8254 as a poly_uint64 too.
8256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8257 Alan Hayward <alan.hayward@arm.com>
8258 David Sherwood <david.sherwood@arm.com>
8260 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8263 Alan Hayward <alan.hayward@arm.com>
8264 David Sherwood <david.sherwood@arm.com>
8266 * combine.c (can_change_dest_mode): Handle polynomial
8267 REGMODE_NATURAL_SIZE.
8268 * expmed.c (store_bit_field_1): Likewise.
8269 * expr.c (store_constructor): Likewise.
8270 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8271 and polynomial REGMODE_NATURAL_SIZE.
8272 (gen_lowpart_common): Likewise.
8273 * reginfo.c (record_subregs_of_mode): Likewise.
8274 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8281 numbers of elements.
8283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8284 Alan Hayward <alan.hayward@arm.com>
8285 David Sherwood <david.sherwood@arm.com>
8287 * match.pd: Cope with polynomial numbers of vector elements.
8289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8290 Alan Hayward <alan.hayward@arm.com>
8291 David Sherwood <david.sherwood@arm.com>
8293 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8294 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
8301 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8302 (ipa_simd_modify_function_body): Likewise.
8304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8305 Alan Hayward <alan.hayward@arm.com>
8306 David Sherwood <david.sherwood@arm.com>
8308 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8309 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8310 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8311 (expand_vector_condition, vector_element): Likewise.
8312 (subparts_gt): New function.
8313 (get_compute_type): Use subparts_gt.
8314 (count_type_subparts): Delete.
8315 (expand_vector_operations_1): Use subparts_gt instead of
8316 count_type_subparts.
8318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8319 Alan Hayward <alan.hayward@arm.com>
8320 David Sherwood <david.sherwood@arm.com>
8322 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8323 (vect_compile_time_alias): ...this new function. Do the calculation
8324 on poly_ints rather than trees.
8325 (vect_prune_runtime_alias_test_list): Update call accordingly.
8327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8328 Alan Hayward <alan.hayward@arm.com>
8329 David Sherwood <david.sherwood@arm.com>
8331 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8333 (vect_schedule_slp_instance): Likewise.
8335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8336 Alan Hayward <alan.hayward@arm.com>
8337 David Sherwood <david.sherwood@arm.com>
8339 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
8340 constant and extern definitions for variable-length vectors.
8341 (vect_get_constant_vectors): Note that the number of units
8342 is known to be constant.
8344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8345 Alan Hayward <alan.hayward@arm.com>
8346 David Sherwood <david.sherwood@arm.com>
8348 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
8349 of units as polynomial. Choose between WIDE and NARROW based
8352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8353 Alan Hayward <alan.hayward@arm.com>
8354 David Sherwood <david.sherwood@arm.com>
8356 * tree-vect-stmts.c (simd_clone_subparts): New function.
8357 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
8359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8360 Alan Hayward <alan.hayward@arm.com>
8361 David Sherwood <david.sherwood@arm.com>
8363 * tree-vect-stmts.c (vectorizable_call): Treat the number of
8364 vectors as polynomial. Use build_index_vector for
8367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8368 Alan Hayward <alan.hayward@arm.com>
8369 David Sherwood <david.sherwood@arm.com>
8371 * tree-vect-stmts.c (get_load_store_type): Treat the number of
8372 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
8373 for variable-length vectors.
8374 (vectorizable_mask_load_store): Treat the number of units as
8375 polynomial, asserting that it is constant if the condition has
8376 already been enforced.
8377 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
8384 of units as polynomial. Punt if we can't tell at compile time
8385 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
8392 of units as polynomial. Punt on SLP inductions. Use an integer
8393 VEC_SERIES_EXPR for variable-length integer reductions. Use a
8394 cast of such a series for variable-length floating-point
8397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8398 Alan Hayward <alan.hayward@arm.com>
8399 David Sherwood <david.sherwood@arm.com>
8401 * tree.h (build_index_vector): Declare.
8402 * tree.c (build_index_vector): New function.
8403 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8404 of units as polynomial, forcibly converting it to a constant if
8405 vectorizable_reduction has already enforced the condition.
8406 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
8407 to create a {1,2,3,...} vector.
8408 (vectorizable_reduction): Treat the number of units as polynomial.
8409 Choose vectype_in based on the largest scalar element size rather
8410 than the smallest number of units. Enforce the restrictions
8413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8414 Alan Hayward <alan.hayward@arm.com>
8415 David Sherwood <david.sherwood@arm.com>
8417 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
8418 number of units as polynomial.
8420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8421 Alan Hayward <alan.hayward@arm.com>
8422 David Sherwood <david.sherwood@arm.com>
8424 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
8425 * target.def (autovectorize_vector_sizes): Return the vector sizes
8426 by pointer, using vector_sizes rather than a bitmask.
8427 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
8428 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
8429 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
8431 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
8432 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
8433 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
8434 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
8435 * omp-general.c (omp_max_vf): Likewise.
8436 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8437 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
8438 * tree-vect-loop.c (vect_analyze_loop): Likewise.
8439 * tree-vect-slp.c (vect_slp_bb): Likewise.
8440 * doc/tm.texi: Regenerate.
8441 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
8443 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
8444 the vector size as a poly_uint64 rather than an unsigned int.
8445 (current_vector_size): Change from an unsigned int to a poly_uint64.
8446 (get_vectype_for_scalar_type): Update accordingly.
8447 * tree.h (build_truth_vector_type): Take the size and number of
8448 units as a poly_uint64 rather than an unsigned int.
8449 (build_vector_type): Add a temporary overload that takes
8450 the number of units as a poly_uint64 rather than an unsigned int.
8451 * tree.c (make_vector_type): Likewise.
8452 (build_truth_vector_type): Take the number of units as a poly_uint64
8453 rather than an unsigned int.
8455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8456 Alan Hayward <alan.hayward@arm.com>
8457 David Sherwood <david.sherwood@arm.com>
8459 * target.def (get_mask_mode): Take the number of units and length
8460 as poly_uint64s rather than unsigned ints.
8461 * targhooks.h (default_get_mask_mode): Update accordingly.
8462 * targhooks.c (default_get_mask_mode): Likewise.
8463 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
8464 * doc/tm.texi: Regenerate.
8466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8467 Alan Hayward <alan.hayward@arm.com>
8468 David Sherwood <david.sherwood@arm.com>
8470 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
8471 * omp-general.c (omp_max_vf): Likewise.
8472 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
8473 (expand_omp_simd): Handle polynomial safelen.
8474 * omp-low.c (omplow_simd_context): Add a default constructor.
8475 (omplow_simd_context::max_vf): Change from int to poly_uint64.
8476 (lower_rec_simd_input_clauses): Update accordingly.
8477 (lower_rec_input_clauses): Likewise.
8479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8480 Alan Hayward <alan.hayward@arm.com>
8481 David Sherwood <david.sherwood@arm.com>
8483 * tree-vectorizer.h (vect_nunits_for_cost): New function.
8484 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
8485 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
8486 (vect_analyze_slp_cost): Likewise.
8487 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
8488 (vect_model_load_cost): Likewise.
8490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8491 Alan Hayward <alan.hayward@arm.com>
8492 David Sherwood <david.sherwood@arm.com>
8494 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
8495 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
8496 from an unsigned int * to a poly_uint64_pod *.
8497 (calculate_unrolling_factor): New function.
8498 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
8500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8501 Alan Hayward <alan.hayward@arm.com>
8502 David Sherwood <david.sherwood@arm.com>
8504 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
8505 from an unsigned int to a poly_uint64.
8506 (_loop_vec_info::slp_unrolling_factor): Likewise.
8507 (_loop_vec_info::vectorization_factor): Change from an int
8509 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
8510 (vect_get_num_vectors): New function.
8511 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
8512 (vect_get_num_copies): Use vect_get_num_vectors.
8513 (vect_analyze_data_ref_dependences): Change max_vf from an int *
8514 to an unsigned int *.
8515 (vect_analyze_data_refs): Change min_vf from an int * to a
8517 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8518 than an unsigned HOST_WIDE_INT.
8519 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8520 (vect_analyze_data_ref_dependence): Change max_vf from an int *
8521 to an unsigned int *.
8522 (vect_analyze_data_ref_dependences): Likewise.
8523 (vect_compute_data_ref_alignment): Handle polynomial vf.
8524 (vect_enhance_data_refs_alignment): Likewise.
8525 (vect_prune_runtime_alias_test_list): Likewise.
8526 (vect_shift_permute_load_chain): Likewise.
8527 (vect_supportable_dr_alignment): Likewise.
8528 (dependence_distance_ge_vf): Take the vectorization factor as a
8529 poly_uint64 rather than an unsigned HOST_WIDE_INT.
8530 (vect_analyze_data_refs): Change min_vf from an int * to a
8532 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
8533 vfm1 as a poly_uint64 rather than an int. Make the same change
8534 for the returned bound_scalar.
8535 (vect_gen_vector_loop_niters): Handle polynomial vf.
8536 (vect_do_peeling): Likewise. Update call to
8537 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
8538 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
8540 * tree-vect-loop.c (vect_determine_vectorization_factor)
8541 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
8542 (vect_get_known_peeling_cost): Likewise.
8543 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
8544 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
8545 (vect_transform_loop): Likewise. Use the lowest possible VF when
8546 updating the upper bounds of the loop.
8547 (vect_min_worthwhile_factor): Make static. Return an unsigned int
8549 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
8550 polynomial unroll factors.
8551 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
8552 (vect_make_slp_decision): Likewise.
8553 (vect_supported_load_permutation_p): Likewise, and polynomial
8555 (vect_analyze_slp_cost): Handle polynomial vf.
8556 (vect_slp_analyze_node_operations): Likewise.
8557 (vect_slp_analyze_bb_1): Likewise.
8558 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
8559 than an unsigned HOST_WIDE_INT.
8560 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
8561 (vectorizable_load): Handle polynomial vf.
8562 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
8564 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
8566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8567 Alan Hayward <alan.hayward@arm.com>
8568 David Sherwood <david.sherwood@arm.com>
8570 * match.pd: Handle bit operations involving three constants
8571 and try to fold one pair.
8573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8575 * tree-vect-loop-manip.c: Include gimple-fold.h.
8576 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
8577 niters_maybe_zero parameters. Handle other cases besides a step of 1.
8578 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
8579 Add a path that uses a step of VF instead of 1, but disable it
8581 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
8582 and niters_no_overflow parameters. Update calls to
8583 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
8584 Create a new SSA name if the latter choses to use a ste other
8585 than zero, and return it via niters_vector_mult_vf_var.
8586 * tree-vect-loop.c (vect_transform_loop): Update calls to
8587 vect_do_peeling, vect_gen_vector_loop_niters and
8588 slpeel_make_loop_iterate_ntimes.
8589 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
8590 (vect_gen_vector_loop_niters): Update declarations after above changes.
8592 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
8594 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
8595 128-bit round to integer instructions.
8596 (ceil<mode>2): Likewise.
8597 (btrunc<mode>2): Likewise.
8598 (round<mode>2): Likewise.
8600 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8602 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
8603 unaligned VSX load/store on P8/P9.
8604 (expand_block_clear): Allow the use of unaligned VSX
8605 load/store on P8/P9.
8607 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8609 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
8611 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
8612 swap associated with both a load and a store.
8614 2018-01-02 Andrew Waterman <andrew@sifive.com>
8616 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
8617 * config/riscv/riscv.md (clear_cache): Use it.
8619 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
8621 * web.c: Remove out-of-date comment.
8623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8625 * expr.c (fixup_args_size_notes): Check that any existing
8626 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
8627 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
8628 (emit_single_push_insn): ...here.
8630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8632 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
8633 (const_vector_encoded_nelts): New function.
8634 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
8635 (const_vector_int_elt, const_vector_elt): Declare.
8636 * emit-rtl.c (const_vector_int_elt_1): New function.
8637 (const_vector_elt): Likewise.
8638 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
8639 of CONST_VECTOR_ELT.
8641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8643 * expr.c: Include rtx-vector-builder.h.
8644 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
8645 directly on the tree encoding.
8646 (const_vector_from_tree): Likewise.
8647 * optabs.c: Include rtx-vector-builder.h.
8648 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
8649 sequence of "u" values.
8650 * vec-perm-indices.c: Include rtx-vector-builder.h.
8651 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
8652 directly on the vec_perm_indices encoding.
8654 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8656 * doc/rtl.texi (const_vector): Describe new encoding scheme.
8657 * Makefile.in (OBJS): Add rtx-vector-builder.o.
8658 * rtx-vector-builder.h: New file.
8659 * rtx-vector-builder.c: Likewise.
8660 * rtl.h (rtx_def::u2): Add a const_vector field.
8661 (CONST_VECTOR_NPATTERNS): New macro.
8662 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
8663 (CONST_VECTOR_DUPLICATE_P): Likewise.
8664 (CONST_VECTOR_STEPPED_P): Likewise.
8665 (CONST_VECTOR_ENCODED_ELT): Likewise.
8666 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8667 (unwrap_const_vec_duplicate): Likewise.
8668 (const_vec_series_p): Check for a non-duplicated vector encoding.
8669 Say that the function only returns true for integer vectors.
8670 * emit-rtl.c: Include rtx-vector-builder.h.
8671 (gen_const_vec_duplicate_1): Delete.
8672 (gen_const_vector): Call gen_const_vec_duplicate instead of
8673 gen_const_vec_duplicate_1.
8674 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8675 (gen_const_vec_duplicate): Use rtx_vector_builder.
8676 (gen_const_vec_series): Likewise.
8677 (gen_rtx_CONST_VECTOR): Likewise.
8678 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8679 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8680 Build a new vector rather than modifying a CONST_VECTOR in-place.
8681 (handle_special_swappables): Update call accordingly.
8682 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8683 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8684 Build a new vector rather than modifying a CONST_VECTOR in-place.
8685 (handle_special_swappables): Update call accordingly.
8687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8689 * simplify-rtx.c (simplify_const_binary_operation): Use
8690 CONST_VECTOR_ELT instead of XVECEXP.
8692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8694 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8695 the selector elements to be different from the data elements
8696 if the selector is a VECTOR_CST.
8697 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8698 ssizetype for the selector.
8700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8702 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8703 before testing each element individually.
8704 * tree-vect-generic.c (lower_vec_perm): Likewise.
8706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8708 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8709 * selftest-run-tests.c (selftest::run_tests): Call it.
8710 * vector-builder.h (vector_builder::operator ==): New function.
8711 (vector_builder::operator !=): Likewise.
8712 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8713 (vec_perm_indices::all_from_input_p): New function.
8714 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8715 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8716 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8717 instead of reading the VECTOR_CST directly. Detect whether both
8718 vector inputs are the same before constructing the vec_perm_indices,
8719 and update the number of inputs argument accordingly. Use the
8720 utility functions added above. Only construct sel2 if we need to.
8722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8724 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8725 the broadcast of the low byte.
8726 (expand_mult_highpart): Use an explicit encoding for the permutes.
8727 * optabs-query.c (can_mult_highpart_p): Likewise.
8728 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8729 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8730 (vectorizable_bswap): Likewise.
8731 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8732 explicit encoding for the power-of-2 permutes.
8733 (vect_permute_store_chain): Likewise.
8734 (vect_grouped_load_supported): Likewise.
8735 (vect_permute_load_chain): Likewise.
8737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8739 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8740 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8741 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8742 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8743 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8744 (vect_gen_perm_mask_any): Likewise.
8746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8748 * int-vector-builder.h: New file.
8749 * vec-perm-indices.h: Include int-vector-builder.h.
8750 (vec_perm_indices): Redefine as an int_vector_builder.
8751 (auto_vec_perm_indices): Delete.
8752 (vec_perm_builder): Redefine as a stand-alone class.
8753 (vec_perm_indices::vec_perm_indices): New function.
8754 (vec_perm_indices::clamp): Likewise.
8755 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8756 (vec_perm_indices::new_vector): New function.
8757 (vec_perm_indices::new_expanded_vector): Update for new
8758 vec_perm_indices class.
8759 (vec_perm_indices::rotate_inputs): New function.
8760 (vec_perm_indices::all_in_range_p): Operate directly on the
8761 encoded form, without computing elided elements.
8762 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8763 encoding. Update for new vec_perm_indices class.
8764 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8765 the given vec_perm_builder.
8766 (expand_vec_perm_var): Update vec_perm_builder constructor.
8767 (expand_mult_highpart): Use vec_perm_builder instead of
8768 auto_vec_perm_indices.
8769 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8770 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8771 or double series encoding as appropriate.
8772 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8773 vec_perm_indices instead of auto_vec_perm_indices.
8774 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8775 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8776 (vect_permute_store_chain): Likewise.
8777 (vect_grouped_load_supported): Likewise.
8778 (vect_permute_load_chain): Likewise.
8779 (vect_shift_permute_load_chain): Likewise.
8780 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8781 (vect_transform_slp_perm_load): Likewise.
8782 (vect_schedule_slp_instance): Likewise.
8783 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8784 (vectorizable_mask_load_store): Likewise.
8785 (vectorizable_bswap): Likewise.
8786 (vectorizable_store): Likewise.
8787 (vectorizable_load): Likewise.
8788 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8789 vec_perm_indices instead of auto_vec_perm_indices. Use
8790 tree_to_vec_perm_builder to read the vector from a tree.
8791 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8792 vec_perm_builder instead of a vec_perm_indices.
8793 (have_whole_vector_shift): Use vec_perm_builder and
8794 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8795 truncation to calc_vec_perm_mask_for_shift.
8796 (vect_create_epilog_for_reduction): Likewise.
8797 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8798 from auto_vec_perm_indices to vec_perm_indices.
8799 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8800 instead of changing individual elements.
8801 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8802 the vector in d.perm.
8803 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8804 from auto_vec_perm_indices to vec_perm_indices.
8805 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8806 instead of changing individual elements.
8807 (arm_vectorize_vec_perm_const): Use new_vector to install
8808 the vector in d.perm.
8809 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8810 Update vec_perm_builder constructor.
8811 (rs6000_expand_interleave): Likewise.
8812 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8813 (rs6000_expand_interleave): Likewise.
8815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8817 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8818 to qimode could truncate the indices.
8819 * optabs.c (expand_vec_perm_var): Likewise.
8821 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8823 * Makefile.in (OBJS): Add vec-perm-indices.o.
8824 * vec-perm-indices.h: New file.
8825 * vec-perm-indices.c: Likewise.
8826 * target.h (vec_perm_indices): Replace with a forward class
8828 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8829 * optabs.h: Include vec-perm-indices.h.
8830 (expand_vec_perm): Delete.
8831 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8832 (expand_vec_perm_const): Declare.
8833 * target.def (vec_perm_const_ok): Replace with...
8834 (vec_perm_const): ...this new hook.
8835 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8836 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8837 * doc/tm.texi: Regenerate.
8838 * optabs.def (vec_perm_const): Delete.
8839 * doc/md.texi (vec_perm_const): Likewise.
8840 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8841 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8842 expand_vec_perm for constant permutation vectors. Assert that
8843 the mode of variable permutation vectors is the integer equivalent
8844 of the mode that is being permuted.
8845 * optabs-query.h (selector_fits_mode_p): Declare.
8846 * optabs-query.c: Include vec-perm-indices.h.
8847 (selector_fits_mode_p): New function.
8848 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8849 is defined, instead of checking whether the vec_perm_const_optab
8850 exists. Use targetm.vectorize.vec_perm_const instead of
8851 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8852 fit in the vector mode before using a variable permute.
8853 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8854 vec_perm_indices instead of an rtx.
8855 (expand_vec_perm): Replace with...
8856 (expand_vec_perm_const): ...this new function. Take the selector
8857 as a vec_perm_indices rather than an rtx. Also take the mode of
8858 the selector. Update call to shift_amt_for_vec_perm_mask.
8859 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8860 Use vec_perm_indices::new_expanded_vector to expand the original
8861 selector into bytes. Check whether the indices fit in the vector
8862 mode before using a variable permute.
8863 (expand_vec_perm_var): Make global.
8864 (expand_mult_highpart): Use expand_vec_perm_const.
8865 * fold-const.c: Includes vec-perm-indices.h.
8866 * tree-ssa-forwprop.c: Likewise.
8867 * tree-vect-data-refs.c: Likewise.
8868 * tree-vect-generic.c: Likewise.
8869 * tree-vect-loop.c: Likewise.
8870 * tree-vect-slp.c: Likewise.
8871 * tree-vect-stmts.c: Likewise.
8872 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8874 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8875 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8876 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8877 (aarch64_vectorize_vec_perm_const): ...this new function.
8878 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8879 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8880 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8881 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8882 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8883 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8884 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8886 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8887 check for NEON modes.
8888 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8889 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8890 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8891 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8893 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8894 the old VEC_PERM_CONST conditions.
8895 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8896 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8897 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8898 (ia64_vectorize_vec_perm_const_ok): Merge into...
8899 (ia64_vectorize_vec_perm_const): ...this new function.
8900 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8901 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8902 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8903 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8904 * config/mips/mips.c (mips_expand_vec_perm_const)
8905 (mips_vectorize_vec_perm_const_ok): Merge into...
8906 (mips_vectorize_vec_perm_const): ...this new function.
8907 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8908 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8909 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8910 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8911 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8912 (rs6000_expand_vec_perm_const): Delete.
8913 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8915 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8916 (altivec_expand_vec_perm_const_le): Take each operand individually.
8917 Operate on constant selectors rather than rtxes.
8918 (altivec_expand_vec_perm_const): Likewise. Update call to
8919 altivec_expand_vec_perm_const_le.
8920 (rs6000_expand_vec_perm_const): Delete.
8921 (rs6000_vectorize_vec_perm_const_ok): Delete.
8922 (rs6000_vectorize_vec_perm_const): New function.
8923 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8924 an element count and rtx array.
8925 (rs6000_expand_extract_even): Update call accordingly.
8926 (rs6000_expand_interleave): Likewise.
8927 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8928 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8929 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8930 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8931 (rs6000_expand_vec_perm_const): Delete.
8932 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8933 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8934 (altivec_expand_vec_perm_const_le): Take each operand individually.
8935 Operate on constant selectors rather than rtxes.
8936 (altivec_expand_vec_perm_const): Likewise. Update call to
8937 altivec_expand_vec_perm_const_le.
8938 (rs6000_expand_vec_perm_const): Delete.
8939 (rs6000_vectorize_vec_perm_const_ok): Delete.
8940 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8941 reference to the SPE evmerge intructions.
8942 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8943 an element count and rtx array.
8944 (rs6000_expand_extract_even): Update call accordingly.
8945 (rs6000_expand_interleave): Likewise.
8946 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8947 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8949 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8951 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8953 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8954 vector mode and that that mode matches the mode of the data
8956 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8957 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8958 directly using expand_vec_perm_1 when forcing selectors into
8960 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8964 * optabs-query.h (can_vec_perm_p): Delete.
8965 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8966 * optabs-query.c (can_vec_perm_p): Split into...
8967 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8968 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8969 particular selector is valid.
8970 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8971 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8972 (vect_grouped_load_supported): Likewise.
8973 (vect_shift_permute_load_chain): Likewise.
8974 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8975 (vect_transform_slp_perm_load): Likewise.
8976 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8977 (vectorizable_bswap): Likewise.
8978 (vect_gen_perm_mask_checked): Likewise.
8979 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8980 implementations of variable permutation vectors into account
8981 when deciding which selector to use.
8982 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8983 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8984 with a false third argument.
8985 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8986 to test whether the constant selector is valid and can_vec_perm_var_p
8987 to test whether a variable selector is valid.
8989 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8991 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8992 * optabs-query.c (can_vec_perm_p): Likewise.
8993 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8994 instead of vec_perm_indices.
8995 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8996 (vect_gen_perm_mask_checked): Likewise,
8997 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8998 (vect_gen_perm_mask_checked): Likewise,
9000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9002 * optabs-query.h (qimode_for_vec_perm): Declare.
9003 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9004 (qimode_for_vec_perm): ...this new function.
9005 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9007 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9009 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9010 does not have a conditional at the top.
9012 2018-01-02 Richard Biener <rguenther@suse.de>
9014 * ipa-inline.c (big_speedup_p): Fix expression.
9016 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9019 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9022 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9026 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9027 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9028 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9029 cond_taken_branch_cost 3->4.
9031 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9033 PR tree-optimization/83581
9034 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9035 TODO_cleanup_cfg if any changes have been made.
9038 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9039 convert_modes if target mode has the right side, but different mode
9043 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9044 last argument when extracting from CONCAT. If either from_real or
9045 from_imag is NULL, use expansion through memory. If result is not
9046 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9047 the parts directly to inner mode, if even that fails, use expansion
9051 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9052 check for bswap in mode rather than HImode and use that in expand_unop
9055 Copyright (C) 2018 Free Software Foundation, Inc.
9057 Copying and distribution of this file, with or without modification,
9058 are permitted in any medium without royalty provided the copyright
9059 notice and this notice are preserved.