1 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/83723
4 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14 PR rtl-optimization/81443
15 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16 from inner REGs to paradoxical SUBREGs.
18 2018-02-16 Richard Biener <rguenther@suse.de>
20 PR tree-optimization/84399
21 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22 For operands we can analyze at their definition make sure we can
23 analyze them at each use as well.
25 2018-02-16 Richard Biener <rguenther@suse.de>
27 PR tree-optimization/84190
28 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
29 volatile accesses if the decl isn't volatile.
31 2018-02-15 Jason Merrill <jason@redhat.com>
33 PR c++/84314 - ICE with templates and fastcall attribute.
34 * attribs.c (build_type_attribute_qual_variant): Don't clobber
35 TYPE_CANONICAL on an existing type.
37 2018-02-15 Jakub Jelinek <jakub@redhat.com>
39 PR tree-optimization/84383
40 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
41 dstoff nor call operand_equal_p if dstbase is NULL.
43 PR tree-optimization/84334
44 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
45 also a CONSTANT_CLASS_P, punt.
47 2018-02-14 Jim Wilson <jimw@sifive.com>
49 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
50 first SMALL_OPERAND check. New local min_second_step. Move assert
51 to where locals are set. Add TARGET_RVC support.
52 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
54 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
56 * doc/invoke.texi: Correct -Wformat-overflow code sample.
58 2018-02-14 Martin Sebor <msebor@redhat.com>
60 PR tree-optimization/83698
61 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
62 arrays constrain the offset range to their bounds.
63 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
64 (builtin_access::overlap): Avoid setting the size of overlap if it's
66 (maybe_diag_overlap): Also consider arrays when deciding what values
67 of offsets to include in diagnostics.
69 2018-02-14 Martin Sebor <msebor@redhat.com>
72 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
73 that correspond to the kind of a declaration.
75 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
78 * config/pa/pa.md: Load address of PIC label using the linkage table
79 if the label is nonlocal.
81 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
83 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
84 warning message if user requests -maltivec=be.
85 * doc/invoke.texi: Document deprecation of -maltivec=be.
87 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
90 * config/rs6000/rs6000-c.c: Update definitions for
91 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
92 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
94 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
97 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
98 add _get_ssp intrinsics. Remove argument from
99 __builtin_ia32_rdssp[d|q].
100 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
101 * config/i386/i386-builtin.def: Remove argument from
102 __builtin_ia32_rdssp[d|q].
103 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
104 ix86_expand_special_args_builtin for _rdssp[d|q].
105 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
106 Clear register before usage.
107 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
108 Add documentation for new _get_ssp and _inc_ssp intrinsics.
110 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
112 PR tree-optimization/84357
113 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
114 operand 1 of an ARRAY_REF too.
116 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
119 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
120 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
122 (set_of_reg): New struct.
123 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
124 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
125 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
127 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
128 Split into bitclr, bitset, bitinvert patterns if appropriate.
129 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
130 use rx_fuse_in_memory_bitop.
131 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
132 to named insn, correct maximum insn length.
134 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
137 * machmode.def: Define a complex mode for PARTIAL_INT.
138 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
140 * doc/rtl.texi: Document CSPImode.
141 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
143 (msp430_hard_regno_nregs_with_padding): Likewise.
145 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
148 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
150 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
152 PR rtl-optimization/84169
153 * combine.c (try_combine): New variable split_i2i3. Set it to true if
154 we generated a parallel as new i3 and we split that to new i2 and i3
155 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
156 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
157 those to i2, not i1. Partially rewrite this scan code.
159 2018-02-13 Jakub Jelinek <jakub@redhat.com>
162 * stor-layout.c (place_field): For variable length fields, adjust
163 offset_align afterwards not just based on the field's alignment,
164 but also on the size.
167 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
168 of exps and logs in the use_exp2 case.
170 2018-02-13 Jeff Law <law@redhat.com>
172 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
175 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
178 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
182 * common.opt (gas-loc-support, gas-locview-support): New.
183 (ginline-points, ginternal-reset-location-views): New.
184 * doc/invoke.texi: Document them. Use @itemx where intended.
185 (gvariable-location-views): Adjust.
186 * target.def (reset_location_view): New.
187 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
188 (TARGET_RESET_LOCATION_VIEW): New.
189 * doc/tm.texi: Rebuilt.
190 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
191 (dwarf2out_default_as_locview_support): New.
192 (output_asm_line_debug_info): Use option variables.
193 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
194 (output_loc_list): Likewise.
195 (add_high_low_attributes): Check option variables.
196 Don't output entry view attribute in strict mode.
197 (gen_inlined_subroutine_die): Check option variables.
198 (dwarf2out_inline_entry): Likewise.
199 (init_sections_and_labels): Likewise.
200 (dwarf2out_early_finish): Likewise.
201 (maybe_reset_location_view): New, from...
202 (dwarf2out_var_location): ... here. Call it.
203 * debug.h (dwarf2out_default_as_loc_support): Declare.
204 (dwarf2out_default_as_locview_support): Declare.
205 * hooks.c (hook_int_rtx_insn_0): New.
206 * hooks.h (hook_int_rtx_insn_0): Declare.
207 * toplev.c (process_options): Take -gas-loc-support and
208 -gas-locview-support from dwarf2out. Enable
209 -gvariable-location-views by default only with locview
210 assembler support. Enable -ginternal-reset-location-views by
211 default only if the target defines the corresponding hook.
212 Enable -ginline-points by default if location views are
213 enabled; force it disabled if statement frontiers are
215 * tree-inline.c (expand_call_inline): Check option variables.
216 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
218 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
220 PR tree-optimization/84321
221 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
222 handling. Also check whether the anti-range contains any values
223 that satisfy the mask; switch to a VR_RANGE if not.
225 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
228 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
230 2018-02-13 Martin Jambor <mjambor@suse.cz>
233 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
234 of call statements, also set location of a load to a temporary.
236 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
238 * config/rl78/rl78.c (add_vector_labels): New function.
239 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
240 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
241 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
242 which checks that no arguments are passed.
243 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
244 * doc/extend.texi: Documentation for the new attribute.
246 2018-02-13 Andreas Schwab <schwab@suse.de>
248 * config/riscv/linux.h (CPP_SPEC): Define.
250 2018-02-13 Jakub Jelinek <jakub@redhat.com>
253 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
254 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
255 OPTION_MASK_ISA_AES as first argument to def_builtin_const
256 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
257 instead of OPTION_MASK_ISA_PCLMUL as first argument to
258 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
259 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
260 temporarily for AES and PCLMUL builtins.
262 PR tree-optimization/84339
263 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
264 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
268 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
269 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
270 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
272 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
274 * omp-simd-clone.h: New file.
275 * omp-simd-clone.c: Include omp-simd-clone.h.
276 (expand_simd_clones): No longer static.
277 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
278 cgraph.h and omp-simd-clone.h.
279 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
280 (vect_recog_widen_shift_pattern): Formatting fix.
281 (vect_pattern_recog_1): Don't check optab for calls.
284 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
285 operands[2] into a REG before using gen_lowpart on it.
287 2018-02-12 Jeff Law <law@redhat.com>
290 * config/sh/sh.c (find_barrier): Consider a sibling call
293 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
294 successfully back substituting a reg.
296 2018-02-12 Richard Biener <rguenther@suse.de>
298 PR tree-optimization/84037
299 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
300 parameter, move visited init to caller.
301 (vect_slp_analyze_operations): Separate cost from validity
302 check, initialize visited once for all instances.
303 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
305 * tree-vect-stmts.c (vect_model_simple_cost): Make early
307 (vect_model_promotion_demotion_cost): Likewise.
308 (vectorizable_bswap): Guard cost modeling with !slp_node
309 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
311 (vectorizable_call): Likewise.
312 (vectorizable_conversion): Likewise.
313 (vectorizable_assignment): Likewise.
314 (vectorizable_shift): Likewise.
315 (vectorizable_operation): Likewise.
316 (vectorizable_store): Likewise.
317 (vectorizable_load): Likewise.
318 (vectorizable_condition): Likewise.
319 (vectorizable_comparison): Likewise.
321 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
324 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
325 (ASAN_MARK): Fix fnspec to account for return value, change pointer
326 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
328 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
331 * params.def (inline-min-speedup): Increase from 8 to 15.
332 (max-inline-insns-auto): Decrease from 40 to 30.
333 * ipa-split.c (consider_split): Add some buffer for function to
334 be considered inlining candidate.
335 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
338 2018-02-12 Richard Biener <rguenther@suse.de>
340 PR tree-optimization/84037
341 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
342 matched stmts if we cannot swap the non-matched ones.
344 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
346 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
347 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
348 _mm_maskz_scalef_round_ss): New intrinsics.
349 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
350 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
351 __builtin_ia32_scalefss_round): Remove.
352 (__builtin_ia32_scalefsd_mask_round,
353 __builtin_ia32_scalefss_mask_round): New intrinsics.
354 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
355 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
356 ((match_operand:VF_128 2 "<round_nimm_predicate>"
357 "<round_constraint>")): Changed to ...
358 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
359 "<round_scalar_constraint>")): ... this.
360 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
361 %0, %1, %2<round_op3>}"): Changed to ...
362 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
363 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
364 %2<round_scalar_mask_op3>}"): ... this.
365 * config/i386/subst.md (round_scalar_nimm_predicate): New.
367 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
369 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
370 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
371 (_mm_maskz_sqrt_round_ss): New intrinsics.
372 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
373 (__builtin_ia32_sqrtsd_mask_round)
374 (__builtin_ia32_sqrtss_mask_round): New builtins.
375 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
376 (__builtin_ia32_sqrtss_round): Remove.
377 (__builtin_ia32_sqrtsd_mask_round)
378 (__builtin_ia32_sqrtss_mask_round): New builtins.
379 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
380 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
381 ((match_operand:VF_128 1 "vector_operand"
382 "xBm,<round_constraint>")): Changed to ...
383 ((match_operand:VF_128 1 "vector_operand"
384 "xBm,<round_scalar_constraint>")): ... this.
385 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
386 %0, %2, %<iptr>1<round_op3>}): Changed to ...
387 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
388 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
389 %<iptr>1<round_scalar_mask_op3>}): ... this.
390 ((set_attr "prefix" "<round_prefix>")): Changed to ...
391 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
393 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
396 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
397 Cast vec_cmpeq result to correct type.
398 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
399 Cast vec_cmpgt result to correct type.
401 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
403 * final.c (final_scan_insn_1): Renamed from...
404 (final_scan_insn): ... this. New wrapper, to recover
405 seen from the outermost call in recursive ones.
406 * config/sparc/sparc.c (output_return): Drop seen from call.
407 (output_sibcall): Likewise.
408 * config/visium/visium.c (output_branch): Likewise.
410 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
412 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
415 2018-02-10 Alan Modra <amodra@gmail.com>
418 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
419 Specify LR as an input.
421 2018-02-10 Jakub Jelinek <jakub@redhat.com>
424 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
425 remove_member_access_dummy_vars): New functions.
426 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
427 lower_omp_1, execute_lower_omp): Use them.
429 PR rtl-optimization/84308
430 * shrink-wrap.c (spread_components): Release todo vector.
432 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
434 PR rtl-optimization/57193
435 * ira-color.c (struct allocno_color_data): Add member
436 conflict_allocno_hard_prefs.
437 (update_conflict_allocno_hard_prefs): New.
438 (bucket_allocno_compare_func): Add a preference based on
439 conflict_allocno_hard_prefs.
440 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
441 (color_allocnos): Remove a dead code. Initiate
442 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
444 2018-02-09 Jakub Jelinek <jakub@redhat.com>
447 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
448 constraint from =wa to wa. Avoid a subreg on the output operand,
449 instead use a pseudo and subreg it in a move.
450 (p9_xxbrd_<mode>): Changed to ...
451 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
452 (p9_xxbrd_v2df): New expander.
453 (p9_xxbrw_<mode>): Changed to ...
454 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
455 (p9_xxbrw_v4sf): New expander.
457 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
459 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
461 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
464 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
465 multiply in 32-bit mode.
466 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
467 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
470 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
472 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
473 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
474 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
475 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
477 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
480 * dwarf2out.c (is_trivial_indirect_ref): New function.
481 (dwarf2out_late_global_decl): Do not generate a location
482 attribute for variables that have a non-trivial DECL_VALUE_EXPR
483 and that are not defined in the current unit.
485 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
487 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
488 instead of a libcall for UNORDERED.
490 2018-02-09 Tamar Christina <tamar.christina@arm.com>
493 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
494 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
496 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
499 * config/s390/s390.c (s390_set_current_function): Invoke
500 s390_indirect_branch_settings also if fndecl didn't change.
502 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
504 * config/rs6000/rs6000.md (blockage): Set length to zero.
506 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
508 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
510 2018-02-09 Jakub Jelinek <jakub@redhat.com>
513 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
514 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
518 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
519 PARALLEL incoming that failed vt_get_decl_and_offset check.
522 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
523 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
525 (get_variable_section): For decls in named .bss* sections pass true as
526 second argument to bss_initializer_p.
528 2018-02-09 Marek Polacek <polacek@redhat.com>
529 Jakub Jelinek <jakub@redhat.com>
532 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
533 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
534 Sync some changes from cxx_fold_indirect_ref.
536 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
538 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
540 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
541 (BLOCK_INLINE_ENTRY_LABEL): New.
542 (dwarf2out_var_location): Disregard inline entry markers.
543 (inline_entry_data): New struct.
544 (inline_entry_data_hasher): New hashtable type.
545 (inline_entry_data_hasher::hash): New.
546 (inline_entry_data_hasher::equal): New.
547 (inline_entry_data_table): New variable.
548 (add_high_low_attributes): Add DW_AT_entry_pc and
549 DW_AT_GNU_entry_view attributes if a pending entry is found
550 in inline_entry_data_table. Add old entry_pc attribute only
551 if debug nonbinding markers are disabled.
552 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
554 (block_within_block_p, dwarf2out_inline_entry): New.
555 (dwarf2out_finish): Check that no entries remained in
556 inline_entry_data_table.
557 * final.c (reemit_insn_block_notes): Handle inline entry notes.
558 (final_scan_insn, notice_source_line): Likewise.
559 (rest_of_clean_state): Skip inline entry markers.
560 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
562 * gimple.c (gimple_build_debug_inline_entry): New.
563 * gimple.h (enum gimple_debug_subcode): Add
564 GIMPLE_DEBUG_INLINE_ENTRY.
565 (gimple_build_debug_inline_entry): Declare.
566 (gimple_debug_inline_entry_p): New.
567 (gimple_debug_nonbind_marker_p): Adjust.
568 * insn-notes.def (INLINE_ENTRY): New.
569 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
570 inline entry marker notes.
571 (print_insn): Likewise.
572 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
573 (INSN_DEBUG_MARKER_KIND): Likewise.
574 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
575 * tree-inline.c (expand_call_inline): Build and insert
576 debug_inline_entry stmt.
577 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
578 inline entry blocks early, if nonbind markers are enabled.
579 (dump_scope_block): Dump fragment info.
580 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
581 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
582 (gimple_build_debug_inline_entry): New.
583 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
584 Enable/disable inline entry points too.
585 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
586 (DEBUG_INSN): Describe inline entry markers.
588 * common.opt (gvariable-location-views): New.
589 (gvariable-location-views=incompat5): New.
590 * config.in: Rebuilt.
591 * configure: Rebuilt.
592 * configure.ac: Test assembler for view support.
593 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
594 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
595 * dwarf2out.c (var_loc_view): New typedef.
596 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
597 (dwarf2out_locviews_in_attribute): New.
598 (dwarf2out_locviews_in_loclist): New.
599 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
600 (enum dw_line_info_opcode): Add LI_adv_address.
601 (struct dw_line_info_table): Add view.
602 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
603 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
604 (zero_view_p): New variable.
605 (ZERO_VIEW_P): New macro.
606 (output_asm_line_debug_info): New.
607 (struct var_loc_node): Add view.
608 (add_AT_view_list, AT_loc_list): New.
609 (add_var_loc_to_decl): Add view param. Test it against last.
610 (new_loc_list): Add view params. Record them.
611 (AT_loc_list_ptr): Handle loc and view lists.
612 (view_list_to_loc_list_val_node): New.
613 (print_dw_val): Handle dw_val_class_view_list.
614 (size_of_die): Likewise.
615 (value_format): Likewise.
616 (loc_list_has_views): New.
617 (gen_llsym): Set vl_symbol too.
618 (maybe_gen_llsym, skip_loc_list_entry): New.
619 (dwarf2out_maybe_output_loclist_view_pair): New.
620 (output_loc_list): Output view list or entries too.
621 (output_view_list_offset): New.
622 (output_die): Handle dw_val_class_view_list.
623 (output_dwarf_version): New.
624 (output_compilation_unit_header): Use it.
625 (output_skeleton_debug_sections): Likewise.
626 (output_rnglists, output_line_info): Likewise.
627 (output_pubnames, output_aranges): Update version comments.
628 (output_one_line_info_table): Output view numbers in asm comments.
629 (dw_loc_list): Determine current endview, pass it to new_loc_list.
630 Call maybe_gen_llsym.
631 (loc_list_from_tree_1): Adjust.
632 (add_AT_location_description): Create view list attribute if
633 needed, check it's absent otherwise.
634 (convert_cfa_to_fb_loc_list): Adjust.
635 (maybe_emit_file): Call output_asm_line_debug_info for test.
636 (dwarf2out_var_location): Reset views as needed. Precompute
637 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
639 (new_line_info_table): Reset next view.
640 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
641 (dwarf2out_source_line): Likewise. Output view resets and labels to
642 the assembler, or select appropriate line info opcodes.
643 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
644 (optimize_string_length): Catch it. Adjust.
645 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
646 dw_val_class_view_list, and remove it if no longer needed.
647 (hash_loc_list): Hash view numbers.
648 (loc_list_hasher::equal): Compare them.
649 (optimize_location_lists): Check whether a view list symbol is
650 needed, and whether the locview attribute is present, and
651 whether they match. Remove the locview attribute if no longer
653 (index_location_lists): Call skip_loc_list_entry for test.
654 (dwarf2out_finish): Call output_asm_line_debug_info for test.
655 Use output_dwarf_version.
656 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
657 (struct dw_val_node): Add val_view_list.
658 * final.c (SEEN_NEXT_VIEW): New.
659 (set_next_view_needed): New.
660 (clear_next_view_needed): New.
661 (maybe_output_next_view): New.
662 (final_start_function): Rename to...
663 (final_start_function_1): ... this. Take pointer to FIRST,
664 add SEEN parameter. Emit param bindings in the initial view.
665 (final_start_function): Reintroduce SEEN-less interface.
666 (final): Rename to...
667 (final_1): ... this. Take SEEN parameter. Output final pending
668 next view at the end.
669 (final): Reintroduce seen-less interface.
670 (final_scan_insn): Output pending next view before switching
671 sections or ending a block. Mark the next view as needed when
672 outputting variable locations. Notify debug backend of section
673 changes, and of location view changes.
674 (rest_of_handle_final): Adjust.
675 * toplev.c (process_options): Autodetect value for debug variable
676 location views option. Warn on incompat5 without -gdwarf-5.
677 * doc/invoke.texi (gvariable-location-views): New.
678 (gvariable-location-views=incompat5): New.
679 (gno-variable-location-views): New.
681 2018-02-08 David Malcolm <dmalcolm@redhat.com>
683 PR tree-optimization/84136
684 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
685 that the result of find_edge is non-NULL.
687 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
690 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
691 storing integer register in SImode. Fix cost of 256 and 512
692 byte aligned SSE register store.
694 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
696 * config/i386/i386.c (ix86_multiplication_cost): Fix
697 multiplication cost for TARGET_AVX512DQ.
699 2018-02-08 Marek Polacek <polacek@redhat.com>
701 PR tree-optimization/84238
702 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
705 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
707 PR tree-optimization/84265
708 * tree-vect-stmts.c (vectorizable_store): Don't treat
709 VMAT_CONTIGUOUS accesses as grouped.
710 (vectorizable_load): Likewise.
712 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
714 PR tree-optimization/81635
715 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
716 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
717 (test_round_for_mask): New functions.
718 (wide_int_cc_tests): Call test_round_for_mask.
719 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
720 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
721 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
722 range returned by get_range_info.
724 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
727 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
728 * symtab.c: Include builtins.h
729 (symtab_node::output_to_lto_symbol_table_p): Move here
730 from lto-streamer-out.c:output_symbol_p.
731 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
732 (output_symbol_p): Move all logic to symtab.c
733 (produce_symtab): Update.
735 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
737 * config/s390/s390-opts.h (enum indirect_branch): Define.
738 * config/s390/s390-protos.h (s390_return_addr_from_memory)
739 (s390_indirect_branch_via_thunk)
740 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
741 (enum s390_indirect_branch_type): Define.
742 * config/s390/s390.c (struct s390_frame_layout, struct
743 machine_function): Remove.
744 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
745 (indirect_branch_table_label_no, indirect_branch_table_name):
747 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
748 (enum s390_indirect_branch_option): Define.
749 (s390_return_addr_from_memory): New function.
750 (s390_handle_string_attribute): New function.
751 (s390_attribute_table): Add new attribute handler.
752 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
753 (s390_indirect_branch_via_thunk): New function.
754 (s390_indirect_branch_via_inline_thunk): New function.
755 (s390_function_ok_for_sibcall): When jumping via thunk disallow
756 sibling call optimization for non z10 compiles.
757 (s390_emit_call): Force indirect branch target to be a single
758 register. Add r1 clobber for non-z10 compiles.
759 (s390_emit_epilogue): Emit return jump via return_use expander.
760 (s390_reorg): Handle JUMP_INSNs as execute targets.
761 (s390_option_override_internal): Perform validity checks for the
762 new command line options.
763 (s390_indirect_branch_attrvalue): New function.
764 (s390_indirect_branch_settings): New function.
765 (s390_set_current_function): Invoke s390_indirect_branch_settings.
766 (s390_output_indirect_thunk_function): New function.
767 (s390_code_end): Implement target hook.
768 (s390_case_values_threshold): Implement target hook.
769 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
771 * config/s390/s390.h (struct s390_frame_layout)
772 (struct machine_function): Move here from s390.c.
773 (TARGET_INDIRECT_BRANCH_NOBP_RET)
774 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
775 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
776 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
777 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
778 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
779 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
780 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
781 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
782 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
783 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
784 (mnemonic attribute): Add values which aren't recognized
786 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
787 pattern for branch conversion. Fix mnemonic attribute.
788 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
789 indirect branch via thunk if requested.
790 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
791 ("*indirect_jump"): Disable for branch conversion using out of
793 ("indirect_jump_via_thunk<mode>_z10")
794 ("indirect_jump_via_thunk<mode>")
795 ("indirect_jump_via_inlinethunk<mode>_z10")
796 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
797 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
798 ("casesi_jump_via_inlinethunk<mode>_z10")
799 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
800 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
801 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
802 ("*indirect2_jump"): Disable for branch conversion.
803 ("casesi_jump"): Turn into expander and expand patterns for branch
805 ("return_use"): New expander.
806 ("*return"): Emit return via thunk and rename it to ...
807 ("*return<mode>"): ... this one.
808 * config/s390/s390.opt: Add new options and and enum for the
811 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
813 * lra-constraints.c (match_reload): Unconditionally use
814 gen_lowpart_SUBREG, rather than selecting between that
815 and equivalent gen_rtx_SUBREG code.
817 2018-02-08 Richard Biener <rguenther@suse.de>
819 PR tree-optimization/84233
820 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
821 changed flag instead of boguously re-using phi_inserted.
823 2018-02-08 Martin Jambor <mjambor@suse.cz>
825 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
826 static local variables.
828 2018-02-08 Richard Biener <rguenther@suse.de>
830 PR tree-optimization/84278
831 * tree-vect-stmts.c (vectorizable_store): When looking for
832 smaller vector types to perform grouped strided loads/stores
833 make sure the mode is supported by the target.
834 (vectorizable_load): Likewise.
836 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
838 * config/aarch64/aarch64.c (aarch64_components_for_bb):
839 Increase LDP/STP opportunities by adding adjacent callee-saves.
841 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
843 PR rtl-optimization/84068
844 PR rtl-optimization/83459
845 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
847 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
849 PR tree-optimization/84224
850 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
851 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
854 2018-02-07 Iain Sandoe <iain@codesourcery.com>
857 * config/rs6000/altivec.md (*restore_world): Remove LR use.
858 * config/rs6000/predicates.md (restore_world_operation): Adjust op
859 count, remove one USE.
861 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
863 * doc/install.texi (Configuration): Document the
864 --with-long-double-format={ibm,ieee} PowerPC configuration
868 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
869 Convert from define_expand to be define_insn_and_split. Rework
870 float/double/_Float128 conversions to QI/HI/SImode to work with
871 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
872 conversions to QI/HImode types did a store and then a load to
873 truncate the value. For conversions to VSX registers, don't split
874 the insn, instead emit the code directly. Use the code iterator
875 any_fix to combine signed and unsigned conversions.
876 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
877 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
878 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
879 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
880 (fix_<mode>di2_hw): Likewise.
881 (fixuns_<mode>di2_hw): Likewise.
882 (fix_<mode>si2_hw): Likewise.
883 (fixuns_<mode>si2_hw): Likewise.
884 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
885 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
886 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
887 fix<uns>_trunc<SFDF:mode>si2_p8.
888 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
890 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
891 (fix<uns>_<mode>_mem): Likewise.
892 (fctiw<u>z_<mode>_mem): Likewise.
893 (fix<uns>_<mode>_mem): Likewise.
894 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
895 the register allocator from doing a direct move to the GPRs to do
896 a store, and instead use the ISA 3.0 store byte/half-word from
897 vector register instruction. For IEEE 128-bit floating point,
898 also optimize stores of 32-bit ints.
899 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
901 2018-02-07 Alan Hayward <alan.hayward@arm.com>
903 * genextract.c (push_pathstr_operand): New function to support
905 (walk_rtx): Call push_pathstr_operand.
906 (print_path): Support [a-zA-Z].
908 2018-02-07 Richard Biener <rguenther@suse.de>
910 PR tree-optimization/84037
911 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
912 (cse_and_gimplify_to_preheader): Declare.
913 (vect_get_place_in_interleaving_chain): Likewise.
914 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
916 (_loop_vec_info::~_loop_vec_info): Delete it.
917 (cse_and_gimplify_to_preheader): New function.
918 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
919 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
920 (vectorizable_load): Likewise. For grouped stores always base
921 the IV on the first element.
922 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
923 condition before gimplifying.
925 2018-02-07 Jakub Jelinek <jakub@redhat.com>
927 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
928 *DIV_EXPR and *MOD_EXPR.
930 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
933 * config/i386/i386.c (ix86_option_override_internal): Mask out
934 the CF_SET bit when checking -fcf-protection.
936 2018-02-07 Tom de Vries <tom@codesourcery.com>
939 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
942 2018-02-07 Richard Biener <rguenther@suse.de>
944 PR tree-optimization/84204
945 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
948 PR tree-optimization/84205
949 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
950 special-case isl_ast_op_zdiv_r.
952 PR tree-optimization/84223
953 * graphite-scop-detection.c (gather_bbs::before_dom_children):
954 Only add conditions from within the region.
955 (gather_bbs::after_dom_children): Adjust.
957 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
960 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
961 * config/avr/avr.md: Only post-reload split REG-REG moves if
962 either register is REGERAL_REG_P.
964 2018-02-07 Jakub Jelinek <jakub@redhat.com>
966 PR tree-optimization/84235
967 * tree-ssa-scopedtables.c
968 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
969 if the subtraction is performed in floating point type where NaNs are
970 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
971 build 1. Formatting fix.
973 2018-02-06 Jakub Jelinek <jakub@redhat.com>
976 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
977 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
978 and skip it regardless of bb boundaries. Use CALL_P macro,
979 don't test INSN_P (insn) together with CALL_P or JUMP_P check
980 unnecessarily, formatting fix.
982 2018-02-06 Michael Collison <michael.collison@arm.com>
984 * config/arm/thumb2.md:
985 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
986 (*thumb_mov_notscc): Ditto.
988 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
991 * config/rs6000/rs6000.md (su code attribute): Use "u" for
992 unsigned_fix, not "s".
994 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
996 * configure.ac (gcc_fn_eh_frame_ro): New function.
997 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
998 correct .eh_frame permissions.
999 * configure: Regenerate.
1001 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1003 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1006 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1008 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1009 Display warning message for -mno-speculate-indirect-jumps.
1011 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1013 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1015 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1017 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1019 PR tree-optimization/84225
1020 * tree-eh.c (find_trapping_overflow): Only call
1021 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1023 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1026 * config/i386/i386.c: Reimplement the check of possible options
1027 -mibt/-mshstk conbination. Change error messages.
1028 * doc/invoke.texi: Fix a typo: remove extra '='.
1030 2018-02-06 Marek Polacek <polacek@redhat.com>
1032 PR tree-optimization/84228
1033 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1035 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1038 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1039 emitted arch directives.
1040 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1041 __ARM_FEATURE_COPROC before changing architectures.
1043 2018-02-06 Richard Biener <rguenther@suse.de>
1045 * config/i386/i386.c (print_reg): Fix typo.
1046 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1048 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1050 * configure: Regenerate.
1052 2018-02-05 Martin Sebor <msebor@redhat.com>
1054 PR tree-optimization/83369
1055 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1058 2018-02-05 Martin Liska <mliska@suse.cz>
1060 * doc/invoke.texi: Cherry-pick upstream r323995.
1062 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1064 * ira.c (ira_init_register_move_cost): Adjust comment.
1066 2018-02-05 Martin Liska <mliska@suse.cz>
1068 PR gcov-profile/84137
1069 * doc/gcov.texi: Fix typo in documentation.
1071 2018-02-05 Martin Liska <mliska@suse.cz>
1073 PR gcov-profile/83879
1074 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1075 using dlopen functionality.
1077 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1079 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1080 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1081 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1082 _mm_maskz_range_round_ss): New intrinsics.
1083 (__builtin_ia32_rangesd128_round)
1084 (__builtin_ia32_rangess128_round): Remove.
1085 (__builtin_ia32_rangesd128_mask_round,
1086 __builtin_ia32_rangess128_mask_round): New builtins.
1087 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1088 __builtin_ia32_rangess128_round): Remove.
1089 (__builtin_ia32_rangesd128_mask_round,
1090 __builtin_ia32_rangess128_mask_round): New builtins.
1091 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1092 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1093 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1094 "<round_saeonly_constraint>")): Changed to ...
1095 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1096 "<round_saeonly_scalar_constraint>")): ... this.
1097 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1098 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1099 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1100 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1101 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1103 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1105 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1107 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1108 Remove all values except native, 8540 and 8548.
1110 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1112 * config/i386/i386.c (ix86_output_function_return): Pass
1113 INVALID_REGNUM, instead of -1, as invalid register number to
1114 indirect_thunk_name and output_indirect_thunk.
1116 2018-02-02 Julia Koval <julia.koval@intel.com>
1118 * config.gcc: Add -march=icelake.
1119 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1120 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1121 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1122 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1123 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1124 (processor_target_table): Add icelake.
1125 (ix86_option_override_internal): Handle new PTAs.
1126 (get_builtin_code_for_version): Handle icelake.
1127 (M_INTEL_COREI7_ICELAKE): New.
1128 (fold_builtin_cpu): Handle icelake.
1129 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1130 * doc/invoke.texi: Add -march=icelake.
1132 2018-02-02 Julia Koval <julia.koval@intel.com>
1134 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1135 to wide_int_bitmask.
1136 * wide-int-bitmask.h: New.
1138 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1141 * config/i386/i386.md: Replace Pmode with word_mode in
1142 builtin_setjmp_setup and builtin_longjmp to support x32.
1144 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1148 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1150 (rs6000_supported_cpu_names): New static variable.
1151 (linux_cpu_translation_table): Likewise.
1152 (elf_platform) <cpu>: Define new static variable and use it.
1153 Translate kernel AT_PLATFORM name to canonical name if needed.
1154 Error if platform name is unknown.
1156 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1159 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1161 2018-02-01 Jeff Law <law@redhat.com>
1164 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1165 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1166 the scratch if RELEASE_VIA_POP is false.
1167 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1168 If we have to save a temporary register, decrement SIZE appropriately.
1169 Pass new arguments to release_scratch_register_on_entry.
1170 (ix86_adjust_stack_and_probe): Likewise.
1171 (ix86_emit_probe_stack_range): Pass new arguments to
1172 release_scratch_register_on_entry.
1174 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1176 PR rtl-optimization/84157
1177 * combine.c (change_zero_ext): Use REG_P predicate in
1178 front of HARD_REGISTER_P predicate.
1180 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1182 * config/avr/avr.c (avr_option_override): Move disabling of
1183 -fdelete-null-pointer-checks to...
1184 * common/config/avr/avr-common.c (avr_option_optimization_table):
1187 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1189 PR tree-optimization/81635
1190 * tree-data-ref.c (split_constant_offset_1): For types that
1191 wrap on overflow, try to use range info to prove that wrapping
1194 2018-02-01 Renlin Li <renlin.li@arm.com>
1197 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1199 (aarch64_register_move_cost): Likewise.
1200 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1202 (REG_CLASS_NAMES): Likewise.
1203 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1204 TAILCALL_ADDR_REGS. Remove IP registers.
1205 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1207 2018-02-01 Richard Biener <rguenther@suse.de>
1209 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1210 for specifying RPO order and allow NULL for that.
1211 * domwalk.c (dom_walker::dom_walker): Likewise.
1212 (dom_walker::walk): Handle NULL RPO order.
1213 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1215 (rewrite_update_dom_walker): Likewise.
1216 (mark_def_dom_walker): Likewise.
1218 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1220 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1221 (aarch64_maybe_expand_sve_subreg_move): Declare.
1222 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1223 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1225 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1226 that are semantically a reverse operation.
1227 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1228 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1229 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1231 (aarch64_can_change_mode_class): For big-endian, forbid changes
1232 between two SVE modes if they have different element sizes.
1234 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1236 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1237 the TImode handling for big-endian targets.
1239 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1241 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1242 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1244 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1245 Remove BSWAP handing for big-endian targets and use the form of
1246 LD1RQ appropariate for the mode.
1248 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1250 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1251 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1254 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1257 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1258 check for operands that need to go through aarch64_sve_reload_be.
1260 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1262 PR tree-optimization/81661
1263 PR tree-optimization/84117
1264 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1265 * tree-eh.c: Include gimplify.h.
1266 (find_trapping_overflow, replace_trapping_overflow,
1267 rewrite_to_non_trapping_overflow): New functions.
1268 * tree-vect-loop.c: Include tree-eh.h.
1269 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1270 * tree-data-ref.c: Include tree-eh.h.
1271 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1273 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1275 PR rtl-optimization/84123
1276 * combine.c (change_zero_ext): Check if hard register satisfies
1277 can_change_dest_mode before calling gen_lowpart_SUBREG.
1279 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
1282 * ira.c (ira_init_register_move_cost): Remove assert.
1284 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1286 PR rtl-optimization/84071
1287 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
1288 * doc/tm.texi: Regenerate.
1290 2018-01-31 Richard Biener <rguenther@suse.de>
1292 PR tree-optimization/84132
1293 * tree-data-ref.c (analyze_miv_subscript): Properly
1294 check whether evolution_function_is_affine_multivariate_p
1295 before calling gcd_of_steps_may_divide_p.
1297 2018-01-31 Julia Koval <julia.koval@intel.com>
1300 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
1301 * config/i386/i386.md (rdpid_rex64) New.
1302 (rdpid): Make 32bit only.
1304 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
1307 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
1308 an IDENTIFIER_NODE for FUNCTION_TYPE's.
1310 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1313 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
1315 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
1317 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1319 PR rtl-optimization/84071
1320 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
1321 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
1323 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1325 * config/arc/arc.c (arc_handle_aux_attribute): New function.
1326 (arc_attribute_table): Add 'aux' attribute.
1327 (arc_in_small_data_p): Consider aux like variables.
1328 (arc_is_aux_reg_p): New function.
1329 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
1330 (arc_get_aux_arg): New function.
1331 (prepare_move_operands): Handle aux-register access.
1332 (arc_handle_aux_attribute): New function.
1333 * doc/extend.texi (ARC Variable attributes): Add subsection.
1335 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1337 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
1338 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
1339 (arc_attribute_table): Add 'uncached' attribute.
1340 (arc_print_operand): Print '.di' flag for uncached memory
1342 (arc_in_small_data_p): Do not consider for small data the uncached
1344 (arc_is_uncached_mem_p): New function.
1345 * config/arc/predicates.md (compact_store_memory_operand): Check
1346 for uncached memory accesses.
1347 (nonvol_nonimm_operand): Likewise.
1348 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
1350 2018-01-31 Jakub Jelinek <jakub@redhat.com>
1353 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
1354 falign-loops=): Add Optimization flag.
1356 2018-01-30 Jeff Law <law@redhat.com>
1359 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
1360 INT_REGISTERS_SAVED. Check it prior to calling
1361 get_scratch_register_on_entry.
1362 (ix86_adjust_stack_and_probe): Similarly.
1363 (ix86_emit_probe_stack_range): Similarly.
1364 (ix86_expand_prologue): Corresponding changes.
1366 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1369 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
1370 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
1372 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
1375 * lra-constraints.c (curr_insn_transform): Process AND in the
1378 2018-01-30 Jakub Jelinek <jakub@redhat.com>
1380 PR rtl-optimization/83986
1381 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
1382 dependence against last_pending_memory_flush in addition to
1385 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
1387 PR tree-optimization/81611
1388 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
1391 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1394 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
1397 2018-01-30 Richard Biener <rguenther@suse.de>
1398 Jakub Jelinek <jakub@redhat.com>
1400 PR tree-optimization/84111
1401 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
1402 inner loops added during recursion, as they don't have up-to-date
1405 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1408 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
1409 (can_inline_edge_by_limits_p): ... here.
1410 (can_early_inline_edge_p, check_callers,
1411 update_caller_keys, update_callee_keys, recursive_inlining,
1412 add_new_edges_to_heap, speculation_useful_p,
1413 inline_small_functions,
1414 inline_small_functions, flatten_function,
1415 inline_to_all_callers_1): Update.
1417 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1419 * profile-count.c (profile_count::combine_with_ipa_count): Handle
1422 2018-01-30 Richard Biener <rguenther@suse.de>
1424 PR tree-optimization/83008
1425 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
1426 invariant and constant vector uses in stmts when they need
1429 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1432 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
1433 * configure: Regenerate.
1435 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1437 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
1439 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
1440 Use gen_rtx_REG rather than gen_lowpart.
1442 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1444 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
1445 rather than 0 when creating partial subregs.
1447 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1449 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
1452 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
1455 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
1456 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
1457 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
1458 flags. This restores the settings used before the 2017-07-24.
1459 Turning off pre increment/decrement/modify allows IVOPTS to
1460 optimize DF/SF loops where the index is an int.
1462 2018-01-29 Richard Biener <rguenther@suse.de>
1463 Kelvin Nilsen <kelvin@gcc.gnu.org>
1466 * tree-vect-stmts.c (vectorizable_call): Don't call
1467 targetm.vectorize_builtin_md_vectorized_function if callee is
1470 2018-01-22 Carl Love <cel@us.ibm.com>
1472 * doc/extend.tex: Fix typo in second arg in
1473 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
1475 2018-01-29 Richard Biener <rguenther@suse.de>
1477 PR tree-optimization/84086
1478 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
1479 (flush_ssaname_freelist): When SSA names were released reset
1480 the SCEV hash table.
1482 2018-01-29 Richard Biener <rguenther@suse.de>
1484 PR tree-optimization/84057
1485 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
1486 removed paths when removing edges.
1488 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
1490 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
1491 -mfunction-return=@var{choice}.
1493 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
1496 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
1498 (layout::print_source_line): Likewise.
1499 (test_get_line_width_without_trailing_whitespace): Add test cases.
1501 2018-01-27 Jakub Jelinek <jakub@redhat.com>
1504 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
1507 2018-01-26 Jim Wilson <jimw@sifive.com>
1509 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
1511 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
1514 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1516 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
1517 and CMP + SUB-immediate -> SUBS.
1519 2018-01-26 Martin Sebor <msebor@redhat.com>
1521 PR tree-optimization/83896
1522 * tree-ssa-strlen.c (get_string_len): Rename...
1523 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
1524 Avoid assuming length is constant.
1525 (handle_char_store): Use HOST_WIDE_INT for string length.
1527 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
1530 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
1531 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
1533 2018-01-26 Richard Biener <rguenther@suse.de>
1535 PR rtl-optimization/84003
1536 * dse.c (record_store): Only record redundant stores when
1537 the earlier store aliases at least all accesses the later one does.
1539 2018-01-26 Jakub Jelinek <jakub@redhat.com>
1541 PR rtl-optimization/83985
1542 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
1543 REG_CFA_RESTORE insns.
1544 (delete_unmarked_insns): Don't ignore separate shrink wrapping
1545 REG_CFA_RESTORE insns here.
1548 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
1549 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
1551 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1553 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
1554 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
1555 (arc_init): Likewise.
1556 (arc_override_options): Likewise.
1557 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
1559 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
1561 * config/arc/arc.h (TARGET_DBNZ): Define.
1562 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
1563 properly set the tune attribute.
1564 (dbnz): Use TARGET_DBNZ guard.
1565 * config/arc/arc.opt (mtune): Add core3 option.
1567 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1569 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
1570 recognize new pic like addresses.
1571 (arc_delegitimize_address): Clean up.
1573 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1575 * config/arc/arc-arches.def: Option mrf16 valid for all
1577 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
1578 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
1579 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
1580 * config/arc/arc-tables.opt: Regenerate.
1581 * config/arc/arc.c (arc_conditional_register_usage): Handle
1582 reduced register file case.
1583 (arc_file_start): Set must have build attributes.
1584 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
1586 * config/arc/arc.opt (mrf16): Add new option.
1587 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
1588 * config/arc/genmultilib.awk: Handle new mrf16 option.
1589 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
1590 * config/arc/t-multilib: Regenerate.
1591 * doc/invoke.texi (ARC Options): Document mrf16 option.
1593 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1595 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
1596 * config/arc/arc.c (arc_handle_secure_attribute): New function.
1597 (arc_attribute_table): Add 'secure_call' attribute.
1598 (arc_print_operand): Print secure call operand.
1599 (arc_function_ok_for_sibcall): Don't optimize tail calls when
1601 (arc_is_secure_call_p): New function. * config/arc/arc.md
1602 (call_i): Add support for sjli instruction.
1603 (call_value_i): Likewise.
1604 * config/arc/constraints.md (Csc): New constraint.
1606 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
1607 John Eric Martin <John.Martin@emmicro-us.com>
1609 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
1610 * config/arc/arc.c (_arc_jli_section): New struct.
1611 (arc_jli_section): New type.
1612 (rc_jli_sections): New static variable.
1613 (arc_handle_jli_attribute): New function.
1614 (arc_attribute_table): Add jli_always and jli_fixed attribute.
1615 (arc_file_end): New function.
1616 (TARGET_ASM_FILE_END): Define.
1617 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
1618 (arc_add_jli_section): New function.
1619 (jli_call_scan): Likewise.
1620 (arc_reorg): Call jli_call_scan.
1621 (arc_output_addsi): Remove 'S' from printing asm operand.
1622 (arc_is_jli_call_p): New function.
1623 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
1625 (movhi_insn): Likewise.
1626 (movsi_insn): Likewise.
1627 (movsi_set_cc_insn): Likewise.
1628 (loadqi_update): Likewise.
1629 (load_zeroextendqisi_update): Likewise.
1630 (load_signextendqisi_update): Likewise.
1631 (loadhi_update): Likewise.
1632 (load_zeroextendhisi_update): Likewise.
1633 (load_signextendhisi_update): Likewise.
1634 (loadsi_update): Likewise.
1635 (loadsf_update): Likewise.
1636 (movsicc_insn): Likewise.
1637 (bset_insn): Likewise.
1638 (bxor_insn): Likewise.
1639 (bclr_insn): Likewise.
1640 (bmsk_insn): Likewise.
1641 (bicsi3_insn): Likewise.
1642 (cmpsi_cc_c_insn): Likewise.
1643 (movsi_ne): Likewise.
1644 (movsi_cond_exec): Likewise.
1645 (clrsbsi2): Likewise.
1656 (call_i): Remove 'S' asm letter, add jli instruction.
1657 (call_value_i): Likewise.
1658 * config/arc/arc.op (mjli-always): New option.
1659 * config/arc/constraints.md (Cji): New constraint.
1660 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
1662 (subsf3_fpx): Likewise.
1663 (mulsf3_fpx): Likewise.
1664 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
1666 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
1668 * doc/invoke.texi (ARC): Document mjli-always option.
1670 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
1672 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
1673 avoid addition with 0 and use incw and decw where possible.
1675 2018-01-26 Richard Biener <rguenther@suse.de>
1677 PR tree-optimization/81082
1678 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
1679 association if it requires casting to unsigned.
1680 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
1681 from fold_plusminus_mult_expr to catch important cases late when
1682 range info is available.
1684 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1686 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
1687 * configure.ac (hidden_linkonce): New test.
1688 * configure: Regenerate.
1689 * config.in: Regenerate.
1691 2018-01-26 Julia Koval <julia.koval@intel.com>
1693 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
1694 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1695 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
1696 _mm_mask_bitshuffle_epi64_mask): Fix type.
1697 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
1698 USI_FTYPE_V4DI_V4DI_USI): Remove.
1699 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
1700 __builtin_ia32_vpshufbitqmb256_mask,
1701 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
1702 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
1703 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
1705 2018-01-26 Alan Modra <amodra@gmail.com>
1708 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
1709 UNSPEC_VBPERMQ. Sort other unspecs.
1711 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
1713 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
1715 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
1718 * predict.c (drop_profile): Do not push/pop cfun; update also
1720 (handle_missing_profiles): Fix logic looking for zero profiles.
1722 2018-01-25 Jakub Jelinek <jakub@redhat.com>
1725 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
1726 on functions with #pragma omp declare simd or functions with simd
1728 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
1729 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
1730 Remove trailing \n from warning_at calls.
1732 2018-01-25 Tom de Vries <tom@codesourcery.com>
1735 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
1736 for neutered workers.
1738 2018-01-24 Joseph Myers <joseph@codesourcery.com>
1741 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
1742 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
1744 2018-01-24 Jeff Law <law@redhat.com>
1747 * i386.c (get_probe_interval): Move to earlier point.
1748 (ix86_compute_frame_layout): If -fstack-clash-protection and
1749 the frame is larger than the probe interval, then use pushes
1750 to save registers rather than reg->mem moves.
1751 (ix86_expand_prologue): Remove conditional for int_registers_saved
1754 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
1757 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
1758 min/max for never referenced object.
1760 2018-01-24 Jakub Jelinek <jakub@redhat.com>
1763 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
1765 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
1766 attributes from DECL_ATTRIBUTES (decl) without affecting
1767 DECL_ATTRIBUTES (current_function_decl).
1768 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
1769 functions with non-NULL DECL_ABSTRACT_ORIGIN.
1771 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
1773 PR tree-optimization/83979
1774 * fold-const.c (fold_comparison): Use constant_boolean_node
1775 instead of boolean_{true,false}_node.
1777 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
1779 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
1782 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1784 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
1785 Simplify the clause that sets the length attribute.
1786 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
1787 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
1788 clause that sets the length attribute.
1789 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
1791 2018-01-24 Tom de Vries <tom@codesourcery.com>
1794 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
1795 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
1796 Add strict parameter.
1797 (prevent_branch_around_nothing): Insert dummy insn between branch to
1798 label and label with no ptx insn inbetween.
1799 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
1801 2018-01-24 Tom de Vries <tom@codesourcery.com>
1804 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
1805 for neutered threads in warp.
1806 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
1808 2018-01-24 Richard Biener <rguenther@suse.de>
1810 PR tree-optimization/83176
1811 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
1814 2018-01-24 Richard Biener <rguenther@suse.de>
1816 PR tree-optimization/82819
1817 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
1818 code generating pluses that are no-ops in the target precision.
1820 2018-01-24 Richard Biener <rguenther@suse.de>
1823 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
1825 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1827 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
1828 to merge probabilities.
1829 * predict.c (probably_never_executed): Also mark as cold functions
1830 with global 0 profile and guessed local profile.
1831 * profile-count.c (profile_probability::combine_with_count): New
1833 * profile-count.h (profile_probability::operator*,
1834 profile_probability::operator*=, profile_probability::operator/,
1835 profile_probability::operator/=): Reduce precision to adjusted
1836 and set value to guessed on contradictory divisions.
1837 (profile_probability::combine_with_freq): Remove.
1838 (profile_probability::combine_wiht_count): Declare.
1839 (profile_count::force_nonzero):: Set to adjusted.
1840 (profile_count::probability_in):: Set quality to adjusted.
1841 * tree-ssa-tail-merge.c (replace_block_by): Use
1844 2018-01-23 Andrew Waterman <andrew@sifive.com>
1845 Jim Wilson <jimw@sifive.com>
1847 * config/riscv/riscv.c (riscv_stack_boundary): New.
1848 (riscv_option_override): Set riscv_stack_boundary. Handle
1849 riscv_preferred_stack_boundary_arg.
1850 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
1851 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
1852 (STACK_BOUNDARY): Set to riscv_stack_boundary.
1853 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
1854 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
1855 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
1857 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
1860 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
1861 of struct ix86_frame.
1862 (ix86_expand_epilogue): Likewise. Add a local variable for
1863 the reg_save_offset field in struct ix86_frame.
1865 2018-01-23 Bin Cheng <bin.cheng@arm.com>
1867 PR tree-optimization/82604
1868 * tree-loop-distribution.c (enum partition_kind): New enum item
1869 PKIND_PARTIAL_MEMSET.
1870 (partition_builtin_p): Support above new enum item.
1871 (generate_code_for_partition): Ditto.
1872 (compute_access_range): Differentiate cases that equality can be
1873 proven at all loops, the innermost loops or no loops.
1874 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
1875 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
1876 (finalize_partitions, distribute_loop): Don't fuse partition of
1877 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
1878 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
1881 2018-01-23 Martin Liska <mliska@suse.cz>
1883 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
1884 order to ignore the predictor.
1885 (PRED_POLYMORPHIC_CALL): Likewise.
1886 (PRED_RECURSIVE_CALL): Likewise.
1888 2018-01-23 Martin Liska <mliska@suse.cz>
1890 * tree-profile.c (tree_profiling): Print function header to
1891 aware reader which function we are working on.
1892 * value-prof.c (gimple_find_values_to_profile): Do not print
1893 not interesting value histograms.
1895 2018-01-23 Martin Liska <mliska@suse.cz>
1897 * profile-count.h (enum profile_quality): Add
1898 profile_uninitialized as the first value. Do not number values
1899 as they are zero based.
1900 (profile_count::verify): Update sanity check.
1901 (profile_probability::verify): Likewise.
1903 2018-01-23 Nathan Sidwell <nathan@acm.org>
1905 * doc/invoke.texi (ffor-scope): Deprecate.
1907 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1909 PR tree-optimization/83510
1910 * domwalk.c (set_all_edges_as_executable): New function.
1911 (dom_walker::dom_walker): Convert bool param
1912 "skip_unreachable_blocks" to enum reachability. Move setup of
1913 edge flags to set_all_edges_as_executable and only do it when
1914 reachability is REACHABLE_BLOCKS.
1915 * domwalk.h (enum dom_walker::reachability): New enum.
1916 (dom_walker::dom_walker): Convert bool param
1917 "skip_unreachable_blocks" to enum reachability.
1918 (set_all_edges_as_executable): New decl.
1919 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
1920 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
1922 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
1923 but converting true to REACHABLE_BLOCKS.
1924 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
1926 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
1927 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
1928 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
1930 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
1931 if check_all_array_refs will be called.
1933 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1935 * tree.c (selftest::test_location_wrappers): Add more test
1938 2018-01-23 David Malcolm <dmalcolm@redhat.com>
1940 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
1941 (selftest::test_bit_in_range): Likewise.
1943 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
1946 * doc/sourcebuild.texi (vect_float): Say that the selector
1947 only describes the situation when -funsafe-math-optimizations is on.
1948 (vect_float_strict): Document.
1950 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
1952 PR tree-optimization/83965
1953 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1954 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
1955 instead of checking only for a reduction.
1956 (vect_recog_widen_sum_pattern): Likewise.
1958 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1960 * predict.c (probably_never_executed): Only use precise profile info.
1961 (compute_function_frequency): Skip after inlining hack since we now
1962 have quality checking.
1964 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
1966 * profile-count.h (profile_probability::very_unlikely,
1967 profile_probability::unlikely, profile_probability::even): Set
1968 precision to guessed.
1970 2018-01-23 Richard Biener <rguenther@suse.de>
1972 PR tree-optimization/83963
1973 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
1974 Properly terminate dominator walk when crossing the exit edge not
1975 when visiting its source block.
1977 2018-01-23 Jakub Jelinek <jakub@redhat.com>
1980 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
1981 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
1983 2018-01-22 Jakub Jelinek <jakub@redhat.com>
1985 PR tree-optimization/83957
1986 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
1987 semicolon after for body surrounded by braces.
1989 PR tree-optimization/83081
1990 * profile-count.h (profile_probability::split): New method.
1991 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
1992 Use profile_probability::split.
1993 (do_compare_rtx_and_jump): Fix adjustment of probabilities
1994 when splitting a single conditional jump into 2.
1996 2018-01-22 David Malcolm <dmalcolm@redhat.com>
1998 PR tree-optimization/69452
1999 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2002 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2004 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2005 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2006 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2010 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2011 * config/rl78/rl78.md (movdi): New define_expand.
2012 * config/rl78/rl78.c (rl78_split_movdi): New function.
2014 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2017 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2019 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2020 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2021 128-bit to produce an UNSPEC move to get the double word with the
2022 signbit and then a shift directly to do signbit.
2023 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2024 implementation with a new version that just does either a direct
2025 move or a regular move. Move memory interface to separate insns.
2026 Move insns so they are next to the expander.
2027 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2028 with signbit move. Split big and little endian case.
2029 (signbit<mode>2_dm_mem_le): Likewise.
2030 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2031 (signbit<mode>2_dm2): Likewise.
2033 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2035 * config/rl78/rl78.md (anddi3): New define_expand.
2037 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2039 * config/rl78/rl78.md (umindi3): New define_expand.
2041 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2043 * config/rl78/rl78.md (smindi3): New define_expand.
2045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2047 * config/rl78/rl78.md (smaxdi3): New define_expand.
2049 2018-01-22 Carl Love <cel@us.ibm.com>
2051 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2052 LVX_V1TI): Add macro expansion.
2053 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2054 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2055 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2056 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2057 Change check to determine if the instruction is a byte reversing
2058 entry. Fix typo in comment.
2059 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2060 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2061 Add def_builtin calls for new builtins.
2062 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2063 Add define_insn expansion.
2065 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2067 * config/rl78/rl78.md (umaxdi3): New define_expand.
2069 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2071 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2072 for non-QImode registers.
2074 2018-01-22 Richard Biener <rguenther@suse.de>
2076 PR tree-optimization/83963
2077 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2078 including the loop exit block.
2079 (scop_detection::merge_sese): Likewise.
2080 (scop_detection::add_scop): Do it here instead.
2082 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2084 * doc/sourcebuild.texi (arm_softfloat): Document.
2086 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2089 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2090 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2091 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2093 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2094 David Edelsohn <dje.gcc@gmail.com>
2097 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2098 Change "crset eq" to "crset 2".
2099 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2100 (*call_indirect_aix<mode>_nospec): Likewise.
2101 (*call_value_indirect_aix<mode>_nospec): Likewise.
2102 (*call_indirect_elfv2<mode>_nospec): Likewise.
2103 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2104 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2105 change assembly output from . to $.
2106 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2107 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2108 (*tablejump<mode>_internal1_nospec): Likewise.
2110 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2113 * config/sh/sh_optimize_sett_clrt.cc:
2114 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2116 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2118 PR tree-optimization/83940
2119 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2120 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2121 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2122 use it to pass back the definition type.
2123 (vect_check_store_rhs): Likewise rhs_dt_out.
2124 (vect_build_gather_load_calls): Add a mask_dt argument and use
2125 it instead of a call to vect_is_simple_use.
2126 (vectorizable_store): Update calls to vect_check_load_store_mask
2127 and vect_check_store_rhs. Use the dt returned by the latter instead
2128 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2129 instead of calls to vect_is_simple_use. Pass the scalar rather
2130 than the vector operand to vect_is_simple_use when handling
2131 second and subsequent copies of an rhs value.
2132 (vectorizable_load): Update calls to vect_check_load_store_mask
2133 and vect_build_gather_load_calls. Use the cached mask_dt and
2134 gs_info.offset_dt instead of calls to vect_is_simple_use.
2136 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2139 * tree-emutls.c: Include gimplify.h.
2140 (lower_emutls_2): New function.
2141 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2142 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2143 it before further processing.
2146 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2147 UINTVAL (trueop1) instead of INTVAL (op1).
2149 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2153 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2154 INCOMING_FRAME_SP_OFFSET if not defined.
2155 (scan_trace): Add ENTRY argument. If true and
2156 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2157 emit a note to adjust the CFA offset.
2158 (create_cfi_notes): Adjust scan_trace callers.
2159 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2160 INCOMING_FRAME_SP_OFFSET in the CIE.
2161 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2162 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2164 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2165 * doc/tm.texi: Regenerated.
2167 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2169 PR rtl-optimization/83147
2170 * lra-constraints.c (remove_inheritance_pseudos): Use
2171 lra_substitute_pseudo_within_insn.
2173 2018-01-19 Tom de Vries <tom@codesourcery.com>
2174 Cesar Philippidis <cesar@codesourcery.com>
2177 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2179 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2182 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2183 spaces for function labels.
2185 2018-01-19 Martin Liska <mliska@suse.cz>
2187 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2188 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2189 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2190 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2191 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2192 (PRED_CONST_RETURN): Change from 69 to 65.
2193 (PRED_NULL_RETURN): Change from 91 to 71.
2194 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2195 (PRED_LOOP_GUARD): Change from 66 to 73.
2197 2018-01-19 Martin Liska <mliska@suse.cz>
2199 * predict.c (predict_insn_def): Add new assert.
2200 (struct branch_predictor): Change type to signed integer.
2201 (test_prediction_value_range): Amend test to cover
2203 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2204 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2205 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2206 (PRED_LOOP_IV_COMPARE): Likewise.
2207 * predict.h (PROB_UNINITIALIZED): Define new constant.
2209 2018-01-19 Martin Liska <mliska@suse.cz>
2211 * predict.c (dump_prediction): Add new format for
2212 analyze_brprob.py script which is enabled with -details
2214 * profile-count.h (precise_p): New function.
2216 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2218 PR tree-optimization/83922
2219 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2220 there are no statements that need masking.
2221 (vect_active_double_reduction_p): New function.
2222 (vect_analyze_loop_operations): Use it when handling phis that
2223 are not in the loop header.
2225 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2227 PR tree-optimization/83914
2228 * tree-vect-loop.c (vectorizable_induction): Don't convert
2229 init_expr or apply the peeling adjustment for inductions
2230 that are nested within the vectorized loop.
2232 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2234 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2237 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2241 * function.h (gimplify_parameters): Add gimple_seq * argument.
2242 * function.c: Include gimple.h and options.h.
2243 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2244 for the added local temporaries if needed.
2245 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2246 if there are any parameter cleanups, wrap whole body into a
2247 try/finally with the cleanups.
2249 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2252 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2253 Use GET_MODE_CLASS for scalar floating point.
2255 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2259 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2260 Fix call of call_cgraph_insertion_hooks.
2262 2018-01-18 Martin Sebor <msebor@redhat.com>
2264 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2266 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2269 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2272 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2275 * common.opt: (-ffile-prefix-map): New option.
2276 * opts.c (common_handle_option): Defer it.
2277 * opts-global.c (handle_common_deferred_options): Handle it.
2278 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
2279 * file-prefix-map.h: New file.
2280 (remap_debug_filename, add_debug_prefix_map): ...here.
2281 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
2282 * final.c (debug_prefix_map, add_debug_prefix_map
2283 remap_debug_filename): Move to...
2284 * file-prefix-map.c: New file.
2285 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
2286 generalize, get rid of alloca(), use strrchr() instead of strchr().
2287 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
2288 Implement in terms of add_prefix_map().
2289 (remap_macro_filename, remap_debug_filename): Implement in term of
2291 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
2292 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
2293 * dbxout.c: Include file-prefix-map.h.
2294 * varasm.c: Likewise.
2295 * vmsdbgout.c: Likewise.
2296 * xcoffout.c: Likewise.
2297 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
2298 * doc/cppopts.texi (-fmacro-prefix-map): Document.
2299 * doc/invoke.texi (-ffile-prefix-map): Document.
2300 (-fdebug-prefix-map): Update description.
2302 2018-01-18 Martin Liska <mliska@suse.cz>
2304 * config/i386/i386.c (indirect_thunk_name): Document that also
2306 (output_indirect_thunk): Document why both instructions
2307 (pause and lfence) are generated.
2309 2018-01-18 Richard Biener <rguenther@suse.de>
2311 PR tree-optimization/83887
2312 * graphite-scop-detection.c
2313 (scop_detection::get_nearest_dom_with_single_entry): Remove.
2314 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
2315 (scop_detection::merge_sese): Re-implement with a flood-fill
2316 algorithm that properly finds a SESE region if it exists.
2318 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2321 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
2322 pointer_diff optimizations use view_convert instead of convert.
2324 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2326 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2327 Generate different code for -mno-speculate-indirect-jumps.
2328 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2329 (*call_indirect_aix<mode>): Disable for
2330 -mno-speculate-indirect-jumps.
2331 (*call_indirect_aix<mode>_nospec): New define_insn.
2332 (*call_value_indirect_aix<mode>): Disable for
2333 -mno-speculate-indirect-jumps.
2334 (*call_value_indirect_aix<mode>_nospec): New define_insn.
2335 (*sibcall_nonlocal_sysv<mode>): Generate different code for
2336 -mno-speculate-indirect-jumps.
2337 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2339 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2341 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
2342 long double type, set the flags for noting the default long double
2343 type, even if we don't pass or return a long double type.
2345 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
2348 * ipa-inline.c (flatten_function): Do not overwrite final inlining
2351 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
2353 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
2354 support for merge[hl].
2355 (fold_mergehl_helper): New helper function.
2356 (tree-vector-builder.h): New #include for tree_vector_builder usage.
2357 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
2358 (altivec_vmrglw_direct): Add xxmrglw insn.
2360 2018-01-17 Andrew Waterman <andrew@sifive.com>
2362 * config/riscv/riscv.c (riscv_conditional_register_usage): If
2363 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
2365 2018-01-17 David Malcolm <dmalcolm@redhat.com>
2368 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
2369 call the lto_location_cache before reading the
2370 DECL_SOURCE_LOCATION of the types.
2372 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2373 Richard Sandiford <richard.sandiford@linaro.org>
2375 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
2376 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
2377 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
2378 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
2379 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
2381 * config/aarch64/constraints.md (aarch64_movti_operand):
2383 * config/aarch64/predicates.md (Uti): Add new constraint.
2385 2018-01-17 Carl Love <cel@us.ibm.com>
2387 * config/rs6000/vsx.md (define_expand xl_len_r,
2388 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
2389 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
2391 (define_expand, define_insn): Move the shift left from the
2392 define_insn to the define_expand for lxvl and stxvl instructions.
2393 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
2394 and XL_LEN_R definitions to PURE.
2396 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
2398 * config/i386/i386.c (indirect_thunk_name): Declare regno
2399 as unsigned int. Compare regno with INVALID_REGNUM.
2400 (output_indirect_thunk): Ditto.
2401 (output_indirect_thunk_function): Ditto.
2402 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
2403 in the call to output_indirect_thunk_function.
2405 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
2408 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
2409 rather than the size of inner_type to determine the stack slot size
2410 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
2412 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
2415 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
2418 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2420 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
2421 endian Linux systems to optionally enable multilibs for selecting
2422 the long double type if the user configured an explicit type.
2423 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
2424 have no long double multilibs if not defined.
2425 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
2426 warn if the user used -mabi={ieee,ibm}longdouble and we built
2427 multilibs for long double.
2428 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
2429 appropriate multilib option.
2430 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
2432 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
2433 for building long double multilibs.
2434 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
2436 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
2438 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
2441 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
2443 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
2446 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
2449 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
2452 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2454 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
2455 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
2457 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2459 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
2460 different rtl trees depending on TARGET_64BIT.
2461 (rs6000_gen_lvx): Likewise.
2463 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2465 * config/visium/visium.md (nop): Tweak comment.
2466 (hazard_nop): Likewise.
2468 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2470 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
2471 -mspeculate-indirect-jumps.
2472 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
2473 for -mno-speculate-indirect-jumps.
2474 (*call_indirect_elfv2<mode>_nospec): New define_insn.
2475 (*call_value_indirect_elfv2<mode>): Disable for
2476 -mno-speculate-indirect-jumps.
2477 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
2478 (indirect_jump): Emit different RTL for
2479 -mno-speculate-indirect-jumps.
2480 (*indirect_jump<mode>): Disable for
2481 -mno-speculate-indirect-jumps.
2482 (*indirect_jump<mode>_nospec): New define_insn.
2483 (tablejump): Emit different RTL for
2484 -mno-speculate-indirect-jumps.
2485 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
2486 (tablejumpsi_nospec): New define_expand.
2487 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
2488 (tablejumpdi_nospec): New define_expand.
2489 (*tablejump<mode>_internal1): Disable for
2490 -mno-speculate-indirect-jumps.
2491 (*tablejump<mode>_internal1_nospec): New define_insn.
2492 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
2495 2018-01-16 Artyom Skrobov tyomitch@gmail.com
2497 * caller-save.c (insert_save): Drop unnecessary parameter. All
2500 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2501 Richard Biener <rguenth@suse.de>
2504 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
2505 return early, inline manually is_gimple_sizepos. Make sure if we
2506 call gimplify_expr we don't end up with a gimple constant.
2507 * tree.c (variably_modified_type_p): Don't return true for
2508 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
2509 * gimplify.h (is_gimple_sizepos): Remove.
2511 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2513 PR tree-optimization/83857
2514 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
2515 vectorizable_live_operation for pure SLP statements.
2516 (vectorizable_live_operation): Handle PHIs.
2518 2018-01-16 Richard Biener <rguenther@suse.de>
2520 PR tree-optimization/83867
2521 * tree-vect-stmts.c (vect_transform_stmt): Precompute
2522 nested_in_vect_loop_p since the scalar stmt may get invalidated.
2524 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2527 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
2528 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
2529 If off is not INTEGER_CST, issue a may not be aligned warning
2530 rather than isn't aligned. Use isn%'t rather than isn't.
2531 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
2533 <case MULT_EXPR>: Improve the case when bottom and one of the
2534 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
2535 operand, in that case check if the other operand is multiple of
2536 bottom divided by the INTEGER_CST operand.
2538 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2541 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
2542 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
2543 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
2544 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
2545 * config/pa/pa.c (pa_function_arg_advance): Likewise.
2546 (pa_function_arg, pa_arg_partial_bytes): Likewise.
2547 (pa_function_arg_size): New function.
2549 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2551 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
2552 in a separate statement.
2554 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2556 PR tree-optimization/83847
2557 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
2558 group gathers and scatters.
2560 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2562 PR rtl-optimization/86620
2563 * params.def (max-sched-ready-insns): Bump minimum value to 1.
2565 PR rtl-optimization/83213
2566 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
2567 to last if both are JUMP_INSNs.
2569 PR tree-optimization/83843
2570 * gimple-ssa-store-merging.c
2571 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
2572 store_immediate_info for bswap/nop orig_stores.
2574 2018-01-15 Andrew Waterman <andrew@sifive.com>
2576 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
2578 <UDIV>: Increase cost if !TARGET_DIV.
2580 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
2582 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
2583 (define_attr "cr_logical_3op"): New.
2584 (cceq_ior_compare): Adjust.
2585 (cceq_ior_compare_complement): Adjust.
2586 (*cceq_rev_compare): Adjust.
2587 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
2588 (is_cracked_insn): Adjust.
2589 (insn_must_be_first_in_group): Adjust.
2590 * config/rs6000/40x.md: Adjust.
2591 * config/rs6000/440.md: Adjust.
2592 * config/rs6000/476.md: Adjust.
2593 * config/rs6000/601.md: Adjust.
2594 * config/rs6000/603.md: Adjust.
2595 * config/rs6000/6xx.md: Adjust.
2596 * config/rs6000/7450.md: Adjust.
2597 * config/rs6000/7xx.md: Adjust.
2598 * config/rs6000/8540.md: Adjust.
2599 * config/rs6000/cell.md: Adjust.
2600 * config/rs6000/e300c2c3.md: Adjust.
2601 * config/rs6000/e500mc.md: Adjust.
2602 * config/rs6000/e500mc64.md: Adjust.
2603 * config/rs6000/e5500.md: Adjust.
2604 * config/rs6000/e6500.md: Adjust.
2605 * config/rs6000/mpc.md: Adjust.
2606 * config/rs6000/power4.md: Adjust.
2607 * config/rs6000/power5.md: Adjust.
2608 * config/rs6000/power6.md: Adjust.
2609 * config/rs6000/power7.md: Adjust.
2610 * config/rs6000/power8.md: Adjust.
2611 * config/rs6000/power9.md: Adjust.
2612 * config/rs6000/rs64.md: Adjust.
2613 * config/rs6000/titan.md: Adjust.
2615 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2617 * config/i386/predicates.md (indirect_branch_operand): Rewrite
2618 ix86_indirect_branch_register logic.
2620 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2622 * config/i386/constraints.md (Bs): Update
2623 ix86_indirect_branch_register check. Don't check
2624 ix86_indirect_branch_register with GOT_memory_operand.
2626 * config/i386/predicates.md (GOT_memory_operand): Don't check
2627 ix86_indirect_branch_register here.
2628 (GOT32_symbol_operand): Likewise.
2630 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2632 * config/i386/predicates.md (constant_call_address_operand):
2633 Rewrite ix86_indirect_branch_register logic.
2634 (sibcall_insn_operand): Likewise.
2636 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2638 * config/i386/constraints.md (Bs): Replace
2639 ix86_indirect_branch_thunk_register with
2640 ix86_indirect_branch_register.
2642 * config/i386/i386.md (indirect_jump): Likewise.
2643 (tablejump): Likewise.
2644 (*sibcall_memory): Likewise.
2645 (*sibcall_value_memory): Likewise.
2646 Peepholes of indirect call and jump via memory: Likewise.
2647 * config/i386/i386.opt: Likewise.
2648 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2649 (GOT_memory_operand): Likewise.
2650 (call_insn_operand): Likewise.
2651 (sibcall_insn_operand): Likewise.
2652 (GOT32_symbol_operand): Likewise.
2654 2018-01-15 Jakub Jelinek <jakub@redhat.com>
2657 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
2658 type rather than type addr's type points to.
2659 (expand_omp_atomic_mutex): Likewise.
2660 (expand_omp_atomic): Likewise.
2662 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
2665 * config/i386/i386.c (output_indirect_thunk_function): Use
2666 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
2667 for __x86_return_thunk.
2669 2018-01-15 Richard Biener <rguenther@suse.de>
2672 * expmed.c (extract_bit_field_1): Fix typo.
2674 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2677 * config/arm/iterators.md (VF): New mode iterator.
2678 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
2679 Remove integer-related logic from pattern.
2680 (neon_vabd<mode>_3): Likewise.
2682 2018-01-15 Jakub Jelinek <jakub@redhat.com>
2685 * common.opt (fstrict-overflow): No longer an alias.
2686 (fwrapv-pointer): New option.
2687 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
2688 also for pointer types based on flag_wrapv_pointer.
2689 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
2690 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
2691 opts->x_flag_wrapv got set.
2692 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
2693 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
2694 POINTER_TYPE_OVERFLOW_UNDEFINED.
2695 * match.pd: Likewise in address comparison pattern.
2696 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
2698 2018-01-15 Richard Biener <rguenther@suse.de>
2701 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
2702 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
2703 Reset type names to their identifier if their TYPE_DECL doesn't
2704 have linkage (and thus is used for ODR and devirt).
2705 (save_debug_info_for_decl): Remove.
2706 (save_debug_info_for_type): Likewise.
2707 (add_tree_to_fld_list): Adjust.
2708 * tree-pretty-print.c (dump_generic_node): Make dumping of
2709 type names more robust.
2711 2018-01-15 Richard Biener <rguenther@suse.de>
2713 * BASE-VER: Bump to 8.0.1.
2715 2018-01-14 Martin Sebor <msebor@redhat.com>
2718 * builtins.c (check_access): Avoid warning when the no-warning bit
2721 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
2723 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
2724 * ira-color (allocno_hard_regs_compare): Likewise.
2726 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
2729 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
2730 Use .pushsection/.popsection.
2732 2018-01-14 Martin Sebor <msebor@redhat.com>
2735 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
2737 2018-01-14 Jakub Jelinek <jakub@redhat.com>
2739 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
2740 entry from extra_headers.
2741 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
2742 extra_headers, make the list bitwise identical to the i?86-*-* one.
2744 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2746 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
2747 -mcmodel=large with -mindirect-branch=thunk,
2748 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
2749 -mfunction-return=thunk-extern.
2750 * doc/invoke.texi: Document -mcmodel=large is incompatible with
2751 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
2752 -mfunction-return=thunk and -mfunction-return=thunk-extern.
2754 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2756 * config/i386/i386.c (print_reg): Print the name of the full
2757 integer register without '%'.
2758 (ix86_print_operand): Handle 'V'.
2759 * doc/extend.texi: Document 'V' modifier.
2761 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2763 * config/i386/constraints.md (Bs): Disallow memory operand for
2764 -mindirect-branch-register.
2766 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2767 (GOT_memory_operand): Likewise.
2768 (call_insn_operand): Likewise.
2769 (sibcall_insn_operand): Likewise.
2770 (GOT32_symbol_operand): Likewise.
2771 * config/i386/i386.md (indirect_jump): Call convert_memory_address
2772 for -mindirect-branch-register.
2773 (tablejump): Likewise.
2774 (*sibcall_memory): Likewise.
2775 (*sibcall_value_memory): Likewise.
2776 Disallow peepholes of indirect call and jump via memory for
2777 -mindirect-branch-register.
2778 (*call_pop): Replace m with Bw.
2779 (*call_value_pop): Likewise.
2780 (*sibcall_pop_memory): Replace m with Bs.
2781 * config/i386/i386.opt (mindirect-branch-register): New option.
2782 * doc/invoke.texi: Document -mindirect-branch-register option.
2784 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2786 * config/i386/i386-protos.h (ix86_output_function_return): New.
2787 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
2788 set function_return_type.
2789 (indirect_thunk_name): Add ret_p to indicate thunk for function
2791 (output_indirect_thunk_function): Pass false to
2792 indirect_thunk_name.
2793 (ix86_output_indirect_branch_via_reg): Likewise.
2794 (ix86_output_indirect_branch_via_push): Likewise.
2795 (output_indirect_thunk_function): Create alias for function
2796 return thunk if regno < 0.
2797 (ix86_output_function_return): New function.
2798 (ix86_handle_fndecl_attribute): Handle function_return.
2799 (ix86_attribute_table): Add function_return.
2800 * config/i386/i386.h (machine_function): Add
2801 function_return_type.
2802 * config/i386/i386.md (simple_return_internal): Use
2803 ix86_output_function_return.
2804 (simple_return_internal_long): Likewise.
2805 * config/i386/i386.opt (mfunction-return=): New option.
2806 (indirect_branch): Mention -mfunction-return=.
2807 * doc/extend.texi: Document function_return function attribute.
2808 * doc/invoke.texi: Document -mfunction-return= option.
2810 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
2812 * config/i386/i386-opts.h (indirect_branch): New.
2813 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
2814 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
2815 with local indirect jump when converting indirect call and jump.
2816 (ix86_set_indirect_branch_type): New.
2817 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
2818 (indirectlabelno): New.
2819 (indirect_thunk_needed): Likewise.
2820 (indirect_thunk_bnd_needed): Likewise.
2821 (indirect_thunks_used): Likewise.
2822 (indirect_thunks_bnd_used): Likewise.
2823 (INDIRECT_LABEL): Likewise.
2824 (indirect_thunk_name): Likewise.
2825 (output_indirect_thunk): Likewise.
2826 (output_indirect_thunk_function): Likewise.
2827 (ix86_output_indirect_branch_via_reg): Likewise.
2828 (ix86_output_indirect_branch_via_push): Likewise.
2829 (ix86_output_indirect_branch): Likewise.
2830 (ix86_output_indirect_jmp): Likewise.
2831 (ix86_code_end): Call output_indirect_thunk_function if needed.
2832 (ix86_output_call_insn): Call ix86_output_indirect_branch if
2834 (ix86_handle_fndecl_attribute): Handle indirect_branch.
2835 (ix86_attribute_table): Add indirect_branch.
2836 * config/i386/i386.h (machine_function): Add indirect_branch_type
2837 and has_local_indirect_jump.
2838 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
2840 (tablejump): Likewise.
2841 (*indirect_jump): Use ix86_output_indirect_jmp.
2842 (*tablejump_1): Likewise.
2843 (simple_return_indirect_internal): Likewise.
2844 * config/i386/i386.opt (mindirect-branch=): New option.
2845 (indirect_branch): New.
2848 (thunk-inline): Likewise.
2849 (thunk-extern): Likewise.
2850 * doc/extend.texi: Document indirect_branch function attribute.
2851 * doc/invoke.texi: Document -mindirect-branch= option.
2853 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
2856 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
2858 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
2860 * ipa-inline.c (want_inline_small_function_p): Return false if
2861 inlining has already failed with CIF_FINAL_ERROR.
2862 (update_caller_keys): Call want_inline_small_function_p before
2864 (update_callee_keys): Likewise.
2866 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2868 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
2870 (rs6000_quadword_masked_address_p): Likewise.
2871 (quad_aligned_load_p): Likewise.
2872 (quad_aligned_store_p): Likewise.
2873 (const_load_sequence_p): Add comment to describe the outer-most loop.
2874 (mimic_memory_attributes_and_flags): New function.
2875 (rs6000_gen_stvx): Likewise.
2876 (replace_swapped_aligned_store): Likewise.
2877 (rs6000_gen_lvx): Likewise.
2878 (replace_swapped_aligned_load): Likewise.
2879 (replace_swapped_load_constant): Capitalize argument name in
2880 comment describing this function.
2881 (rs6000_analyze_swaps): Add a third pass to search for vector loads
2882 and stores that access quad-word aligned addresses and replace
2883 with stvx or lvx instructions when appropriate.
2884 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
2885 New function prototype.
2886 (rs6000_quadword_masked_address_p): Likewise.
2887 (rs6000_gen_lvx): Likewise.
2888 (rs6000_gen_stvx): Likewise.
2889 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
2890 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
2891 when memory address is aligned.
2892 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
2893 this split to select lvx instruction when memory address is aligned.
2894 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
2895 instruction when memory address is aligned.
2896 (*vsx_le_perm_load_v16qi): Likewise.
2897 (four unnamed splitters): Modify to select the stvx instruction
2898 when memory is aligned.
2900 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
2902 * predict.c (determine_unlikely_bbs): Handle correctly BBs
2903 which appears in the queue multiple times.
2905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2906 Alan Hayward <alan.hayward@arm.com>
2907 David Sherwood <david.sherwood@arm.com>
2909 * tree-vectorizer.h (vec_lower_bound): New structure.
2910 (_loop_vec_info): Add check_nonzero and lower_bounds.
2911 (LOOP_VINFO_CHECK_NONZERO): New macro.
2912 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
2913 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
2914 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
2915 fields. Make seg_len the distance travelled, not including the
2917 (dr_direction_indicator): Declare.
2918 (dr_zero_step_indicator): Likewise.
2919 (dr_known_forward_stride_p): Likewise.
2920 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
2922 (runtime_alias_check_p): Allow runtime alias checks with
2924 (operator ==): Compare access_size and align.
2925 (prune_runtime_alias_test_list): Rework for new distinction between
2926 the access_size and seg_len.
2927 (create_intersect_range_checks_index): Likewise. Cope with polynomial
2929 (get_segment_min_max): New function.
2930 (create_intersect_range_checks): Use it.
2931 (dr_step_indicator): New function.
2932 (dr_direction_indicator): Likewise.
2933 (dr_zero_step_indicator): Likewise.
2934 (dr_known_forward_stride_p): Likewise.
2935 * tree-loop-distribution.c (data_ref_segment_size): Return
2936 DR_STEP * (niters - 1).
2937 (compute_alias_check_pairs): Update call to the dr_with_seg_len
2939 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
2940 (vect_preserves_scalar_order_p): New function, split out from...
2941 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
2942 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
2943 (vect_vfa_access_size): New function.
2944 (vect_vfa_align): Likewise.
2945 (vect_compile_time_alias): Take access_size_a and access_b arguments.
2946 (dump_lower_bound): New function.
2947 (vect_check_lower_bound): Likewise.
2948 (vect_small_gap_p): Likewise.
2949 (vectorizable_with_step_bound_p): Likewise.
2950 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
2951 depencies if the vectorization factor is 1. Convert the checks
2952 for nonzero steps into checks on the bounds of DR_STEP. Try using
2953 a bunds check for variable steps if the minimum required step is
2954 relatively small. Update calls to the dr_with_seg_len
2955 constructor and to vect_compile_time_alias.
2956 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
2958 (vect_loop_versioning): Call it.
2959 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
2961 (vect_estimate_min_profitable_iters): Account for any bounds checks.
2963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
2964 Alan Hayward <alan.hayward@arm.com>
2965 David Sherwood <david.sherwood@arm.com>
2967 * doc/sourcebuild.texi (vect_scatter_store): Document.
2968 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
2970 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
2972 * genopinit.c (main): Add supports_vec_scatter_store and
2973 supports_vec_scatter_store_cached to target_optabs.
2974 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
2975 IFN_MASK_SCATTER_STORE.
2976 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
2978 * internal-fn.h (internal_store_fn_p): Declare.
2979 (internal_fn_stored_value_index): Likewise.
2980 * internal-fn.c (scatter_store_direct): New macro.
2981 (expand_scatter_store_optab_fn): New function.
2982 (direct_scatter_store_optab_supported_p): New macro.
2983 (internal_store_fn_p): New function.
2984 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
2985 IFN_MASK_SCATTER_STORE.
2986 (internal_fn_mask_index): Likewise.
2987 (internal_fn_stored_value_index): New function.
2988 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
2990 * optabs-query.h (supports_vec_scatter_store_p): Declare.
2991 * optabs-query.c (supports_vec_scatter_store_p): New function.
2992 * tree-vectorizer.h (vect_get_store_rhs): Declare.
2993 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
2994 true for scatter stores.
2995 (vect_gather_scatter_fn_p): Handle scatter stores too.
2996 (vect_check_gather_scatter): Consider using scatter stores if
2997 supports_vec_scatter_store_p.
2998 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3000 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3001 internal_fn_stored_value_index.
3002 (check_load_store_masking): Handle scatter stores too.
3003 (vect_get_store_rhs): Make public.
3004 (vectorizable_call): Use internal_store_fn_p.
3005 (vectorizable_store): Handle scatter store internal functions.
3006 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3007 when deciding whether the end of the group has been reached.
3008 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3009 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3010 (mask_scatter_store<mode>): New insns.
3012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3013 Alan Hayward <alan.hayward@arm.com>
3014 David Sherwood <david.sherwood@arm.com>
3016 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3017 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3018 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3020 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3021 Use vect_truncate_gather_scatter_offset if we can't treat the
3022 operation as a normal gather load or scatter store.
3023 (get_group_load_store_type): Take the gather_scatter_info
3024 as argument. Try using a gather load or scatter store for
3025 single-element groups.
3026 (get_load_store_type): Update calls to get_group_load_store_type
3027 and vect_use_strided_gather_scatters_p.
3029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3030 Alan Hayward <alan.hayward@arm.com>
3031 David Sherwood <david.sherwood@arm.com>
3033 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3034 optional tree argument.
3035 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3037 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3038 but continue to use the current value as a fallback.
3039 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3040 to compare the updates.
3041 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3042 (get_load_store_type): Use it when handling a strided access.
3043 (vect_get_strided_load_store_ops): New function.
3044 (vect_get_data_ptr_increment): Likewise.
3045 (vectorizable_load): Handle strided gather loads. Always pass
3046 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3049 Alan Hayward <alan.hayward@arm.com>
3050 David Sherwood <david.sherwood@arm.com>
3052 * doc/md.texi (gather_load@var{m}): Document.
3053 (mask_gather_load@var{m}): Likewise.
3054 * genopinit.c (main): Add supports_vec_gather_load and
3055 supports_vec_gather_load_cached to target_optabs.
3056 * optabs-tree.c (init_tree_optimization_optabs): Use
3057 ggc_cleared_alloc to allocate target_optabs.
3058 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3059 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3061 * internal-fn.h (internal_load_fn_p): Declare.
3062 (internal_gather_scatter_fn_p): Likewise.
3063 (internal_fn_mask_index): Likewise.
3064 (internal_gather_scatter_fn_supported_p): Likewise.
3065 * internal-fn.c (gather_load_direct): New macro.
3066 (expand_gather_load_optab_fn): New function.
3067 (direct_gather_load_optab_supported_p): New macro.
3068 (direct_internal_fn_optab): New function.
3069 (internal_load_fn_p): Likewise.
3070 (internal_gather_scatter_fn_p): Likewise.
3071 (internal_fn_mask_index): Likewise.
3072 (internal_gather_scatter_fn_supported_p): Likewise.
3073 * optabs-query.c (supports_at_least_one_mode_p): New function.
3074 (supports_vec_gather_load_p): Likewise.
3075 * optabs-query.h (supports_vec_gather_load_p): Declare.
3076 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3077 and memory_type field.
3078 (NUM_PATTERNS): Bump to 15.
3079 * tree-vect-data-refs.c: Include internal-fn.h.
3080 (vect_gather_scatter_fn_p): New function.
3081 (vect_describe_gather_scatter_call): Likewise.
3082 (vect_check_gather_scatter): Try using internal functions for
3083 gather loads. Recognize existing calls to a gather load function.
3084 (vect_analyze_data_refs): Consider using gather loads if
3085 supports_vec_gather_load_p.
3086 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3087 (vect_get_gather_scatter_offset_type): Likewise.
3088 (vect_convert_mask_for_vectype): Likewise.
3089 (vect_add_conversion_to_patterm): Likewise.
3090 (vect_try_gather_scatter_pattern): Likewise.
3091 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3092 (vect_vect_recog_func_ptrs): Add it.
3093 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3094 internal_fn_mask_index and internal_gather_scatter_fn_p.
3095 (check_load_store_masking): Take the gather_scatter_info as an
3096 argument and handle gather loads.
3097 (vect_get_gather_scatter_ops): New function.
3098 (vectorizable_call): Check internal_load_fn_p.
3099 (vectorizable_load): Likewise. Handle gather load internal
3101 (vectorizable_store): Update call to check_load_store_masking.
3102 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3103 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3104 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3105 (aarch64_gather_scale_operand_d): New predicates.
3106 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3107 (mask_gather_load<mode>): New insns.
3109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3110 Alan Hayward <alan.hayward@arm.com>
3111 David Sherwood <david.sherwood@arm.com>
3113 * optabs.def (fold_left_plus_optab): New optab.
3114 * doc/md.texi (fold_left_plus_@var{m}): Document.
3115 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3116 * internal-fn.c (fold_left_direct): Define.
3117 (expand_fold_left_optab_fn): Likewise.
3118 (direct_fold_left_optab_supported_p): Likewise.
3119 * fold-const-call.c (fold_const_fold_left): New function.
3120 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3121 * tree-parloops.c (valid_reduction_p): New function.
3122 (gather_scalar_reductions): Use it.
3123 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3124 (vect_finish_replace_stmt): Declare.
3125 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3126 (needs_fold_left_reduction_p): New function, split out from...
3127 (vect_is_simple_reduction): ...here. Accept reductions that
3128 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3129 (vect_force_simple_reduction): Also store the reduction type in
3130 the assignment's STMT_VINFO_REDUC_TYPE.
3131 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3132 (merge_with_identity): New function.
3133 (vect_expand_fold_left): Likewise.
3134 (vectorize_fold_left_reduction): Likewise.
3135 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3136 scalar phi in place for it. Check for target support and reject
3137 cases that would reassociate the operation. Defer the transform
3138 phase to vectorize_fold_left_reduction.
3139 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3140 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3141 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3145 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3146 call to ifc_temp_var.
3148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3149 Alan Hayward <alan.hayward@arm.com>
3150 David Sherwood <david.sherwood@arm.com>
3152 * target.def (legitimize_address_displacement): Take the original
3153 offset as a poly_int.
3154 * targhooks.h (default_legitimize_address_displacement): Update
3156 * targhooks.c (default_legitimize_address_displacement): Likewise.
3157 * doc/tm.texi: Regenerate.
3158 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3159 as an argument, moving assert of ad->disp == ad->disp_term to...
3160 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3161 Try calling targetm.legitimize_address_displacement before expanding
3162 the address rather than afterwards, and adjust for the new interface.
3163 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3164 Match the new hook interface. Handle SVE addresses.
3165 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3170 * Makefile.in (OBJS): Add early-remat.o.
3171 * target.def (select_early_remat_modes): New hook.
3172 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3173 * doc/tm.texi: Regenerate.
3174 * targhooks.h (default_select_early_remat_modes): Declare.
3175 * targhooks.c (default_select_early_remat_modes): New function.
3176 * timevar.def (TV_EARLY_REMAT): New timevar.
3177 * passes.def (pass_early_remat): New pass.
3178 * tree-pass.h (make_pass_early_remat): Declare.
3179 * early-remat.c: New file.
3180 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3182 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3185 Alan Hayward <alan.hayward@arm.com>
3186 David Sherwood <david.sherwood@arm.com>
3188 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3189 vfm1 with a bound_epilog parameter.
3190 (vect_do_peeling): Update calls accordingly, and move the prologue
3191 call earlier in the function. Treat the base bound_epilog as 0 for
3192 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3193 this base when peeling for gaps.
3194 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3195 with fully-masked loops.
3196 (vect_estimate_min_profitable_iters): Handle the single peeled
3197 iteration in that case.
3199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3200 Alan Hayward <alan.hayward@arm.com>
3201 David Sherwood <david.sherwood@arm.com>
3203 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3204 single-element interleaving even if the size is not a power of 2.
3205 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3206 accesses for single-element interleaving if the group size is
3209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3210 Alan Hayward <alan.hayward@arm.com>
3211 David Sherwood <david.sherwood@arm.com>
3213 * doc/md.texi (fold_extract_last_@var{m}): Document.
3214 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3215 * optabs.def (fold_extract_last_optab): New optab.
3216 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3217 * internal-fn.c (fold_extract_direct): New macro.
3218 (expand_fold_extract_optab_fn): Likewise.
3219 (direct_fold_extract_optab_supported_p): Likewise.
3220 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3221 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3222 EXTRACT_LAST_REDUCTION.
3223 (get_initial_def_for_reduction): Do not create an initial vector
3224 for EXTRACT_LAST_REDUCTION reductions.
3225 (vectorizable_reduction): Leave the scalar phi in place for
3226 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3227 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3228 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3229 transform phase to vectorizable_condition.
3230 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3232 (vect_finish_stmt_generation): ...here.
3233 (vect_finish_replace_stmt): New function.
3234 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3235 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3237 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3240 Alan Hayward <alan.hayward@arm.com>
3241 David Sherwood <david.sherwood@arm.com>
3243 * doc/md.texi (extract_last_@var{m}): Document.
3244 * optabs.def (extract_last_optab): New optab.
3245 * internal-fn.def (EXTRACT_LAST): New internal function.
3246 * internal-fn.c (cond_unary_direct): New macro.
3247 (expand_cond_unary_optab_fn): Likewise.
3248 (direct_cond_unary_optab_supported_p): Likewise.
3249 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3250 loops using EXTRACT_LAST.
3251 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3252 (extract_last_<mode>): ...this optab.
3253 (vec_extract<mode><Vel>): Update accordingly.
3255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3256 Alan Hayward <alan.hayward@arm.com>
3257 David Sherwood <david.sherwood@arm.com>
3259 * target.def (empty_mask_is_expensive): New hook.
3260 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3261 * doc/tm.texi: Regenerate.
3262 * targhooks.h (default_empty_mask_is_expensive): Declare.
3263 * targhooks.c (default_empty_mask_is_expensive): New function.
3264 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3265 if the target says that empty masks are expensive.
3266 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3268 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3271 Alan Hayward <alan.hayward@arm.com>
3272 David Sherwood <david.sherwood@arm.com>
3274 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3275 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3276 (vect_use_loop_mask_for_alignment_p): New function.
3277 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3278 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
3279 niters_skip argument. Make sure that the first niters_skip elements
3280 of the first iteration are inactive.
3281 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
3282 Update call to vect_set_loop_masks_directly.
3283 (get_misalign_in_elems): New function, split out from...
3284 (vect_gen_prolog_loop_niters): ...here.
3285 (vect_update_init_of_dr): Take a code argument that specifies whether
3286 the adjustment should be added or subtracted.
3287 (vect_update_init_of_drs): Likewise.
3288 (vect_prepare_for_masked_peels): New function.
3289 (vect_do_peeling): Skip prologue peeling if we're using a mask
3290 instead. Update call to vect_update_inits_of_drs.
3291 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3293 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
3294 alignment. Do not include the number of peeled iterations in
3295 the minimum threshold in that case.
3296 (vectorizable_induction): Adjust the start value down by
3297 LOOP_VINFO_MASK_SKIP_NITERS iterations.
3298 (vect_transform_loop): Call vect_prepare_for_masked_peels.
3299 Take the number of skipped iterations into account when calculating
3301 * tree-vect-stmts.c (vect_gen_while_not): New function.
3303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3304 Alan Hayward <alan.hayward@arm.com>
3305 David Sherwood <david.sherwood@arm.com>
3307 * doc/sourcebuild.texi (vect_fully_masked): Document.
3308 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
3310 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
3312 (vect_analyze_loop_2): ...here. Don't check the vectorization
3313 factor against the number of loop iterations if the loop is
3316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3317 Alan Hayward <alan.hayward@arm.com>
3318 David Sherwood <david.sherwood@arm.com>
3320 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
3321 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
3322 (dump_groups): Update accordingly.
3323 (iv_use::mem_type): New member variable.
3324 (address_p): New function.
3325 (record_use): Add a mem_type argument and initialize the new
3327 (record_group_use): Add a mem_type argument. Use address_p.
3328 Remove obsolete null checks of base_object. Update call to record_use.
3329 (find_interesting_uses_op): Update call to record_group_use.
3330 (find_interesting_uses_cond): Likewise.
3331 (find_interesting_uses_address): Likewise.
3332 (get_mem_type_for_internal_fn): New function.
3333 (find_address_like_use): Likewise.
3334 (find_interesting_uses_stmt): Try find_address_like_use before
3335 calling find_interesting_uses_op.
3336 (addr_offset_valid_p): Use the iv mem_type field as the type
3337 of the addressed memory.
3338 (add_autoinc_candidates): Likewise.
3339 (get_address_cost): Likewise.
3340 (split_small_address_groups_p): Use address_p.
3341 (split_address_groups): Likewise.
3342 (add_iv_candidate_for_use): Likewise.
3343 (autoinc_possible_for_pair): Likewise.
3344 (rewrite_groups): Likewise.
3345 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
3346 (determine_group_iv_cost): Update after split of USE_ADDRESS.
3347 (get_alias_ptr_type_for_ptr_address): New function.
3348 (rewrite_use_address): Rewrite address uses in calls that were
3349 identified by find_address_like_use.
3351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3352 Alan Hayward <alan.hayward@arm.com>
3353 David Sherwood <david.sherwood@arm.com>
3355 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
3357 * gimple-expr.h (is_gimple_addressable: Likewise.
3358 * gimple-expr.c (is_gimple_address): Likewise.
3359 * internal-fn.c (expand_call_mem_ref): New function.
3360 (expand_mask_load_optab_fn): Use it.
3361 (expand_mask_store_optab_fn): Likewise.
3363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3364 Alan Hayward <alan.hayward@arm.com>
3365 David Sherwood <david.sherwood@arm.com>
3367 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
3368 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
3369 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
3370 (cond_umax@var{mode}): Document.
3371 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
3372 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
3373 (cond_umin_optab, cond_umax_optab): New optabs.
3374 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
3375 (COND_IOR, COND_XOR): New internal functions.
3376 * internal-fn.h (get_conditional_internal_fn): Declare.
3377 * internal-fn.c (cond_binary_direct): New macro.
3378 (expand_cond_binary_optab_fn): Likewise.
3379 (direct_cond_binary_optab_supported_p): Likewise.
3380 (get_conditional_internal_fn): New function.
3381 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
3382 Cope with reduction statements that are vectorized as calls rather
3384 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
3385 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
3386 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
3387 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3388 (UNSPEC_COND_EOR): New unspecs.
3389 (optab): Add mappings for them.
3390 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
3391 (sve_int_op, sve_fp_op): New int attributes.
3393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3394 Alan Hayward <alan.hayward@arm.com>
3395 David Sherwood <david.sherwood@arm.com>
3397 * optabs.def (while_ult_optab): New optab.
3398 * doc/md.texi (while_ult@var{m}@var{n}): Document.
3399 * internal-fn.def (WHILE_ULT): New internal function.
3400 * internal-fn.h (direct_internal_fn_supported_p): New override
3401 that takes two types as argument.
3402 * internal-fn.c (while_direct): New macro.
3403 (expand_while_optab_fn): New function.
3404 (convert_optab_supported_p): Likewise.
3405 (direct_while_optab_supported_p): New macro.
3406 * wide-int.h (wi::udiv_ceil): New function.
3407 * tree-vectorizer.h (rgroup_masks): New structure.
3408 (vec_loop_masks): New typedef.
3409 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
3411 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
3412 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
3413 (vect_max_vf): New function.
3414 (slpeel_make_loop_iterate_ntimes): Delete.
3415 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
3416 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
3417 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
3418 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
3419 internal-fn.h, stor-layout.h and optabs-query.h.
3420 (vect_set_loop_mask): New function.
3421 (add_preheader_seq): Likewise.
3422 (add_header_seq): Likewise.
3423 (interleave_supported_p): Likewise.
3424 (vect_maybe_permute_loop_masks): Likewise.
3425 (vect_set_loop_masks_directly): Likewise.
3426 (vect_set_loop_condition_masked): Likewise.
3427 (vect_set_loop_condition_unmasked): New function, split out from
3428 slpeel_make_loop_iterate_ntimes.
3429 (slpeel_make_loop_iterate_ntimes): Rename to..
3430 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
3431 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
3432 (vect_do_peeling): Update call accordingly.
3433 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
3435 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3436 mask_compare_type, can_fully_mask_p and fully_masked_p.
3437 (release_vec_loop_masks): New function.
3438 (_loop_vec_info): Use it to free the loop masks.
3439 (can_produce_all_loop_masks_p): New function.
3440 (vect_get_max_nscalars_per_iter): Likewise.
3441 (vect_verify_full_masking): Likewise.
3442 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
3443 retries, and free the mask rgroups before retrying. Check loop-wide
3444 reasons for disallowing fully-masked loops. Make the final decision
3445 about whether use a fully-masked loop or not.
3446 (vect_estimate_min_profitable_iters): Do not assume that peeling
3447 for the number of iterations will be needed for fully-masked loops.
3448 (vectorizable_reduction): Disable fully-masked loops.
3449 (vectorizable_live_operation): Likewise.
3450 (vect_halve_mask_nunits): New function.
3451 (vect_double_mask_nunits): Likewise.
3452 (vect_record_loop_mask): Likewise.
3453 (vect_get_loop_mask): Likewise.
3454 (vect_transform_loop): Handle the case in which the final loop
3455 iteration might handle a partial vector. Call vect_set_loop_condition
3456 instead of slpeel_make_loop_iterate_ntimes.
3457 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
3458 (check_load_store_masking): New function.
3459 (prepare_load_store_mask): Likewise.
3460 (vectorizable_store): Handle fully-masked loops.
3461 (vectorizable_load): Likewise.
3462 (supportable_widening_operation): Use vect_halve_mask_nunits for
3464 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
3465 (vect_gen_while): New function.
3466 * config/aarch64/aarch64.md (umax<mode>3): New expander.
3467 (aarch64_uqdec<mode>): New insn.
3469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3470 Alan Hayward <alan.hayward@arm.com>
3471 David Sherwood <david.sherwood@arm.com>
3473 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
3474 (reduc_xor_scal_optab): New optabs.
3475 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
3476 (reduc_xor_scal_@var{m}): Document.
3477 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
3478 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
3480 * fold-const-call.c (fold_const_call): Handle them.
3481 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
3482 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
3483 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
3484 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
3485 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
3486 (UNSPEC_XORV): New unspecs.
3487 (optab): Add entries for them.
3488 (BITWISEV): New int iterator.
3489 (bit_reduc_op): New int attributes.
3491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3492 Alan Hayward <alan.hayward@arm.com>
3493 David Sherwood <david.sherwood@arm.com>
3495 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
3496 * internal-fn.def (VEC_SHL_INSERT): New internal function.
3497 * optabs.def (vec_shl_insert_optab): New optab.
3498 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
3499 (duplicate_and_interleave): Likewise.
3500 * tree-vect-loop.c: Include internal-fn.h.
3501 (neutral_op_for_slp_reduction): New function, split out from
3502 get_initial_defs_for_reduction.
3503 (get_initial_def_for_reduction): Handle option 2 for variable-length
3504 vectors by loading the neutral value into a vector and then shifting
3505 the initial value into element 0.
3506 (get_initial_defs_for_reduction): Replace the code argument with
3507 the neutral value calculated by neutral_op_for_slp_reduction.
3508 Use gimple_build_vector for constant-length vectors.
3509 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
3510 but the first group_size elements have a neutral value.
3511 Use duplicate_and_interleave otherwise.
3512 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
3513 Update call to get_initial_defs_for_reduction. Handle SLP
3514 reductions for variable-length vectors by creating one vector
3515 result for each scalar result, with the elements associated
3516 with other scalar results stubbed out with the neutral value.
3517 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
3518 Require IFN_VEC_SHL_INSERT for double reductions on
3519 variable-length vectors, or SLP reductions that have
3520 a neutral value. Require can_duplicate_and_interleave_p
3521 support for variable-length unchained SLP reductions if there
3522 is no neutral value, such as for MIN/MAX reductions. Also require
3523 the number of vector elements to be a multiple of the number of
3524 SLP statements when doing variable-length unchained SLP reductions.
3525 Update call to vect_create_epilog_for_reduction.
3526 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
3527 and remove initial values.
3528 (duplicate_and_interleave): Make public.
3529 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
3530 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
3532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3533 Alan Hayward <alan.hayward@arm.com>
3534 David Sherwood <david.sherwood@arm.com>
3536 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
3537 (can_duplicate_and_interleave_p): New function.
3538 (vect_get_and_check_slp_defs): Take the vector of statements
3539 rather than just the current one. Remove excess parentheses.
3540 Restriction rejectinon of vect_constant_def and vect_external_def
3541 for variable-length vectors to boolean types, or types for which
3542 can_duplicate_and_interleave_p is false.
3543 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
3544 (duplicate_and_interleave): New function.
3545 (vect_get_constant_vectors): Use gimple_build_vector for
3546 constant-length vectors and suitable variable-length constant
3547 vectors. Use duplicate_and_interleave for other variable-length
3548 vectors. Don't defer the update when inserting new statements.
3550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3551 Alan Hayward <alan.hayward@arm.com>
3552 David Sherwood <david.sherwood@arm.com>
3554 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
3555 min_profitable_iters doesn't go negative.
3557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3558 Alan Hayward <alan.hayward@arm.com>
3559 David Sherwood <david.sherwood@arm.com>
3561 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
3562 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
3563 * optabs.def (vec_mask_load_lanes_optab): New optab.
3564 (vec_mask_store_lanes_optab): Likewise.
3565 * internal-fn.def (MASK_LOAD_LANES): New internal function.
3566 (MASK_STORE_LANES): Likewise.
3567 * internal-fn.c (mask_load_lanes_direct): New macro.
3568 (mask_store_lanes_direct): Likewise.
3569 (expand_mask_load_optab_fn): Handle masked operations.
3570 (expand_mask_load_lanes_optab_fn): New macro.
3571 (expand_mask_store_optab_fn): Handle masked operations.
3572 (expand_mask_store_lanes_optab_fn): New macro.
3573 (direct_mask_load_lanes_optab_supported_p): Likewise.
3574 (direct_mask_store_lanes_optab_supported_p): Likewise.
3575 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
3577 (vect_load_lanes_supported): Likewise.
3578 * tree-vect-data-refs.c (strip_conversion): New function.
3579 (can_group_stmts_p): Likewise.
3580 (vect_analyze_data_ref_accesses): Use it instead of checking
3581 for a pair of assignments.
3582 (vect_store_lanes_supported): Take a masked_p parameter.
3583 (vect_load_lanes_supported): Likewise.
3584 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
3585 vect_store_lanes_supported and vect_load_lanes_supported.
3586 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3587 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
3588 parameter. Don't allow gaps for masked accesses.
3589 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
3590 and vect_load_lanes_supported.
3591 (get_load_store_type): Take a masked_p parameter and update
3592 call to get_group_load_store_type.
3593 (vectorizable_store): Update call to get_load_store_type.
3594 Handle IFN_MASK_STORE_LANES.
3595 (vectorizable_load): Update call to get_load_store_type.
3596 Handle IFN_MASK_LOAD_LANES.
3598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3599 Alan Hayward <alan.hayward@arm.com>
3600 David Sherwood <david.sherwood@arm.com>
3602 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
3604 * config/aarch64/aarch64-protos.h
3605 (aarch64_sve_struct_memory_operand_p): Declare.
3606 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
3607 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
3608 (VPRED, vpred): Handle SVE structure modes.
3609 * config/aarch64/constraints.md (Utx): New constraint.
3610 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
3611 (aarch64_sve_struct_nonimmediate_operand): New predicates.
3612 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
3613 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
3614 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
3615 structure modes. Split into pieces after RA.
3616 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
3617 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
3619 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
3620 SVE structure modes.
3621 (aarch64_classify_address): Likewise.
3622 (sizetochar): Move earlier in file.
3623 (aarch64_print_operand): Handle SVE register lists.
3624 (aarch64_array_mode): New function.
3625 (aarch64_sve_struct_memory_operand_p): Likewise.
3626 (TARGET_ARRAY_MODE): Redefine.
3628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3629 Alan Hayward <alan.hayward@arm.com>
3630 David Sherwood <david.sherwood@arm.com>
3632 * target.def (array_mode): New target hook.
3633 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
3634 * doc/tm.texi: Regenerate.
3635 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
3636 * hooks.c (hook_optmode_mode_uhwi_none): New function.
3637 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
3639 * stor-layout.c (mode_for_array): Likewise. Support polynomial
3642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3643 Alan Hayward <alan.hayward@arm.com>
3644 David Sherwood <david.sherwood@arm.com>
3646 * fold-const.c (fold_binary_loc): Check the argument types
3647 rather than the result type when testing for a vector operation.
3649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3651 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
3652 * doc/tm.texi: Regenerate.
3654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3655 Alan Hayward <alan.hayward@arm.com>
3656 David Sherwood <david.sherwood@arm.com>
3658 * doc/invoke.texi (-msve-vector-bits=): Document new option.
3659 (sve): Document new AArch64 extension.
3660 * doc/md.texi (w): Extend the description of the AArch64
3661 constraint to include SVE vectors.
3662 (Upl, Upa): Document new AArch64 predicate constraints.
3663 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
3665 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
3666 (msve-vector-bits=): New option.
3667 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
3668 SVE when these are disabled.
3669 (sve): New extension.
3670 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
3671 modes. Adjust their number of units based on aarch64_sve_vg.
3672 (MAX_BITSIZE_MODE_ANY_MODE): Define.
3673 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
3674 aarch64_addr_query_type.
3675 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
3676 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
3677 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
3678 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
3679 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
3680 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
3681 (aarch64_simd_imm_zero_p): Delete.
3682 (aarch64_check_zero_based_sve_index_immediate): Declare.
3683 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
3684 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
3685 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
3686 (aarch64_sve_float_mul_immediate_p): Likewise.
3687 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
3689 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
3690 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
3691 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
3692 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
3693 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
3694 (aarch64_regmode_natural_size): Likewise.
3695 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
3696 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
3698 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
3699 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
3700 for VG and the SVE predicate registers.
3701 (V_ALIASES): Add a "z"-prefixed alias.
3702 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
3703 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
3704 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
3705 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
3706 (REG_CLASS_NAMES): Add entries for them.
3707 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
3708 and the predicate registers.
3709 (aarch64_sve_vg): Declare.
3710 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
3711 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
3712 (REGMODE_NATURAL_SIZE): Define.
3713 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3715 * config/aarch64/aarch64.c: Include cfgrtl.h.
3716 (simd_immediate_info): Add a constructor for series vectors,
3717 and an associated step field.
3718 (aarch64_sve_vg): New variable.
3719 (aarch64_dbx_register_number): Handle VG and the predicate registers.
3720 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
3721 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
3722 (VEC_ANY_DATA, VEC_STRUCT): New constants.
3723 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
3724 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
3725 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
3726 (aarch64_get_mask_mode): New functions.
3727 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
3728 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
3729 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
3730 predicate modes and predicate registers. Explicitly restrict
3731 GPRs to modes of 16 bytes or smaller. Only allow FP registers
3732 to store a vector mode if it is recognized by
3733 aarch64_classify_vector_mode.
3734 (aarch64_regmode_natural_size): New function.
3735 (aarch64_hard_regno_caller_save_mode): Return the original mode
3737 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
3738 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
3739 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
3740 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
3742 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
3743 does not overlap dest if the function is frame-related. Handle
3745 (aarch64_split_add_offset): New function.
3746 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
3747 them aarch64_add_offset.
3748 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
3749 and update call to aarch64_sub_sp.
3750 (aarch64_add_cfa_expression): New function.
3751 (aarch64_expand_prologue): Pass extra temporary registers to the
3752 functions above. Handle the case in which we need to emit new
3753 DW_CFA_expressions for registers that were originally saved
3754 relative to the stack pointer, but now have to be expressed
3755 relative to the frame pointer.
3756 (aarch64_output_mi_thunk): Pass extra temporary registers to the
3758 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
3759 IP0 and IP1 values for SVE frames.
3760 (aarch64_expand_vec_series): New function.
3761 (aarch64_expand_sve_widened_duplicate): Likewise.
3762 (aarch64_expand_sve_const_vector): Likewise.
3763 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
3764 Handle SVE constants. Use emit_move_insn to move a force_const_mem
3765 into the register, rather than emitting a SET directly.
3766 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
3767 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
3768 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
3769 (offset_9bit_signed_scaled_p): New functions.
3770 (aarch64_replicate_bitmask_imm): New function.
3771 (aarch64_bitmask_imm): Use it.
3772 (aarch64_cannot_force_const_mem): Reject expressions involving
3773 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
3774 (aarch64_classify_index): Handle SVE indices, by requiring
3775 a plain register index with a scale that matches the element size.
3776 (aarch64_classify_address): Handle SVE addresses. Assert that
3777 the mode of the address is VOIDmode or an integer mode.
3778 Update call to aarch64_classify_symbol.
3779 (aarch64_classify_symbolic_expression): Update call to
3780 aarch64_classify_symbol.
3781 (aarch64_const_vec_all_in_range_p): New function.
3782 (aarch64_print_vector_float_operand): Likewise.
3783 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
3784 "vN" for FP registers with SVE modes. Handle (const ...) vectors
3785 and the FP immediates 1.0 and 0.5.
3786 (aarch64_print_address_internal): Handle SVE addresses.
3787 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
3788 (aarch64_regno_regclass): Handle predicate registers.
3789 (aarch64_secondary_reload): Handle big-endian reloads of SVE
3791 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
3792 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
3793 (aarch64_convert_sve_vector_bits): New function.
3794 (aarch64_override_options): Use it to handle -msve-vector-bits=.
3795 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
3797 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
3798 Handle SVE vector and predicate modes. Accept VL-based constants
3799 that need only one temporary register, and VL offsets that require
3800 no temporary registers.
3801 (aarch64_conditional_register_usage): Mark the predicate registers
3802 as fixed if SVE isn't available.
3803 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
3804 Return true for SVE vector and predicate modes.
3805 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
3806 rather than an unsigned int. Handle SVE modes.
3807 (aarch64_preferred_simd_mode): Update call accordingly. Handle
3809 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
3811 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
3812 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
3813 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
3814 (aarch64_sve_float_mul_immediate_p): New functions.
3815 (aarch64_sve_valid_immediate): New function.
3816 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
3817 Explicitly reject structure modes. Check for INDEX constants.
3818 Handle PTRUE and PFALSE constants.
3819 (aarch64_check_zero_based_sve_index_immediate): New function.
3820 (aarch64_simd_imm_zero_p): Delete.
3821 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
3822 vector modes. Accept constants in the range of CNT[BHWD].
3823 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
3824 ask for an Advanced SIMD mode.
3825 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
3826 (aarch64_simd_vector_alignment): Handle SVE predicates.
3827 (aarch64_vectorize_preferred_vector_alignment): New function.
3828 (aarch64_simd_vector_alignment_reachable): Use it instead of
3830 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
3831 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
3833 (MAX_VECT_LEN): Delete.
3834 (expand_vec_perm_d): Add a vec_flags field.
3835 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
3836 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
3837 (aarch64_evpc_ext): Don't apply a big-endian lane correction
3839 (aarch64_evpc_rev): Rename to...
3840 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
3841 (aarch64_evpc_rev_global): New function.
3842 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
3843 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
3845 (aarch64_evpc_sve_tbl): New function.
3846 (aarch64_expand_vec_perm_const_1): Update after rename of
3847 aarch64_evpc_rev. Handle SVE permutes too, trying
3848 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
3849 than aarch64_evpc_tbl.
3850 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
3851 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
3852 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
3853 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
3854 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
3855 (aarch64_expand_sve_vcond): New functions.
3856 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
3857 of aarch64_vector_mode_p.
3858 (aarch64_dwarf_poly_indeterminate_value): New function.
3859 (aarch64_compute_pressure_classes): Likewise.
3860 (aarch64_can_change_mode_class): Likewise.
3861 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
3862 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
3863 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
3864 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
3865 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
3866 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
3867 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
3868 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
3870 (Dn, Dl, Dr): Accept const as well as const_vector.
3871 (Dz): Likewise. Compare against CONST0_RTX.
3872 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
3873 of "vector" where appropriate.
3874 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
3875 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
3876 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
3877 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
3878 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
3879 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
3880 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
3881 (v_int_equiv): Extend to SVE modes.
3882 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
3884 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
3885 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
3886 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
3887 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
3888 (SVE_COND_FP_CMP): New int iterators.
3889 (perm_hilo): Handle the new unpack unspecs.
3890 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
3892 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
3893 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
3894 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
3895 (aarch64_equality_operator, aarch64_constant_vector_operand)
3896 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
3897 (aarch64_sve_nonimmediate_operand): Likewise.
3898 (aarch64_sve_general_operand): Likewise.
3899 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
3900 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
3901 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
3902 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
3903 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
3904 (aarch64_sve_float_arith_immediate): Likewise.
3905 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
3906 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
3907 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
3908 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
3909 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
3910 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
3911 (aarch64_sve_float_arith_operand): Likewise.
3912 (aarch64_sve_float_arith_with_sub_operand): Likewise.
3913 (aarch64_sve_float_mul_operand): Likewise.
3914 (aarch64_sve_vec_perm_operand): Likewise.
3915 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
3916 (aarch64_mov_operand): Accept const_poly_int and const_vector.
3917 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
3918 as well as const_vector.
3919 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
3920 in file. Use CONST0_RTX and CONSTM1_RTX.
3921 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
3922 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
3923 Use aarch64_simd_imm_zero.
3924 * config/aarch64/aarch64-sve.md: New file.
3925 * config/aarch64/aarch64.md: Include it.
3926 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
3927 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
3928 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
3929 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
3930 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
3931 (sve): New attribute.
3932 (enabled): Disable instructions with the sve attribute unless
3934 (movqi, movhi): Pass CONST_POLY_INT operaneds through
3935 aarch64_expand_mov_immediate.
3936 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
3937 CNT[BHSD] immediates.
3938 (movti): Split CONST_POLY_INT moves into two halves.
3939 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
3940 Split additions that need a temporary here if the destination
3941 is the stack pointer.
3942 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
3943 (*add<mode>3_poly_1): New instruction.
3944 (set_clobber_cc): New expander.
3946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3948 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
3949 parameter and use it instead of GET_MODE_SIZE (innermode). Use
3950 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
3951 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
3952 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
3953 Change innermode from fixed_mode_size to machine_mode.
3954 (simplify_subreg): Update call accordingly. Handle a constant-sized
3955 subreg of a variable-length CONST_VECTOR.
3957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3958 Alan Hayward <alan.hayward@arm.com>
3959 David Sherwood <david.sherwood@arm.com>
3961 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
3962 (add_offset_to_base): New function, split out from...
3963 (create_mem_ref): ...here. When handling a scale other than 1,
3964 check first whether the address is valid without the offset.
3965 Add it into the base if so, leaving the index and scale as-is.
3967 2018-01-12 Jakub Jelinek <jakub@redhat.com>
3970 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
3971 fold_for_warn before checking if arg2 is INTEGER_CST.
3973 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
3975 * config/rs6000/predicates.md (load_multiple_operation): Delete.
3976 (store_multiple_operation): Delete.
3977 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
3978 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
3979 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
3980 guarded by TARGET_STRING.
3981 (rs6000_output_load_multiple): Delete.
3982 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
3983 OPTION_MASK_STRING / TARGET_STRING handling.
3984 (print_operand) <'N', 'O'>: Add comment that these are unused now.
3985 (const rs6000_opt_masks) <"string">: Change mask to 0.
3986 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
3987 (MASK_STRING): Delete.
3988 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
3990 (load_multiple): Delete.
3997 (store_multiple): Delete.
4004 (movmemsi_8reg): Delete.
4005 (corresponding unnamed define_insn): Delete.
4006 (movmemsi_6reg): Delete.
4007 (corresponding unnamed define_insn): Delete.
4008 (movmemsi_4reg): Delete.
4009 (corresponding unnamed define_insn): Delete.
4010 (movmemsi_2reg): Delete.
4011 (corresponding unnamed define_insn): Delete.
4012 (movmemsi_1reg): Delete.
4013 (corresponding unnamed define_insn): Delete.
4014 * config/rs6000/rs6000.opt (mno-string): New.
4015 (mstring): Replace by deprecation warning stub.
4016 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4018 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4020 * regrename.c (regrename_do_replace): If replacing the same
4021 reg multiple times, try to reuse last created gen_raw_REG.
4024 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4025 main to workaround a bug in GDB.
4027 2018-01-12 Tom de Vries <tom@codesourcery.com>
4030 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4032 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4034 PR rtl-optimization/80481
4035 * ira-color.c (get_cap_member): New function.
4036 (allocnos_conflict_by_live_ranges_p): Use it.
4037 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4038 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4040 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4043 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4044 (*saddl_se_1): Ditto.
4046 (*ssubl_se_1): Ditto.
4048 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4050 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4051 rather than wi::to_widest for DR_INITs.
4052 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4053 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4054 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4056 (vect_analyze_group_access_1): Note that here.
4058 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4060 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4061 polynomial type sizes.
4063 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4065 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4066 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4067 (gimple_add_tmp_var): Likewise.
4069 2018-01-12 Martin Liska <mliska@suse.cz>
4071 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4072 (gimple_alloc_sizes): Likewise.
4073 (dump_gimple_statistics): Use PRIu64 in printf format.
4074 * gimple.h: Change uint64_t to int.
4076 2018-01-12 Martin Liska <mliska@suse.cz>
4078 * tree-core.h: Use uint64_t instead of int.
4079 * tree.c (tree_node_counts): Likewise.
4080 (tree_node_sizes): Likewise.
4081 (dump_tree_statistics): Use PRIu64 in printf format.
4083 2018-01-12 Martin Liska <mliska@suse.cz>
4085 * Makefile.in: As qsort_chk is implemented in vec.c, add
4086 vec.o to linkage of gencfn-macros.
4087 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4088 passing the info to record_node_allocation_statistics.
4089 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4091 * ggc-common.c (struct ggc_usage): Add operator== and use
4092 it in operator< and compare function.
4093 * mem-stats.h (struct mem_usage): Likewise.
4094 * vec.c (struct vec_usage): Remove operator< and compare
4095 function. Can be simply inherited.
4097 2018-01-12 Martin Jambor <mjambor@suse.cz>
4100 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4101 * tree-ssa-math-opts.c: Include domwalk.h.
4102 (convert_mult_to_fma_1): New function.
4103 (fma_transformation_info): New type.
4104 (fma_deferring_state): Likewise.
4105 (cancel_fma_deferring): New function.
4106 (result_of_phi): Likewise.
4107 (last_fma_candidate_feeds_initial_phi): Likewise.
4108 (convert_mult_to_fma): Added deferring logic, split actual
4109 transformation to convert_mult_to_fma_1.
4110 (math_opts_dom_walker): New type.
4111 (math_opts_dom_walker::after_dom_children): New method, body moved
4112 here from pass_optimize_widening_mul::execute, added deferring logic
4114 (pass_optimize_widening_mul::execute): Moved most of code to
4115 math_opts_dom_walker::after_dom_children.
4116 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4117 * config/i386/i386.c (ix86_option_override_internal): Added
4118 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4120 2018-01-12 Richard Biener <rguenther@suse.de>
4123 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4124 inline instance vars.
4126 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4129 * config/rx/rx.c (rx_is_restricted_memory_address):
4132 2018-01-12 Richard Biener <rguenther@suse.de>
4134 PR tree-optimization/80846
4135 * target.def (split_reduction): New target hook.
4136 * targhooks.c (default_split_reduction): New function.
4137 * targhooks.h (default_split_reduction): Declare.
4138 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4139 target requests first reduce vectors by combining low and high
4141 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4142 (get_vectype_for_scalar_type_and_size): Export.
4143 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4144 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4145 * doc/tm.texi: Regenerate.
4146 * config/i386/i386.c (ix86_split_reduction): Implement
4147 TARGET_VECTORIZE_SPLIT_REDUCTION.
4149 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4152 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4153 in PIC mode except for TARGET_VXWORKS_RTP.
4154 * config/sparc/sparc.c: Include cfgrtl.h.
4155 (TARGET_INIT_PIC_REG): Define.
4156 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4157 (sparc_pic_register_p): New predicate.
4158 (sparc_legitimate_address_p): Use it.
4159 (sparc_legitimize_pic_address): Likewise.
4160 (sparc_delegitimize_address): Likewise.
4161 (sparc_mode_dependent_address_p): Likewise.
4162 (gen_load_pcrel_sym): Remove 4th parameter.
4163 (load_got_register): Adjust call to above. Remove obsolete stuff.
4164 (sparc_expand_prologue): Do not call load_got_register here.
4165 (sparc_flat_expand_prologue): Likewise.
4166 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4167 (sparc_use_pseudo_pic_reg): New function.
4168 (sparc_init_pic_reg): Likewise.
4169 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4170 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4172 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4174 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4175 Add item for branch_cost.
4177 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4179 PR rtl-optimization/83565
4180 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4181 not extend the result to a larger mode for rotate operations.
4182 (num_sign_bit_copies1): Likewise.
4184 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4187 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4189 Use values-Xc.o for -pedantic.
4190 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4192 2018-01-12 Martin Liska <mliska@suse.cz>
4195 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4197 (possible_polymorphic_call_targets): Use it.
4198 (ipa_devirt): Likewise.
4200 2018-01-12 Martin Liska <mliska@suse.cz>
4202 * profile-count.h (enum profile_quality): Use 0 as invalid
4203 enum value of profile_quality.
4205 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4207 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4208 -mext-string options.
4210 2018-01-12 Richard Biener <rguenther@suse.de>
4212 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4213 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4214 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4216 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4218 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4220 * configure.ac (--with-long-double-format): Add support for the
4221 configuration option to change the default long double format on
4223 * config.gcc (powerpc*-linux*-*): Likewise.
4224 * configure: Regenerate.
4225 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4226 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4227 used without modification.
4229 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4231 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4232 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4233 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4234 MISC_BUILTIN_SPEC_BARRIER.
4235 (rs6000_init_builtins): Likewise.
4236 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4238 (speculation_barrier): New define_insn.
4239 * doc/extend.texi: Document __builtin_speculation_barrier.
4241 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4244 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4245 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4246 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4248 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4249 integral modes instead of "ss" and "sd".
4250 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4251 vectors with 32-bit and 64-bit elements.
4252 (vecdupssescalarmodesuffix): New mode attribute.
4253 (vec_dup<mode>): Use it.
4255 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4258 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4259 frame if argument is passed on stack.
4261 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4264 * ree.c (combine_reaching_defs): Optimize also
4265 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4266 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4268 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4271 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4273 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4276 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4277 after they are computed.
4279 2018-01-11 Bin Cheng <bin.cheng@arm.com>
4281 PR tree-optimization/83695
4282 * gimple-loop-linterchange.cc
4283 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
4284 reset cached scev information after interchange.
4285 (pass_linterchange::execute): Remove call to scev_reset_htab.
4287 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4289 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
4290 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
4291 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
4292 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
4293 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
4294 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
4295 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
4296 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
4297 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
4298 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
4299 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
4300 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
4301 (V_lane_reg): Likewise.
4302 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
4304 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
4305 (vfmal_lane_low<mode>_intrinsic,
4306 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
4307 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
4308 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
4309 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
4310 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
4311 vfmsl_lane_high<mode>_intrinsic): New define_insns.
4313 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4315 * config/arm/arm-cpus.in (fp16fml): New feature.
4316 (ALL_SIMD): Add fp16fml.
4317 (armv8.2-a): Add fp16fml as an option.
4318 (armv8.3-a): Likewise.
4319 (armv8.4-a): Add fp16fml as part of fp16.
4320 * config/arm/arm.h (TARGET_FP16FML): Define.
4321 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
4323 * config/arm/arm-modes.def (V2HF): Define.
4324 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
4325 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
4326 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
4327 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
4328 vfmsl_low, vfmsl_high): New set of builtins.
4329 * config/arm/iterators.md (PLUSMINUS): New code iterator.
4330 (vfml_op): New code attribute.
4331 (VFMLHALVES): New int iterator.
4332 (VFML, VFMLSEL): New mode attributes.
4333 (V_reg): Define mapping for V2HF.
4334 (V_hi, V_lo): New mode attributes.
4335 (VF_constraint): Likewise.
4336 (vfml_half, vfml_half_selector): New int attributes.
4337 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
4339 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
4340 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
4342 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
4343 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
4344 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
4345 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
4347 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
4348 Document new effective target and option set.
4350 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4352 * config/arm/arm-cpus.in (armv8_4): New feature.
4353 (ARMv8_4a): New fgroup.
4354 (armv8.4-a): New arch.
4355 * config/arm/arm-tables.opt: Regenerate.
4356 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
4357 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
4358 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
4359 Add matching rules for -march=armv8.4-a and extensions.
4360 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
4362 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
4365 * config/rx/rx.md (BW): New mode attribute.
4366 (sync_lock_test_and_setsi): Add mode suffix to insn output.
4368 2018-01-11 Richard Biener <rguenther@suse.de>
4370 PR tree-optimization/83435
4371 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
4372 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
4373 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
4375 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4376 Alan Hayward <alan.hayward@arm.com>
4377 David Sherwood <david.sherwood@arm.com>
4379 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
4381 (aarch64_classify_address): Initialize it. Track polynomial offsets.
4382 (aarch64_print_address_internal): Use it to check for a zero offset.
4384 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4385 Alan Hayward <alan.hayward@arm.com>
4386 David Sherwood <david.sherwood@arm.com>
4388 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
4389 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
4390 Return a poly_int64 rather than a HOST_WIDE_INT.
4391 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
4392 rather than a HOST_WIDE_INT.
4393 * config/aarch64/aarch64.h (aarch64_frame): Protect with
4394 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
4395 hard_fp_offset, frame_size, initial_adjust, callee_offset and
4396 final_offset from HOST_WIDE_INT to poly_int64.
4397 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
4398 to_constant when getting the number of units in an Advanced SIMD
4400 (aarch64_builtin_vectorized_function): Check for a constant number
4402 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
4404 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
4405 attribute instead of GET_MODE_NUNITS.
4406 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
4407 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
4408 GET_MODE_SIZE for fixed-size registers.
4409 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
4410 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
4411 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
4412 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
4413 (aarch64_print_operand, aarch64_print_address_internal)
4414 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
4415 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
4416 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
4417 Handle polynomial GET_MODE_SIZE.
4418 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
4419 wider than SImode without modification.
4420 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
4421 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
4422 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
4423 passing and returning SVE modes.
4424 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
4425 rather than GEN_INT.
4426 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
4427 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
4428 (aarch64_allocate_and_probe_stack_space): Likewise.
4429 (aarch64_layout_frame): Cope with polynomial offsets.
4430 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
4431 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
4433 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
4434 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
4435 poly_int64 rather than a HOST_WIDE_INT.
4436 (aarch64_get_separate_components, aarch64_process_components)
4437 (aarch64_expand_prologue, aarch64_expand_epilogue)
4438 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
4439 (aarch64_anchor_offset): New function, split out from...
4440 (aarch64_legitimize_address): ...here.
4441 (aarch64_builtin_vectorization_cost): Handle polynomial
4442 TYPE_VECTOR_SUBPARTS.
4443 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
4445 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
4446 number of elements from the PARALLEL rather than the mode.
4447 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
4448 rather than GET_MODE_BITSIZE.
4449 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
4450 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
4451 (aarch64_expand_vec_perm_const_1): Handle polynomial
4452 d->perm.length () and d->perm elements.
4453 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
4454 Apply to_constant to d->perm elements.
4455 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
4456 polynomial CONST_VECTOR_NUNITS.
4457 (aarch64_move_pointer): Take amount as a poly_int64 rather
4459 (aarch64_progress_pointer): Avoid temporary variable.
4460 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
4461 the mode attribute instead of GET_MODE.
4463 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4464 Alan Hayward <alan.hayward@arm.com>
4465 David Sherwood <david.sherwood@arm.com>
4467 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
4468 x exists before using it.
4469 (aarch64_add_constant_internal): Rename to...
4470 (aarch64_add_offset_1): ...this. Replace regnum with separate
4471 src and dest rtxes. Handle the case in which they're different,
4472 including when the offset is zero. Replace scratchreg with an rtx.
4473 Use 2 additions if there is no spare register into which we can
4474 move a 16-bit constant.
4475 (aarch64_add_constant): Delete.
4476 (aarch64_add_offset): Replace reg with separate src and dest
4477 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
4478 Use aarch64_add_offset_1.
4479 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
4480 an rtx rather than an int. Take the delta as a poly_int64
4481 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
4482 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
4483 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
4484 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
4485 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
4487 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
4488 aarch64_add_constant.
4490 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4492 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
4493 Use scalar_float_mode.
4495 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4497 * config/aarch64/aarch64-simd.md
4498 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
4499 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
4500 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
4501 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
4502 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
4503 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
4504 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
4505 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
4506 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
4507 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
4509 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4512 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
4513 targ_options->x_arm_arch_string is non NULL.
4515 2018-01-11 Tamar Christina <tamar.christina@arm.com>
4517 * config/aarch64/aarch64.h
4518 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
4520 2018-01-11 Sudakshina Das <sudi.das@arm.com>
4523 * expmed.c (emit_store_flag_force): Swap if const op0
4524 and change VOIDmode to mode of op0.
4526 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4528 PR rtl-optimization/83761
4529 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
4530 than bytes to mode_for_size.
4532 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4535 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
4536 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
4539 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4542 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
4543 when in layout mode.
4544 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
4545 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
4548 2018-01-10 Michael Collison <michael.collison@arm.com>
4550 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
4551 * config/aarch64/aarch64-option-extension.def: Add
4552 AARCH64_OPT_EXTENSION of 'fp16fml'.
4553 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4554 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
4555 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
4556 * config/aarch64/constraints.md (Ui7): New constraint.
4557 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
4558 (VFMLA_SEL_W): Ditto.
4561 (VFMLA16_LOW): New int iterator.
4562 (VFMLA16_HIGH): Ditto.
4563 (UNSPEC_FMLAL): New unspec.
4564 (UNSPEC_FMLSL): Ditto.
4565 (UNSPEC_FMLAL2): Ditto.
4566 (UNSPEC_FMLSL2): Ditto.
4567 (f16mac): New code attribute.
4568 * config/aarch64/aarch64-simd-builtins.def
4569 (aarch64_fmlal_lowv2sf): Ditto.
4570 (aarch64_fmlsl_lowv2sf): Ditto.
4571 (aarch64_fmlalq_lowv4sf): Ditto.
4572 (aarch64_fmlslq_lowv4sf): Ditto.
4573 (aarch64_fmlal_highv2sf): Ditto.
4574 (aarch64_fmlsl_highv2sf): Ditto.
4575 (aarch64_fmlalq_highv4sf): Ditto.
4576 (aarch64_fmlslq_highv4sf): Ditto.
4577 (aarch64_fmlal_lane_lowv2sf): Ditto.
4578 (aarch64_fmlsl_lane_lowv2sf): Ditto.
4579 (aarch64_fmlal_laneq_lowv2sf): Ditto.
4580 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
4581 (aarch64_fmlalq_lane_lowv4sf): Ditto.
4582 (aarch64_fmlsl_lane_lowv4sf): Ditto.
4583 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
4584 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
4585 (aarch64_fmlal_lane_highv2sf): Ditto.
4586 (aarch64_fmlsl_lane_highv2sf): Ditto.
4587 (aarch64_fmlal_laneq_highv2sf): Ditto.
4588 (aarch64_fmlsl_laneq_highv2sf): Ditto.
4589 (aarch64_fmlalq_lane_highv4sf): Ditto.
4590 (aarch64_fmlsl_lane_highv4sf): Ditto.
4591 (aarch64_fmlalq_laneq_highv4sf): Ditto.
4592 (aarch64_fmlsl_laneq_highv4sf): Ditto.
4593 * config/aarch64/aarch64-simd.md:
4594 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
4595 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4596 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
4597 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
4598 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
4599 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
4600 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
4601 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
4602 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
4603 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
4604 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
4605 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
4606 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
4607 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
4608 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
4609 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
4610 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
4611 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
4612 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
4613 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
4614 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
4615 (vfmlsl_low_u32): Ditto.
4616 (vfmlalq_low_u32): Ditto.
4617 (vfmlslq_low_u32): Ditto.
4618 (vfmlal_high_u32): Ditto.
4619 (vfmlsl_high_u32): Ditto.
4620 (vfmlalq_high_u32): Ditto.
4621 (vfmlslq_high_u32): Ditto.
4622 (vfmlal_lane_low_u32): Ditto.
4623 (vfmlsl_lane_low_u32): Ditto.
4624 (vfmlal_laneq_low_u32): Ditto.
4625 (vfmlsl_laneq_low_u32): Ditto.
4626 (vfmlalq_lane_low_u32): Ditto.
4627 (vfmlslq_lane_low_u32): Ditto.
4628 (vfmlalq_laneq_low_u32): Ditto.
4629 (vfmlslq_laneq_low_u32): Ditto.
4630 (vfmlal_lane_high_u32): Ditto.
4631 (vfmlsl_lane_high_u32): Ditto.
4632 (vfmlal_laneq_high_u32): Ditto.
4633 (vfmlsl_laneq_high_u32): Ditto.
4634 (vfmlalq_lane_high_u32): Ditto.
4635 (vfmlslq_lane_high_u32): Ditto.
4636 (vfmlalq_laneq_high_u32): Ditto.
4637 (vfmlslq_laneq_high_u32): Ditto.
4638 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
4639 (AARCH64_FL_FOR_ARCH8_4): New.
4640 (AARCH64_ISA_F16FML): New ISA flag.
4641 (TARGET_F16FML): New feature flag for fp16fml.
4642 (doc/invoke.texi): Document new fp16fml option.
4644 2018-01-10 Michael Collison <michael.collison@arm.com>
4646 * config/aarch64/aarch64-builtins.c:
4647 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
4648 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4649 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
4650 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
4651 (AARCH64_ISA_SHA3): New ISA flag.
4652 (TARGET_SHA3): New feature flag for sha3.
4653 * config/aarch64/iterators.md (sha512_op): New int attribute.
4654 (CRYPTO_SHA512): New int iterator.
4655 (UNSPEC_SHA512H): New unspec.
4656 (UNSPEC_SHA512H2): Ditto.
4657 (UNSPEC_SHA512SU0): Ditto.
4658 (UNSPEC_SHA512SU1): Ditto.
4659 * config/aarch64/aarch64-simd-builtins.def
4660 (aarch64_crypto_sha512hqv2di): New builtin.
4661 (aarch64_crypto_sha512h2qv2di): Ditto.
4662 (aarch64_crypto_sha512su0qv2di): Ditto.
4663 (aarch64_crypto_sha512su1qv2di): Ditto.
4664 (aarch64_eor3qv8hi): Ditto.
4665 (aarch64_rax1qv2di): Ditto.
4666 (aarch64_xarqv2di): Ditto.
4667 (aarch64_bcaxqv8hi): Ditto.
4668 * config/aarch64/aarch64-simd.md:
4669 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
4670 (aarch64_crypto_sha512su0qv2di): Ditto.
4671 (aarch64_crypto_sha512su1qv2di): Ditto.
4672 (aarch64_eor3qv8hi): Ditto.
4673 (aarch64_rax1qv2di): Ditto.
4674 (aarch64_xarqv2di): Ditto.
4675 (aarch64_bcaxqv8hi): Ditto.
4676 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
4677 (vsha512h2q_u64): Ditto.
4678 (vsha512su0q_u64): Ditto.
4679 (vsha512su1q_u64): Ditto.
4680 (veor3q_u16): Ditto.
4681 (vrax1q_u64): Ditto.
4683 (vbcaxq_u16): Ditto.
4684 * config/arm/types.md (crypto_sha512): New type attribute.
4685 (crypto_sha3): Ditto.
4686 (doc/invoke.texi): Document new sha3 option.
4688 2018-01-10 Michael Collison <michael.collison@arm.com>
4690 * config/aarch64/aarch64-builtins.c:
4691 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
4692 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4693 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
4694 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
4695 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
4696 (AARCH64_ISA_SM4): New ISA flag.
4697 (TARGET_SM4): New feature flag for sm4.
4698 * config/aarch64/aarch64-simd-builtins.def
4699 (aarch64_sm3ss1qv4si): Ditto.
4700 (aarch64_sm3tt1aq4si): Ditto.
4701 (aarch64_sm3tt1bq4si): Ditto.
4702 (aarch64_sm3tt2aq4si): Ditto.
4703 (aarch64_sm3tt2bq4si): Ditto.
4704 (aarch64_sm3partw1qv4si): Ditto.
4705 (aarch64_sm3partw2qv4si): Ditto.
4706 (aarch64_sm4eqv4si): Ditto.
4707 (aarch64_sm4ekeyqv4si): Ditto.
4708 * config/aarch64/aarch64-simd.md:
4709 (aarch64_sm3ss1qv4si): Ditto.
4710 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
4711 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
4712 (aarch64_sm4eqv4si): Ditto.
4713 (aarch64_sm4ekeyqv4si): Ditto.
4714 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
4715 (sm3part_op): Ditto.
4716 (CRYPTO_SM3TT): Ditto.
4717 (CRYPTO_SM3PART): Ditto.
4718 (UNSPEC_SM3SS1): New unspec.
4719 (UNSPEC_SM3TT1A): Ditto.
4720 (UNSPEC_SM3TT1B): Ditto.
4721 (UNSPEC_SM3TT2A): Ditto.
4722 (UNSPEC_SM3TT2B): Ditto.
4723 (UNSPEC_SM3PARTW1): Ditto.
4724 (UNSPEC_SM3PARTW2): Ditto.
4725 (UNSPEC_SM4E): Ditto.
4726 (UNSPEC_SM4EKEY): Ditto.
4727 * config/aarch64/constraints.md (Ui2): New constraint.
4728 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
4729 * config/arm/types.md (crypto_sm3): New type attribute.
4730 (crypto_sm4): Ditto.
4731 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
4732 (vsm3tt1aq_u32): Ditto.
4733 (vsm3tt1bq_u32): Ditto.
4734 (vsm3tt2aq_u32): Ditto.
4735 (vsm3tt2bq_u32): Ditto.
4736 (vsm3partw1q_u32): Ditto.
4737 (vsm3partw2q_u32): Ditto.
4738 (vsm4eq_u32): Ditto.
4739 (vsm4ekeyq_u32): Ditto.
4740 (doc/invoke.texi): Document new sm4 option.
4742 2018-01-10 Michael Collison <michael.collison@arm.com>
4744 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
4745 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
4746 (AARCH64_FL_FOR_ARCH8_4): New.
4747 (AARCH64_FL_V8_4): New flag.
4748 (doc/invoke.texi): Document new armv8.4-a option.
4750 2018-01-10 Michael Collison <michael.collison@arm.com>
4752 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
4753 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
4754 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
4755 * config/aarch64/aarch64-option-extension.def: Add
4756 AARCH64_OPT_EXTENSION of 'sha2'.
4757 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
4758 (crypto): Disable sha2 and aes if crypto disabled.
4759 (crypto): Enable aes and sha2 if enabled.
4760 (simd): Disable sha2 and aes if simd disabled.
4761 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
4763 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
4764 (TARGET_SHA2): New feature flag for sha2.
4765 (TARGET_AES): New feature flag for aes.
4766 * config/aarch64/aarch64-simd.md:
4767 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
4768 conditional on TARGET_AES.
4769 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
4770 (aarch64_crypto_sha1hsi): Make pattern conditional
4772 (aarch64_crypto_sha1hv4si): Ditto.
4773 (aarch64_be_crypto_sha1hv4si): Ditto.
4774 (aarch64_crypto_sha1su1v4si): Ditto.
4775 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
4776 (aarch64_crypto_sha1su0v4si): Ditto.
4777 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
4778 (aarch64_crypto_sha256su0v4si): Ditto.
4779 (aarch64_crypto_sha256su1v4si): Ditto.
4780 (doc/invoke.texi): Document new aes and sha2 options.
4782 2018-01-10 Martin Sebor <msebor@redhat.com>
4784 PR tree-optimization/83781
4785 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
4788 2018-01-11 Martin Sebor <msebor@gmail.com>
4789 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4791 PR tree-optimization/83501
4792 PR tree-optimization/81703
4794 * tree-ssa-strlen.c (get_string_cst): Rename...
4795 (get_string_len): ...to this. Handle global constants.
4796 (handle_char_store): Adjust.
4798 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
4799 Jim Wilson <jimw@sifive.com>
4801 * config/riscv/riscv-protos.h (riscv_output_return): New.
4802 * config/riscv/riscv.c (struct machine_function): New naked_p field.
4803 (riscv_attribute_table, riscv_output_return),
4804 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
4805 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
4806 (riscv_compute_frame_info): Only compute frame->mask if not a naked
4808 (riscv_expand_prologue): Add early return for naked function.
4809 (riscv_expand_epilogue): Likewise.
4810 (riscv_function_ok_for_sibcall): Return false for naked function.
4811 (riscv_set_current_function): New.
4812 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
4813 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
4814 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
4815 * doc/extend.texi (RISC-V Function Attributes): New.
4817 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4819 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
4820 check for 128-bit long double before checking TCmode.
4821 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
4822 128-bit long doubles before checking TFmode or TCmode.
4823 (FLOAT128_IBM_P): Likewise.
4825 2018-01-10 Martin Sebor <msebor@redhat.com>
4827 PR tree-optimization/83671
4828 * builtins.c (c_strlen): Unconditionally return zero for the empty
4830 Use -Warray-bounds for warnings.
4831 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
4832 for non-constant array indices with COMPONENT_REF, arrays of
4833 arrays, and pointers to arrays.
4834 (gimple_fold_builtin_strlen): Determine and set length range for
4835 non-constant character arrays.
4837 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
4840 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
4843 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
4845 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
4847 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
4850 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
4851 VECTOR_MEM_ALTIVEC_OR_VSX_P.
4852 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
4853 indexed_or_indirect_operand predicate.
4854 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
4855 (*vsx_le_perm_load_v8hi): Likewise.
4856 (*vsx_le_perm_load_v16qi): Likewise.
4857 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
4858 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
4859 (*vsx_le_perm_store_v8hi): Likewise.
4860 (*vsx_le_perm_store_v16qi): Likewise.
4861 (eight unnamed splitters): Likewise.
4863 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
4865 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
4866 * config/rs6000/emmintrin.h: Likewise.
4867 * config/rs6000/mmintrin.h: Likewise.
4868 * config/rs6000/xmmintrin.h: Likewise.
4870 2018-01-10 David Malcolm <dmalcolm@redhat.com>
4873 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
4875 * tree.c (tree_nop_conversion): Return true for location wrapper
4877 (maybe_wrap_with_location): New function.
4878 (selftest::check_strip_nops): New function.
4879 (selftest::test_location_wrappers): New function.
4880 (selftest::tree_c_tests): Call it.
4881 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
4882 (maybe_wrap_with_location): New decl.
4883 (EXPR_LOCATION_WRAPPER_P): New macro.
4884 (location_wrapper_p): New inline function.
4885 (tree_strip_any_location_wrapper): New inline function.
4887 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
4890 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
4891 stack_realign_offset for the largest alignment of stack slot
4893 (ix86_find_max_used_stack_alignment): New function.
4894 (ix86_finalize_stack_frame_flags): Use it. Set
4895 max_used_stack_alignment if we don't realign stack.
4896 * config/i386/i386.h (machine_function): Add
4897 max_used_stack_alignment.
4899 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
4901 * config/arm/arm.opt (-mbranch-cost): New option.
4902 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
4905 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
4908 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
4909 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
4911 2018-01-10 Richard Biener <rguenther@suse.de>
4914 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
4915 early out so it also covers the case where we have a non-NULL
4918 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
4920 PR tree-optimization/83753
4921 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
4922 for non-strided grouped accesses if the number of elements is 1.
4924 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
4927 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
4928 * i386.h (TARGET_USE_GATHER): Define.
4929 * x86-tune.def (X86_TUNE_USE_GATHER): New.
4931 2018-01-10 Martin Liska <mliska@suse.cz>
4934 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
4935 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
4937 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
4938 CLEANUP_NO_PARTITIONING is not set.
4940 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
4942 * doc/rtl.texi: Remove documentation of (const ...) wrappers
4943 for vectors, as a partial revert of r254296.
4944 * rtl.h (const_vec_p): Delete.
4945 (const_vec_duplicate_p): Don't test for vector CONSTs.
4946 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
4947 * expmed.c (make_tree): Likewise.
4950 * common.md (E, F): Use CONSTANT_P instead of checking for
4952 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
4953 checking for CONST_VECTOR.
4955 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
4958 * predict.c (force_edge_cold): Handle in more sane way edges
4961 2018-01-09 Carl Love <cel@us.ibm.com>
4963 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
4965 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
4966 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
4967 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
4968 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
4969 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
4970 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
4971 * config/rs6000/rs6000-protos.h: Add extern defition for
4972 rs6000_generate_float2_double_code.
4973 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
4975 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
4976 (float2_v2df): Add define_expand.
4978 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
4981 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
4982 op_mode in the force_to_mode call.
4984 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
4986 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
4987 instead of checking each element individually.
4988 (aarch64_evpc_uzp): Likewise.
4989 (aarch64_evpc_zip): Likewise.
4990 (aarch64_evpc_ext): Likewise.
4991 (aarch64_evpc_rev): Likewise.
4992 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
4993 instead of checking each element individually. Return true without
4995 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
4996 whether all selected elements come from the same input, instead of
4997 checking each element individually. Remove calls to gen_rtx_REG,
4998 start_sequence and end_sequence and instead assert that no rtl is
5001 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5003 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5004 order of HIGH and CONST checks.
5006 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5008 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5009 if the destination isn't an SSA_NAME.
5011 2018-01-09 Richard Biener <rguenther@suse.de>
5013 PR tree-optimization/83668
5014 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5016 (canonicalize_loop_form): ... here, renamed from ...
5017 (canonicalize_loop_closed_ssa_form): ... this and amended to
5018 swap successor edges for loop exit blocks to make us use
5019 the RPO order we need for initial schedule generation.
5021 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5023 PR tree-optimization/64811
5024 * match.pd: When optimizing comparisons with Inf, avoid
5025 introducing or losing exceptions from comparisons with NaN.
5027 2018-01-09 Martin Liska <mliska@suse.cz>
5030 * asan.c (shadow_mem_size): Add gcc_assert.
5032 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5034 Don't save registers in main().
5037 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5038 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5039 * config/avr/avr.c (avr_set_current_function): Don't error if
5040 naked, OS_task or OS_main are specified at the same time.
5041 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5043 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5045 * common/config/avr/avr-common.c (avr_option_optimization_table):
5046 Switch on -mmain-is-OS_task for optimizing compilations.
5048 2018-01-09 Richard Biener <rguenther@suse.de>
5050 PR tree-optimization/83572
5051 * graphite.c: Include cfganal.h.
5052 (graphite_transform_loops): Connect infinite loops to exit
5053 and remove fake edges at the end.
5055 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5057 * ipa-inline.c (edge_badness): Revert accidental checkin.
5059 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5062 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5063 symbols; not inline clones.
5065 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5068 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5069 hard registers. Formatting fixes.
5071 PR preprocessor/83722
5072 * gcc.c (try_generate_repro): Pass
5073 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5074 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5077 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5078 Kito Cheng <kito.cheng@gmail.com>
5080 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5081 (riscv_leaf_function_p): Delete.
5082 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5084 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5086 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5088 (do_ifelse): New function.
5089 (do_isel): New function.
5090 (do_sub3): New function.
5091 (do_add3): New function.
5092 (do_load_mask_compare): New function.
5093 (do_overlap_load_compare): New function.
5094 (expand_compare_loop): New function.
5095 (expand_block_compare): Call expand_compare_loop() when appropriate.
5096 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5098 (-mblock-compare-inline-loop-limit): New option.
5100 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5103 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5104 Reverse order of second and third operands in first alternative.
5105 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5106 of first and second elements in UNSPEC_VPERMR vector.
5107 (altivec_expand_vec_perm_le): Likewise.
5109 2018-01-08 Jeff Law <law@redhat.com>
5111 PR rtl-optimizatin/81308
5112 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5113 (process_switch): If group_case_labels makes a change, then set
5115 (pass_convert_switch::execute): If a switch is converted, then
5116 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5118 PR rtl-optimization/81308
5119 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5122 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5124 PR target/83663 - Revert r255946
5125 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5126 generation for cases where splatting a value is not useful.
5127 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5128 across a vec_duplicate and a paradoxical subreg forming a vector
5129 mode to a vec_concat.
5131 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5133 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5134 -march=armv8.3-a variants.
5135 * config/arm/t-multilib: Likewise.
5136 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5138 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5140 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5142 (cceq_ior_compare_complement): Give it a name so I can use it, and
5143 change boolean_or_operator predicate to boolean_operator so it can
5144 be used to generate a crand.
5145 (eqne): New code iterator.
5146 (bd/bd_neg): New code_attrs.
5147 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5148 a single define_insn.
5149 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5150 decrement (bdnzt/bdnzf/bdzt/bdzf).
5151 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5152 with the new names of the branch decrement patterns, and added the
5153 names of the branch decrement conditional patterns.
5155 2018-01-08 Richard Biener <rguenther@suse.de>
5157 PR tree-optimization/83563
5158 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5161 2018-01-08 Richard Biener <rguenther@suse.de>
5164 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5166 2018-01-08 Richard Biener <rguenther@suse.de>
5168 PR tree-optimization/83685
5169 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5170 references to abnormals.
5172 2018-01-08 Richard Biener <rguenther@suse.de>
5175 * dwarf2out.c (output_indirect_strings): Handle empty
5176 skeleton_debug_str_hash.
5177 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5179 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5181 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5182 (emit_store_direct): Likewise.
5183 (arc_trampoline_adjust_address): Likewise.
5184 (arc_asm_trampoline_template): New function.
5185 (arc_initialize_trampoline): Use asm_trampoline_template.
5186 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5187 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5188 * config/arc/arc.md (flush_icache): Delete pattern.
5190 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5192 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5193 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5196 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5199 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5200 by not USED_FOR_TARGET.
5201 (make_pass_resolve_sw_modes): Likewise.
5203 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5205 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5208 2018-01-08 Richard Biener <rguenther@suse.de>
5211 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5213 2018-01-08 Richard Biener <rguenther@suse.de>
5216 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5218 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5221 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5222 basic blocks with a small number of successors.
5223 (convert_control_dep_chain_into_preds): Improve handling of
5225 (dump_predicates): Split apart into...
5226 (dump_pred_chain): ...here...
5227 (dump_pred_info): ...and here.
5228 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5229 (can_chain_union_be_invalidated_p): Improve check for invalidation
5231 (uninit_uses_cannot_happen): Avoid unnecessary if
5232 convert_control_dep_chain_into_preds yielded nothing.
5234 2018-01-06 Martin Sebor <msebor@redhat.com>
5236 PR tree-optimization/83640
5237 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5238 subtracting negative offset from size.
5239 (builtin_access::overlap): Adjust offset bounds of the access to fall
5240 within the size of the object if possible.
5242 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5244 PR rtl-optimization/83699
5245 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5246 extract_bit_field_as_subreg to cases in which the extracted
5247 value is also a vector.
5249 * lra-constraints.c (process_alt_operands): Test for the equivalence
5250 substitutions when detecting a possible reload cycle.
5252 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5255 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5256 by default if flag_selective_schedling{,2}. Formatting fixes.
5258 PR rtl-optimization/83682
5259 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5260 if it has non-VECTOR_MODE element mode.
5261 (vec_duplicate_p): Likewise.
5264 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5265 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5267 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5270 * config/i386/i386-builtin.def
5271 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5272 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5273 Require also OPTION_MASK_ISA_AVX512F in addition to
5274 OPTION_MASK_ISA_GFNI.
5275 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5276 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5277 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5278 to OPTION_MASK_ISA_GFNI.
5279 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
5280 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
5281 OPTION_MASK_ISA_AVX512BW.
5282 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
5283 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
5284 addition to OPTION_MASK_ISA_GFNI.
5285 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
5286 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
5287 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
5288 to OPTION_MASK_ISA_GFNI.
5289 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
5290 a requirement for all ISAs rather than any of them with a few
5292 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
5294 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
5295 bitmasks to be enabled with 3 exceptions, instead of requiring any
5296 enabled ISA with lots of exceptions.
5297 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
5298 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
5299 Change avx512bw in isa attribute to avx512f.
5300 * config/i386/sgxintrin.h: Add license boilerplate.
5301 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
5302 to __AVX512F__ and __AVX512VL to __AVX512VL__.
5303 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
5304 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
5306 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
5307 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
5308 temporarily sse2 rather than sse if not enabled already.
5311 * config/i386/sse.md (VI248_VLBW): Rename to ...
5312 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
5313 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
5314 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
5315 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
5316 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
5317 mode iterator instead of VI248_VLBW.
5319 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
5321 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
5322 (record_modified): Skip clobbers; add debug output.
5323 (param_change_prob): Use sreal frequencies.
5325 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5327 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
5328 punt for user-aligned variables.
5330 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5332 * tree-chrec.c (chrec_contains_symbols): Return true for
5335 2018-01-05 Sudakshina Das <sudi.das@arm.com>
5338 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
5339 of (x|y) == x for BICS pattern.
5341 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5343 PR tree-optimization/83605
5344 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
5345 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
5348 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
5350 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
5351 * config/epiphany/rtems.h: New file.
5353 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5354 Uros Bizjak <ubizjak@gmail.com>
5357 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
5358 QIreg_operand instead of register_operand predicate.
5359 * config/i386/i386.c (ix86_rop_should_change_byte_p,
5360 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
5361 comments instead of -fmitigate[-_]rop.
5363 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5366 * cgraphunit.c (symbol_table::compile): Switch to text_section
5367 before calling assembly_start debug hook.
5368 * run-rtl-passes.c (run_rtl_passes): Likewise.
5371 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5373 * tree-vrp.c (extract_range_from_binary_expr_1): Check
5374 range_int_cst_p rather than !symbolic_range_p before calling
5375 extract_range_from_multiplicative_op_1.
5377 2018-01-04 Jeff Law <law@redhat.com>
5379 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
5380 redundant test in assertion.
5382 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5384 * doc/rtl.texi: Document machine_mode wrapper classes.
5386 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5388 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
5391 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5393 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
5394 the VEC_PERM_EXPR fold to fail.
5396 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5399 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
5400 to switched_sections.
5402 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5405 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
5408 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
5411 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
5412 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
5414 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5417 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
5418 is BLKmode and bitpos not zero or mode change is needed.
5420 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5423 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
5426 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
5429 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
5430 instead of MULT rtx. Update all corresponding splitters.
5432 (*ssub<modesuffix>): Ditto.
5434 (*cmp_sadd_di): Update split patterns.
5435 (*cmp_sadd_si): Ditto.
5436 (*cmp_sadd_sidi): Ditto.
5437 (*cmp_ssub_di): Ditto.
5438 (*cmp_ssub_si): Ditto.
5439 (*cmp_ssub_sidi): Ditto.
5440 * config/alpha/predicates.md (const23_operand): New predicate.
5441 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
5442 Look for ASHIFT, not MULT inner operand.
5443 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
5445 2018-01-04 Martin Liska <mliska@suse.cz>
5447 PR gcov-profile/83669
5448 * gcov.c (output_intermediate_file): Add version to intermediate
5450 * doc/gcov.texi: Document new field 'version' in intermediate
5451 file format. Fix location of '-k' option of gcov command.
5453 2018-01-04 Martin Liska <mliska@suse.cz>
5456 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
5458 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5460 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
5462 2018-01-03 Martin Sebor <msebor@redhat.com>
5464 PR tree-optimization/83655
5465 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
5466 checking calls with invalid arguments.
5468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5470 * tree-vect-stmts.c (vect_get_store_rhs): New function.
5471 (vectorizable_mask_load_store): Delete.
5472 (vectorizable_call): Return false for masked loads and stores.
5473 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
5474 instead of gimple_assign_rhs1.
5475 (vectorizable_load): Handle IFN_MASK_LOAD.
5476 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
5478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5480 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
5482 (vectorizable_mask_load_store): ...here.
5483 (vectorizable_load): ...and here.
5485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5487 * tree-vect-stmts.c (vect_build_all_ones_mask)
5488 (vect_build_zero_merge_argument): New functions, split out from...
5489 (vectorizable_load): ...here.
5491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5493 * tree-vect-stmts.c (vect_check_store_rhs): New function,
5495 (vectorizable_mask_load_store): ...here.
5496 (vectorizable_store): ...and here.
5498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5500 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
5502 (vectorizable_mask_load_store): ...here.
5504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5506 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
5507 (vect_model_store_cost): Take a vec_load_store_type instead of a
5509 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
5510 (vect_model_store_cost): Take a vec_load_store_type instead of a
5512 (vectorizable_mask_load_store): Update accordingly.
5513 (vectorizable_store): Likewise.
5514 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
5516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5518 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
5519 IFN_MASK_LOAD calls here rather than...
5520 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
5522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5523 Alan Hayward <alan.hayward@arm.com>
5524 David Sherwood <david.sherwood@arm.com>
5526 * expmed.c (extract_bit_field_1): For vector extracts,
5527 fall back to extract_bit_field_as_subreg if vec_extract
5530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5531 Alan Hayward <alan.hayward@arm.com>
5532 David Sherwood <david.sherwood@arm.com>
5534 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
5535 they are variable or constant sized.
5536 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
5537 slots for constant-sized data.
5539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5540 Alan Hayward <alan.hayward@arm.com>
5541 David Sherwood <david.sherwood@arm.com>
5543 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
5544 handling COND_EXPRs with boolean comparisons, try to find a better
5545 basis for the mask type than the boolean itself.
5547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5549 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
5550 is calculated and how it can be overridden.
5551 * genmodes.c (max_bitsize_mode_any_mode): New variable.
5552 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
5554 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
5557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5558 Alan Hayward <alan.hayward@arm.com>
5559 David Sherwood <david.sherwood@arm.com>
5561 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
5562 Remove the mode argument.
5563 (aarch64_simd_valid_immediate): Remove the mode and inverse
5565 * config/aarch64/iterators.md (bitsize): New iterator.
5566 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
5567 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
5568 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
5569 aarch64_simd_valid_immediate.
5570 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
5571 (aarch64_reg_or_bic_imm): Likewise.
5572 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
5573 with an insn_type enum and msl with a modifier_type enum.
5574 Replace element_width with a scalar_mode. Change the shift
5575 to unsigned int. Add constructors for scalar_float_mode and
5576 scalar_int_mode elements.
5577 (aarch64_vect_float_const_representable_p): Delete.
5578 (aarch64_can_const_movi_rtx_p)
5579 (aarch64_simd_scalar_immediate_valid_for_move)
5580 (aarch64_simd_make_constant): Update call to
5581 aarch64_simd_valid_immediate.
5582 (aarch64_advsimd_valid_immediate_hs): New function.
5583 (aarch64_advsimd_valid_immediate): Likewise.
5584 (aarch64_simd_valid_immediate): Remove mode and inverse
5585 arguments. Rewrite to use the above. Use const_vec_duplicate_p
5586 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
5587 and aarch64_float_const_representable_p on the result.
5588 (aarch64_output_simd_mov_immediate): Remove mode argument.
5589 Update call to aarch64_simd_valid_immediate and use of
5590 simd_immediate_info.
5591 (aarch64_output_scalar_simd_mov_immediate): Update call
5594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5595 Alan Hayward <alan.hayward@arm.com>
5596 David Sherwood <david.sherwood@arm.com>
5598 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
5599 (mode_nunits): Likewise CONST_MODE_NUNITS.
5600 * machmode.def (ADJUST_NUNITS): Document.
5601 * genmodes.c (mode_data::need_nunits_adj): New field.
5602 (blank_mode): Update accordingly.
5603 (adj_nunits): New variable.
5604 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
5606 (emit_mode_size_inline): Set need_bytesize_adj for all modes
5607 listed in adj_nunits.
5608 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
5609 listed in adj_nunits. Don't emit case statements for such modes.
5610 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
5611 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
5612 nothing if adj_nunits is nonnull.
5613 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
5614 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
5615 (emit_mode_fbit): Update use of print_maybe_const_decl.
5616 (emit_move_size): Likewise. Treat the array as non-const
5618 (emit_mode_adjustments): Handle adj_nunits.
5620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5622 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
5623 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
5624 (VECTOR_MODES): Use it.
5625 (make_vector_modes): Take the prefix as an argument.
5627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5628 Alan Hayward <alan.hayward@arm.com>
5629 David Sherwood <david.sherwood@arm.com>
5631 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
5632 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
5633 for MODE_VECTOR_BOOL.
5634 * machmode.def (VECTOR_BOOL_MODE): Document.
5635 * genmodes.c (VECTOR_BOOL_MODE): New macro.
5636 (make_vector_bool_mode): New function.
5637 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
5639 * lto-streamer-in.c (lto_input_mode_table): Likewise.
5640 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
5642 * stor-layout.c (int_mode_for_mode): Likewise.
5643 * tree.c (build_vector_type_for_mode): Likewise.
5644 * varasm.c (output_constant_pool_2): Likewise.
5645 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
5646 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
5647 for MODE_VECTOR_BOOL.
5648 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
5649 of mode class checks.
5650 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
5651 instead of a list of mode class checks.
5652 (expand_vector_scalar_condition): Likewise.
5653 (type_for_widest_vector_mode): Handle BImode as an inner mode.
5655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5656 Alan Hayward <alan.hayward@arm.com>
5657 David Sherwood <david.sherwood@arm.com>
5659 * machmode.h (mode_size): Change from unsigned short to
5661 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
5662 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
5663 or if measurement_type is not polynomial.
5664 (fixed_size_mode::includes_p): Check for constant-sized modes.
5665 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
5666 return a poly_uint16 rather than an unsigned short.
5667 (emit_mode_size): Change the type of mode_size from unsigned short
5668 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
5669 (emit_mode_adjustments): Cope with polynomial vector sizes.
5670 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5672 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5674 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
5675 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
5676 * caller-save.c (setup_save_areas): Likewise.
5677 (replace_reg_with_saved_mem): Likewise.
5678 * calls.c (emit_library_call_value_1): Likewise.
5679 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
5680 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
5681 (gen_lowpart_for_combine): Likewise.
5682 * convert.c (convert_to_integer_1): Likewise.
5683 * cse.c (equiv_constant, cse_insn): Likewise.
5684 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
5685 (cselib_subst_to_values): Likewise.
5686 * dce.c (word_dce_process_block): Likewise.
5687 * df-problems.c (df_word_lr_mark_ref): Likewise.
5688 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
5689 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
5690 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
5691 (rtl_for_decl_location): Likewise.
5692 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
5693 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
5694 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
5695 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
5696 (expand_expr_real_1): Likewise.
5697 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
5698 (pad_below): Likewise.
5699 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
5700 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
5701 * ira.c (get_subreg_tracking_sizes): Likewise.
5702 * ira-build.c (ira_create_allocno_objects): Likewise.
5703 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
5704 (ira_sort_regnos_for_alter_reg): Likewise.
5705 * ira-costs.c (record_operand_costs): Likewise.
5706 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
5707 (resolve_simple_move): Likewise.
5708 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
5709 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
5710 (lra_constraints): Likewise.
5711 (CONST_POOL_OK_P): Reject variable-sized modes.
5712 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
5713 (add_pseudo_to_slot, lra_spill): Likewise.
5714 * omp-low.c (omp_clause_aligned_alignment): Likewise.
5715 * optabs-query.c (get_best_extraction_insn): Likewise.
5716 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
5717 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
5718 (expand_mult_highpart, valid_multiword_target_p): Likewise.
5719 * recog.c (offsettable_address_addr_space_p): Likewise.
5720 * regcprop.c (maybe_mode_change): Likewise.
5721 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
5722 * regrename.c (build_def_use): Likewise.
5723 * regstat.c (dump_reg_info): Likewise.
5724 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
5725 (find_reloads, find_reloads_subreg_address): Likewise.
5726 * reload1.c (eliminate_regs_1): Likewise.
5727 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
5728 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
5729 (simplify_binary_operation_1, simplify_subreg): Likewise.
5730 * targhooks.c (default_function_arg_padding): Likewise.
5731 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
5732 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
5733 (verify_gimple_assign_ternary): Likewise.
5734 * tree-inline.c (estimate_move_cost): Likewise.
5735 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
5736 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
5737 (get_address_cost_ainc): Likewise.
5738 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
5739 (vect_supportable_dr_alignment): Likewise.
5740 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5741 (vectorizable_reduction): Likewise.
5742 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
5743 (vectorizable_operation, vectorizable_load): Likewise.
5744 * tree.c (build_same_sized_truth_vector_type): Likewise.
5745 * valtrack.c (cleanup_auto_inc_dec): Likewise.
5746 * var-tracking.c (emit_note_insn_var_location): Likewise.
5747 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
5748 (ADDR_VEC_ALIGN): Likewise.
5750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5751 Alan Hayward <alan.hayward@arm.com>
5752 David Sherwood <david.sherwood@arm.com>
5754 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
5756 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
5757 or if measurement_type is polynomial.
5758 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
5759 * combine.c (make_extraction): Likewise.
5760 * dse.c (find_shift_sequence): Likewise.
5761 * dwarf2out.c (mem_loc_descriptor): Likewise.
5762 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
5763 (extract_bit_field, extract_low_bits): Likewise.
5764 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
5765 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
5766 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
5767 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
5768 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
5769 * reload.c (find_reloads): Likewise.
5770 * reload1.c (alter_reg): Likewise.
5771 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
5772 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
5773 * tree-if-conv.c (predicate_mem_writes): Likewise.
5774 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
5775 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
5776 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
5777 * valtrack.c (dead_debug_insert_temp): Likewise.
5778 * varasm.c (mergeable_constant_section): Likewise.
5779 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
5781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5782 Alan Hayward <alan.hayward@arm.com>
5783 David Sherwood <david.sherwood@arm.com>
5785 * expr.c (expand_assignment): Cope with polynomial mode sizes
5786 when assigning to a CONCAT.
5788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5789 Alan Hayward <alan.hayward@arm.com>
5790 David Sherwood <david.sherwood@arm.com>
5792 * machmode.h (mode_precision): Change from unsigned short to
5794 (mode_to_precision): Return a poly_uint16 rather than an unsigned
5796 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
5797 or if measurement_type is not polynomial.
5798 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
5799 in which the mode is already known to be a scalar_int_mode.
5800 * genmodes.c (emit_mode_precision): Change the type of mode_precision
5801 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
5803 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5804 for GET_MODE_PRECISION.
5805 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5806 for GET_MODE_PRECISION.
5807 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
5809 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
5810 (expand_field_assignment, make_extraction): Likewise.
5811 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
5812 (get_last_value): Likewise.
5813 * convert.c (convert_to_integer_1): Likewise.
5814 * cse.c (cse_insn): Likewise.
5815 * expr.c (expand_expr_real_1): Likewise.
5816 * lra-constraints.c (simplify_operand_subreg): Likewise.
5817 * optabs-query.c (can_atomic_load_p): Likewise.
5818 * optabs.c (expand_atomic_load): Likewise.
5819 (expand_atomic_store): Likewise.
5820 * ree.c (combine_reaching_defs): Likewise.
5821 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
5822 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
5823 * tree.h (type_has_mode_precision_p): Likewise.
5824 * ubsan.c (instrument_si_overflow): Likewise.
5826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5827 Alan Hayward <alan.hayward@arm.com>
5828 David Sherwood <david.sherwood@arm.com>
5830 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
5831 polynomial numbers of units.
5832 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
5833 (valid_vector_subparts_p): New function.
5834 (build_vector_type): Remove temporary shim and take the number
5835 of units as a poly_uint64 rather than an int.
5836 (build_opaque_vector_type): Take the number of units as a
5837 poly_uint64 rather than an int.
5838 * tree.c (build_vector_from_ctor): Handle polynomial
5839 TYPE_VECTOR_SUBPARTS.
5840 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
5841 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
5842 (build_vector_from_val): If the number of units is variable,
5843 use build_vec_duplicate_cst for constant operands and
5844 VEC_DUPLICATE_EXPR otherwise.
5845 (make_vector_type): Remove temporary is_constant ().
5846 (build_vector_type, build_opaque_vector_type): Take the number of
5847 units as a poly_uint64 rather than an int.
5848 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
5850 * cfgexpand.c (expand_debug_expr): Likewise.
5851 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
5852 (store_constructor, expand_expr_real_1): Likewise.
5853 (const_scalar_mask_from_tree): Likewise.
5854 * fold-const-call.c (fold_const_reduction): Likewise.
5855 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
5856 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
5857 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
5858 (fold_relational_const): Likewise.
5859 (native_interpret_vector): Likewise. Change the size from an
5860 int to an unsigned int.
5861 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
5862 TYPE_VECTOR_SUBPARTS.
5863 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
5864 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
5865 duplicating a non-constant operand into a variable-length vector.
5866 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
5867 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
5868 * ipa-icf.c (sem_variable::equals): Likewise.
5869 * match.pd: Likewise.
5870 * omp-simd-clone.c (simd_clone_subparts): Likewise.
5871 * print-tree.c (print_node): Likewise.
5872 * stor-layout.c (layout_type): Likewise.
5873 * targhooks.c (default_builtin_vectorization_cost): Likewise.
5874 * tree-cfg.c (verify_gimple_comparison): Likewise.
5875 (verify_gimple_assign_binary): Likewise.
5876 (verify_gimple_assign_ternary): Likewise.
5877 (verify_gimple_assign_single): Likewise.
5878 * tree-pretty-print.c (dump_generic_node): Likewise.
5879 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
5880 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
5881 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
5882 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
5883 (vect_shift_permute_load_chain): Likewise.
5884 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
5885 (expand_vector_condition, optimize_vector_constructor): Likewise.
5886 (lower_vec_perm, get_compute_type): Likewise.
5887 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5888 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
5889 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
5890 (vect_recog_mask_conversion_pattern): Likewise.
5891 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
5892 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
5893 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
5894 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
5895 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
5896 (vectorizable_shift, vectorizable_operation, vectorizable_store)
5897 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
5898 (supportable_widening_operation): Likewise.
5899 (supportable_narrowing_operation): Likewise.
5900 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
5902 * varasm.c (output_constant): Likewise.
5904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5905 Alan Hayward <alan.hayward@arm.com>
5906 David Sherwood <david.sherwood@arm.com>
5908 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
5909 so that both the length == 3 and length != 3 cases set up their
5910 own permute vectors. Add comments explaining why we know the
5911 number of elements is constant.
5912 (vect_permute_load_chain): Likewise.
5914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5915 Alan Hayward <alan.hayward@arm.com>
5916 David Sherwood <david.sherwood@arm.com>
5918 * machmode.h (mode_nunits): Change from unsigned char to
5920 (ONLY_FIXED_SIZE_MODES): New macro.
5921 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
5922 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
5923 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
5925 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
5926 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
5927 or if measurement_type is not polynomial.
5928 * genmodes.c (ZERO_COEFFS): New macro.
5929 (emit_mode_nunits_inline): Make mode_nunits_inline return a
5931 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
5932 Use ZERO_COEFFS when emitting initializers.
5933 * data-streamer.h (bp_pack_poly_value): New function.
5934 (bp_unpack_poly_value): Likewise.
5935 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
5936 for GET_MODE_NUNITS.
5937 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
5938 for GET_MODE_NUNITS.
5939 * tree.c (make_vector_type): Remove temporary shim and make
5940 the real function take the number of units as a poly_uint64
5942 (build_vector_type_for_mode): Handle polynomial nunits.
5943 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
5944 * emit-rtl.c (const_vec_series_p_1): Likewise.
5945 (gen_rtx_CONST_VECTOR): Likewise.
5946 * fold-const.c (test_vec_duplicate_folding): Likewise.
5947 * genrecog.c (validate_pattern): Likewise.
5948 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
5949 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
5950 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
5951 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
5952 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
5953 * rtlanal.c (subreg_get_info): Likewise.
5954 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
5955 (vect_grouped_load_supported): Likewise.
5956 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
5957 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
5958 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
5959 (simplify_const_unary_operation, simplify_binary_operation_1)
5960 (simplify_const_binary_operation, simplify_ternary_operation)
5961 (test_vector_ops_duplicate, test_vector_ops): Likewise.
5962 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
5963 instead of CONST_VECTOR_NUNITS.
5964 * varasm.c (output_constant_pool_2): Likewise.
5965 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
5966 explicit-encoded elements in the XVEC for variable-length vectors.
5968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5970 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
5972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5973 Alan Hayward <alan.hayward@arm.com>
5974 David Sherwood <david.sherwood@arm.com>
5976 * coretypes.h (fixed_size_mode): Declare.
5977 (fixed_size_mode_pod): New typedef.
5978 * builtins.h (target_builtins::x_apply_args_mode)
5979 (target_builtins::x_apply_result_mode): Change type to
5980 fixed_size_mode_pod.
5981 * builtins.c (apply_args_size, apply_result_size, result_vector)
5982 (expand_builtin_apply_args_1, expand_builtin_apply)
5983 (expand_builtin_return): Update accordingly.
5985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5987 * cse.c (hash_rtx_cb): Hash only the encoded elements.
5988 * cselib.c (cselib_hash_rtx): Likewise.
5989 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
5990 CONST_VECTOR encoding.
5992 2018-01-03 Jakub Jelinek <jakub@redhat.com>
5993 Jeff Law <law@redhat.com>
5996 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
5997 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
5998 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
5999 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6002 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6003 explicitly probe *sp in a noreturn function if there were any callee
6004 register saves or frame pointer is needed.
6006 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6009 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6010 BLKmode for ternary, binary or unary expressions.
6013 * var-tracking.c (delete_vta_debug_insn): New inline function.
6014 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6015 insns from get_insns () to NULL instead of each bb separately.
6016 Use delete_vta_debug_insn. No longer static.
6017 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6018 delete_vta_debug_insns callers.
6019 * rtl.h (delete_vta_debug_insns): Declare.
6020 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6021 instead of variable_tracking_main.
6023 2018-01-03 Martin Sebor <msebor@redhat.com>
6025 PR tree-optimization/83603
6026 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6027 arguments past the endof the argument list in functions declared
6028 without a prototype.
6029 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6030 Avoid checking when arguments are null.
6032 2018-01-03 Martin Sebor <msebor@redhat.com>
6035 * doc/extend.texi (attribute const): Fix a typo.
6036 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6037 issuing -Wsuggest-attribute for void functions.
6039 2018-01-03 Martin Sebor <msebor@redhat.com>
6041 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6042 offset_int::from instead of wide_int::to_shwi.
6043 (maybe_diag_overlap): Remove assertion.
6044 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6045 * gimple-ssa-sprintf.c (format_directive): Same.
6046 (parse_directive): Same.
6047 (sprintf_dom_walker::compute_format_length): Same.
6048 (try_substitute_return_value): Same.
6050 2018-01-03 Jeff Law <law@redhat.com>
6053 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6054 non-constant residual for zero at runtime and avoid probing in
6055 that case. Reorganize code for trailing problem to mirror handling
6058 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6060 PR tree-optimization/83501
6061 * tree-ssa-strlen.c (get_string_cst): New.
6062 (handle_char_store): Call get_string_cst.
6064 2018-01-03 Martin Liska <mliska@suse.cz>
6066 PR tree-optimization/83593
6067 * tree-ssa-strlen.c: Include tree-cfg.h.
6068 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6069 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6070 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6072 (strlen_dom_walker::before_dom_children): Call
6073 gimple_purge_dead_eh_edges. Dump tranformation with details
6075 (strlen_dom_walker::before_dom_children): Update call by adding
6076 new argument cleanup_eh.
6077 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6079 2018-01-03 Martin Liska <mliska@suse.cz>
6082 * cif-code.def (VARIADIC_THUNK): New enum value.
6083 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6086 2018-01-03 Jan Beulich <jbeulich@suse.com>
6088 * sse.md (mov<mode>_internal): Tighten condition for when to use
6089 vmovdqu<ssescalarsize> for TI and OI modes.
6091 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6093 Update copyright years.
6095 2018-01-03 Martin Liska <mliska@suse.cz>
6098 * ipa-visibility.c (function_and_variable_visibility): Skip
6099 functions with noipa attribure.
6101 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6103 * gcc.c (process_command): Update copyright notice dates.
6104 * gcov-dump.c (print_version): Ditto.
6105 * gcov.c (print_version): Ditto.
6106 * gcov-tool.c (print_version): Ditto.
6107 * gengtype.c (create_file): Ditto.
6108 * doc/cpp.texi: Bump @copying's copyright year.
6109 * doc/cppinternals.texi: Ditto.
6110 * doc/gcc.texi: Ditto.
6111 * doc/gccint.texi: Ditto.
6112 * doc/gcov.texi: Ditto.
6113 * doc/install.texi: Ditto.
6114 * doc/invoke.texi: Ditto.
6116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6118 * vector-builder.h (vector_builder::m_full_nelts): Change from
6119 unsigned int to poly_uint64.
6120 (vector_builder::full_nelts): Update prototype accordingly.
6121 (vector_builder::new_vector): Likewise.
6122 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6123 (vector_builder::operator ==): Likewise.
6124 (vector_builder::finalize): Likewise.
6125 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6126 Take the number of elements as a poly_uint64 rather than an
6128 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6129 from unsigned int to poly_uint64.
6130 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6131 (vec_perm_indices::new_vector): Likewise.
6132 (vec_perm_indices::length): Likewise.
6133 (vec_perm_indices::nelts_per_input): Likewise.
6134 (vec_perm_indices::input_nelts): Likewise.
6135 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6136 number of elements per input as a poly_uint64 rather than an
6137 unsigned int. Use the original encoding for variable-length
6138 vectors, rather than clamping each individual element.
6139 For the second and subsequent elements in each pattern,
6140 clamp the step and base before clamping their sum.
6141 (vec_perm_indices::series_p): Handle polynomial element counts.
6142 (vec_perm_indices::all_in_range_p): Likewise.
6143 (vec_perm_indices_to_tree): Likewise.
6144 (vec_perm_indices_to_rtx): Likewise.
6145 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6146 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6147 (tree_vector_builder::new_binary_operation): Handle polynomial
6148 element counts. Return false if we need to know the number
6149 of elements at compile time.
6150 * fold-const.c (fold_vec_perm): Punt if the number of elements
6151 isn't known at compile time.
6153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6155 * vec-perm-indices.h (vec_perm_builder): Change element type
6156 from HOST_WIDE_INT to poly_int64.
6157 (vec_perm_indices::element_type): Update accordingly.
6158 (vec_perm_indices::clamp): Handle polynomial element_types.
6159 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6160 (vec_perm_indices::all_in_range_p): Likewise.
6161 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6163 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6164 polynomial vec_perm_indices element types.
6165 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6166 * fold-const.c (fold_vec_perm): Likewise.
6167 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6168 * tree-vect-generic.c (lower_vec_perm): Likewise.
6169 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6170 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6171 element type to HOST_WIDE_INT.
6173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6174 Alan Hayward <alan.hayward@arm.com>
6175 David Sherwood <david.sherwood@arm.com>
6177 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6178 rather than an int. Use plus_constant.
6179 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6180 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6183 Alan Hayward <alan.hayward@arm.com>
6184 David Sherwood <david.sherwood@arm.com>
6186 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6187 a HOST_WIDE_INT to a poly_int64.
6189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6190 Alan Hayward <alan.hayward@arm.com>
6191 David Sherwood <david.sherwood@arm.com>
6193 * calls.c (load_register_parameters): Cope with polynomial
6194 mode sizes. Require a constant size for BLKmode parameters
6195 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6196 forces a parameter to be padded at the lsb end in order to
6197 fill a complete number of words, require the parameter size
6198 to be ordered wrt UNITS_PER_WORD.
6200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6201 Alan Hayward <alan.hayward@arm.com>
6202 David Sherwood <david.sherwood@arm.com>
6204 * reload1.c (spill_stack_slot_width): Change element type
6205 from unsigned int to poly_uint64_pod.
6206 (alter_reg): Treat mode sizes as polynomial.
6208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6209 Alan Hayward <alan.hayward@arm.com>
6210 David Sherwood <david.sherwood@arm.com>
6212 * reload.c (complex_word_subreg_p): New function.
6213 (reload_inner_reg_of_subreg, push_reload): Use it.
6215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6216 Alan Hayward <alan.hayward@arm.com>
6217 David Sherwood <david.sherwood@arm.com>
6219 * lra-constraints.c (process_alt_operands): Reject matched
6220 operands whose sizes aren't ordered.
6221 (match_reload): Refer to this check here.
6223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6224 Alan Hayward <alan.hayward@arm.com>
6225 David Sherwood <david.sherwood@arm.com>
6227 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6228 that the mode size is in the set {1, 2, 4, 8, 16}.
6230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6231 Alan Hayward <alan.hayward@arm.com>
6232 David Sherwood <david.sherwood@arm.com>
6234 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6235 Use plus_constant instead of gen_rtx_PLUS.
6237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6238 Alan Hayward <alan.hayward@arm.com>
6239 David Sherwood <david.sherwood@arm.com>
6241 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6242 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6243 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6244 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6245 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6246 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6247 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6248 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6249 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6250 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6252 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6253 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6254 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6255 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6256 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6257 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6258 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6259 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6260 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6261 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6263 * expr.c (emit_move_resolve_push): Treat the input and result
6264 of PUSH_ROUNDING as a poly_int64.
6265 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6266 (emit_push_insn): Likewise.
6267 * lra-eliminations.c (mark_not_eliminable): Likewise.
6268 * recog.c (push_operand): Likewise.
6269 * reload1.c (elimination_effects): Likewise.
6270 * rtlanal.c (nonzero_bits1): Likewise.
6271 * calls.c (store_one_arg): Likewise. Require the padding to be
6272 known at compile time.
6274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6275 Alan Hayward <alan.hayward@arm.com>
6276 David Sherwood <david.sherwood@arm.com>
6278 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
6279 Use plus_constant instead of gen_rtx_PLUS.
6281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6282 Alan Hayward <alan.hayward@arm.com>
6283 David Sherwood <david.sherwood@arm.com>
6285 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
6288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6289 Alan Hayward <alan.hayward@arm.com>
6290 David Sherwood <david.sherwood@arm.com>
6292 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
6293 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
6294 via stack temporaries. Treat the mode size as polynomial too.
6296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6297 Alan Hayward <alan.hayward@arm.com>
6298 David Sherwood <david.sherwood@arm.com>
6300 * expr.c (expand_expr_real_2): When handling conversions involving
6301 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
6302 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
6303 as a poly_uint64 too.
6305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6306 Alan Hayward <alan.hayward@arm.com>
6307 David Sherwood <david.sherwood@arm.com>
6309 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
6311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6312 Alan Hayward <alan.hayward@arm.com>
6313 David Sherwood <david.sherwood@arm.com>
6315 * combine.c (can_change_dest_mode): Handle polynomial
6316 REGMODE_NATURAL_SIZE.
6317 * expmed.c (store_bit_field_1): Likewise.
6318 * expr.c (store_constructor): Likewise.
6319 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
6320 and polynomial REGMODE_NATURAL_SIZE.
6321 (gen_lowpart_common): Likewise.
6322 * reginfo.c (record_subregs_of_mode): Likewise.
6323 * rtlanal.c (read_modify_subreg_p): Likewise.
6325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6326 Alan Hayward <alan.hayward@arm.com>
6327 David Sherwood <david.sherwood@arm.com>
6329 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6330 numbers of elements.
6332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6333 Alan Hayward <alan.hayward@arm.com>
6334 David Sherwood <david.sherwood@arm.com>
6336 * match.pd: Cope with polynomial numbers of vector elements.
6338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6339 Alan Hayward <alan.hayward@arm.com>
6340 David Sherwood <david.sherwood@arm.com>
6342 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
6343 in a POINTER_PLUS_EXPR.
6345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6346 Alan Hayward <alan.hayward@arm.com>
6347 David Sherwood <david.sherwood@arm.com>
6349 * omp-simd-clone.c (simd_clone_subparts): New function.
6350 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
6351 (ipa_simd_modify_function_body): Likewise.
6353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6354 Alan Hayward <alan.hayward@arm.com>
6355 David Sherwood <david.sherwood@arm.com>
6357 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
6358 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
6359 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
6360 (expand_vector_condition, vector_element): Likewise.
6361 (subparts_gt): New function.
6362 (get_compute_type): Use subparts_gt.
6363 (count_type_subparts): Delete.
6364 (expand_vector_operations_1): Use subparts_gt instead of
6365 count_type_subparts.
6367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6368 Alan Hayward <alan.hayward@arm.com>
6369 David Sherwood <david.sherwood@arm.com>
6371 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
6372 (vect_compile_time_alias): ...this new function. Do the calculation
6373 on poly_ints rather than trees.
6374 (vect_prune_runtime_alias_test_list): Update call accordingly.
6376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6377 Alan Hayward <alan.hayward@arm.com>
6378 David Sherwood <david.sherwood@arm.com>
6380 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
6382 (vect_schedule_slp_instance): Likewise.
6384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6385 Alan Hayward <alan.hayward@arm.com>
6386 David Sherwood <david.sherwood@arm.com>
6388 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6389 constant and extern definitions for variable-length vectors.
6390 (vect_get_constant_vectors): Note that the number of units
6391 is known to be constant.
6393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6394 Alan Hayward <alan.hayward@arm.com>
6395 David Sherwood <david.sherwood@arm.com>
6397 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
6398 of units as polynomial. Choose between WIDE and NARROW based
6401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6402 Alan Hayward <alan.hayward@arm.com>
6403 David Sherwood <david.sherwood@arm.com>
6405 * tree-vect-stmts.c (simd_clone_subparts): New function.
6406 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
6408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6409 Alan Hayward <alan.hayward@arm.com>
6410 David Sherwood <david.sherwood@arm.com>
6412 * tree-vect-stmts.c (vectorizable_call): Treat the number of
6413 vectors as polynomial. Use build_index_vector for
6416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6417 Alan Hayward <alan.hayward@arm.com>
6418 David Sherwood <david.sherwood@arm.com>
6420 * tree-vect-stmts.c (get_load_store_type): Treat the number of
6421 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
6422 for variable-length vectors.
6423 (vectorizable_mask_load_store): Treat the number of units as
6424 polynomial, asserting that it is constant if the condition has
6425 already been enforced.
6426 (vectorizable_store, vectorizable_load): Likewise.
6428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6429 Alan Hayward <alan.hayward@arm.com>
6430 David Sherwood <david.sherwood@arm.com>
6432 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
6433 of units as polynomial. Punt if we can't tell at compile time
6434 which vector contains the final result.
6436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6437 Alan Hayward <alan.hayward@arm.com>
6438 David Sherwood <david.sherwood@arm.com>
6440 * tree-vect-loop.c (vectorizable_induction): Treat the number
6441 of units as polynomial. Punt on SLP inductions. Use an integer
6442 VEC_SERIES_EXPR for variable-length integer reductions. Use a
6443 cast of such a series for variable-length floating-point
6446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6447 Alan Hayward <alan.hayward@arm.com>
6448 David Sherwood <david.sherwood@arm.com>
6450 * tree.h (build_index_vector): Declare.
6451 * tree.c (build_index_vector): New function.
6452 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
6453 of units as polynomial, forcibly converting it to a constant if
6454 vectorizable_reduction has already enforced the condition.
6455 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
6456 to create a {1,2,3,...} vector.
6457 (vectorizable_reduction): Treat the number of units as polynomial.
6458 Choose vectype_in based on the largest scalar element size rather
6459 than the smallest number of units. Enforce the restrictions
6462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6463 Alan Hayward <alan.hayward@arm.com>
6464 David Sherwood <david.sherwood@arm.com>
6466 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
6467 number of units as polynomial.
6469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6470 Alan Hayward <alan.hayward@arm.com>
6471 David Sherwood <david.sherwood@arm.com>
6473 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
6474 * target.def (autovectorize_vector_sizes): Return the vector sizes
6475 by pointer, using vector_sizes rather than a bitmask.
6476 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
6477 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
6478 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
6480 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
6481 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
6482 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
6483 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
6484 * omp-general.c (omp_max_vf): Likewise.
6485 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6486 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
6487 * tree-vect-loop.c (vect_analyze_loop): Likewise.
6488 * tree-vect-slp.c (vect_slp_bb): Likewise.
6489 * doc/tm.texi: Regenerate.
6490 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
6492 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
6493 the vector size as a poly_uint64 rather than an unsigned int.
6494 (current_vector_size): Change from an unsigned int to a poly_uint64.
6495 (get_vectype_for_scalar_type): Update accordingly.
6496 * tree.h (build_truth_vector_type): Take the size and number of
6497 units as a poly_uint64 rather than an unsigned int.
6498 (build_vector_type): Add a temporary overload that takes
6499 the number of units as a poly_uint64 rather than an unsigned int.
6500 * tree.c (make_vector_type): Likewise.
6501 (build_truth_vector_type): Take the number of units as a poly_uint64
6502 rather than an unsigned int.
6504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6505 Alan Hayward <alan.hayward@arm.com>
6506 David Sherwood <david.sherwood@arm.com>
6508 * target.def (get_mask_mode): Take the number of units and length
6509 as poly_uint64s rather than unsigned ints.
6510 * targhooks.h (default_get_mask_mode): Update accordingly.
6511 * targhooks.c (default_get_mask_mode): Likewise.
6512 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
6513 * doc/tm.texi: Regenerate.
6515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6516 Alan Hayward <alan.hayward@arm.com>
6517 David Sherwood <david.sherwood@arm.com>
6519 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
6520 * omp-general.c (omp_max_vf): Likewise.
6521 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
6522 (expand_omp_simd): Handle polynomial safelen.
6523 * omp-low.c (omplow_simd_context): Add a default constructor.
6524 (omplow_simd_context::max_vf): Change from int to poly_uint64.
6525 (lower_rec_simd_input_clauses): Update accordingly.
6526 (lower_rec_input_clauses): Likewise.
6528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6529 Alan Hayward <alan.hayward@arm.com>
6530 David Sherwood <david.sherwood@arm.com>
6532 * tree-vectorizer.h (vect_nunits_for_cost): New function.
6533 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
6534 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
6535 (vect_analyze_slp_cost): Likewise.
6536 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
6537 (vect_model_load_cost): Likewise.
6539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6540 Alan Hayward <alan.hayward@arm.com>
6541 David Sherwood <david.sherwood@arm.com>
6543 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
6544 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
6545 from an unsigned int * to a poly_uint64_pod *.
6546 (calculate_unrolling_factor): New function.
6547 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
6549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6550 Alan Hayward <alan.hayward@arm.com>
6551 David Sherwood <david.sherwood@arm.com>
6553 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
6554 from an unsigned int to a poly_uint64.
6555 (_loop_vec_info::slp_unrolling_factor): Likewise.
6556 (_loop_vec_info::vectorization_factor): Change from an int
6558 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
6559 (vect_get_num_vectors): New function.
6560 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
6561 (vect_get_num_copies): Use vect_get_num_vectors.
6562 (vect_analyze_data_ref_dependences): Change max_vf from an int *
6563 to an unsigned int *.
6564 (vect_analyze_data_refs): Change min_vf from an int * to a
6566 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6567 than an unsigned HOST_WIDE_INT.
6568 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6569 (vect_analyze_data_ref_dependence): Change max_vf from an int *
6570 to an unsigned int *.
6571 (vect_analyze_data_ref_dependences): Likewise.
6572 (vect_compute_data_ref_alignment): Handle polynomial vf.
6573 (vect_enhance_data_refs_alignment): Likewise.
6574 (vect_prune_runtime_alias_test_list): Likewise.
6575 (vect_shift_permute_load_chain): Likewise.
6576 (vect_supportable_dr_alignment): Likewise.
6577 (dependence_distance_ge_vf): Take the vectorization factor as a
6578 poly_uint64 rather than an unsigned HOST_WIDE_INT.
6579 (vect_analyze_data_refs): Change min_vf from an int * to a
6581 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
6582 vfm1 as a poly_uint64 rather than an int. Make the same change
6583 for the returned bound_scalar.
6584 (vect_gen_vector_loop_niters): Handle polynomial vf.
6585 (vect_do_peeling): Likewise. Update call to
6586 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
6587 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
6589 * tree-vect-loop.c (vect_determine_vectorization_factor)
6590 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
6591 (vect_get_known_peeling_cost): Likewise.
6592 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
6593 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
6594 (vect_transform_loop): Likewise. Use the lowest possible VF when
6595 updating the upper bounds of the loop.
6596 (vect_min_worthwhile_factor): Make static. Return an unsigned int
6598 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
6599 polynomial unroll factors.
6600 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
6601 (vect_make_slp_decision): Likewise.
6602 (vect_supported_load_permutation_p): Likewise, and polynomial
6604 (vect_analyze_slp_cost): Handle polynomial vf.
6605 (vect_slp_analyze_node_operations): Likewise.
6606 (vect_slp_analyze_bb_1): Likewise.
6607 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
6608 than an unsigned HOST_WIDE_INT.
6609 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
6610 (vectorizable_load): Handle polynomial vf.
6611 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
6613 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
6615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6616 Alan Hayward <alan.hayward@arm.com>
6617 David Sherwood <david.sherwood@arm.com>
6619 * match.pd: Handle bit operations involving three constants
6620 and try to fold one pair.
6622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6624 * tree-vect-loop-manip.c: Include gimple-fold.h.
6625 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
6626 niters_maybe_zero parameters. Handle other cases besides a step of 1.
6627 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
6628 Add a path that uses a step of VF instead of 1, but disable it
6630 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
6631 and niters_no_overflow parameters. Update calls to
6632 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
6633 Create a new SSA name if the latter choses to use a ste other
6634 than zero, and return it via niters_vector_mult_vf_var.
6635 * tree-vect-loop.c (vect_transform_loop): Update calls to
6636 vect_do_peeling, vect_gen_vector_loop_niters and
6637 slpeel_make_loop_iterate_ntimes.
6638 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
6639 (vect_gen_vector_loop_niters): Update declarations after above changes.
6641 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
6643 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
6644 128-bit round to integer instructions.
6645 (ceil<mode>2): Likewise.
6646 (btrunc<mode>2): Likewise.
6647 (round<mode>2): Likewise.
6649 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6651 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
6652 unaligned VSX load/store on P8/P9.
6653 (expand_block_clear): Allow the use of unaligned VSX
6654 load/store on P8/P9.
6656 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6658 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
6660 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
6661 swap associated with both a load and a store.
6663 2018-01-02 Andrew Waterman <andrew@sifive.com>
6665 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
6666 * config/riscv/riscv.md (clear_cache): Use it.
6668 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
6670 * web.c: Remove out-of-date comment.
6672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6674 * expr.c (fixup_args_size_notes): Check that any existing
6675 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
6676 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
6677 (emit_single_push_insn): ...here.
6679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6681 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
6682 (const_vector_encoded_nelts): New function.
6683 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
6684 (const_vector_int_elt, const_vector_elt): Declare.
6685 * emit-rtl.c (const_vector_int_elt_1): New function.
6686 (const_vector_elt): Likewise.
6687 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
6688 of CONST_VECTOR_ELT.
6690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6692 * expr.c: Include rtx-vector-builder.h.
6693 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
6694 directly on the tree encoding.
6695 (const_vector_from_tree): Likewise.
6696 * optabs.c: Include rtx-vector-builder.h.
6697 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
6698 sequence of "u" values.
6699 * vec-perm-indices.c: Include rtx-vector-builder.h.
6700 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
6701 directly on the vec_perm_indices encoding.
6703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6705 * doc/rtl.texi (const_vector): Describe new encoding scheme.
6706 * Makefile.in (OBJS): Add rtx-vector-builder.o.
6707 * rtx-vector-builder.h: New file.
6708 * rtx-vector-builder.c: Likewise.
6709 * rtl.h (rtx_def::u2): Add a const_vector field.
6710 (CONST_VECTOR_NPATTERNS): New macro.
6711 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
6712 (CONST_VECTOR_DUPLICATE_P): Likewise.
6713 (CONST_VECTOR_STEPPED_P): Likewise.
6714 (CONST_VECTOR_ENCODED_ELT): Likewise.
6715 (const_vec_duplicate_p): Check for a duplicated vector encoding.
6716 (unwrap_const_vec_duplicate): Likewise.
6717 (const_vec_series_p): Check for a non-duplicated vector encoding.
6718 Say that the function only returns true for integer vectors.
6719 * emit-rtl.c: Include rtx-vector-builder.h.
6720 (gen_const_vec_duplicate_1): Delete.
6721 (gen_const_vector): Call gen_const_vec_duplicate instead of
6722 gen_const_vec_duplicate_1.
6723 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
6724 (gen_const_vec_duplicate): Use rtx_vector_builder.
6725 (gen_const_vec_series): Likewise.
6726 (gen_rtx_CONST_VECTOR): Likewise.
6727 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
6728 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
6729 Build a new vector rather than modifying a CONST_VECTOR in-place.
6730 (handle_special_swappables): Update call accordingly.
6731 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
6732 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
6733 Build a new vector rather than modifying a CONST_VECTOR in-place.
6734 (handle_special_swappables): Update call accordingly.
6736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6738 * simplify-rtx.c (simplify_const_binary_operation): Use
6739 CONST_VECTOR_ELT instead of XVECEXP.
6741 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6743 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
6744 the selector elements to be different from the data elements
6745 if the selector is a VECTOR_CST.
6746 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
6747 ssizetype for the selector.
6749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6751 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
6752 before testing each element individually.
6753 * tree-vect-generic.c (lower_vec_perm): Likewise.
6755 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6757 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
6758 * selftest-run-tests.c (selftest::run_tests): Call it.
6759 * vector-builder.h (vector_builder::operator ==): New function.
6760 (vector_builder::operator !=): Likewise.
6761 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
6762 (vec_perm_indices::all_from_input_p): New function.
6763 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6764 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
6765 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
6766 instead of reading the VECTOR_CST directly. Detect whether both
6767 vector inputs are the same before constructing the vec_perm_indices,
6768 and update the number of inputs argument accordingly. Use the
6769 utility functions added above. Only construct sel2 if we need to.
6771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6773 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
6774 the broadcast of the low byte.
6775 (expand_mult_highpart): Use an explicit encoding for the permutes.
6776 * optabs-query.c (can_mult_highpart_p): Likewise.
6777 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
6778 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6779 (vectorizable_bswap): Likewise.
6780 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
6781 explicit encoding for the power-of-2 permutes.
6782 (vect_permute_store_chain): Likewise.
6783 (vect_grouped_load_supported): Likewise.
6784 (vect_permute_load_chain): Likewise.
6786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6788 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
6789 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
6790 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
6791 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6792 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
6793 (vect_gen_perm_mask_any): Likewise.
6795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6797 * int-vector-builder.h: New file.
6798 * vec-perm-indices.h: Include int-vector-builder.h.
6799 (vec_perm_indices): Redefine as an int_vector_builder.
6800 (auto_vec_perm_indices): Delete.
6801 (vec_perm_builder): Redefine as a stand-alone class.
6802 (vec_perm_indices::vec_perm_indices): New function.
6803 (vec_perm_indices::clamp): Likewise.
6804 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
6805 (vec_perm_indices::new_vector): New function.
6806 (vec_perm_indices::new_expanded_vector): Update for new
6807 vec_perm_indices class.
6808 (vec_perm_indices::rotate_inputs): New function.
6809 (vec_perm_indices::all_in_range_p): Operate directly on the
6810 encoded form, without computing elided elements.
6811 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
6812 encoding. Update for new vec_perm_indices class.
6813 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
6814 the given vec_perm_builder.
6815 (expand_vec_perm_var): Update vec_perm_builder constructor.
6816 (expand_mult_highpart): Use vec_perm_builder instead of
6817 auto_vec_perm_indices.
6818 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
6819 vec_perm_indices instead of auto_vec_perm_indices. Use a single
6820 or double series encoding as appropriate.
6821 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
6822 vec_perm_indices instead of auto_vec_perm_indices.
6823 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6824 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6825 (vect_permute_store_chain): Likewise.
6826 (vect_grouped_load_supported): Likewise.
6827 (vect_permute_load_chain): Likewise.
6828 (vect_shift_permute_load_chain): Likewise.
6829 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
6830 (vect_transform_slp_perm_load): Likewise.
6831 (vect_schedule_slp_instance): Likewise.
6832 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6833 (vectorizable_mask_load_store): Likewise.
6834 (vectorizable_bswap): Likewise.
6835 (vectorizable_store): Likewise.
6836 (vectorizable_load): Likewise.
6837 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
6838 vec_perm_indices instead of auto_vec_perm_indices. Use
6839 tree_to_vec_perm_builder to read the vector from a tree.
6840 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
6841 vec_perm_builder instead of a vec_perm_indices.
6842 (have_whole_vector_shift): Use vec_perm_builder and
6843 vec_perm_indices instead of auto_vec_perm_indices. Leave the
6844 truncation to calc_vec_perm_mask_for_shift.
6845 (vect_create_epilog_for_reduction): Likewise.
6846 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
6847 from auto_vec_perm_indices to vec_perm_indices.
6848 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
6849 instead of changing individual elements.
6850 (aarch64_vectorize_vec_perm_const): Use new_vector to install
6851 the vector in d.perm.
6852 * config/arm/arm.c (expand_vec_perm_d::perm): Change
6853 from auto_vec_perm_indices to vec_perm_indices.
6854 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
6855 instead of changing individual elements.
6856 (arm_vectorize_vec_perm_const): Use new_vector to install
6857 the vector in d.perm.
6858 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
6859 Update vec_perm_builder constructor.
6860 (rs6000_expand_interleave): Likewise.
6861 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
6862 (rs6000_expand_interleave): Likewise.
6864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6866 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
6867 to qimode could truncate the indices.
6868 * optabs.c (expand_vec_perm_var): Likewise.
6870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
6872 * Makefile.in (OBJS): Add vec-perm-indices.o.
6873 * vec-perm-indices.h: New file.
6874 * vec-perm-indices.c: Likewise.
6875 * target.h (vec_perm_indices): Replace with a forward class
6877 (auto_vec_perm_indices): Move to vec-perm-indices.h.
6878 * optabs.h: Include vec-perm-indices.h.
6879 (expand_vec_perm): Delete.
6880 (selector_fits_mode_p, expand_vec_perm_var): Declare.
6881 (expand_vec_perm_const): Declare.
6882 * target.def (vec_perm_const_ok): Replace with...
6883 (vec_perm_const): ...this new hook.
6884 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
6885 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
6886 * doc/tm.texi: Regenerate.
6887 * optabs.def (vec_perm_const): Delete.
6888 * doc/md.texi (vec_perm_const): Likewise.
6889 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
6890 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
6891 expand_vec_perm for constant permutation vectors. Assert that
6892 the mode of variable permutation vectors is the integer equivalent
6893 of the mode that is being permuted.
6894 * optabs-query.h (selector_fits_mode_p): Declare.
6895 * optabs-query.c: Include vec-perm-indices.h.
6896 (selector_fits_mode_p): New function.
6897 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
6898 is defined, instead of checking whether the vec_perm_const_optab
6899 exists. Use targetm.vectorize.vec_perm_const instead of
6900 targetm.vectorize.vec_perm_const_ok. Check whether the indices
6901 fit in the vector mode before using a variable permute.
6902 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
6903 vec_perm_indices instead of an rtx.
6904 (expand_vec_perm): Replace with...
6905 (expand_vec_perm_const): ...this new function. Take the selector
6906 as a vec_perm_indices rather than an rtx. Also take the mode of
6907 the selector. Update call to shift_amt_for_vec_perm_mask.
6908 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
6909 Use vec_perm_indices::new_expanded_vector to expand the original
6910 selector into bytes. Check whether the indices fit in the vector
6911 mode before using a variable permute.
6912 (expand_vec_perm_var): Make global.
6913 (expand_mult_highpart): Use expand_vec_perm_const.
6914 * fold-const.c: Includes vec-perm-indices.h.
6915 * tree-ssa-forwprop.c: Likewise.
6916 * tree-vect-data-refs.c: Likewise.
6917 * tree-vect-generic.c: Likewise.
6918 * tree-vect-loop.c: Likewise.
6919 * tree-vect-slp.c: Likewise.
6920 * tree-vect-stmts.c: Likewise.
6921 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
6923 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
6924 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
6925 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
6926 (aarch64_vectorize_vec_perm_const): ...this new function.
6927 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6928 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6929 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
6930 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
6931 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6932 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6933 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
6935 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
6936 check for NEON modes.
6937 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
6938 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
6939 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
6940 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
6942 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
6943 the old VEC_PERM_CONST conditions.
6944 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
6945 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
6946 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
6947 (ia64_vectorize_vec_perm_const_ok): Merge into...
6948 (ia64_vectorize_vec_perm_const): ...this new function.
6949 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
6950 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
6951 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
6952 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
6953 * config/mips/mips.c (mips_expand_vec_perm_const)
6954 (mips_vectorize_vec_perm_const_ok): Merge into...
6955 (mips_vectorize_vec_perm_const): ...this new function.
6956 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
6957 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
6958 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
6959 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
6960 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
6961 (rs6000_expand_vec_perm_const): Delete.
6962 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
6964 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6965 (altivec_expand_vec_perm_const_le): Take each operand individually.
6966 Operate on constant selectors rather than rtxes.
6967 (altivec_expand_vec_perm_const): Likewise. Update call to
6968 altivec_expand_vec_perm_const_le.
6969 (rs6000_expand_vec_perm_const): Delete.
6970 (rs6000_vectorize_vec_perm_const_ok): Delete.
6971 (rs6000_vectorize_vec_perm_const): New function.
6972 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
6973 an element count and rtx array.
6974 (rs6000_expand_extract_even): Update call accordingly.
6975 (rs6000_expand_interleave): Likewise.
6976 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
6977 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
6978 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
6979 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
6980 (rs6000_expand_vec_perm_const): Delete.
6981 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
6982 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
6983 (altivec_expand_vec_perm_const_le): Take each operand individually.
6984 Operate on constant selectors rather than rtxes.
6985 (altivec_expand_vec_perm_const): Likewise. Update call to
6986 altivec_expand_vec_perm_const_le.
6987 (rs6000_expand_vec_perm_const): Delete.
6988 (rs6000_vectorize_vec_perm_const_ok): Delete.
6989 (rs6000_vectorize_vec_perm_const): New function. Remove stray
6990 reference to the SPE evmerge intructions.
6991 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
6992 an element count and rtx array.
6993 (rs6000_expand_extract_even): Update call accordingly.
6994 (rs6000_expand_interleave): Likewise.
6995 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
6996 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
6998 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7002 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7003 vector mode and that that mode matches the mode of the data
7005 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7006 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7007 directly using expand_vec_perm_1 when forcing selectors into
7009 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7013 * optabs-query.h (can_vec_perm_p): Delete.
7014 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7015 * optabs-query.c (can_vec_perm_p): Split into...
7016 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7017 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7018 particular selector is valid.
7019 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7020 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7021 (vect_grouped_load_supported): Likewise.
7022 (vect_shift_permute_load_chain): Likewise.
7023 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7024 (vect_transform_slp_perm_load): Likewise.
7025 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7026 (vectorizable_bswap): Likewise.
7027 (vect_gen_perm_mask_checked): Likewise.
7028 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7029 implementations of variable permutation vectors into account
7030 when deciding which selector to use.
7031 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7032 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7033 with a false third argument.
7034 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7035 to test whether the constant selector is valid and can_vec_perm_var_p
7036 to test whether a variable selector is valid.
7038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7040 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7041 * optabs-query.c (can_vec_perm_p): Likewise.
7042 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7043 instead of vec_perm_indices.
7044 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7045 (vect_gen_perm_mask_checked): Likewise,
7046 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7047 (vect_gen_perm_mask_checked): Likewise,
7049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7051 * optabs-query.h (qimode_for_vec_perm): Declare.
7052 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7053 (qimode_for_vec_perm): ...this new function.
7054 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7056 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7058 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7059 does not have a conditional at the top.
7061 2018-01-02 Richard Biener <rguenther@suse.de>
7063 * ipa-inline.c (big_speedup_p): Fix expression.
7065 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7068 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7071 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7075 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7076 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7077 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7078 cond_taken_branch_cost 3->4.
7080 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7082 PR tree-optimization/83581
7083 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7084 TODO_cleanup_cfg if any changes have been made.
7087 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7088 convert_modes if target mode has the right side, but different mode
7092 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7093 last argument when extracting from CONCAT. If either from_real or
7094 from_imag is NULL, use expansion through memory. If result is not
7095 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7096 the parts directly to inner mode, if even that fails, use expansion
7100 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7101 check for bswap in mode rather than HImode and use that in expand_unop
7104 Copyright (C) 2018 Free Software Foundation, Inc.
7106 Copying and distribution of this file, with or without modification,
7107 are permitted in any medium without royalty provided the copyright
7108 notice and this notice are preserved.