2018-07-10 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blobf8717804b2aee377ee6239edd4f121718ce9c855
1 2018-07-10  Richard Biener  <rguenther@suse.de>
3         * hash-map.h (hash_map::iterator::operator*): Return
4         references to key and value.
6 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
8         PR c++/86443
9         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
11         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12         (gimplify_omp_for): For composite loops, move outer
13         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
14         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
15         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
16         TREE_LIST for both the original class iterator and the "last" helper
17         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
18         loop, remember has_decl_expr from outer composite loops for the
19         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
21 2018-07-09  Martin Sebor  <msebor@redhat.com>
23         PR middle-end/77357
24         PR middle-end/86428
25         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
26         accessing implicitly initialized array elements.
27         * expr.c (string_constant): Handle string initializers of
28         character arrays within aggregates.
29         * gimple-fold.c (fold_array_ctor_reference): Add argument.
30         Store element offset.  As a special case, handle zero size.
31         (fold_nonarray_ctor_reference): Same.
32         (fold_ctor_reference): Add argument.  Store subobject offset.
33         * gimple-fold.h (fold_ctor_reference): Add argument.
35 2018-07-09  Paul Koning  <ni1d@arrl.net>
37         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
38         (pdp11_insn_cost): New function.
39         (pdp11_md_asm_adjust): New function.
40         (TARGET_INVALID_WITHIN_DOLOOP): Define.
41         (pdp11_rtx_costs): Update to match machine better.
42         (output_addr_const_pdp11): Correct format mismatch warnings.
43         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
44         * config/pdp11/pdp11.md: General change to add base_cost and/or
45         length attributes for use by new pdp11_insn_cost function.
46         (MIN_BRANCH): Correct definition.
47         (MIN_SOB): Ditto.
48         (doloop_end): Use standard pattern name for looping pattern.
49         (doloop_end_nocc): New.
50         (movsf): Add another constraint alternative.
51         (zero_extendqihi2): Add constraint alternatives for not in place
52         extend.
53         (zero_extendhisi2): Remove.
54         (shift patterns): Add CC handling variants.
55         (bswaphi2): New.
56         (bswapsi2): New.
57         (rothi3): New.
58         (define_peephole2): New peephole to recognize mov that sets CC for
59         subsequent test.
61 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
63         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
64         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
65         wi::add.
67 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
69         PR c/86420
70         * real.c (real_nextafter): Return true if result is denormal.
72 2018-07-09  Martin Liska  <mliska@suse.cz>
74         * common.opt: Add back wrongly removed attribute.
76 2018-07-09  Richard Biener  <rguenther@suse.de>
78         PR debug/86413
79         * dwarf2out.c (gen_block_die): For an early generated DIE
80         always output high/low PC attributes.
82 2018-07-09  Tom de Vries  <tdevries@suse.de>
84         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
85         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
86         onto VAR_DECL with abstract origin.
88 2018-07-07  Jim Wilson  <jimw@sifive.com>
90         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
92 2018-07-07  Tom de Vries  <tdevries@suse.de>
94         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
96 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
98         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
99         overflow_type.
100         (combine_bound): Use wide-int overflow calculation instead of
101         rolling our own.
102         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
103         overflow_type.
104         * fold-const.c (int_const_binop_2): Same.
105         (extract_muldiv_1): Same.
106         (fold_div_compare): Same.
107         (fold_abs_const): Same.
108         * match.pd: Same.
109         * poly-int.h (add): Same.
110         (sub): Same.
111         (neg): Same.
112         (mul): Same.
113         * predict.c (predict_iv_comparison): Same.
114         * profile-count.c (slow_safe_scale_64bit): Same.
115         * simplify-rtx.c (simplify_const_binary_operation): Same.
116         * tree-chrec.c (tree_fold_binomial): Same.
117         * tree-data-ref.c (split_constant_offset_1): Same.
118         * tree-if-conv.c (idx_within_array_bound): Same.
119         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
120         * tree-ssa-phiopt.c (minmax_replacement): Same.
121         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
122         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
123         * vr-values.c (vr_values::adjust_range_with_scev): Same.
124         * wide-int.cc (wi::add_large): Same.
125         (wi::mul_internal): Same.
126         (wi::sub_large): Same.
127         (wi::divmod_internal): Same.
128         * wide-int.h: Change overflow type to overflow_type for neg, add,
129         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
130         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
131         mul_internal, divmod_internal.
132         (overflow_type): New enum.
133         (accumulate_overflow): New.
135 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
137         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
138         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
140 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
142         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
143         argument is checked for zero before entering loop, avoid checking again.
145 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
147         * gimplify.h (generic_expr_could_trap_p): Set as global function.
148         * gimplify.h (generic_expr_could_trap_p): Likwise.
149         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
151 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
153         PR tree-optimization/86401
154         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
155         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
156         (fold_bit_and_mask): ... here.  New helper function for match.pd.
157         * fold-const.h (fold_bit_and_mask): Declare.
158         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
160 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
162         PR target/86324
163         * target.def (translate_mode_attribute): New hook.
164         * targhooks.h (default_translate_mode_attribute): Declare.
165         * targhooks.c (default_translate_mode_attribute): New function.
166         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
167         * doc/tm.texi: Regenerate.
168         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
169         (rs6000_translate_mode_attribute): New function.
171 2018-07-06  Paul Koning  <ni1d@arrl.net>
173         * doc/md.texi (define_split): Document DONE and FAIL.
174         (define_peephole2): Ditto.
176 2018-07-05  Jeff Law  <law@redhat.com>
178         PR tree-optimization/86010
179         * tree-ssa-dse.c (compute_trims): More aggressively trim at
180         both the head and tail of mem* and str* calls.
182 2018-07-05  Jim Wilson  <jimw@sifive.com>
184         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
186 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com> 
188         * config/aarch64/aarch64-simd.md: correct flags text for 
189         MIN_EXPR replacement.
191 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
193         * configure: Regenerated.
195 2018-07-05  Carl Love  <cel@us.ibm.com>
197         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
198         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
199         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
200         VSX_BUILTIN_DOUBLEL_V4SF.
202 2018-07-05  Martin Sebor  <msebor@redhat.com>
204         PR c++/86400
205         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
206         than its domain to compute its the upper bound of a char array.
208 2018-07-05  Nathan Sidwell  <nathan@acm.org>
210         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
211         * doc/cpp.texi: Update comment.
212         * doc/tm.texi: Rebuilt.
213         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
214         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
215         * doc/extend.texi (Backwards Compatibility): Clarify it is system
216         headers affected by extern "C".
217         * system.h: Poison NO_IMPLICIT_EXTERN_C.
218         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
219         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
220         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
221         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
222         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
223         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
224         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
225         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
226         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
227         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
228         config/sparc/sp64-elf.h, config/spu/spu.h,
229         config/stormy16/stormy16.h, config/v850/v850.h,
230         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
231         define NO_IMPLICIT_EXTERN_C.
232         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
234 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
236         PR target/84711
237         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
238         instead of GET_MODE_SIZE when comparing Units.
240 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
242         PR target/84711
243         * rtlanal.c (set_noop_p): Constrain on mode change,
244         include hard-reg-set.h
246 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
248         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
250 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
252         Revert
253         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
255         PR sanitizer/84250
256         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
257         libasan.
258         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
260 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
262         PR sanitizer/84250
263         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
264         libasan.
265         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
267 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
269         PR middle-end/86380
270         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
272 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
274         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
275         neg_*_op* variables.
277 2018-07-04  Martin Liska  <mliska@suse.cz>
279         * tree-switch-conversion.c: Define
280         max_ratio_for_speed and max_ratio_for_size constants.
282 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
283             Martin Liska  <mliska@suse.cz>
285         PR middle-end/66240
286         PR target/45996
287         PR c/84100
288         * common.opt: Rename align options with 'str_' prefix.
289         * common/config/i386/i386-common.c (set_malign_value): New
290         function.
291         (ix86_handle_option): Use it to set -falign-* options/
292         * config/aarch64/aarch64-protos.h (struct tune_params): Change
293         type from int to string.
294         * config/aarch64/aarch64.c: Update default values from int
295         to string.
296         * config/alpha/alpha.c (alpha_override_options_after_change):
297         Likewise.
298         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
299         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
300         max skip conditionally.
301         * config/i386/freebsd.h (SUBALIGN_LOG): New.
302         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
303         max skip conditionally.
304         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
305         max skip conditionally.
306         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
307         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
308         max skip conditionally.
309         * config/i386/i386.c (struct ptt): Change type from int to
310         string.
311         (ix86_default_align): Set default values.
312         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
313         max skip conditionally.
314         * config/i386/iamcu.h (SUBALIGN_LOG): New.
315         (ASM_OUTPUT_MAX_SKIP_ALIGN):
316         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
317         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
318         max skip conditionally.
319         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
320         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
321         * config/i386/x86-64.h (SUBALIGN_LOG): New.
322         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
323         max skip conditionally.
324         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
325         * config/ia64/ia64.c (ia64_option_override): Set default values
326         for alignment options.
327         * config/m68k/m68k.c: Handle new str_align_* options.
328         * config/mips/mips.c (mips_set_compression_mode): Change
329         type of constants.
330         (mips_option_override): Set default values for options.
331         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
332         Likewise.
333         * config/rs6000/rs6000.c (rs6000_option_override_internal):
334         Likewise.
335         * config/rx/rx.c (rx_option_override): Likewise.
336         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
337         (LABEL_ALIGN): Use align_labels_log.
338         (LOOP_ALIGN): Use align_loops_align.
339         * config/s390/s390.c (s390_asm_output_function_label): Use new
340         macros.
341         * config/sh/sh.c (sh_override_options_after_change):
342         Change type of constants.
343         * config/spu/spu.c (spu_sched_init): Likewise.
344         * config/sparc/sparc.c (sparc_option_override): Set default
345         values for options.
346         * config/visium/visium.c (visium_option_override): Likewise.
347         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
348         emit p2align format with last argument if it's not needed.
349         * doc/invoke.texi: Document extended format of -falign-*.
350         * final.c: Use align_labels alignment.
351         * flags.h (struct target_flag_state): Change type to use
352         align_flags.
353         (struct align_flags_tuple): New.
354         (struct align_flags): Likewise.
355         (align_loops_log): Redefine macro to use new types.
356         (align_loops_max_skip): Redefine macro to use new types.
357         (align_jumps_log): Redefine macro to use new types.
358         (align_jumps_max_skip): Redefine macro to use new types.
359         (align_labels_log): Redefine macro to use new types.
360         (align_labels_max_skip): Redefine macro to use new types.
361         (align_functions_log): Redefine macro to use new types.
362         (align_loops): Redefine macro to use new types.
363         (align_jumps): Redefine macro to use new types.
364         (align_labels): Redefine macro to use new types.
365         (align_functions): Redefine macro to use new types.
366         (align_functions_max_skip): Redefine macro to use new types.
367         (align_loops_value): New macro.
368         (align_jumps_value): New macro.
369         (align_labels_value): New macro.
370         (align_functions_value): New macro.
371         * function.c (invoke_set_current_function_hook): Propagate
372         alignment values from flags to global variables default in
373         topleev.h.
374         * ipa-icf.c (sem_function::equals_wpa): Use
375         cl_optimization_option_eq instead of memcmp.
376         * lto-streamer.h (cl_optimization_stream_out): Support streaming
377         of string types.
378         (cl_optimization_stream_in): Likewise.
379         * optc-save-gen.awk: Support strings in cl_optimization.
380         * opth-gen.awk: Likewise.
381         * opts.c (finish_options): Remove error checking of invalid
382         value ranges.
383         (MAX_CODE_ALIGN): Remove.
384         (MAX_CODE_ALIGN_VALUE): Likewise.
385         (parse_and_check_align_values): New function.
386         (check_alignment_argument): Likewise.
387         (common_handle_option): Use check_alignment_argument.
388         * opts.h (parse_and_check_align_values): Declare.
389         * toplev.c (init_alignments): Remove.
390         (read_log_maxskip): New.
391         (parse_N_M): Likewise.
392         (parse_alignment_opts): Likewise.
393         (backend_init_target): Remove usage of init_alignments.
394         * toplev.h (parse_alignment_opts): Declare.
395         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
396         argument.
397         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
398         * tree.c (cl_option_hasher::equal): New.
399         * varasm.c: Use new global macros.
401 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
403         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
404         Use a simpler align directive also if MAXSKIP = ALIGN-1.
405         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
406         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
407         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
408         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
409         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
410         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
411         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
412         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
413         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
414         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
415         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
417 2018-07-04  Martin Liska  <mliska@suse.cz>
418             Jonathan Wakely  <jwakely@redhat.com>
420         * coverage.c: Use correct type.
421         * doc/invoke.texi: Language correction.
423 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
425         PR target/85620
426         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
427         ENDBRANCH for non-tail call which may return via indirect branch.
428         * doc/extend.texi: Document indirect_return attribute.
430 2018-07-03  Martin Sebor  <msebor@redhat.com>
432         PR tree-optimization/86274
433         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
434         precondition.
435         (format_floating): Correct handling of infinities and NaNs.
437 2018-07-03  Martin Sebor  <msebor@redhat.com>
439         * print-tree.c (print_real_cst): New function.
440         (print_node_brief): Call it.
441         (print_node): Ditto.
443 2018-07-03  Jeff Law  <law@redhat.com>
445         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
446         into a single pattern.
448         * config/h8300/h8300.md (ors code_iterator): New.
449         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
450         a single pattern and single splitter.
451         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
452         (iorqi3_1, xorqi3_1): Likewise.
453         (iorqi3, xorqi3 expanders): Similarly.
455         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
456         (movmd_internal) into a single pattern using the P mode iterator.
457         (movmd splitters): Similarly.
458         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
459         (movsd splitters): Similarly.
461         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
462         ADDB, ADDW and ADDL into a single ADD attribute which selects the
463         right table based on the size of the operand.
464         * config/h8300/h8300.md (length_table): Corresponding changes. All
465         references to "addb", "addw" and "addl" changed to "add".
466         (btst patterns): Merge two variants into a single pattern.
467         (tstqi, tsthi): Likewise.
468         (addhi3_incdec, addsi3_incdec): Likewise.
469         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
470         (mulhi3, mulsi3): Likewise.
471         (udivhi3, udivsi3): Likewise.
472         (divhi3, divsi3): Likewise.
473         (andorqi3, andorhi3, andorsi3): Likewise.
475 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
477         PR target/85694
478         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
479         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
481 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
483         PR tree-optimization/85694
484         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
485         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
486         UNSPEC_URHADD.
487         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
488         (<u>avg<mode>3_ceil): New patterns.
490 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
492         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
493         scan-tree-dump directive.
494         * gcc.dg/vect/slp-perm-2.c: Likewise.
495         * gcc.dg/vect/slp-perm-3.c: Likewise.
496         * gcc.dg/vect/slp-perm-5.c: Likewise.
497         * gcc.dg/vect/slp-perm-6.c: Likewise.
498         * gcc.dg/vect/slp-perm-7.c: Likewise.
499         * gcc.dg/vect/slp-perm-8.c: Likewise.
501 2018-07-03  Marek Polacek  <polacek@redhat.com>
503         PR middle-end/86202
504         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
506 2018-07-03  Richard Biener  <rguenther@suse.de>
508         PR ipa/86389
509         * tree-ssa-structalias.c (find_func_clobbers): Properly
510         handle indirect calls.
512 2018-07-03  Jeff Law  <law@redhat.com>
514         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
515         (shifts): New code iterator.
516         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
517         expander.  Fix HImode handling on H8/SX.
518         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
519         (subqi3, subhi3, subsi3 expanders): Likewise.
520         (andqi3, andhi3, andsi3 expanders): Likewise.
521         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
522         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
523         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
524         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
525         (zero_extendqihi2, zero_extendqisi2): Likewise.
526         (extendqihi2, extendqisi2): Likewise.
527         (rotlqi3, rotlhi3, rotlsi3): Likewise.
528         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
529         (rotlqi3_1, rotlhi3_1): Likewise.
530         (logicalhi3_sn, logicalsi3_sn): Likewise.
531         (logicalhi3, logicalsi3): Likewise.
533 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
535         * tree-vect-patterns.c (vect_recog_rotate_pattern)
536         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
537         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
538         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
539         type to append_pattern_def_seq instead of creating a stmt_vec_info
540         directly.
541         (build_mask_conversion): Likewise.  Remove vinfo argument.
542         (vect_add_conversion_to_patterm): Likewise, renaming to...
543         (vect_add_conversion_to_pattern): ...this.
544         (vect_recog_mask_conversion_pattern): Update call to
545         build_mask_conversion.  Pass the vector type to
546         append_pattern_def_seq here too.
547         (vect_recog_gather_scatter_pattern): Update call to
548         vect_add_conversion_to_pattern.
550 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
552         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
553         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
554         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
555         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
556         STMT_VINFO_PATTERN_DEF_SEQ to null here.
557         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
558         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
559         append_pattern_def_seq instead of new_pattern_def_seq.
560         (vect_recog_divmod_pattern): Do both of the above.
561         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
562         is null.
564 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
566         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
567         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
568         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
569         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
570         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
571         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
572         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
573         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
574         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
575         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
576         parameter with a single stmt_vec_info.
577         (vect_recog_func_ptr): Likewise.
578         (vect_recog_gather_scatter_pattern): Likewise, folding in...
579         (vect_try_gather_scatter_pattern): ...this.
580         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
581         the stmt_vec_info of the statement to be matched.  Don't clear
582         STMT_VINFO_RELATED_STMT.
583         (vect_pattern_recog): Update call accordingly.
585 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
587         PR tree-optimization/85694
588         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
589         (uavgM3_ceil): Document new optabs.
590         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
591         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
592         functions.
593         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
594         (savg_ceil_optab): New optabs.
595         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
596         (vect_vect_recog_func_ptrs): Add it.
597         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
598         constant directly from the associated lhs.
600 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
602         * tree-vect-patterns.c (vect_split_statement): New function.
603         (vect_convert_input): Use it to try to split an existing cast.
605 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
607         * poly-int.h (print_hex): New function.
608         * dumpfile.h (dump_dec, dump_hex): Declare.
609         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
610         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
611         min_input_precision, operation_precision and operation_sign.
612         * tree-vect-patterns.c (vect_get_range_info): New function.
613         (vect_same_loop_or_bb_p, vect_single_imm_use)
614         (vect_operation_fits_smaller_type): Delete.
615         (vect_look_through_possible_promotion): Add an optional
616         single_use_p parameter.
617         (vect_recog_over_widening_pattern): Rewrite to use new
618         stmt_vec_info infomration.  Handle one operation at a time.
619         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
620         (vect_truncatable_operation_p, vect_set_operation_type)
621         (vect_set_min_input_precision): New functions.
622         (vect_determine_min_output_precision_1): Likewise.
623         (vect_determine_min_output_precision): Likewise.
624         (vect_determine_precisions_from_range): Likewise.
625         (vect_determine_precisions_from_users): Likewise.
626         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
627         (vect_vect_recog_func_ptrs): Put over_widening first.
628         Add cast_forwprop.
629         (vect_pattern_recog): Call vect_determine_precisions.
631 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
633         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
634         statements that have been replaced by further pattern statements.
635         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
637 2018-07-03  Richard Biener  <rguenther@suse.de>
639         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
640         always set *dt.  Dump vectype in vectype overload.
641         * dumpfile.h (dump_gimple_expr): New function.
642         (dump_gimple_expr_loc): Likewise.
643         * dumpfile.c (dump_gimple_expr): New function.
644         (dump_gimple_expr_loc): Likewise.
646 2018-07-02  Jeff Law  <law@redhat.com>
648         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
649         the H8/300, H8/300H and H8/S variants into a single pattern.
650         (movhi_h8300, movqi_h8300hs): Similarly.
651         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
652         (QHI mode iterator): New.
654         * config/h8300/h8300.md: Remove trailing whitespace.
656 2018-07-02  Jim Wilson  <jimw@sifive.com>
658         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
659         instead of emit_insn for interrupt returns.
660         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
661         (riscv_sret, riscv_uret): Likewise.
663 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
665         * pretty-print.c (selftest::test_pp_format): Move save and restore
666         of quotes to class auto_fix_quotes, and add an instance.
667         * selftest.c: Include "intl.h".
668         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
669         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
670         * selftest.h (selftest::auto_fix_quotes): New class.
672 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
674         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
675         (aarch64_sve_prepare_conditional_op): Remove.
676         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
677         Allow aarch64_simd_reg_or_zero as select operand; remove
678         the aarch64_sve_prepare_conditional_op call.
679         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
680         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
681         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
682         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
683         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
684         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
685         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
686         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
687         and a splitters to match all of the *_any patterns.
688         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
690         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
691         (SVE_COND_FP_BINARY_REV): Remove.
692         (sve_int_op_rev, sve_fp_op_rev): New.
693         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
694         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
695         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
696         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
697         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
698         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
699         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
700         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
701         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
703         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
704         Remove match_dup 1 from the inner unspec.
705         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
707         * config/aarch64/aarch64.md (movprfx): New attr.
708         (length): Default movprfx to 8.
709         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
710         (*madd<SVE_I>, *msub<SVE_I): Likewise.
711         (*<su>mul<SVE_I>3_highpart): Likewise.
712         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
713         (*v<ASHIFT><SVE_I>3): Likewise.
714         (*<su><MAXMIN><SVE_I>3): Likewise.
715         (*<su><MAXMIN><SVE_F>3): Likewise.
716         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
717         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
718         (*div<SVE_F>4): Likewise.
720 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
722         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
723         in dump string.
725 2018-07-02  Richard Biener  <rguenther@suse.de>
727         PR tree-optimization/86363
728         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
729         memset argument refers to a non-variable address.
731 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
733         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
734         {PLUS,MINUS}_EXPR code to...
735         (adjust_symbolic_bound): ...here,
736         (combine_bound): ...here,
737         (set_value_range_with_overflow): ...and here.
739 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
741         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
742         code...
743         (extract_range_from_abs_expr): ...here.
745 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
747         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
748         -fno-omit-frame-pointer when not optimizing.
750 2018-07-02  Martin Liska  <mliska@suse.cz>
752         PR ipa/86279
753         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
754         (propagate_nothrow): Likewise.
756 2018-07-02  Martin Liska  <mliska@suse.cz>
758         PR ipa/86323
759         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
761 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
763         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
764         function in r262149, changing "loc" param from source_location to
765         const dump_location_t &.
766         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
767         declaration, as above.
769 2018-07-01  Paul Koning  <ni1d@arrl.net>
771         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
772         -munit-asm, -mgnu-asm, -mdec-asm.
773         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
774         (pdp11_output_labelref): New.
775         (pdp11_output_def): New.
776         (pdp11_output_addr_vec_elt): New.
777         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
778         %# and %@ format codes.
779         (pdp11_option_override): New.
780         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
781         (pdp11_output_ident): New.
782         (pdp11_asm_named_section): New.
783         (pdp11_asm_init_sections): New.
784         (pdp11_file_start): New.
785         (pdp11_file_end): New.
786         (output_ascii): Use .ascii/.asciz for -mdec-asm.
787         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
788         %o, like %c but octal.
789         (pdp11_option_override): New.
790         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
791         -mdec-asm.
792         (DATA_SECTION_ASM_OP): Ditto.
793         (READONLY_DATA_SECTION_ASM_OP): New.
794         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
795         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
796         (ASM_OUTPUT_LABELREF): Ditto.
797         (ASM_OUTPUT_DEF): Ditto.
798         (ASM_OUTPUT_EXTERNAL): New.
799         (ASM_OUTPUT_SOURCE_FILENAME): New.
800         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
801         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
802         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
803         %# and %@ format codes.
804         * config/pdp11/pdp11.opt (mgnu-asm): New.
805         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
806         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
807         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
809 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
811         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
812         dereferencing path[] beyond its length.
813         (debug_path): New.
814         (debug_all_paths): New.
815         (rewire_first_differing_edge): New.
816         (adjust_paths_after_duplication): New.
817         (duplicate_thread_path): Call adjust_paths_after_duplication.
818         Add new argument.
819         (thread_through_all_blocks): Add new argument to
820         duplicate_thread_path.
822 2018-06-30  Jim Wilson  <jimw@sifive.com>
824         * config/riscv/predicates.md (p2m1_shift_operand): New.
825         (high_mask_shift_operand): New.
826         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
827         pattern using p2m1_shift_operand.
828         (lshsi3_zero_extend_3+2): New combiner pattern using
829         high_mask_shift_operand.
831 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
833         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
834         split out from...
835         (vect_recog_rotate_pattern): ...here.
836         (vect_convert_input): Try to insert casts of invariants in the
837         preheader.
838         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
839         preheader to be empty.
841 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
843         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
844         vector type.  If given, install it in the new statement's
845         STMT_VINFO_VECTYPE.
846         (vect_element_precision): New function.
847         (vect_unpromoted_value): New struct.
848         (vect_unpromoted_value::vect_unpromoted_value): New function.
849         (vect_unpromoted_value::set_op): Likewise.
850         (vect_look_through_possible_promotion): Likewise.
851         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
852         (vect_widened_op_tree, vect_convert_input): Likewise.
853         (vect_convert_inputs, vect_convert_output): Likewise.
854         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
855         to handle the optional cast of the multiplication result and
856         vect_widened_op_tree to detect the widened multiplication itself.
857         Do not require the input and output of promotion casts to have
858         the same sign, but base the signedness of the operation on the
859         input rather than the result.  If the pattern includes two
860         promotions, check that those promotions have the same sign.
861         Do not restrict the MULT_EXPR handling to a double-width result;
862         handle quadruple-width results and wider.  Use vect_convert_inputs
863         to convert the inputs to the common type.
864         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
865         to handle the optional cast of the ABS result.  Also allow a sign
866         change or a sign extension between the ABS and MINUS.
867         Use vect_widened_op_tree to detect the widened subtraction and use
868         vect_convert_inputs to convert the inputs to the common type.
869         (vect_handle_widen_op_by_const): Delete.
870         (vect_recog_widen_op_pattern): New function.
871         (vect_recog_widen_mult_pattern): Use it.
872         (vect_recog_widen_shift_pattern): Likewise.
873         (vect_recog_widen_sum_pattern): Use
874         vect_look_through_possible_promotion to handle the promoted
875         PLUS_EXPR operand.
877 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
879         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
880         the containing gimple_seq *.
881         * gimple-iterator.h (gsi_for_stmt): Declare it.
882         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
883         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
884         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
885         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
886         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
887         checks.
888         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
889         split out from...
890         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
891         statement being replaced is part of an existing pattern
892         definition sequence, inserting the new pattern statements before
893         the original one.
894         (vect_pattern_recog_1): Don't return a bool.  If the statement
895         is already part of a pattern, instead apply pattern matching
896         to the pattern definition statements.  Don't clear the
897         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
898         (vect_pattern_recog): Don't break after the first match;
899         continue processing the pattern definition statements instead.
900         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
902 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
904         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
905         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
906         (vect_recog_widen_sum_pattern): Use it.
908 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
910         * tree-vect-loop.c (vectorizable_reduction): Assert that the
911         phi is not a pattern statement and has not been replaced by
912         a pattern statement.
913         * tree-vect-patterns.c (type_conversion_p): Don't check
914         STMT_VINFO_IN_PATTERN_P.
915         (vect_recog_vector_vector_shift_pattern): Likewise.
916         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
917         the pattern statement rather than the original statement; check
918         directly for a WIDEN_MULT_EXPR here.
919         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
920         vect_is_simple_use to return the pattern statement rather
921         than the original statement; use is_pattern_stmt_p to check
922         for such a pattern statement.
923         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
924         to return the pattern statement rather than the original statement;
925         don't do the same transformation here.
926         (vect_is_simple_use): If the defining statement has been replaced
927         by a pattern statement, return the pattern statement instead.
928         Remove the corresponding (local) transformation from the vectype
929         overload.
931 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
933         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
934         end and default to null.
935         * tree-vect-loop.c (vect_create_epilog_for_reduction)
936         (vectorizable_reduction): Update calls accordingly, dropping the
937         gimple ** argument if the passed-back statement isn't needed.
938         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
939         (vect_recog_rotate_pattern): Likewise.
940         (vect_recog_mask_conversion_pattern): Likewise.
941         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
942         (vect_mask_constant_operand_p): Likewise.
943         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
944         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
945         (get_group_load_store_type, get_load_store_type): Likewise.
946         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
947         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
948         (vectorizable_conversion, vectorizable_assignment): Likewise.
949         (vectorizable_shift, vectorizable_operation): Likewise.
950         (vectorizable_store, vect_is_simple_cond): Likewise.
951         (vectorizable_condition, vectorizable_comparison): Likewise.
952         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
953         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
954         and move it to the end.  Cope with null def_stmt_outs.
956 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
958         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
960 2018-06-29  Jeff Law  <law@redhat.com>
962         * config/v850/v850.c (v850_legitimate_address_p): Handle large
963         displacements for TARGET_V850E2V3 and newer.
964         (TARGET_LRA_P): Remove.  Defaults to LRA now.
965         * config/v850/v850.md (sign23byte_load): Remove.
966         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
967         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
969 2018-06-29  Martin Liska  <mliska@suse.cz>
971         PR lto/85759
972         * coverage.c (coverage_init): Mangle full path name.
973         * doc/invoke.texi: Document the change.
974         * gcov-io.c (mangle_path): New.
975         * gcov-io.h (mangle_path): Likewise.
976         * gcov.c (mangle_name): Use mangle_path for path mangling.
978 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
980         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
981         if starting source register is not even.
983 2018-06-29  Martin Liska  <mliska@suse.cz>
985         PR tree-optimization/86263
986         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
987         Make edge redirection.
989 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
991         * dumpfile.c (dump_loc): Add indentation based on scope depth.
992         (dump_scope_depth): New variable.
993         (get_dump_scope_depth): New function.
994         (dump_begin_scope): New function.
995         (dump_end_scope): New function.
996         * dumpfile.h (get_dump_scope_depth): New declaration.
997         (dump_begin_scope): New declaration.
998         (dump_end_scope): New declaration.
999         (class auto_dump_scope): New class.
1000         (AUTO_DUMP_SCOPE): New macro.
1001         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1002         AUTO_DUMP_SCOPE.
1004 2018-06-29  Richard Biener  <rguenther@suse.de>
1006         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1007         compute_all_dependences succeeds.
1008         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1009         exceed --param loop-max-datarefs-for-datadeps.
1011 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
1013         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1015 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
1017         PR target/86348
1018         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1019         alternative 0 in preferred_for_speed attribute.
1021 2018-06-28  Paul Koning  <ni1d@arrl.net>
1023         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1024         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1025         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1026         * config/pdp11/pdp11.md: Correct "length" attribute calculation
1027         for shift insn patterns.
1029 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
1031         * cgraph.c (cgraph_node::get_body): Replace assignments to
1032         "dump_file" with calls to set_dump_file.
1033         * dumpfile.c (alt_dump_file): Make static, and group with...
1034         (alt_flags): ...this definition.
1035         (dumps_are_enabled): New variable.
1036         (refresh_dumps_are_enabled): New function.
1037         (set_dump_file): New function.
1038         (set_alt_dump_file): New function.
1039         (gcc::dump_manager::dump_start): Replace assignments to
1040         "dump_file" and "alt_dump_file" with calls to set_dump_file and
1041         set_alt_dump_file.
1042         (gcc::dump_manager::dump_finish): Likewise.
1043         * dumpfile.h (alt_dump_file): Delete decl.
1044         (dumps_are_enabled): New variable decl.
1045         (set_dump_file): New function decl.
1046         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1047         global.
1048         * tree-nested.c (lower_nested_functions): Replace assignments to
1049         "dump_file" with calls to set_dump_file.
1051 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
1053         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1054         goto_locus of each outgoing edge of each basic block.
1056 2018-06-28  Richard Biener  <rguenther@suse.de>
1058         * dwarf2out.c (decl_scope_table): Remove.
1059         (push_decl_scope): Likewise.
1060         (pop_decl_scope): Likewise.
1061         (gen_type_die_for_member): Do not call push/pop_decl_scope.
1062         (gen_struct_or_union_type_die): Likewise.
1063         (gen_tagged_type_die): Likewise.
1064         (dwarf2out_init): Do not initialize decl_scope_table.
1065         (dwarf2out_c_finalize): Do not free it.
1067 2018-06-28  Richard Biener  <rguenther@suse.de>
1069         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1070         deciding whether to not re-use a DIE.
1072 2018-06-28  Richard Biener  <rguenther@suse.de>
1074         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1075         DW_AT_abstract_origin attribute.
1077 2018-06-28  Martin Liska  <mliska@suse.cz>
1079         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1080         Use newly introduced constants.
1081         * tree-switch-conversion.h (struct jump_table_cluster):
1082         Define max_ratio_for_size and max_ratio_for_speed.
1084 2018-06-28  Martin Liska  <mliska@suse.cz>
1086         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1087         Add new checking assert to catch invalid state.
1088         (jump_table_cluster::can_be_handled): Handle single case
1089         clusters.
1090         (jump_table_cluster::is_beneficial): Bail out for such case.
1091         (bit_test_cluster::find_bit_tests):
1092         Add new checking assert to catch invalid state.
1093         (bit_test_cluster::can_be_handled): Handle single case
1094         clusters.
1095         (bit_test_cluster::is_beneficial): Bail out for such case.
1096         (switch_decision_tree::analyze_switch_statement):
1097         Fix comment.
1099 2018-06-28  Martin Liska  <mliska@suse.cz>
1101         * common.opt: Introduce -completion option.
1102         * gcc.c (driver_handle_option): Handle it.
1103         (driver::main): Print completions if completion
1104         is set.
1105         * opt-suggestions.c (option_proposer::get_completions):
1106         New function.
1107         (option_proposer::suggest_completion): Likewise.
1108         (option_proposer::find_param_completions): Likewise.
1109         (verify_autocompletions): Likewise.
1110         (test_completion_valid_options): Likewise.
1111         (test_completion_valid_params): Likewise.
1112         (in_completion_p): Likewise.
1113         (empty_completion_p): Likewise.
1114         (test_completion_partial_match): Likewise.
1115         (test_completion_garbage): Likewise.
1116         (opt_proposer_c_tests): Likewise.
1117         * opt-suggestions.h: Declare new functions.
1118         * opts.c (common_handle_option): Handle OPT__completion_.
1119         * selftest-run-tests.c (selftest::run_tests): Add
1120         opt_proposer_c_tests.
1121         * selftest.c (assert_str_startswith): New.
1122         * selftest.h (assert_str_startswith): Likewise.
1123         (opt_proposer_c_tests): New.
1124         (ASSERT_STR_STARTSWITH): Likewise.
1126 2018-06-28  Martin Liska  <mliska@suse.cz>
1128         * Makefile.in: Add opt-suggestions.o.
1129         * gcc-main.c: Include opt-suggestions.h.
1130         * gcc.c (driver::driver): Likewise.
1131         (driver::~driver): Remove m_option_suggestions.
1132         (driver::build_option_suggestions): Moved to option_proposer.
1133         (driver::suggest_option): Likewise.
1134         (driver::handle_unrecognized_options): Use option_proposer.
1135         * gcc.h (class driver): Add new memver m_option_proposer.
1136         * opt-suggestions.c: New file.
1137         * opt-suggestions.h: New file.
1139 2018-06-28  Martin Liska  <mliska@suse.cz>
1141         * vec.h (class auto_string_vec): New (moved from auto_argvec).
1142         (auto_string_vec::~auto_string_vec): Likewise.
1144 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
1146         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1147         prevent_decl_creation_for_types fields up and add reset_location field.
1148         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1149         statement if id->reset_location is true.
1150         (copy_edges_for_bb): Do not set goto_locus on the new edges if
1151         id->reset_location is true.
1152         (copy_phis_for_bb): Force input_location on the arguments if
1153         id->reset_location is true.
1154         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1155         is set on the function to be inlined.
1157 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
1159         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1161 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
1163         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1164         registers for Pmode.
1165         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1166         hard registers for the clobbered pseudo.
1168 2018-06-27  Paul Koning  <ni1d@arrl.net>
1170         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1171         mutually exclusive options.
1172         * config/pdp11/constraints.md (h): New constraint.
1173         (O): Update definition to match shift code generation.
1174         (D): New constraint.
1175         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1176         (CCFP): Remove.
1177         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1178         function.
1179         (output_jump): Change arguments.
1180         (pdp11_fixed_cc_regs): New function.
1181         (pdp11_cc_mode): Ditto.
1182         (pdp11_expand_shift): Ditto.
1183         (pdp11_assemble_shift): Ditto.
1184         (pdp11_small_shift): Ditto.
1185         (pdp11_branch_cost): Remove.
1186         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1187         from output.
1188         (pdp11_register_move_cost): Update for CC registers.
1189         (pdp11_rtx_costs): Add case for LSHIFTRT.
1190         (pdp11_output_jump): Add CCNZ mode conditional branches.
1191         (notice_update_cc_on_set): Remove.
1192         (pdp11_cc_mode): New function.
1193         (simple_memory_operand): Correct pre/post decrement case.
1194         (no_side_effect_operand): New function.
1195         (pdp11_regno_reg_class): Add CC_REGS class.
1196         (pdp11_fixed_cc_regs): New function.
1197         (pdp11_small_shift): New function.
1198         (pdp11_expand_shift): New function to expand shift insns.
1199         (pdp11_assemble_shift): New function to output shifts.
1200         (pdp11_branch_cost): Remove.
1201         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1202         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1203         (WCHAR_TYPE): Ditto.
1204         (PTRDIFF_TYPE): Ditto.
1205         (ADJUST_INSN_LENGTH): New macro.
1206         (FIXED_REGISTERS): Add CC registers.
1207         (CALL_USED_REGISTERS): Ditto.
1208         (reg_class): Ditto.
1209         (REG_CLASS_NAMES): Ditto.
1210         (REG_CLASS_CONTENTS): Ditto.
1211         (SELECT_CC_MODE): Use new function.
1212         (TARGET_FLAGS_REGNUM): New macro.
1213         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1214         (cc0_reg_rtx): Remove.
1215         (CC_STATUS_MDEP): Remove.
1216         (CC_STATUS_MDEFP_INIT): Remove.
1217         (CC_IN_FPU): Remove.
1218         (NOTICE_UPDATE_CC): Remove.
1219         (REGISTER_NAMES): Add CC registers.
1220         (BRANCH_COST): Change to constant 1.
1221         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1222         handling.
1223         * config/pdp11/pdp11.opt (mbcopy): Remove.
1224         (mbcopy-builtin): Remove.
1225         (mbranch-cheap): Remove.
1226         (mbranch-expensive): Remove.
1227         * config/pdp11/predicates.md (expand_shift_operand): Update to
1228         match shift code generation.
1229         (ccnz_operator): New predicate.
1230         * doc/invoke.texi (PDP-11 Options): Remove deleted options
1231         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1232         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
1233         exclusive options.
1234         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
1235         description of O constraint.
1237 2018-06-27  Jeff Law  <law@redhat.com>
1238             Austin Law  <austinklaw@gmail.com>
1240         * config/v850/v850.md (addsi3_set_flags): New pattern.
1241         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1242         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1243         (zero_extendhisi2_v850_set_flags): Likewise.
1244         (zero_extendqisi2_v850_set_flags): Likewise.
1245         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1246         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1247         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1249         * config/v850/v850-protos.h (notice_update_cc): Remove.
1250         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1251         (v850_print_operand): Handle 'D' and "d".
1252         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1253         Add handling of arithmetic/logical operations compared against zero.
1254         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1255         Do not look at v850_compare_op, instead get mode from last argument.
1256         (v850_gen_compare): Remove
1257         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1258         after reload for prologue insns.
1259         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1260         patterns.
1261         (construct_save_jarl): Likewise.
1262         (TARGET_FLAGS_REGNUM): Define.
1263         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1264         (NOTICE_UPDATE_CC): Remove.
1265         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1266         than cc0.  Conditionalize on reload_completed.
1267         (cmpsi_insn, setfcc_insn): Likewise.
1268         (tst1 splitter): Turn into define_and_split which sets the flags
1269         after reload.
1270         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1271         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1272         (cstoresf4, cstoredf4): Clobber the flags.
1273         (cmpsi, cmpsf, cmpdf): Remove expanders.
1274         (setf_insn): Remove pattern.
1275         (addsi3): Turn into define_and_split which clobbers the flags after
1276         reload and a suitable pattern (addsi3_clobber_flags) for use after
1277         reload.
1278         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1279         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1280         (ashrsi3, ashrsi3_v850e2): Likewise.
1281         (bins): Clobber the flags.
1282         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1283         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1284         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1285         (call_value_internal_short, call_value_internal_long): Likewise.
1286         (callt_save_interrupt, callt_return_interrupt): Likewise.
1287         (save_interrupt, return_interrupt): Likewise.
1288         (callt_save_all_interrupt, save_all_interrupt): Likewise.
1289         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1290         (restore_all_interrupt, _restore_all_interrupt): Likewise.
1291         (All FP comparisons): Only allow after reload has completed.
1292         (trfsr): Likewise.
1293         (divh, divhu): Tweak output template.
1294         (branch_z_normal, branch_z_invert): Remove
1295         (branch_nz_normal, branch_nz_invert): Likewise.
1296         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1298         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1299         * config/v850/v850.c (notice_update_cc): Remove.
1300         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
1301         (CC_NO_CARRY): Likewise.
1302         (NOTICE_UPDATE_CC): Define to nothing.
1303         * config/v850/v850.md: Remove block comment on cc0 handling
1304         Remove "cc" attribute from all patterns.  Remove cc_status handling
1305         from all patterns.  Minor formatting fixes.
1307 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1309         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1310         (cortex-a76.cortex-a55): Likewise.
1311         * config/aarch64/aarch64-tune.md: Regenerate.
1312         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1313         cortex-a76.cortex-a55.
1315 2018-06-27  Jeff Law  <law@redhat.com>
1317         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1318         (MULTILIB_DIRNAMES): Similarly.
1320 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
1322         * gimple.h (gimple_return_retbnd): Delete.
1323         (gimple_return_set_retbnd): Likewise.
1324         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1325         gimple_return_set_retbnd.
1326         * gimple-pretty-print.c (dump_gimple_return): Remove call to
1327         gimple_return_retbnd and adjust.
1328         * tree-inline.h (struct copy_body_data): Remove retbnd field.
1329         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1330         Explicitly return NULL in a couple more cases.  Move assertion
1331         on debug statements and remove unreachable code.
1332         (reset_debug_binding): Do not test id->retbnd.
1333         (expand_call_inline): Do not set it.
1335 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
1337         * configure.ac: Add --disable-gcov option.
1338         * configure: Regenerate.
1339         * Makefile.in: Honour @enable_gcov@.
1340         * doc/install.texi: Document --disable-gcov.
1342 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1344         * config/arm/arm-cpus.in (cortex-a76): New entry.
1345         (cortex-a76.cortex-a55): Likewise.
1346         * config/arm/arm-tables.opt: Regenerate.
1347         * config/arm/arm-tune.md: Likewise.
1348         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1349         * doc/invoke.texi (ARM Options): Document cortex-a76 and
1350         cortex-a76.cortex-a55.
1352 2018-06-27 Tamar Christina  <tamar.christina@arm.com>
1354         PR target/85769
1355         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1357 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
1359         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1360         comment.
1361         (EPILOGUE_USES): Likewise.
1363 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
1365         * tree-inline.c (remap_location): New function extracted from...
1366         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
1367         (copy_phis_for_bb): ...here.  Call remap_location.
1368         (copy_cfg_body): Adjust call to copy_edges_for_bb.
1370 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1372         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1373         unaligned vsx for 16B memset.
1375 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
1377         PR target/86285
1378         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1379         ieee128_float_type_node to long_double_type_node unless
1380         TARGET_LONG_DOUBLE_128 is set.
1382 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
1384         * cfgloop.c (get_loop_location): Convert return type from
1385         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1386         by implicit construction from rtx_insn *, and using
1387         dump_user_location_t::from_function_decl for the fallback case.
1388         * cfgloop.h (get_loop_location): Convert return type from
1389         location_t to dump_user_location_t.
1390         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1391         dump_printf_loc to pass in a dump_location_t rather than a
1392         location_t, via the gimple stmt.
1393         * coverage.c (get_coverage_counts): Update calls to
1394         dump_printf_loc to pass in dump_location_t rather than a
1395         location_t.
1396         * doc/optinfo.texi (Dump types): Convert example of
1397         dump_printf_loc from taking "locus" to taking "insn".  Update
1398         description of the "_loc" calls to cover dump_location_t.
1399         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1400         "selftest.h".
1401         (dump_user_location_t::dump_user_location_t): New constructors,
1402         from gimple *stmt and rtx_insn *.
1403         (dump_user_location_t::from_function_decl): New function.
1404         (dump_loc): Make static.
1405         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1406         const dump_location_t &.
1407         (dump_generic_expr_loc): Delete.
1408         (dump_printf_loc): Convert param "loc" from location_t to
1409         const dump_location_t &.
1410         (selftest::test_impl_location): New function.
1411         (selftest::dumpfile_c_tests): New function.
1412         * dumpfile.h: Include "profile-count.h".
1413         (class dump_user_location_t): New class.
1414         (struct dump_impl_location_t): New struct.
1415         (class dump_location_t): New class.
1416         (dump_printf_loc): Convert 2nd param from source_location to
1417         const dump_location_t &.
1418         (dump_generic_expr_loc): Delete.
1419         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1420         const dump_location_t &.
1421         * gimple-fold.c (fold_gimple_assign): Update call to
1422         dump_printf_loc to pass in a dump_location_t rather than a
1423         location_t, via the gimple stmt.
1424         (gimple_fold_call): Likewise.
1425         * gimple-loop-interchange.cc
1426         (loop_cand::analyze_iloop_reduction_var): Update for change to
1427         check_reduction_path.
1428         (tree_loop_interchange::interchange): Update for change to
1429         find_loop_location.
1430         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1431         change in return-type of find_loop_location.
1432         (graphite_regenerate_ast_isl): Likewise.
1433         * graphite-optimize-isl.c (optimize_isl): Likewise.
1434         * graphite.c (graphite_transform_loops): Likewise.
1435         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1436         pass in a dump_location_t rather than a location_t, via the
1437         gimple stmt.
1438         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1439         * ipa.c (walk_polymorphic_call_targets): Likewise.
1440         * loop-unroll.c (report_unroll): Convert "locus" param from
1441         location_t to dump_location_t.
1442         (decide_unrolling): Update for change to get_loop_location's
1443         return type.
1444         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1445         location_t to dump_user_location_t.
1446         (grid_find_single_omp_among_assignments_1): Updates calls to
1447         dump_printf_loc to pass in a dump_location_t rather than a
1448         location_t, via the gimple stmt.
1449         (grid_parallel_clauses_gridifiable): Convert "tloc" from
1450         location_t to dump_location_t.  Updates calls to dump_printf_loc
1451         to pass in a dump_location_t rather than a location_t, via the
1452         gimple stmt.
1453         (grid_inner_loop_gridifiable_p): Likewise.
1454         (grid_dist_follows_simple_pattern): Likewise.
1455         (grid_gfor_follows_tiling_pattern): Likewise.
1456         (grid_target_follows_gridifiable_pattern): Likewise.
1457         (grid_attempt_target_gridification): Convert initialization
1458         of local "grid" from memset to zero-initialization; FIXME: does
1459         this require C++11?  Update call to dump_printf_loc to pass in a
1460         optinfo_location rather than a location_t, via the gimple stmt.
1461         * profile.c (read_profile_edge_counts): Updates call to
1462         dump_printf_loc to pass in a dump_location_t rather than a
1463         location_t
1464         (compute_branch_probabilities): Likewise.
1465         * selftest-run-tests.c (selftest::run_tests): Call
1466         dumpfile_c_tests.
1467         * selftest.h (dumpfile_c_tests): New decl.
1468         * tree-loop-distribution.c (pass_loop_distribution::execute):
1469         Update for change in return type of find_loop_location.
1470         * tree-parloops.c (parallelize_loops): Likewise.
1471         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1472         "locus" from location_t to dump_user_location_t.
1473         (canonicalize_loop_induction_variables): Likewise.
1474         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1475         for change in return type of find_loop_location.
1476         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1477         to dump_printf_loc to pass in a dump_location_t rather than a
1478         location_t, via the stmt.
1479         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1480         Likewise.
1481         * tree-vect-loop-manip.c (find_loop_location): Convert return
1482         type from source_location to dump_user_location_t.
1483         (vect_do_peeling): Update for above change.
1484         (vect_loop_versioning): Update for change in type of
1485         vect_location.
1486         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1487         from location_t to dump_user_location_t.
1488         (vect_estimate_min_profitable_iters): Update for change in type
1489         of vect_location.
1490         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1491         location_t to dump_location_t.
1492         (vect_slp_bb): Update for change in type of vect_location.
1493         * tree-vectorizer.c (vect_location): Convert from source_location
1494         to dump_user_location_t.
1495         (try_vectorize_loop_1): Update for change in vect_location's type.
1496         (vectorize_loops): Likewise.
1497         (increase_alignment): Likewise.
1498         * tree-vectorizer.h (vect_location): Convert from source_location
1499         to dump_user_location_t.
1500         (find_loop_location): Convert return type from source_location to
1501         dump_user_location_t.
1502         (check_reduction_path): Convert 1st param from location_t to
1503         dump_user_location_t.
1504         * value-prof.c (check_counter): Update call to dump_printf_loc to
1505         pass in a dump_user_location_t rather than a location_t; update
1506         call to error_at for change in type of "locus".
1507         (check_ic_target): Update call to dump_printf_loc to
1508         pass in a dump_user_location_t rather than a location_t, via the
1509         call_stmt.
1511 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
1513         * config/s390/s390.h (enum processor_flags): Do not use
1514         default tune parameter when -march was specified.
1516 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
1518         PR target/86314
1519         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1520         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1521         operands.
1523 2018-06-26  Richard Biener  <rguenther@suse.de>
1525         PR tree-optimization/86287
1526         PR bootstrap/86316
1527         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1528         (vect_analyze_loop): Initialize n_stmts.
1530 2018-06-26  Richard Biener  <rguenther@suse.de>
1532         PR middle-end/86271
1533         * fold-const.c (fold_convertible_p): Pointer extension
1534         isn't valid.
1536 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
1538         PR debug/86064
1539         * dwarf2out.c (loc_list_has_views): Adjust comments.
1540         (dw_loc_list): Split single cross-partition range with
1541         nonzero locview.
1543 2018-06-25  Jeff Law  <law@redhat.com>
1545         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1546         on -mbig-switch by default.
1548         * config/v850/predicates.md (const_float_1_operand): Fix match_code
1549         test.
1550         (const_float_0_operand): Remove unused predicate.
1551         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1552         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1553         (recipsf2): New expander.  Original pattern now called
1554         (recipsf2_insn).
1555         (recipdf2, recipdf2_insn): Similarly.
1556         (rsqrtsf2, rsqrtsf2_insn): Similarly
1557         (rsqrtdf2, rsqrtdf2_insn): Similarly
1559 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
1561         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1562         Simplify logic for FreeBSD (twice).
1564 2018-06-25  Martin Sebor  <msebor@redhat.com>
1566         PR tree-optimization/86204
1567         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1568         a strnlen result if it's less than the length of the string.
1570 2018-06-25  Martin Sebor  <msebor@redhat.com>
1572         PR tree-optimization/85700
1573         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1574         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1575         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1577 2018-06-25  Martin Sebor  <msebor@redhat.com>
1579         * doc/extend.texi (Zero-length arrays): Update and clarify.
1581 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
1583         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1584         added IEEE/IBM long double multilib support on PowerPC little
1585         endian Linux systems.
1586         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1587         (MULTILIB_DEFAULTS): Likewise.
1588         * config/rs6000/rs6000.c (rs6000_option_override_internal):
1589         Likewise.
1590         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1591         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1592         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1594 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
1596         PR middle-end/86311
1597         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1598         (REORDER_45): Likewise.
1600 2018-06-25  Jeff Law  <law@redhat.com>
1602         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1603         dividend to 32 bits.  Adjust length.
1604         (udivmodhi4): Cleanup output template.  Fix length.
1606 2018-06-25  Carl Love  <cel@us.ibm.com>
1608         * config/rs6000/vsx.md: Change word selector to prefered location.
1610 2018-06-25  Richard Biener  <rguenther@suse.de>
1612         PR tree-optimization/86304
1613         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1614         epilogue-if-converted loops as well.
1616 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
1618         * lto-section-out.c (lto_begin_section): Do not print section
1619         name for noaddr and unnumbered dumps.
1621 2018-06-25  Richard Biener  <rguenther@suse.de>
1623         * tree-vectorizer.h (struct vec_info_shared): New structure
1624         with parts split out from struct vec_info and loop_nest from
1625         struct _loop_vec_info.
1626         (struct vec_info): Adjust accordingly.
1627         (struct _loop_vec_info): Likewise.
1628         (LOOP_VINFO_LOOP_NEST): Adjust.
1629         (LOOP_VINFO_DATAREFS): Likewise.
1630         (LOOP_VINFO_DDRS): Likewise.
1631         (struct _bb_vec_info): Likewise.
1632         (BB_VINFO_DATAREFS): Likewise.
1633         (BB_VINFO_DDRS): Likewise.
1634         (struct _stmt_vec_info): Add dr_aux member.
1635         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1636         (DR_MISALIGNMENT_UNINITIALIZED): New.
1637         (set_dr_misalignment): Adjust.
1638         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1639         (vect_analyze_loop): Adjust prototype.
1640         (vect_analyze_loop_form): Likewise.
1641         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1642         Compute dependences lazily.
1643         (vect_record_base_alignments): Use shared datarefs/ddrs.
1644         (vect_verify_datarefs_alignment): Likewise.
1645         (vect_analyze_data_refs_alignment): Likewise.
1646         (vect_analyze_data_ref_accesses): Likewise.
1647         (vect_analyze_data_refs): Likewise.
1648         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1649         constructor parameter for shared part.
1650         (vect_analyze_loop_form): Pass in shared part and adjust.
1651         (vect_analyze_loop_2): Pass in storage for the number of
1652         stmts.  Move loop nest finding to the caller.  Compute
1653         datarefs lazily.
1654         (vect_analyze_loop): Pass in shared part.
1655         (vect_transform_loop): Verify shared datarefs are unchanged.
1656         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1657         constructor parameter for shared part.
1658         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1659         (vect_slp_bb): Verify shared datarefs are unchanged before
1660         transform.
1661         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1662         change.
1663         (new_stmt_vec_info): Initialize DR_AUX misalignment to
1664         DR_MISALIGNMENT_UNINITIALIZED.
1665         * tree-vectorizer.c (vec_info::vec_info): Add constructor
1666         parameter for shared part.
1667         (vec_info::~vec_info): Adjust.
1668         (vec_info_shared::vec_info_shared): New.
1669         (vec_info_shared::~vec_info_shared): Likewise.
1670         (vec_info_shared::save_datarefs): Likewise.
1671         (vec_info_shared::check_datarefs): Likewise.
1672         (try_vectorize_loop_1): Construct shared part live for analyses
1673         of a single loop for multiple vector sizes.
1674         * tree-parloops.c (gather_scalar_reductions): Adjust.
1676 2018-06-25  Richard Biener  <rguenther@suse.de>
1678         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1679         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1680         (vect_analyze_data_refs): Remove similar code from here and
1681         simplify accordingly.
1683 2018-06-25  Richard Biener  <rguenther@suse.de>
1685         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1686         for reverse storage order accesses rather than asserting
1687         they cannot happen here.
1689 2018-06-25  Tom de Vries  <tdevries@suse.de>
1691         PR debug/86257
1692         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1693         Use data16 instead of .byte for insn prefix.
1695 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
1697         PR C++/86082
1698         * parser.c (make_char_string_pack): Pass this literal chars
1699         through cpp_interpret_string.
1700         (cp_parser_userdef_numeric_literal): Check the result of
1701         make_char_string_pack.
1703 2018-06-24  Maya Rashish  <coypu@sdf.org>
1705         * ginclude/stddef.h: Simplify conditions around avoiding
1706         re-definition of __size_t.
1708 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
1710         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1711         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1713 2018-06-22  Maya Rashish  <coypu@sdf.org>
1715         * doc/invoke.texi (mno-fancy-math-387): Update for changes
1716         made to OpenBSD and NetBSD through the years.
1718 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1720         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1721         behavior of vec_pack (vector double, vector double) to match
1722         behavior of vec_float2 (vector double, vector double).
1724 2018-06-22  Olivier Hainque  <hainque@adacore.com>
1726         * gimplify.c (gimplify_function_tree): Prevent creation
1727         of a trampoline for the address of the current function
1728         passed to entry/exit instrumentation hooks.
1730 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
1732         PR target/86222
1733         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1734         correctly.
1736 2018-06-22  Martin Liska  <mliska@suse.cz>
1738         PR tree-optimization/86263
1739         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1740         Bail out if is_enabled is false.
1741         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1742         New declaration.
1743         (jump_table_cluster::is_enabled): New function.
1745 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
1747         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1748         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1749         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1750         (lto_input_ts_binfo_tree_pointers): Likewise.
1751         * tree-streamer-out.c (streamer_write_tree_bitfields,
1752         write_ts_binfo_tree_pointers): Likewise.
1753         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1755 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
1757         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1759 2018-06-22  Martin Liska  <mliska@suse.cz>
1761         * symbol-summary.h (get): Make it pure and inline move
1762         functionality from ::get function.
1763         (get): Remove and inline into ::get and ::get_create.
1764         (get_create): Move code from ::get function.
1766 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1768         PR target/85994
1769         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1770         -x assembler-with-cpp.
1772 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1774         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1775         _FILE_OFFSET_BITS=64 for C++.
1777 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
1779         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1780         conversion insn that shows up when pr85657-3.c is compiled using
1781         IEEE 128-bit long double.
1782         (neg<mode>2_internal): Use the correct mode to check whether the
1783         mode is IBM extended.
1784         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1785         multiply and divide external functions from being created more
1786         than once.
1788 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
1790         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1791         functions.
1792         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
1793         the edge can be forwarded.
1794         (cfg_layout_merge_blocks): Likewise.
1796 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
1798         * except.c (finish_eh_generation): Commit edge insertions only after
1799         the EH edges have been redirected from post-landing to landing pads.
1801 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
1803         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1804         create_tmp_var_for to create the FRAME decl.
1805         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1807 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
1809         * tree-inline.c (copy_edges_for_bb): Minor tweak.
1810         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1811         debug statement when resetting its value.
1812         (expand_call_inline): Copy the locus of the call onto the assignment
1813         of the return value, if any.  Use local variable in more cases.
1815 2018-06-21  Martin Liska  <mliska@suse.cz>
1817         * ipa-pure-const.c (propagate_nothrow): Use
1818         funct_state_summaries->get.
1819         (dump_malloc_lattice): Likewise.
1820         (propagate_malloc): Likewise.
1822 2018-06-21  Richard Biener  <rguenther@suse.de>
1824         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1825         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1826         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1827         comment.
1828         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1829         BLOCK_ABSTRACT_ORIGIN unconditionally.
1831 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
1833         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1834         deleting it.
1835         * ipa-reference.c (ipa_reference_c_finalize): Delete
1836         ipa_ref_opt_sum_summaries and set it to NULL.
1838 2018-06-21  Tom de Vries  <tdevries@suse.de>
1840         PR tree-optimization/85859
1841         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1842         test with comment from bb_no_side_effects_p.
1844 2018-06-21  Richard Biener  <rguenther@suse.de>
1846         PR tree-optimization/86232
1847         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1848         max for constant niter.
1850 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1852         * config/aarch64/aarch64-simd.md
1853         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1855 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1857         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1858         Make opernads of the unspec commutative.
1860 2018-06-21  Richard Biener  <rguenther@suse.de>
1862         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1863         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1864         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1865         (vect_analyze_data_ref_dependence): Re-order checks to deal with
1866         NULL DR_STEP.
1867         (vect_record_base_alignments): Do not record base alignment
1868         for gathers or scatters.
1869         (vect_compute_data_ref_alignment): Drop return value that is always
1870         true.  Bail out early for gathers or scatters.
1871         (vect_enhance_data_refs_alignment): Bail out early for gathers
1872         or scatters.
1873         (vect_find_same_alignment_drs): Likewise.
1874         (vect_analyze_data_refs_alignment): Remove dead code.
1875         (vect_slp_analyze_and_verify_node_alignment): Likewise.
1876         (vect_analyze_data_refs): For possible gathers or scatters do
1877         not create an alternate DR, just check their possible validity
1878         and mark them.  Adjust DECL_NONALIASED handling to not rely
1879         on DR_BASE_ADDRESS.
1880         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1881         update inits of gathers or scatters.
1882         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1883         Also copy gather/scatter flag to pattern vinfo.
1885 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1887         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1888         behavior of vec_packsu (vector unsigned long long, vector unsigned
1889         long long) to match behavior of vec_packs with same signature.
1891 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
1892             Thomas Schwinge <thomas@codesourcery.com>
1893             Cesar Philippidis  <cesar@codesourcery.com>
1895         * gimplify.c (gimplify_scan_omp_clauses): Add support for
1896         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1897         (gimplify_adjust_omp_clauses): Likewise.
1898         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1899         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1900         (gimplify_omp_target_update): Update handling of acc update and
1901         enter/exit data.
1902         * omp-low.c (install_var_field): Remove unused parameter
1903         base_pointers_restrict.
1904         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1905         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1906         FINALIZE}
1907         (omp_target_base_pointers_restrict_p): Delete.
1908         (scan_omp_target): Update call to scan_sharing_clauses.
1909         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1910         FINALIZE}.
1911         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1912         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1913         (convert_local_omp_clauses): Likewise.
1914         * tree-pretty-print.c (dump_omp_clause): Likewise.
1915         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
1916         FINALIZE}.
1917         (omp_clause_code_name): Likewise.
1919 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
1921         PR debug/86194
1922         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1923         be narrowed.
1925         PR tree-optimization/86231
1926         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
1927         anti-range don't overwrite *vr0min before using it to compute *vr0max.
1929 2018-06-20  Tom de Vries  <tdevries@suse.de>
1931         PR tree-optimization/86097
1932         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1933         iv type if signedness of iv type is not the same as that of *nit.
1935 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
1937         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
1938         EDGE_EH edges, verify they are all EDGE_EH.
1940 2018-06-20  Maya Rashish  <coypu@sdf.org>
1942         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1944 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1946         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1947         * config/aarch64/aarch64.c (xgene1_tunings): Add
1948         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1949         (aarch64_mode_valid_for_sched_fusion_p):
1950         Allow 16-byte modes.
1951         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1952         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1953         128-bit modes.
1954         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1955         New pattern.
1956         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1957         * config/aarch64/iterators.md (VQ2): New mode iterator.
1959 2018-06-20  Martin Liska  <mliska@suse.cz>
1961         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1962         Change default ratio from 10 to 8.
1964 2018-06-20  Martin Liska  <mliska@suse.cz>
1966         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1967         New.
1968         (bit_test_cluster::find_bit_tests): Likewise.
1969         (switch_decision_tree::analyze_switch_statement): Find clusters.
1970         * tree-switch-conversion.h (struct jump_table_cluster): Document
1971         hierarchy.
1973 2018-06-20  Martin Liska  <mliska@suse.cz>
1975         * tree-switch-conversion.c (switch_conversion::collect):
1976         Record m_uniq property.
1977         (switch_conversion::expand): Bail out for special conditions.
1978         (group_cluster::~group_cluster): New.
1979         (group_cluster::group_cluster): Likewise.
1980         (group_cluster::dump): Likewise.
1981         (jump_table_cluster::emit): New.
1982         (switch_decision_tree::fix_phi_operands_for_edges): New.
1983         (struct case_node): Remove struct.
1984         (jump_table_cluster::can_be_handled): New.
1985         (case_values_threshold): Moved to header.
1986         (reset_out_edges_aux): Likewise.
1987         (jump_table_cluster::is_beneficial): New.
1988         (bit_test_cluster::can_be_handled): Likewise.
1989         (add_case_node): Remove.
1990         (bit_test_cluster::is_beneficial): New.
1991         (case_bit_test::cmp): New.
1992         (bit_test_cluster::emit): New.
1993         (expand_switch_as_decision_tree_p): Remove.
1994         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1995         (fix_phi_operands_for_edge): Likewise.
1996         (switch_decision_tree::analyze_switch_statement): New.
1997         (compute_cases_per_edge): Move ...
1998         (switch_decision_tree::compute_cases_per_edge): ... here.
1999         (try_switch_expansion): Likewise.
2000         (switch_decision_tree::try_switch_expansion): Likewise.
2001         (record_phi_operand_mapping): Likewise.
2002         (switch_decision_tree::record_phi_operand_mapping): Likewise.
2003         (emit_case_decision_tree): Likewise.
2004         (switch_decision_tree::emit): Likewise.
2005         (balance_case_nodes): Likewise.
2006         (switch_decision_tree::balance_case_nodes): Likewise.
2007         (dump_case_nodes): Likewise.
2008         (switch_decision_tree::dump_case_nodes): Likewise.
2009         (emit_jump): Likewise.
2010         (switch_decision_tree::emit_jump): Likewise.
2011         (emit_cmp_and_jump_insns): Likewise.
2012         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2013         (emit_case_nodes): Likewise.
2014         (switch_decision_tree::emit_case_nodes): Likewise.
2015         (conditional_probability): Remove.
2016         * tree-switch-conversion.h (enum cluster_type): New.
2017         (PRINT_CASE): New.
2018         (struct cluster): Likewise.
2019         (cluster::cluster): Likewise.
2020         (struct simple_cluster): Likewise.
2021         (simple_cluster::simple_cluster): Likewise.
2022         (struct group_cluster): Likewise.
2023         (struct jump_table_cluster): Likewise.
2024         (struct bit_test_cluster): Likewise.
2025         (struct min_cluster_item): Likewise.
2026         (struct case_tree_node): Likewise.
2027         (case_tree_node::case_tree_node): Likewise.
2028         (jump_table_cluster::case_values_threshold): Likewise.
2029         (struct case_bit_test): Likewise.
2030         (struct switch_decision_tree): Likewise.
2031         (struct switch_conversion): Likewise.
2032         (switch_decision_tree::reset_out_edges_aux): Likewise.
2034 2018-06-20  Martin Liska  <mliska@suse.cz>
2036         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2037         (hoist_edge_and_branch_if_true): Likewise.
2038         (expand_switch_using_bit_tests_p): Likewise.
2039         (struct case_bit_test): Likewise.
2040         (case_bit_test_cmp): Likewise.
2041         (emit_case_bit_tests): Likewise.
2042         (switch_conversion::switch_conversion): New class.
2043         (struct switch_conv_info): Remove old struct.
2044         (collect_switch_conv_info): More to ...
2045         (switch_conversion::collect): ... this.
2046         (check_range): Likewise.
2047         (switch_conversion::check_range): Likewise.
2048         (check_all_empty_except_final): Likewise.
2049         (switch_conversion::check_all_empty_except_final): Likewise.
2050         (check_final_bb): Likewise.
2051         (switch_conversion::check_final_bb): Likewise.
2052         (create_temp_arrays): Likewise.
2053         (switch_conversion::create_temp_arrays): Likewise.
2054         (free_temp_arrays): Likewise.
2055         (gather_default_values): Likewise.
2056         (switch_conversion::gather_default_values): Likewise.
2057         (build_constructors): Likewise.
2058         (switch_conversion::build_constructors): Likewise.
2059         (constructor_contains_same_values_p): Likewise.
2060         (switch_conversion::contains_same_values_p): Likewise.
2061         (array_value_type): Likewise.
2062         (switch_conversion::array_value_type): Likewise.
2063         (build_one_array): Likewise.
2064         (switch_conversion::build_one_array): Likewise.
2065         (build_arrays): Likewise.
2066         (switch_conversion::build_arrays): Likewise.
2067         (gen_def_assigns): Likewise.
2068         (switch_conversion::gen_def_assigns): Likewise.
2069         (prune_bbs): Likewise.
2070         (switch_conversion::prune_bbs): Likewise.
2071         (fix_phi_nodes): Likewise.
2072         (switch_conversion::fix_phi_nodes): Likewise.
2073         (gen_inbound_check): Likewise.
2074         (switch_conversion::gen_inbound_check): Likewise.
2075         (process_switch): Use the newly created class.
2076         (switch_conversion::expand): New.
2077         (switch_conversion::~switch_conversion): New.
2078         * tree-switch-conversion.h: New file.
2080 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2082         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2083         tree-vect-patterns.c.
2084         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2085         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
2086         (vect_recog_sad_pattern): Likewise.
2087         (vect_recog_widen_sum_pattern): Likewise.
2088         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
2089         (vect_recog_widen_shift_pattern): Remove the type_in argument.
2090         (vect_recog_rotate_pattern): Likewise.
2091         (vect_recog_mult_pattern): Likewise.
2092         (vect_recog_vector_vector_shift_pattern): Likewise.
2093         (vect_recog_divmod_pattern): Likewise.
2094         (vect_recog_mixed_size_cond_pattern): Likewise.
2095         (vect_recog_bool_pattern): Likewise.
2096         (vect_recog_mask_conversion_pattern): Likewise.
2097         (vect_try_gather_scatter_pattern): Likewise.
2098         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
2099         (vect_recog_over_widening_pattern): Likewise.
2100         (vect_recog_gather_scatter_pattern): Likewise.
2101         (vect_recog_func_ptr): Move from tree-vectorizer.h
2102         (vect_vect_recog_func_ptrs): Move further down the file.
2103         (vect_recog_func): Likewise.  Remove the third argument.
2104         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2105         (vect_pattern_recog_1): Expect the pattern function to do any
2106         necessary target tests.  Also expect it to provide a vector type.
2107         Remove the type_in handling.
2109 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2111         * tree-vect-patterns.c (vect_pattern_detected): New function.
2112         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2113         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2114         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2115         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2116         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2117         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2118         (vect_recog_mask_conversion_pattern)
2119         (vect_try_gather_scatter_pattern): Likewise.
2121 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2123         * tree-vect-patterns.c (vect_get_internal_def): New function.
2124         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2125         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2126         (search_type_for_mask_1): Use it.
2128 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2130         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2131         redundant WIDEN_SUM_EXPR handling.
2132         (vect_recog_sad_pattern): Likewise.
2134 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2136         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2137         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2138         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2139         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2140         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2142 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2144         * tree-vect-stmts.c (vectorizable_call): Make sure that we
2145         use the stmt_vec_info of the original bb statement for the
2146         new zero assignment, even if the call is part of a pattern.
2148 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2150         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2151         that the sequence is attached to the original statement rather
2152         than the pattern statement.
2153         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2154         PATTERN_DEF_SEQ from the original statement rather than
2155         the main pattern statement.
2156         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2157         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2158         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2160 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
2162         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2163         definition statements before the early exit for statements that aren't
2164         live or relevant.
2165         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2166         split out from...
2167         (vect_transform_loop): ...here.  Process pattern definition
2168         statements without first checking whether the main pattern
2169         statement is live or relevant.
2171 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
2173         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2174         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2176 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
2178         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2179         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2180         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2181         (expand_block_compare): Change select_block_compare_mode call.
2182         (expand_strncmp_align_check): Use new functions, fix comment.
2183         (emit_final_str_compare_gpr): New function.
2184         (expand_strn_compare): Refactor and clean up code.
2185         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2187 2018-06-19  Tony Reix  <tony.reix@atos.com>
2188             Damien Bergamini  <damien.bergamini@atos.com>
2189             David Edelsohn  <dje.gcc@gmail.com>
2191         * collect2.c (static_obj): New variable.
2192         (static_libs): New variable.
2193         (is_in_list): Uncomment declaration.
2194         (main): Track AIX libraries linked statically.
2195         (is_in_list): Uncomment definition.
2196         (scan_prog_file): Don't add AIX shared libraries initializer
2197         to constructor list if linking statically.
2199 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
2201         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2202         constant.
2203         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2205 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
2207         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2208         blocks.
2210 2018-06-19  Martin Liska  <mliska@suse.cz>
2212         * config/i386/i386.c (ix86_can_inline_p): Do not use
2213         ipa_fn_summaries::get_create.
2214         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2215         get.
2216         (devirtualization_time_bonus): Likewise.
2217         (ipcp_propagate_stage): Likewise.
2218         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2219         (edge_set_predicate): Likewise.
2220         (evaluate_conditions_for_known_args): Likewise.
2221         (evaluate_properties_for_edge): Likewise.
2222         (ipa_call_summary::reset): Tranform to ...
2223         (ipa_call_summary::~ipa_call_summary): ... this.
2224         (ipa_fn_summary::reset): Transform to ...
2225         (ipa_fn_summary::~ipa_fn_summary): ... this.
2226         (ipa_fn_summary_t::remove): Rename to ...
2227         (ipa_fn_summary_t::remove_callees): ... this.
2228         (ipa_fn_summary_t::duplicate): Use placement new
2229         instead of memory copy.
2230         (ipa_call_summary_t::duplicate): Likewise.
2231         (ipa_call_summary_t::remove): Remove.
2232         (dump_ipa_call_summary): Change get_create to get.
2233         (ipa_dump_fn_summary): Dump only when summary exists.
2234         (analyze_function_body): Use symbol_summary::get instead
2235         of get_create.
2236         (compute_fn_summary): Likewise.
2237         (estimate_edge_devirt_benefit): Likewise.
2238         (estimate_edge_size_and_time): Likewise.
2239         (inline_update_callee_summaries): Likewise.
2240         (remap_edge_change_prob): Likewise.
2241         (remap_edge_summaries): Likewise.
2242         (ipa_merge_fn_summary_after_inlining): Likewise.
2243         (write_ipa_call_summary): Likewise.
2244         (ipa_fn_summary_write): Likewise.
2245         (ipa_free_fn_summary): Likewise.
2246         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2247         (struct ipa_call_summary): Likewise.
2248         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2249         of get_create.
2250         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2251         (estimate_size_after_inlining): Likewise.
2252         (estimate_growth): Likewise.
2253         (growth_likely_positive): Likewise.
2254         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2255         (inline_call): Likewise.
2256         * ipa-inline.c (caller_growth_limits): Likewise.
2257         (can_inline_edge_p): Likewise.
2258         (can_inline_edge_by_limits_p): Likewise.
2259         (compute_uninlined_call_time): Likewise.
2260         (compute_inlined_call_time): Likewise.
2261         (want_inline_small_function_p): Likewise.
2262         (edge_badness): Likewise.
2263         (update_caller_keys): Likewise.
2264         (update_callee_keys): Likewise.
2265         (inline_small_functions): Likewise.
2266         (inline_to_all_callers_1): Likewise.
2267         (dump_overall_stats): Likewise.
2268         (early_inline_small_functions): Likewise.
2269         (early_inliner): Likewise.
2270         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2271         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2272         * ipa-pure-const.c (malloc_candidate_p): Likewise.
2273         * ipa-split.c (execute_split_functions): Likewise.
2274         * symbol-summary.h: Likewise.
2275         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2277 2018-06-19  Richard Biener  <rguenther@suse.de>
2279         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2280         (vectorize_loops): ... here.  Fix dbgcnt handling.
2281         (try_vectorize_loop): Wrap try_vectorize_loop_1.
2283 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
2285         PR target/86197
2286         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2287         ieee128 argument takes up only one (vector) register, not two (floating
2288         point) registers.
2290 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
2292         * gimplify.c (gimplify_init_constructor): Really never clear for an
2293         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2295 2018-06-19  Richard Biener  <rguenther@suse.de>
2297         PR tree-optimization/86179
2298         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2299         after failed recognition.
2301 2018-06-18  Martin Sebor  <msebor@redhat.com>
2303         PR middle-end/85602
2304         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2305         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2306         Handle integer subtraction.
2307         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2308         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2310 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
2312         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2313         param from rtx to rtx_insn *.
2314         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2315         param.
2316         (frv_ifcvt_modify_insn): Likwise.
2317         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2318         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
2319         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2320         as_a <rtx_insn *> cast to local "unprotected_region" once
2321         it's been established that it's not NULL or pc_rtx.
2322         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2323         param "sethi" from rtx to rtx_insn *.
2324         (nds32_group_float_insns): Likewise for param "insn".
2325         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2326         param.
2327         (vax_output_int_subtract): Likewise.
2328         * config/vax/vax.c (vax_output_int_add): Likewise for param
2329         "insn".
2330         (vax_output_int_subtract): Likewise.
2331         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2332         (emit_pattern_after): Likewise for param "after".
2333         (emit_insn_after): Likewise.
2334         (emit_jump_insn_after): Likewise.
2335         (emit_call_insn_after): Likewise.
2336         (emit_debug_insn_after): Likewise.
2337         (emit_pattern_before): Likewise for param "before".
2338         (emit_insn_before): Likewise.
2339         (emit_jump_insn_before): Likewise.
2340         * final.c (get_insn_template): Likewise for param "insn", removing
2341         a cast.
2342         * output.h (get_insn_template): Likewise for 2nd param.
2343         * rtl.h (emit_insn_before): Likewise.
2344         (emit_jump_insn_before): Likewise.
2345         (emit_debug_insn_before_noloc): Likewise.
2346         (emit_insn_after): Likewise.
2347         (emit_jump_insn_after): Likewise.
2348         (emit_call_insn_after): Likewise.
2349         (emit_debug_insn_after): Likewise.
2350         (set_insn_deleted): Likewise for param.
2352 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
2354         PR target/85358
2355         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2356         floating point modes, so that IFmode is numerically greater than
2357         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2358         to declare the ordering.  This prevents IFmode from being
2359         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2360         machine.  Include rs6000-modes.h to share the fractional values
2361         between genmodes* and the rest of the compiler.
2362         (IFmode): Likewise.
2363         (KFmode): Likewise.
2364         (TFmode): Likewise.
2365         * config/rs6000/rs6000-modes.h: New file.
2366         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2367         meaning of rs6000_long_double_size so that 126..128 selects an
2368         appropriate 128-bit floating point type.
2369         (rs6000_option_override_internal): Likewise.
2370         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2371         (TARGET_LONG_DOUBLE_128): Change the meaning of
2372         rs6000_long_double_size so that 126..128 selects an appropriate
2373         128-bit floating point type.
2374         (LONG_DOUBLE_TYPE_SIZE): Update comment.
2375         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2376         source and destination to match the standard usage.
2377         (truncifkf2): Likewise.
2378         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2379         ISA 2.07 to use an explicit clobber, instead of passing in a
2380         temporary.
2381         (copysign<mode>3_soft): Likewise.
2383 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
2385         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2386         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2387         (vect_slp_analyze_instance_dependence): Likewise.
2388         (vect_enhance_data_refs_alignment): Likewise.
2389         (vect_analyze_data_refs_alignment): Likewise.
2390         (vect_slp_analyze_and_verify_instance_alignment
2391         (vect_analyze_data_ref_accesses): Likewise.
2392         (vect_prune_runtime_alias_test_list): Likewise.
2393         (vect_analyze_data_refs): Likewise.
2394         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2395         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2396         (vect_analyze_scalar_cycles_1): Likewise.
2397         (vect_get_loop_niters): Likewise.
2398         (vect_analyze_loop_form_1): Likewise.
2399         (vect_update_vf_for_slp): Likewise.
2400         (vect_analyze_loop_operations): Likewise.
2401         (vect_analyze_loop): Likewise.
2402         (vectorizable_induction): Likewise.
2403         (vect_transform_loop): Likewise.
2404         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2405         * tree-vect-slp.c (vect_analyze_slp): Likewise.
2406         (vect_make_slp_decision): Likewise.
2407         (vect_detect_hybrid_slp): Likewise.
2408         (vect_slp_analyze_operations): Likewise.
2409         (vect_slp_bb): Likewise.
2410         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2411         (vectorizable_bswap): Likewise.
2412         (vectorizable_call): Likewise.
2413         (vectorizable_simd_clone_call): Likewise.
2414         (vectorizable_conversion): Likewise.
2415         (vectorizable_assignment): Likewise.
2416         (vectorizable_shift): Likewise.
2417         (vectorizable_operation): Likewise.
2418         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2420 2018-06-18  Martin Sebor  <msebor@redhat.com>
2422         PR tree-optimization/81384
2423         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2424         * builtins.c (expand_builtin_strnlen): New function.
2425         (expand_builtin): Call it.
2426         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2427         * builtins.def (BUILT_IN_STRNLEN): New.
2428         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2429         Warn for bounds in excess of maximum object size.
2430         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2431         single-value ranges.  Handle strnlen.
2432         (handle_builtin_strlen): Handle strnlen.
2433         (strlen_check_and_optimize_stmt): Same.
2434         * doc/extend.texi (Other Builtins): Document strnlen.
2436 2018-06-18  Maya Rashish  <coypu@sdf.org>
2438         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2439         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2440         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2442         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2443         here to ...
2444         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2446 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2448         * tree.c (escaped_string::escape): Replace cast to char * by
2449         const_cast<char *> (unescaped).
2451 2018-06-18  Nick Clifton  <nickc@redhat.com>
2453         PR 84195
2454         * tree.c (escaped_string): New class.  Converts an unescaped
2455         string into its escaped equivalent.
2456         (warn_deprecated_use): Use the new class to convert the
2457         deprecation message, if present.
2458         (test_escaped_strings): New self test.
2459         (test_c_tests): Add test_escaped_strings.
2460         * doc/extend.texi (deprecated): Add a note that the
2461         deprecation message is affected by the -fmessage-length
2462         option, and that control characters will be escaped.
2463         (#pragma GCC error): Document this pragma.
2464         (#pragma GCC warning): Likewise.
2465         * doc/invoke.texi (-fmessage-length): Document this option's
2466         effect on the #warning and #error preprocessor directives and
2467         the deprecated attribute.
2469 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
2471         * tree.c (decl_value_expr_lookup): Revert latest change.
2472         (decl_value_expr_insert): Likewise.
2474 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
2476         * gimplify.c (nonlocal_vlas): Delete.
2477         (nonlocal_vla_vars): Likewise.
2478         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2479         referenced VLAs.
2480         (gimplify_body): Do not create and destroy nonlocal_vlas.
2481         * tree-nested.c: Include diagnostic.h.
2482         (use_pointer_in_frame): Tweak.
2483         (lookup_field_for_decl): Add assertion and declare the transformation.
2484         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2485         internal error when the reference is in a wrong context.  Do not
2486         create a debug decl by default.
2487         (note_nonlocal_block_vlas): Delete.
2488         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2489         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
2490         create a debug decl by default.
2491         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2492         call is in a wrong context.
2493         (fixup_vla_decls): New function.
2494         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
2495         debug variables were created.
2496         * tree.c (decl_value_expr_lookup): Add checking assertion.
2497         (decl_value_expr_insert): Likewise.
2499 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
2501         PR middle-end/82479
2502         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2503         * tree-scalar-evolution.c (interpret_expr): Likewise.
2504         (expression_expensive_p): Likewise.
2505         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2506         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2507         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2508         (ssa_defined_by_minus_one_stmt_p): New.
2510 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
2512         PR middle-end/64946
2513         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2514         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2515         * dojump.c (do_jump): Likewise.
2516         * expr.c (expand_expr_real_2): Check operand type's sign.
2517         * fold-const.c (const_unop): Handle ABSU_EXPR.
2518         (fold_abs_const): Likewise.
2519         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2520         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2521         (strip_sign_op_1): Likesise.
2522         * match.pd: Add new pattern to generate ABSU_EXPR.
2523         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2524         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2525         * tree-eh.c (operation_could_trap_helper_p): Likewise.
2526         * tree-inline.c (estimate_operator_cost): Likewise.
2527         * tree-pretty-print.c (dump_generic_node): Likewise.
2528         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2529         * tree.def (ABSU_EXPR): New.
2531 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
2533         PR middle-end/86095
2534         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2535         documented as preserved for backward compatibility only.
2536         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2538         PR rtl-optimization/86108
2539         * bb-reorder.c (create_forwarder_block): Renamed to ...
2540         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
2541         jump from new landing pad to the second part.
2542         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2543         Adjust callers.
2545 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
2547         PR middle-end/85878
2548         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2549         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2550         Only call store_expr for halves if the mode is the same.
2552         PR middle-end/86123
2553         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2554         Fix up comment formatting.
2556 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2558         * typed-splay-tree.h (typed_splay_tree::remove): New function.
2559         (typed_splay_tree::closure,
2560         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2561         (typed_splay_tree::typed_splay_tree,
2562         typed_splay_tree::operator =): Declared private.
2563         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2564         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2565         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2566         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2567         typed_splay_tree::splay_tree_splay,
2568         typed_splay_tree::splay_tree_foreach_helper,
2569         typed_splay_tree::splay_tree_insert,
2570         typed_splay_tree::splay_tree_remove,
2571         typed_splay_tree::splay_tree_lookup,
2572         typed_splay_tree::splay_tree_predecessor,
2573         typed_splay_tree::splay_tree_successor,
2574         typed_splay_tree::splay_tree_min,
2575         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2576         (typed_splay_tree::root, typed_splay_tree::comp,
2577         typed_splay_tree::delete_key,
2578         typed_splay_tree::delete_value): New data members.
2579         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2580         typed_splay_tree::remove.
2582 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
2584         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2585         -mginv and -mno-ginv to the assembler.
2586         * config/mips/mips.opt (-mcrc): New option.
2587         (-mginv): Likewise.
2588         * doc/invoke.text (-mcrc): Document.
2589         (-mginv): Likewise.
2591 2018-06-15  Nick Clifton  <nickc@redhat.com>
2593         PR 84195
2594         * tree.c (escaped_string): New class.  Converts an unescaped
2595         string into its escaped equivalent.
2596         (warn_deprecated_use): Use the new class to convert the
2597         deprecation message, if present.
2598         (test_escaped_strings): New self test.
2599         (test_c_tests): Add test_escaped_strings.
2600         * doc/extend.texi (deprecated): Add a note that the
2601         deprecation message is affected by the -fmessage-length
2602         option, and that control characters will be escaped.
2603         (#pragma GCC error): Document this pragma.
2604         (#pragma GCC warning): Likewise.
2605         * doc/invoke.texi (-fmessage-length): Document this option's
2606         effect on the #warning and #error preprocessor directives and
2607         the deprecated attribute.
2609 2018-06-15  Richard Biener  <rguenther@suse.de>
2611         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2612         here, also noting vector size used.
2613         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
2614         size used in MSG_OPTIMIZED_LOCATIONS dump.
2615         (pass_slp_vectorize::execute): Adjust.
2617 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
2619         PR target/85968
2620         * config/arc/arc.c (arc_return_address_register): Fix
2621         if-condition.
2623 2018-06-15  Richard Biener  <rguenther@suse.de>
2625         PR middle-end/86159
2626         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2627         leave useless conversion stripping to force_gimple_operand_gsi.
2628         (gimplify_build2): Likewise.
2629         (gimplify_build1): Likewise.
2631 2018-06-15  Richard Biener  <rguenther@suse.de>
2633         PR middle-end/86076
2634         * tree-cfg.c (move_stmt_op): unshare invariant addresses
2635         before adjusting their block.
2637 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2639         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2640         multilibs for *-*-rtems*.
2641         * config/riscv/t-rtems: New file.
2643 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
2645         PR middle-end/86122
2646         * match.pd ((A +- CST1) +- CST2): Punt if last resort
2647         unsigned_type_for returns NULL.
2649         PR target/85945
2650         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2651         subregs of multi-word pseudos unless the float mode has word size.
2653 2018-06-14  Richard Biener  <rguenther@suse.de>
2655         PR middle-end/86139
2656         * tree-vect-generic.c (build_word_mode_vector_type): Remove
2657         duplicate and harmful type_hash_canon.
2658         * tree.c (type_hash_canon): Assert we didn't find ourselves.
2660 2018-06-14  Richard Biener  <rguenther@suse.de>
2662         PR ipa/86124
2663         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2664         NULL cgraph_node.
2666 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2668         * config/rtems.h (STDINT_LONG32): Define.
2670 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
2671             Prachi Godbole  <prachi.godbole@imgtec.com>
2673         * config/mips/mips-cpus.def: Define P6600.
2674         * config/mips/mips-tables.opt: Regenerate.
2675         * config/mips/mips.c (mips_ucbranch_type): New enum.
2676         (mips_rtx_cost_data): Add support for P6600.
2677         (mips_issue_rate): Likewise.
2678         (mips_multipass_dfa_lookahead): Likewise.
2679         (mips_avoid_hazard): Likewise.
2680         (mips_reorg_process_insns): Likewise.
2681         (mips_classify_branch_p6600): New function.
2682         * config/mips/mips.h (TUNE_P6600): New define.
2683         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2684         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2685         * config/mips/mips.md: Include p6600.md.
2686         (processor): Add p6600.
2687         * config/mips/p6600.md: New file.
2688         * doc/invoke.texi: Add p6600 to supported architectures.
2690 2018-06-13  Martin Sebor  <msebor@redhat.com>
2692         PR tree-optimization/86114
2693         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2694         of integer types.
2695         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2697 2018-06-13  Richard Biener  <rguenther@suse.de>
2699         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2700         Properly set vector type of the intermediate stmt.
2701         * tree-vect-stmts.c (vectorizable_operation): The destination
2702         var always has vectype_out type.
2704 2018-06-13  Jeff Law  <law@redhat.com>
2706         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2707         integer 0 for argument to print_rtl_with_bb.
2708         (rl78_reorg): Likewise.
2710 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
2712         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2713         from rtx to rtx_insn *.
2714         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2715         "label".
2716         (add_sched_insns_for_speculation): Likewise for local "target",
2717         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2718         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2719         from rtx_insn ** to rtx_code_label **.
2720         (reorg_emit_nops): Likewise.
2721         (c6x_reorg): Likewise for local "call_labels".
2722         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2723         rtx to rtx_insn *.
2724         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2725         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2726         the loops over LABEL_REFS.
2727         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2728         braf_label.
2729         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2730         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2731         (split_branches): Strengthen local "olabel" from rtx to
2732         rtx_insn *, adding a safe_as_a cast.
2733         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2734         to "rtx_insn *".
2735         (add_insn_after): Likewise for first two params.
2736         (add_insn_before): Likewise.
2737         (remove_insn): Likewise for param.
2738         (emit_pattern_before_noloc): Likewise for second and third params.
2739         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2740         (emit_call_insn_before_noloc): Likewise.
2741         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2742         to "rtx_insn *".
2743         (emit_barrier_before): Likewise.
2744         (emit_label_before): Strengthen "label" param from "rtx" to
2745         "rtx_code_label *".  Strengthen "before" param from "rtx" to
2746         "rtx_insn *".
2747         (emit_insn_after_1): Strengthen "after" param from "rtx" to
2748         "rtx_insn *".
2749         (emit_pattern_after_noloc): Likewise.
2750         (emit_insn_after_noloc): Likewise.
2751         (emit_jump_insn_after_noloc): Likewise.
2752         (emit_call_insn_after_noloc): Likewise.
2753         (emit_debug_insn_after_noloc): Likewise.
2754         (emit_barrier_after): Likewise.
2755         (emit_label_after): Likewise for both params.
2756         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
2757         "loc" param from "int" to "location_t".
2758         (emit_insn_after_setloc): Likewise.
2759         (emit_jump_insn_after_setloc): Likewise.
2760         (emit_call_insn_after_setloc): Likewise.
2761         (emit_debug_insn_after_setloc): Likewise.
2762         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
2763         "loc" param from "int" to "location_t".
2764         (emit_pattern_before): Convert NULL_RTX to NULL.
2765         (emit_insn_before_setloc): Convert "loc" param from "int" to
2766         "location_t".
2767         (emit_jump_insn_before_setloc): Likewise.
2768         (emit_call_insn_before_setloc): Likewise.
2769         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2770         rtx_insn *.  Convert "loc" param from "int" to "location_t".
2771         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2772         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2773         Convert 3rd param from "int" to "location_t".
2774         (emit_barrier_before, emit_barrier_after, next_real_insn):
2775         Strengthen param from rtx to rtx_insn *.
2776         (emit_label_before): Strengthen 1st param from "rtx" to
2777         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
2778         "rtx_insn *".
2779         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2780         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2781         Strengthen 2nd param from "rtx" to "rtx_insn *".
2782         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2783         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2784         Likewise. Convert 3rd param from "int" to "location_t".
2785         (emit_label_after): Strengthen 1st param from "rtx" to
2786         "rtx_code_label *".
2787         (next_real_insn, remove_insn): Strengthen param from "rtx" to
2788         "rtx_insn *".
2789         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2790         from "rtx" to "rtx_insn *".
2792 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
2794         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2795         bodies streamed in with -Q.
2796         * dumpfile.c (dump_files): Add lto-stream-out dump file.
2797         * dumpfile.h (tree_dump_index): Add lto_stream_out.
2798         * gimple-streamer-out.c: Include gimple-pretty-print.h
2799         (output_bb): Dump stmts streamed.
2800         * lto-section-out.c: Include print-tree.h
2801         (lto_begin_section): Dump sections created.
2802         (lto_output_decl_index): Dump decl encoded.
2803         * lto-streamer-out.c: Include print-tree.h
2804         (create_output_block): Dump output block created.
2805         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2806         (output_function): Dump function output.
2807         (output_constructor): Dump constructor streamed.
2808         (write_global_stream): Output indexes encoded.
2809         (produce_asm_for_decls): Dump streams encoded.
2810         * lto-streamer.c (streamer_dump_file): New global var.
2811         * lto-streamer.h (streamer_dump_file): Declare.
2812         * passes.c (ipa_write_summaries): Initialize streamer dump.
2813         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2814         in.
2816 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
2818         PR target/86048
2819         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2820         offsets for register save directives.  Emit a second batch of save
2821         directives, if need be, when the function accesses prior frames.
2823 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2825         * config/arc/fpu.md (fmasf4): Force operand to register.
2826         (fnmasf4): Likewise.
2828 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2830         * config/arc/arc-protos.h (arc_pad_return): Remove.
2831         * config/arc/arc.c (machine_function): Remove force_short_suffix
2832         and size_reason.
2833         (arc_print_operand): Adjust printing of '&'.
2834         (arc_verify_short): Remove conditional printing of short suffix.
2835         (arc_final_prescan_insn): Remove reference to size_reason.
2836         (pad_return): New function.
2837         (arc_reorg): Call pad_return.
2838         (arc_pad_return): Remove.
2839         (arc_init_machine_status): Remove reference to force_short_suffix.
2840         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2841         (attr length): When attribute iscompact is true force to 2
2842         regardless; in the case of maybe check if we want to force the
2843         instruction to have 4 bytes length.
2844         (nopv): Change it to generate 4 byte long nop as well.
2845         (blockage): New pattern.
2846         (simple_return): Remove call to arc_pad_return.
2847         (p_return_i): Likewise.
2849 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2851         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2853 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2855         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2856         ARC cores.
2858 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2860         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2861         for ARC700 and ARCv2.
2863 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
2865         PR target/86076
2866         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2867         operands[2] instead of operands[1].
2870 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
2872         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2873         case, check whether the outer register overlaps an unallocatable
2874         register, not just whether it fits the required class.
2876 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
2878         * poly-int.h (can_div_trunc_p): Add new overload in which all values
2879         are poly_ints.
2880         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2881         (memrefs_conflict_p): Likewise.
2882         (init_alias_analysis): Likewise.
2883         * cfgexpand.c (expand_debug_expr): Likewise.
2884         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2885         * cse.c (fold_rtx): Likewise.
2886         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2887         * expr.c (emit_block_move_hints): Likewise.
2888         (clear_storage_hints, push_block, emit_push_insn): Likewise.
2889         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2890         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2891         (emit_group_store): Likewise.
2892         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
2893         to read the PRE/POST_MODIFY increment.
2894         * calls.c (store_one_arg): Use strip_offset.
2895         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2896         poly_int_rtx_p.
2897         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2898         by a VEC_SELECT.
2899         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2900         (simplify_binary_operation_1): Extend CONST_INT handling to
2901         poly_int_rtx_p.
2902         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2903         than a HOST_WIDE_INT.
2904         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2905         poly_int64.
2906         (adjust_mems, add_stores): Update accodingly.
2907         (vt_canonicalize_addr): Track polynomial offsets.
2908         (emit_note_insn_var_location): Likewise.
2909         (vt_add_function_parameter): Likewise.
2910         (vt_initialize): Likewise.
2912 2018-06-12  Jeff Law  <law@redhat.com>
2914         * config.gcc (alpha*-*-freebsd*): Remove.
2915         * config/alpha/freebsd.h: Remove.
2917 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
2919         PR other/69968
2920         * spellcheck-tree.c (levenshtein_distance): Rename to...
2921         (get_edit_distance): ...this, and update for underlying renaming.
2922         * spellcheck-tree.h (levenshtein_distance): Rename to...
2923         (get_edit_distance): ...this.
2924         * spellcheck.c (levenshtein_distance): Rename to...
2925         (get_edit_distance): ...this.  Convert from Levenshtein distance
2926         to Damerau-Levenshtein distance by supporting transpositions of
2927         adjacent characters.  Rename "v1" to "v_next" and "v0" to
2928         "v_one_ago".
2929         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2930         (selftest::test_edit_distance_unit_test_oneway): ...this, and
2931         update for underlying renaming.
2932         (selftest::levenshtein_distance_unit_test): Rename to...
2933         (selftest::test_get_edit_distance_unit): ...this, and update for
2934         underlying renaming.
2935         (selftest::test_find_closest_string): Add example from PR 69968
2936         where transposition helps
2937         (selftest::test_metric_conditions): Update for renaming.
2938         (selftest::test_metric_conditions): Likewise.
2939         (selftest::spellcheck_c_tests): Likewise.
2940         * spellcheck.h (levenshtein_distance): Rename both overloads to...
2941         (get_edit_distance): ...this.
2942         (best_match::consider): Update for renaming.
2944 2018-06-12  Martin Sebor  <msebor@redhat.com>
2946         PR tree-optimization/85259
2947         * builtins.c (compute_objsize): Handle constant offsets.
2948         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2949         true iff a warning has been issued.
2950         * gimple.h (gimple_nonartificial_location): New function.
2951         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2952         gimple_nonartificial_location and handle -Wno-system-headers.
2953         (handle_builtin_stxncpy): Same.
2955 2018-06-12  Martin Sebor  <msebor@redhat.com>
2957         PR c/85931
2958         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2960 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
2962         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2963         BUILTIN_VEC_XST entries for pointer to double and long long.
2965 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
2967         PR target/85990
2968         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2969         Update comments.
2970         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2971         Likewise.
2973 2018-06-12  Martin Liska  <mliska@suse.cz>
2975         * doc/options.texi: Document IntegerRange.
2977 2018-06-12  Martin Liska  <mliska@suse.cz>
2979         * config/i386/i386.opt: Make MPX-related options as Deprecated.
2980         * opt-functions.awk: Handle Deprecated flag.
2981         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2982         and report error.
2983         (read_cmdline_option): Report warning for a deprecated option.
2984         * opts.h (struct cl_option): Add new field cl_deprecated.
2985         (CL_ERR_DEPRECATED): New.
2987 2018-06-12  Martin Liska  <mliska@suse.cz>
2989         * doc/options.texi: Document Deprecated option flag.
2991 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
2993         * config/arc/arc-arch.h (arc_extras): New enum.
2994         (arc_cpu_t):Add field extra.
2995         (arc_cpu_types): Consider the extras.
2996         * config/arc/arc-cpus.def: Add extras info.
2997         * config/arc/arc-opts.h (processor_type): Consider extra field.
2998         * config/arc/arc.c (arc_override_options): Handle extra field.
3000 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
3002         * config/arc/arc-arch.h: Update ARC_OPTX macro.
3003         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3004         field.
3005         * config/arc/arc.c (arc_init): Update pic warning.
3006         (irq_range): Update irq range parsing warnings.
3007         (arc_override_options): Update various warning messages.
3008         (arc_handle_aux_attribute): Likewise.
3010 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
3012         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3014 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3016         * doc/sourcebuild.texi: Document usage of line number 0 in verify
3017         compiler messages directives.
3019 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
3021         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3022         * config/mips/mips-tables.opt: Regenerate.
3023         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3024         mips64r6.
3025         * doc/invoke.texi: Document -march=i6500.
3027 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
3029         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3030         (i6400_gpmul): Add cpu_unit.
3031         (i6400_gpdiv): Likewise.
3032         (i6400_msa_add_d): Update reservations.
3033         (i6400_msa_int_add) Likewise.
3034         (i6400_msa_short_logic3) Likewise.
3035         (i6400_msa_short_logic2) Likewise.
3036         (i6400_msa_short_logic) Likewise.
3037         (i6400_msa_move) Likewise.
3038         (i6400_msa_cmp) Likewise.
3039         (i6400_msa_short_float2) Likewise.
3040         (i6400_msa_div_d) Likewise.
3041         (i6400_msa_long_logic1) Likewise.
3042         (i6400_msa_long_logic2) Likewise.
3043         (i6400_msa_mult) Likewise.
3044         (i6400_msa_long_float2) Likewise.
3045         (i6400_msa_long_float4) Likewise.
3046         (i6400_msa_long_float5) Likewise.
3047         (i6400_msa_long_float8) Likewise.
3048         (i6400_fpu_fadd): Include frint type.
3049         (i6400_fpu_store): New define_insn_reservation.
3050         (i6400_fpu_load): Likewise.
3051         (i6400_fpu_move): Likewise.
3052         (i6400_fpu_fcmp): Likewise.
3053         (i6400_fpu_fmadd): Likewise.
3054         (i6400_int_mult): Include imul3nc type and update reservation.
3055         (i6400_int_div): Include idiv3 type and update reservation.
3056         (i6400_int_load): Update to check type not move_type.
3057         (i6400_int_store): Likewise.
3058         (i6400_int_prefetch): Set zero latency.
3060 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
3062         * gcc.c: Document new %@{...} sequence.
3063         (LINK_COMMAND_SPEC): Use it for the -L switches.
3064         (cpp_unique_options): Use it for the -I switches.
3065         (at_file_argbuf): New global variable.
3066         (in_at_file): Likewise.
3067         (alloc_args): Create at_file_argbuf.
3068         (clear_args): Truncate at_file_argbuf.
3069         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3070         (open_at_file): New function.
3071         (close_at_file): Likewise.
3072         (create_at_file): Delete.
3073         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3074         <'o'>: Likewise.
3075         <'@'>: New case.
3076         (validate_switches_from_spec): Deal with %@{...} sequence.
3077         (validate_switches): Likewise.
3078         (driver::finalize): Call clear_args.
3080 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
3082         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3084 2018-06-11  Martin Sebor  <msebor@redhat.com>
3086         * doc/invoke.texi (-Wall): List -Wc++17-compat.
3087         (Wno-class-memaccess): Add @opindex.
3088         (Wno-templates, Wno-multiple-inheritance): Same.
3089         (Wno-virtual-inheritance, Wno-namespaces): Same.
3090         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3091         (Wno-format-overflow, Wno-format-truncation): Same.
3092         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3093         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3094         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3095         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3096         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3097         (Wno-misspelled-isr): Same.
3099 2018-06-11  Martin Sebor  <msebor@redhat.com>
3101         * PR tree-optimization/86083
3102         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3104 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3106         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3108 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
3110         PR target/85755
3111         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3112         on the correct operand.
3113         (*movdi_internal64): Ditto.
3115 2018-06-11  Martin Liska  <mliska@suse.cz>
3117         PR tree-optimization/86089
3118         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3120 2018-06-11  Julia Koval  <julia.koval@intel.com>
3122         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3123         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3124         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3126 2018-06-11  Olivier Hainque  <hainque@adacore.com>
3128         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3129         for Ada with strict dwarf2.
3131 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
3133         PR target/85755
3134         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3135         addresses.
3137 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
3139         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3141 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
3143         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3144         TARGET_ELF.
3146 2018-06-08  Martin Liska  <mliska@suse.cz>
3148         * tree-cfg.h (debug_function): Fix argument type to match
3149         implementation.
3151 2018-06-08  Martin Liska  <mliska@suse.cz>
3153         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3154         Remove usage of MPX-related (and removed) fields.
3155         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3157 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
3159         * cfg.c (debug): Use TDF_NONE rather than 0.
3160         * cfghooks.c (debug): Likewise.
3161         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3162         (struct dump_option_value_info): Convert to...
3163         (struct kv_pair): ...this template type.
3164         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3165         rather than 0.
3166         (optinfo_verbosity_options): Likewise.
3167         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3168         OPTGROUP_NONE.
3169         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3170         than int for "optgroup_flags" param.
3171         (dump_generic_expr_loc): Use dump_flags_t rather than int for
3172         "dump_kind" param.
3173         (dump_dec): Likewise.
3174         (dump_finish): Use TDF_NONE rather than 0.
3175         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3176         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
3177         than 0.  Update for change to option_ptr.
3178         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3179         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
3180         0.  Update for changes to optinfo_verbosity_options and
3181         optgroup_options.
3182         (opt_info_switch_p): Convert optgroup_flags from int to
3183         optgroup_flags_t.
3184         (dump_basic_block): Use dump_flags_t rather than int
3185         for "dump_kind" param.
3186         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3187         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3188         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3189         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3190         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3191         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3192         TDF_NONE): Convert from macros to...
3193         (enum dump_flag): ...this new enum.
3194         (dump_flags_t): Update to use enum.
3195         (operator|, operator&, operator~, operator|=, operator&=):
3196         Implement for dump_flags_t.
3197         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3198         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3199         Convert from macros to...
3200         (enum optgroup_flag): ...this new enum.
3201         (optgroup_flags_t): New typedef.
3202         (operator|, operator|=): Implement for optgroup_flags_t.
3203         (struct dump_file_info): Convert field "alt_flags" to
3204         dump_flags_t.  Convert field "optgroup_flags" to
3205         optgroup_flags_t.
3206         (dump_basic_block): Use dump_flags_t rather than int for param.
3207         (dump_generic_expr_loc): Likewise.
3208         (dump_dec): Likewise.
3209         (dump_register): Convert param "optgroup_flags" to
3210         optgroup_flags_t.
3211         (opt_info_enable_passes): Likewise.
3212         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3213         than 0.
3214         * gimple-pretty-print.c (debug): Likewise.
3215         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3216         (merged_store_group::apply_stores): Likewise.
3217         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3218         * gimple.c (verify_gimple_pp): Likewise.
3219         * graphite-poly.c (print_pbb_body): Likewise.
3220         * passes.c (pass_manager::register_one_dump_file): Convert
3221         local "optgroup_flags" to optgroup_flags_t.
3222         * print-tree.c (print_node): Use TDF_NONE rather than 0.
3223         (debug): Likewise.
3224         (debug_body): Likewise.
3225         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3226         to optgroup_flags_t.
3227         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3228         than 0.
3229         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3230         (convert_mult_to_fma): Likewise.
3231         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3232         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3233         * tree-vect-data-refs.c (dump_lower_bound): Convert param
3234         "dump_kind" to dump_flags_t.
3236 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
3238         * config/rs6000/rs6000.c (min, max): Delete.
3240 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
3242         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3243         -mabi=spe and -mabi=no-spe.
3245 2018-06-08  Martin Liska  <mliska@suse.cz>
3247         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3248         where we expect an existing summary.
3250 2018-06-08  Martin Liska  <mliska@suse.cz>
3252         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3253         * ipa-inline.h (estimate_edge_growth): Likewise.
3255 2018-06-08  Martin Liska  <mliska@suse.cz>
3257         * cgraph.c (function_version_hasher::hash): Use
3258         cgraph_node::get_uid ().
3259         (function_version_hasher::equal):
3260         * cgraph.h (cgraph_node::get_uid): New method.
3261         * ipa-inline.c (update_caller_keys): Use
3262         cgraph_node::get_uid ().
3263         (update_callee_keys): Likewise.
3264         * ipa-utils.c (searchc): Likewise.
3265         (ipa_reduced_postorder): Likewise.
3266         * lto-cgraph.c (input_node): Likewise.
3267         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3268         * symbol-summary.h (symtab_insertion): Likewise.
3269         (symtab_removal): Likewise.
3270         (symtab_duplication): Likewise.
3271         * tree-pretty-print.c (dump_function_header): Likewise.
3272         * tree-sra.c (convert_callers_for_node): Likewise.
3274 2018-06-08  Martin Liska  <mliska@suse.cz>
3276         * cgraph.c (symbol_table::create_edge): Always assign a new
3277         unique number.
3278         (symbol_table::free_edge): Do not recycle numbers.
3279         * cgraph.h (cgraph_edge::get): New method.
3280         * symbol-summary.h (symtab_removal): Use it.
3281         (symtab_duplication): Likewise.
3282         (call_summary::hashable_uid): Remove.
3284 2018-06-08  Martin Liska  <mliska@suse.cz>
3286         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3287         (initialize_growth_caches): Remove.
3288         (free_growth_caches): Likewise.
3289         (do_estimate_edge_time): Use edge_growth_cache.
3290         (do_estimate_edge_size): Likewise.
3291         (do_estimate_edge_hints): Likewise.
3292         * ipa-inline.c (reset_edge_caches): Likewise.
3293         (recursive_inlining): Likewise.
3294         (inline_small_functions): Likewise.
3295         * ipa-inline.h (initialize_growth_caches): Remove.
3296         (estimate_edge_size): Likewise.
3297         (estimate_edge_time): Likewise.
3298         (estimate_edge_hints): Likewise.
3299         (reset_edge_growth_cache): Likewise.
3300         * symbol-summary.h (call_summary::remove): New method.
3302 2018-06-08  Martin Liska  <mliska@suse.cz>
3304         * ipa-cp.c (class edge_clone_summary): New summary.
3305         (grow_edge_clone_vectors): Remove.
3306         (ipcp_edge_duplication_hook): Remove.
3307         (class edge_clone_summary_t): New call_summary class.
3308         (ipcp_edge_removal_hook): Remove.
3309         (edge_clone_summary_t::duplicate): New function.
3310         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3311         (create_specialized_node): Likewise.
3312         (ipcp_driver): Initialize edge_clone_summaries and do not
3313         register hooks.
3315 2018-06-08  Martin Liska  <mliska@suse.cz>
3317         * symbol-summary.h (get): New function.
3318         (call_summary::m_initialize_when_cloning): New class member.
3320 2018-06-08  Martin Liska  <mliska@suse.cz>
3322         * cgraph.c (cgraph_node::remove): Do not recycle uid.
3323         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3324         (symbol_table::allocate_cgraph_symbol): Do not set uid.
3325         * passes.c (uid_hash_t): Record removed_nodes by their uids.
3326         (remove_cgraph_node_from_order): Use the removed_nodes set.
3327         (do_per_function_toporder): Likwise.
3328         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3329         instead of summary_uid.
3330         (symtab_removal): Likewise.
3331         (symtab_duplication): Likewise.
3333 2018-06-08  Martin Liska  <mliska@suse.cz>
3335         * ipa-cp.c (ipcp_store_bits_results): Use
3336         ipcp_transformation_sum.
3337         (ipcp_store_vr_results): Likewise.
3338         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3339         to ...
3340         (ipcp_transformation_initialize): ... this.
3341         (ipa_set_node_agg_value_chain):
3342         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3343         (write_ipcp_transformation_info): Likewise.
3344         (read_ipcp_transformation_info): Likewise.
3345         (ipcp_update_bits): Likewise.
3346         (ipcp_update_vr): Likewise.
3347         (ipcp_transform_function): Likewise.
3348         * ipa-prop.h: Rename ipcp_transformation_summary to
3349         ipcp_transformation.
3350         (class ipcp_transformation_t): New function summary.
3351         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3352         (ipa_get_agg_replacements_for_node): Likewise.
3354 2018-06-08  Martin Liska  <mliska@suse.cz>
3356         * ipa-pure-const.c (struct funct_state_d): Do it class instead
3357         of struct.
3358         (class funct_state_summary_t): New function_summary class.
3359         (has_function_state): Remove.
3360         (get_function_state): Likewise.
3361         (set_function_state): Likewise.
3362         (add_new_function): Likewise.
3363         (funct_state_summary_t::insert): New function.
3364         (duplicate_node_data): Remove.
3365         (remove_node_data): Remove.
3366         (funct_state_summary_t::duplicate): New function.
3367         (register_hooks): Create new funct_state_summaries.
3368         (pure_const_generate_summary): Use it.
3369         (pure_const_write_summary): Likewise.
3370         (pure_const_read_summary): Likewise.
3371         (propagate_pure_const): Likewise.
3372         (propagate_nothrow): Likewise.
3373         (dump_malloc_lattice): Likewise.
3374         (propagate_malloc): Likewise.
3375         (execute): Do not register hooks, just remove summary
3376         instead.
3377         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3378         constructor.
3380 2018-06-08  Martin Liska  <mliska@suse.cz>
3382         * ipa-reference.c (remove_node_data): Remove.
3383         (duplicate_node_data): Likewise.
3384         (class ipa_ref_var_info_summary_t): New class.
3385         (class ipa_ref_opt_summary_t): Likewise.
3386         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3387         (get_reference_optimization_summary): Use
3388         ipa_ref_opt_sum_summaries.
3389         (set_reference_vars_info): Remove.
3390         (set_reference_optimization_summary): Likewise.
3391         (ipa_init): Create summaries.
3392         (init_function_info): Use function summary.
3393         (ipa_ref_opt_summary_t::duplicate): New function.
3394         (ipa_ref_opt_summary_t::remove): New function.
3395         (get_read_write_all_from_node): Fix GNU coding style.
3396         (propagate): Use function summary.
3397         (write_node_summary_p): Fix GNU coding style.
3398         (stream_out_bitmap): Likewise.
3399         (ipa_reference_read_optimization_summary): Use function summary.
3400         (ipa_reference_c_finalize): Do not release hooks.
3402 2018-06-08  Martin Liska  <mliska@suse.cz>
3404         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3405         (analyze_function_body): Extract multiple calls of get_create.
3406         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3407         * ipa-inline.c (recursive_inlining): Use ::get method.
3408         * ipa-inline.h (estimate_edge_growth): Likewise.
3410 2018-06-08  Martin Liska  <mliska@suse.cz>
3412         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3413         HSA_INVALID.
3414         (hsa_function_summary::hsa_function_summary): Use the new enum
3415         value.
3416         (hsa_gpu_implementation_p): Use hsa_summaries::get.
3417         * hsa-gen.c (hsa_get_host_function): Likewise.
3418         (get_brig_function_name): Likewise.
3419         * ipa-hsa.c (process_hsa_functions): Likewise.
3420         (ipa_hsa_write_summary): Likewise.
3421         * symbol-summary.h (symtab_duplication): Use ::get function/
3422         (get): New function.
3424 2018-06-08  Martin Liska  <mliska@suse.cz>
3426         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3427         of get.
3428         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3429         (hsa_register_kernel): Likewise.
3430         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3431         * hsa-gen.c (hsa_get_host_function): Likewise.
3432         (get_brig_function_name): Likewise.
3433         (generate_hsa): Likewise.
3434         (pass_gen_hsail::execute): Likewise.
3435         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3436         (devirtualization_time_bonus): Likewise.
3437         (ipcp_propagate_stage): Likewise.
3438         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3439         (edge_set_predicate): Likewise.
3440         (evaluate_conditions_for_known_args): Likewise.
3441         (evaluate_properties_for_edge): Likewise.
3442         (ipa_fn_summary::reset): Likewise.
3443         (ipa_fn_summary_t::duplicate): Likewise.
3444         (dump_ipa_call_summary): Likewise.
3445         (ipa_dump_fn_summary): Likewise.
3446         (analyze_function_body): Likewise.
3447         (compute_fn_summary): Likewise.
3448         (estimate_edge_devirt_benefit): Likewise.
3449         (estimate_edge_size_and_time): Likewise.
3450         (estimate_calls_size_and_time): Likewise.
3451         (estimate_node_size_and_time): Likewise.
3452         (inline_update_callee_summaries): Likewise.
3453         (remap_edge_change_prob): Likewise.
3454         (remap_edge_summaries): Likewise.
3455         (ipa_merge_fn_summary_after_inlining): Likewise.
3456         (ipa_update_overall_fn_summary): Likewise.
3457         (read_ipa_call_summary): Likewise.
3458         (inline_read_section): Likewise.
3459         (write_ipa_call_summary): Likewise.
3460         (ipa_fn_summary_write): Likewise.
3461         (ipa_free_fn_summary): Likewise.
3462         * ipa-hsa.c (process_hsa_functions): Likewise.
3463         (ipa_hsa_write_summary): Likewise.
3464         (ipa_hsa_read_section): Likewise.
3465         * ipa-icf.c (sem_function::merge): Likewise.
3466         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3467         (do_estimate_edge_time): Likewise.
3468         (estimate_size_after_inlining): Likewise.
3469         (estimate_growth): Likewise.
3470         (growth_likely_positive): Likewise.
3471         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3472         (inline_call): Likewise.
3473         * ipa-inline.c (caller_growth_limits): Likewise.
3474         (can_inline_edge_p): Likewise.
3475         (can_inline_edge_by_limits_p): Likewise.
3476         (compute_uninlined_call_time): Likewise.
3477         (compute_inlined_call_time): Likewise.
3478         (want_inline_small_function_p): Likewise.
3479         (edge_badness): Likewise.
3480         (update_caller_keys): Likewise.
3481         (update_callee_keys): Likewise.
3482         (recursive_inlining): Likewise.
3483         (inline_small_functions): Likewise.
3484         (inline_to_all_callers_1): Likewise.
3485         (dump_overall_stats): Likewise.
3486         (early_inline_small_functions): Likewise.
3487         (early_inliner): Likewise.
3488         * ipa-inline.h (estimate_edge_growth): Likewise.
3489         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3490         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3491         * ipa-prop.h (IPA_NODE_REF): Likewise.
3492         (IPA_EDGE_REF): Likewise.
3493         * ipa-pure-const.c (malloc_candidate_p): Likewise.
3494         (propagate_malloc): Likewise.
3495         * ipa-split.c (execute_split_functions): Likewise.
3496         * symbol-summary.h: Rename get to get_create.
3497         (get): Likewise.
3498         (get_create): Likewise.
3499         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3501 2018-06-08  Martin Liska  <mliska@suse.cz>
3503         * symbol-summary.h (release): Move definition out of class
3504         declaration.
3505         (symtab_removal): Likewise.
3506         (symtab_duplication): Likewise.
3508 2018-06-08  Martin Liska  <mliska@suse.cz>
3510         * symbol-summary.h (function_summary): Move constructor
3511         implementation out of class declaration.
3512         (release): Likewise.
3513         (symtab_insertion): Likewise.
3514         (symtab_removal): Likewise.
3515         (symtab_duplication): Likewise.
3516         (get): Likewise.
3518 2018-06-08  Martin Liska  <mliska@suse.cz>
3520         * Makefile.in: Remove support for MPX (macros, related functions,
3521         fields in cgraph_node, ...).
3522         * builtin-types.def (BT_BND): Likewise.
3523         (BT_FN_BND_CONST_PTR): Likewise.
3524         (BT_FN_CONST_PTR_BND): Likewise.
3525         (BT_FN_VOID_PTR_BND): Likewise.
3526         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3527         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3528         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3529         (expand_builtin_mempcpy_with_bounds): Likewise.
3530         (expand_builtin_memset_with_bounds): Likewise.
3531         (expand_builtin_memset_args): Likewise.
3532         (std_expand_builtin_va_start): Likewise.
3533         (expand_builtin): Likewise.
3534         (expand_builtin_with_bounds): Likewise.
3535         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3536         (DEF_LIB_BUILTIN_CHKP): Likewise.
3537         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3538         (DEF_CHKP_BUILTIN): Likewise.
3539         (BUILT_IN_MEMCPY): Likewise.
3540         (BUILT_IN_MEMMOVE): Likewise.
3541         (BUILT_IN_MEMPCPY): Likewise.
3542         (BUILT_IN_MEMSET): Likewise.
3543         (BUILT_IN_STPCPY): Likewise.
3544         (BUILT_IN_STRCAT): Likewise.
3545         (BUILT_IN_STRCHR): Likewise.
3546         (BUILT_IN_STRCPY): Likewise.
3547         (BUILT_IN_STRLEN): Likewise.
3548         (BUILT_IN_MEMCPY_CHK): Likewise.
3549         (BUILT_IN_MEMMOVE_CHK): Likewise.
3550         (BUILT_IN_MEMPCPY_CHK): Likewise.
3551         (BUILT_IN_MEMSET_CHK): Likewise.
3552         (BUILT_IN_STPCPY_CHK): Likewise.
3553         (BUILT_IN_STRCAT_CHK): Likewise.
3554         (BUILT_IN_STRCPY_CHK): Likewise.
3555         * calls.c (store_bounds): Likewise.
3556         (emit_call_1): Likewise.
3557         (special_function_p): Likewise.
3558         (maybe_warn_nonstring_arg): Likewise.
3559         (initialize_argument_information): Likewise.
3560         (finalize_must_preallocate): Likewise.
3561         (compute_argument_addresses): Likewise.
3562         (expand_call): Likewise.
3563         * cfgexpand.c (expand_call_stmt): Likewise.
3564         (expand_return): Likewise.
3565         (expand_gimple_stmt_1): Likewise.
3566         (pass_expand::execute): Likewise.
3567         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3568         (cgraph_node::remove): Likewise.
3569         (cgraph_node::dump): Likewise.
3570         (cgraph_node::verify_node): Likewise.
3571         * cgraph.h (chkp_function_instrumented_p): Likewise.
3572         (symtab_node::get_alias_target): Likewise.
3573         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3574         (cgraph_local_p): Likewise.
3575         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3576         (cgraph_edge::rebuild_references): Likewise.
3577         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3578         (walk_polymorphic_call_targets): Likewise.
3579         (cgraph_node::expand_thunk): Likewise.
3580         (symbol_table::output_weakrefs): Likewise.
3581         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3582         (ix86_handle_option): Likewise.
3583         * config/i386/constraints.md: Likewise.
3584         * config/i386/i386-builtin-types.def (BND): Likewise.
3585         (VOID): Likewise.
3586         (PVOID): Likewise.
3587         (ULONG): Likewise.
3588         * config/i386/i386-builtin.def (BDESC_END): Likewise.
3589         (BDESC_FIRST): Likewise.
3590         (BDESC): Likewise.
3591         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3592         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3593         * config/i386/i386.c (enum reg_class): Likewise.
3594         (ix86_target_string): Likewise.
3595         (ix86_option_override_internal): Likewise.
3596         (ix86_conditional_register_usage): Likewise.
3597         (ix86_valid_target_attribute_inner_p): Likewise.
3598         (ix86_set_indirect_branch_type): Likewise.
3599         (ix86_set_current_function): Likewise.
3600         (ix86_function_arg_regno_p): Likewise.
3601         (init_cumulative_args): Likewise.
3602         (ix86_function_arg_advance): Likewise.
3603         (ix86_function_arg): Likewise.
3604         (ix86_pass_by_reference): Likewise.
3605         (ix86_function_value_regno_p): Likewise.
3606         (ix86_function_value_1): Likewise.
3607         (ix86_function_value_bounds): Likewise.
3608         (ix86_return_in_memory): Likewise.
3609         (ix86_setup_incoming_vararg_bounds): Likewise.
3610         (ix86_va_start): Likewise.
3611         (indirect_thunk_need_prefix): Likewise.
3612         (print_reg): Likewise.
3613         (ix86_print_operand): Likewise.
3614         (ix86_expand_call): Likewise.
3615         (ix86_output_function_return): Likewise.
3616         (reg_encoded_number): Likewise.
3617         (BDESC_VERIFYS): Likewise.
3618         (ix86_init_mpx_builtins): Likewise.
3619         (ix86_init_builtins): Likewise.
3620         (ix86_emit_cmove): Likewise.
3621         (ix86_emit_move_max): Likewise.
3622         (ix86_expand_builtin): Likewise.
3623         (ix86_builtin_mpx_function): Likewise.
3624         (ix86_get_arg_address_for_bt): Likewise.
3625         (ix86_load_bounds): Likewise.
3626         (ix86_store_bounds): Likewise.
3627         (ix86_load_returned_bounds): Likewise.
3628         (ix86_store_returned_bounds): Likewise.
3629         (ix86_class_likely_spilled_p): Likewise.
3630         (ix86_hard_regno_mode_ok): Likewise.
3631         (x86_order_regs_for_local_alloc): Likewise.
3632         (ix86_mitigate_rop): Likewise.
3633         (ix86_bnd_prefixed_insn_p): Likewise.
3634         (ix86_mpx_bound_mode): Likewise.
3635         (ix86_make_bounds_constant): Likewise.
3636         (ix86_initialize_bounds): Likewise.
3637         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3638         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3639         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3640         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3641         (TARGET_CHKP_BOUND_MODE): Likewise.
3642         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3643         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3644         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3645         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3646         * config/i386/i386.h (TARGET_MPX): Likewise.
3647         (TARGET_MPX_P): Likewise.
3648         (VALID_BND_REG_MODE): Likewise.
3649         (FIRST_BND_REG): Likewise.
3650         (LAST_BND_REG): Likewise.
3651         (enum reg_class): Likewise.
3652         (BND_REG_P): Likewise.
3653         (BND_REGNO_P): Likewise.
3654         (BNDmode): Likewise.
3655         (ADJUST_INSN_LENGTH): Likewise.
3656         * config/i386/i386.md: Likewise.
3657         * config/i386/i386.opt: Likewise.
3658         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3659         (defined): Likewise.
3660         (LINK_MPX): Likewise.
3661         (MPX_SPEC): Likewise.
3662         (LIBMPX_SPEC): Likewise.
3663         (LIBMPXWRAPPERS_SPEC): Likewise.
3664         (CHKP_SPEC): Likewise.
3665         * config/i386/predicates.md: Likewise.
3666         * dbxout.c (dbxout_type): Likewise.
3667         * doc/extend.texi: Likewise.
3668         * doc/invoke.texi: Likewise.
3669         * doc/md.texi: Likewise.
3670         * doc/tm.texi: Likewise.
3671         * doc/tm.texi.in: Likewise.
3672         * dwarf2out.c (is_base_type): Likewise.
3673         (gen_formal_types_die): Likewise.
3674         (gen_subprogram_die): Likewise.
3675         (gen_type_die_with_usage): Likewise.
3676         (gen_decl_die): Likewise.
3677         (dwarf2out_late_global_decl): Likewise.
3678         * expr.c (expand_assignment): Likewise.
3679         (emit_storent_insn): Likewise.
3680         (store_expr_with_bounds): Likewise.
3681         (store_expr): Likewise.
3682         (expand_expr_real_1): Likewise.
3683         * expr.h (store_expr_with_bounds): Likewise.
3684         * function.c (use_register_for_decl): Likewise.
3685         (struct bounds_parm_data): Likewise.
3686         (assign_parms_augmented_arg_list): Likewise.
3687         (assign_parm_find_entry_rtl): Likewise.
3688         (assign_parm_is_stack_parm): Likewise.
3689         (assign_parm_load_bounds): Likewise.
3690         (assign_bounds): Likewise.
3691         (assign_parms): Likewise.
3692         (expand_function_start): Likewise.
3693         * gcc.c (CHKP_SPEC): Likewise.
3694         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3695         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3696         (wrestrict_dom_walker::check_call): Likewise.
3697         * gimple.c (gimple_build_call_from_tree): Likewise.
3698         * gimple.h (enum gf_mask): Likewise.
3699         (gimple_call_with_bounds_p): Likewise.
3700         (gimple_call_set_with_bounds): Likewise.
3701         * gimplify.c (gimplify_init_constructor): Likewise.
3702         * ipa-cp.c (initialize_node_lattices): Likewise.
3703         (propagate_constants_across_call): Likewise.
3704         (find_more_scalar_values_for_callers_subset): Likewise.
3705         * ipa-hsa.c (process_hsa_functions): Likewise.
3706         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3707         * ipa-icf.c (sem_function::merge): Likewise.
3708         * ipa-inline.c (early_inliner): Likewise.
3709         * ipa-pure-const.c (warn_function_noreturn): Likewise.
3710         (warn_function_cold): Likewise.
3711         (propagate_pure_const): Likewise.
3712         * ipa-ref.h (enum GTY): Likewise.
3713         * ipa-split.c (find_retbnd): Likewise.
3714         (consider_split): Likewise.
3715         (split_function): Likewise.
3716         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3717         * ipa.c (walk_polymorphic_call_targets): Likewise.
3718         (symbol_table::remove_unreachable_nodes): Likewise.
3719         (process_references): Likewise.
3720         (cgraph_build_static_cdtor_1): Likewise.
3721         * lto-cgraph.c (lto_output_node): Likewise.
3722         (output_refs): Likewise.
3723         (compute_ltrans_boundary): Likewise.
3724         (input_overwrite_node): Likewise.
3725         (input_node): Likewise.
3726         (input_cgraph_1): Likewise.
3727         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3728         * passes.c (pass_manager::execute_early_local_passes): Likewise.
3729         (class pass_chkp_instrumentation_passes): Likewise.
3730         (make_pass_chkp_instrumentation_passes): Likewise.
3731         * passes.def: Likewise.
3732         * rtl.h (struct GTY): Likewise.
3733         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3734         * stor-layout.c (layout_type): Likewise.
3735         * symtab.c: Likewise.
3736         * target.def: Likewise.
3737         * targhooks.c (default_chkp_bound_type): Likewise.
3738         (default_chkp_bound_mode): Likewise.
3739         (default_builtin_chkp_function): Likewise.
3740         (default_chkp_function_value_bounds): Likewise.
3741         (default_chkp_make_bounds_constant): Likewise.
3742         (default_chkp_initialize_bounds): Likewise.
3743         * targhooks.h (default_chkp_bound_type): Likewise.
3744         (default_chkp_bound_mode): Likewise.
3745         (default_builtin_chkp_function): Likewise.
3746         (default_chkp_function_value_bounds): Likewise.
3747         (default_chkp_make_bounds_constant): Likewise.
3748         (default_chkp_initialize_bounds): Likewise.
3749         * toplev.c (compile_file): Likewise.
3750         (process_options): Likewise.
3751         * tree-core.h (DEF_BUILTIN): Likewise.
3752         (DEF_BUILTIN_CHKP): Likewise.
3753         * tree-inline.c (declare_return_variable): Likewise.
3754         (remap_gimple_stmt): Likewise.
3755         (copy_bb): Likewise.
3756         (initialize_inlined_parameters): Likewise.
3757         (expand_call_inline): Likewise.
3758         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3759         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3760         (make_pass_ipa_chkp_produce_thunks): Likewise.
3761         (make_pass_chkp): Likewise.
3762         (make_pass_chkp_opt): Likewise.
3763         (make_pass_chkp_instrumentation_passes): Likewise.
3764         * tree-pretty-print.c (dump_generic_node): Likewise.
3765         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3766         * tree-ssa-dce.c (propagate_necessity): Likewise.
3767         (eliminate_unnecessary_stmts): Likewise.
3768         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3769         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3770         * tree-ssa-sccvn.h: Likewise.
3771         * tree-ssa-strlen.c (get_string_length): Likewise.
3772         (valid_builtin_call): Likewise.
3773         (adjust_last_stmt): Likewise.
3774         (handle_builtin_strchr): Likewise.
3775         (handle_builtin_strcpy): Likewise.
3776         (handle_builtin_stxncpy): Likewise.
3777         (handle_builtin_memcpy): Likewise.
3778         (handle_builtin_strcat): Likewise.
3779         (strlen_check_and_optimize_stmt): Likewise.
3780         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3781         * tree-streamer-in.c: Likewise.
3782         * tree-streamer.c (record_common_node): Likewise.
3783         * tree.c (tree_code_size): Likewise.
3784         (wide_int_to_tree_1): Likewise.
3785         (type_contains_placeholder_1): Likewise.
3786         (build_common_tree_nodes): Likewise.
3787         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3788         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3789         (POINTER_BOUNDS_P): Likewise.
3790         (BOUNDED_TYPE_P): Likewise.
3791         (BOUNDED_P): Likewise.
3792         (CALL_WITH_BOUNDS_P): Likewise.
3793         (pointer_bounds_type_node): Likewise.
3794         * value-prof.c (gimple_ic): Likewise.
3795         * var-tracking.c (vt_add_function_parameters): Likewise.
3796         * varasm.c (make_decl_rtl): Likewise.
3797         (assemble_start_function): Likewise.
3798         (output_constant): Likewise.
3799         (maybe_assemble_visibility): Likewise.
3800         * varpool.c (ctor_for_folding): Likewise.
3801         * chkp-builtins.def: Remove.
3802         * ipa-chkp.c: Remove.
3803         * ipa-chkp.h: Remove.
3804         * rtl-chkp.c: Remove.
3805         * rtl-chkp.h: Remove.
3806         * tree-chkp-opt.c: Remove.
3807         * tree-chkp.c: Remove.
3808         * tree-chkp.h: Remove.
3810 2018-06-07  Carl Love  <cel@us.ibm.com>
3812         * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3813         vextract_fp_from_shortl): Add BE support.
3815 2018-06-07  Paul Koning  <ni1d@arrl.net>
3817         * compare-elim.c (try_merge_compare): Don't merge compare if
3818         address contains a side effect.
3819         (try_eliminate_compare): Likewise.
3821 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
3823         * config.gcc: Support "tremont".
3824         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3825         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3826         PROCESSOR_TREMONT.
3827         * config/i386/i386.c (m_TREMONT): Define.
3828         (processor_target_table): Add "tremont".
3829         (PTA_TREMONT): Define.
3830         (ix86_lea_outperforms): Add TARGET_TREMONT.
3831         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3832         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3833         and M_INTEL_GOLDMONT_PLUS.
3834         (fold_builtin_cpu): Add "tremont".
3835         (ix86_add_stmt_cost): Add TARGET_TREMONT.
3836         (ix86_option_override_internal): Add "tremont".
3837         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3838         (processor_type): Add PROCESSOR_TREMONT.
3839         * config/i386/x86-tune.def: Add m_TREMONT.
3840         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3842 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3844         * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3845         symbol defined for msp430i* devices to be lower case.
3847         gcc/testsuite/gcc.target/msp430/
3848         * msp430i-device-symbol.c: New test.
3849         * msp430f-device-symbol.c: New test.
3850         * msp430.h: New test header file.
3852 2018-06-07  Richard Biener  <rguenther@suse.de>
3854         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3855         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
3856         Properly wrap signed arithmetic if overflow wraps.
3858 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
3860         PR tree-optimization/69615
3861         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3862         of a cast from a same precision integral SSA_NAME in a bb dominated
3863         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
3864         cast to utype if rhs2 has already a compatible type.
3866 2018-06-07  Richard Biener  <rguenther@suse.de>
3868         PR tree-optimization/85935
3869         * graphite-scop-detection.c (find_params_in_bb): Analyze
3870         condition operands with respect to the correct loop.  Assert
3871         the analysis doesn't fail.
3873 2018-06-04  Carl Love  <cel@us.ibm.com>
3875         * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3876         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
3877         as it is slightly cheaper.
3878         (first_match_or_eos_index_<mode>):
3879         Calculate index using natural element order.
3880         (first_match_index_<mode>):
3881         Calculate index using natural element order.
3882         (first_match_or_eos_index_<mode>):
3883         Calculate index using natural order.
3884         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3885         for BE and LE modes.
3886         * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3887         P9V_BUILTIN_VCLZLSBB_V16QI.
3888         * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3889         specific.
3891 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3893         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3894         indentation and line wrap for many prototypes.  Add missing
3895         @smallexample directives around block of prototypes for vec_xl and
3896         vec_xst.
3898 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
3900         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3901         track if we pass or return IEEE 128-bit floating point.
3902         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3903         C++ mangling that is compatible with GCC 8.1.
3904         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3905         (init_cumulative_args): Note if we pass or return IEEE 128-bit
3906         floating point types.
3907         (rs6000_function_arg_advance_1): Likewise.
3908         (rs6000_mangle_type): Optionally generate mangled names that match
3909         what GCC 8.1 generated for IEEE 128-bit floating point types.
3910         (rs6000_globalize_decl_name): If we have an external function that
3911         passes or returns IEEE 128-bit types, generate a weak reference
3912         from the mangled name used in GCC 8.1 to the current mangled
3913         name.
3914         (rs6000_init_builtins): Make __ibm128 use the long double type if
3915         long double is IBM extended double.  Make __float128 use the long
3916         double type if long double is IEEE 128-bit.
3918         PR target/85657
3919         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3920         macro for __ibm128 built-in functions.
3921         (PACK_IF): Add __ibm128 pack/unpack functions.
3922         (UNPACK_IF): Likewise.
3923         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3924         enable long double built-in functions if long double is IEEE
3925         128-bit floating point.
3926         (rs6000_invalid_builtin): Update long double built-in function
3927         error message.
3928         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3929         functions, adjust the built-in function to use the long double
3930         built-in function if __ibm128 and long double are the same type.
3931         * doc/extend.texi (PowerPC builtins): Update documention for
3932         __builtin_{,un}pack_longdouble.  Add documentation for
3933         __builtin_{,un}pack_ibm128.
3935 2018-06-06  Jim Wilson  <jimw@sifive.com>
3937         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3938         (struct machine_function): New field interrupt_mode.
3939         (riscv_handle_type_attribute): New function.  Add forward declaration.
3940         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3941         (riscv_expand_epilogue): Check interrupt_mode field.
3942         (riscv_set_current_function): Check interrupt attribute args and
3943         set interrupt_mode field.
3944         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3945         (riscv_sret, riscv_uret): New.
3946         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3947         new arguments to interrupt attribute.
3949 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
3951         PR target/63177
3952         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3953         Don't handle -mcpu=power8 if -mpower9-vector is also used.
3955 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3957         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3958         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3959         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3960         several redundant entries.
3962 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
3964         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3965         type from "rtx" to "rtx_insn *".
3966         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3967         for local "call_insn", removing cast.
3968         (ix86_expand_call): Likewise, introducing a "call_insn" local.
3970 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
3972         PR tree-optimization/86066
3973         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3974         for BIT_INSERT_EXPR stores.
3976 2018-06-06  Richard Biener  <rguenther@suse.de>
3978         PR tree-optimization/86062
3979         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3980         component refs ontop
3981         of to be offsetted base.
3983 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3985         * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3986         to be static and remove check on interrupt attribute name.
3988         gcc/testsuite/gcc.target/msp430/
3989         * function-attributes-4.c: New test.
3990         * static-interrupts.c: New test.
3992 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3994         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3995         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3997 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
3999         PR target/79924
4000         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4001         second argument.
4002         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4003         Remove second argument, change how error is called.
4004         (aarch64_layout_arg): Remove second argument from
4005         aarch64_err_no_fpadvsimd call.
4006         (aarch64_init_cumulative_args): Ditto.
4007         (aarch64_gimplify_va_arg_expr): Ditto.
4008         * config/aarch64/aarch64.md (mov<mode>): Ditto.
4010 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
4012         * config/i386/i386.md (simple_return_indirect_internal): New expander.
4013         (*simple_return_indirect_internal<mode>): Rename from
4014         simple_return_indirect_internal.  Use W mode iterator.
4015         (rstorssp): New expander.
4016         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
4017         (clrssbsy): New expander.
4018         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
4020 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4022         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4023         __typeof__.
4024         (cmse_check_pointed_object): Likewise.
4026 2018-06-05  Martin Liska  <mliska@suse.cz>
4028         PR gcov-profile/47618
4029         * doc/invoke.texi: Document how -fprofile-dir format
4030         is extended.
4032 2018-06-05  Richard Biener  <rguenther@suse.de>
4034         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4035         removal pretend DOM info isn't available so we do not update
4036         it and only remove edges, not dominated blocks.  Actually free
4037         DOM info in case we removed something.  Remove unreachable blocks.
4038         (mfb_keep_latches): Work with either DOM info or marked backedges.
4039         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4040         first.  Mark backedges if DOM info isn't available.
4041         (Re-)compute DOM info after cleanup_control_flow_pre.
4043 2018-06-05  Richard Biener  <rguenther@suse.de>
4045         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4046         (locus_discrim_hasher::hash): Adjust.
4047         (locus_discrim_hasher::equal): Likewise.
4048         (next_discriminator_for_locus): Work on line directly.
4049         (same_line_p): Pass in expanded locus1 as well.
4050         (assign_discriminators): Avoid redundant location expansions.
4052 2018-06-05  Richard Biener  <rguenther@suse.de>
4054         PR tree-optimization/86046
4055         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4056         if required after clearing TREE_ADDRESSABLE.
4058 2018-06-05  Richard Biener  <rguenther@suse.de>
4060         PR tree-optimization/86047
4061         * tree-ssa-loop.c (for_each_index): Glob handling of all
4062         decls and constants and really handle all of them.
4064 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4066         PR target/81497
4067         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4068         qualifier_void_pointer and qualifier_const_void_pointer.
4069         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4070         (arm_init_builtins): Handle the above.
4071         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4072         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4073         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4074         void intrinsics.
4076 2018-06-05  Martin Liska  <mliska@suse.cz>
4078         * auto-profile.c (read_autofdo_file): Do not use
4079         gcov_ctr_summary struct.
4080         (afdo_callsite_hot_enough_for_early_inline): Likewise.
4081         * coverage.c (struct counts_entry): Likewise.
4082         (read_counts_file): Read just single summary entry.
4083         (get_coverage_counts): Use gcov_summary struct.
4084         * coverage.h (get_coverage_counts): Likewise.
4085         * gcov-dump.c (dump_working_sets): Likewise.
4086         (tag_summary): Dump just single summary.
4087         * gcov-io.c (gcov_write_summary): Write just histogram
4088         summary.
4089         (gcov_read_summary): Read just single summary.
4090         (compute_working_sets): Use gcov_summary struct.
4091         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4092         of GCOV_COUNTERS_SUMMABLE.
4093         (GCOV_COUNTERS_SUMMABLE): Remove.
4094         (GCOV_FIRST_VALUE_COUNTER): Replace with
4095         GCOV_COUNTER_V_INTERVAL.
4096         (struct gcov_ctr_summary): Remove.
4097         (struct gcov_summary): Directly use fields of former
4098         gcov_ctr_summary.
4099         (compute_working_sets): Use gcov_summary struct.
4100         * gcov.c (read_count_file): Do not use ctrs fields.
4101         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4102         struct.
4103         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4104         struct.
4105         * profile.c: Likewise.
4106         * profile.h: Likewise.
4108 2018-06-05  Martin Liska  <mliska@suse.cz>
4110         PR gcov-profile/84846
4111         * gcov.c (output_lines): Print working directory only
4112         in intermediate format.
4114 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
4116         * config/s390/s390-builtin-types.def: Add void function type.
4117         * config/s390/s390-builtins.def: Use the function type for the
4118         tbeginc builtin.
4120 2018-06-04  Jim Wilson  <jimw@sifive.com>
4122         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4123         to int.
4124         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4125         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
4126         handle EH_RETURN_DATA_REGNO registers properly.
4127         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4128         (riscv_expand_epilogue): Update comment.  Change argument name and
4129         type.  Update code to use new name and type.  Pass new args to
4130         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
4131         EXCEPTION_RETURN.
4132         * config/riscv/riscv.md (NORMAL_RETURN): New.
4133         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4134         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4135         (eh_return): Call gen_eh_return_internal and emit barrier.
4136         (eh_return_internal): Call riscv_expand_epilogue.
4138 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
4140         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4141         bit_insertion field and declare can_be_merged_into method.
4142         (merged_store_group::can_be_merged_into): New method.
4143         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4144         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
4145         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4147 2018-06-04  Richard Biener  <rguenther@suse.de>
4149         PR tree-optimization/85955
4150         * builtins.c (fold_builtin_sincos): Convert pointers to
4151         destination to appropriate type before dereferencing.
4153 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
4155         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4157 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
4159         * expr.c (expand_expr_real_1): Force the operand into memory if
4160         its TYPE_MODE is BLKmode and if there is no integer mode for
4161         the number of bits being extracted.
4163 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
4165         PR target/85832
4166         PR target/86036
4167         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4168         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4170 2018-06-04  Richard Biener  <rguenther@suse.de>
4172         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4173         (cleanup_tree_cfg_noloop): ... single caller.  Do
4174         start_recording_case_labels later.
4176 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
4178         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4179         to _IMMINTRIN_H_INCLUDED.
4180         * config/i386/pconfigintrin.h: Ditto.
4181         * config/i386/waitpkgintrin.h: Ditto.
4182         * config/i386/immintrin.h: Add includes for sgxintrin.h,
4183         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4184         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4185         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4186         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4187         waitpkgintrin.h and cldemoteintrin.h.
4189 2018-06-04  Richard Biener  <rguenther@suse.de>
4191         PR tree-optimization/86038
4192         * tracer.c (find_best_successor): Check probability for
4193         being initialized, bail out if not.
4195 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
4197         PR target/86003
4198         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4199         of bits to ignore when comparing architectures.
4201 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
4203         PR tree-optimization/69615
4204         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4205         maximum or minimum of the type, try to merge it also as if
4206         range1 is + [-, x - 1] or + [x + 1, -].
4208         PR c++/86025
4209         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4211 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
4213         PR tree-optimization/86034
4214         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4215         the unsigned bitfield type in a bit insertion sequence if it does not
4216         have a larger precision than the bitfield size.
4217         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4219 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
4221         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4223 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
4225         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4226         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4227         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4228         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4230 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
4232         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4233         Disable -fdelete-null-pointer-checks for ELF toolchain.
4235 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
4236             Kito Cheng  <kito.cheng@gmail.com>
4238         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4239         (nds32le-*-*, nds32be-*-*): Integrate checking process.
4240         (nds32*-*-*): Add glibc and uclibc conditions.
4241         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4242         (TARGET_EXCEPT_UNWIND_INFO): Define.
4243         * config/nds32/elf.h: New file.
4244         * config/nds32/linux.h: New file.
4245         * config/nds32/nds32-elf.opt: New file.
4246         * config/nds32/nds32-linux.opt: New file.
4247         * config/nds32/nds32-fp-as-gp.c
4248         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4249         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4250         TARGET_LINUX_ABI.
4251         (nds32_asm_file_end): Ditto.
4252         (nds32_print_operand): Ditto.
4253         (nds32_insert_attributes): Ditto.
4254         (nds32_init_libfuncs): New function.
4255         (TARGET_HAVE_TLS): Define.
4256         (TARGET_INIT_LIBFUNCS): Define.
4257         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4258         spec content.
4259         (TARGET_ELF): Apply different mcmodel setting.
4260         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4261         been migrated into elf.h and linux.h files.
4262         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4263         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4264         (mcmodel): The content has been migrated into nds32-elf.opt and
4265         nds32-linux.opt files.
4266         * config/nds32/t-elf: New file.
4267         * config/nds32/t-linux: New file.
4269 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
4270             Shiva Chen  <shiva0217@gmail.com>
4272         * config/nds32/constants.md (unspec_volatile_element): Add
4273         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4274         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4275         optimization.
4276         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4277         (make_pass_nds32_fp_as_gp): Declare.
4278         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4279         optmization pass.
4280         (nds32_asm_function_end_prologue): Remove unused asm output.
4281         (nds32_asm_function_begin_epilogue): Remove unused asm output.
4282         (nds32_asm_file_start): Output necessary fp_as_gp information.
4283         (nds32_option_override): Adjust register usage.
4284         (nds32_expand_prologue): Consider fp_as_gp situation.
4285         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4286         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4287         (epilogue): Ditto.
4288         (return): Ditto.
4289         (simple_return): Ditto.
4290         (omit_fp_begin): Output special directive for fp_as_gp.
4291         (omit_fp_end): Output special directive for fp_as_gp.
4292         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4293         mforbid-fp-as-gp): New options.
4295 2018-06-01  Mark Wielaard  <mark@klomp.org>
4297         * dwarf2out.c (dwarf2out_finish): Remove generation of
4298         DW_AT_loclists_base.
4300 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
4302         * gimple-ssa-store-merging.c: Include gimple-fold.h.
4303         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4304         (struct merged_store_group): Add bit_insertion field.
4305         (dump_char_array): Use standard hexadecimal format.
4306         (merged_store_group::merged_store_group): Set bit_insertion to false.
4307         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
4308         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
4309         also print the mask in the dump file.
4310         (pass_store_merging::gate): Minor tweak.
4311         (imm_store_chain_info::coalesce_immediate): Fix wrong association
4312         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
4313         stores with INTEGER_CST stores.
4314         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4315         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4316         and use it throughout.  Generate bit insertion sequences if need be.
4317         (pass_store_merging::process_store): Remove redundant condition.
4318         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4320 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
4322         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4323         the 128-bit floating point types.  Fix function comment.
4325 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4327         * config/aarch64/aarch64-simd.md
4328         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4329         mnemonics.
4330         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4331         mnemonics.
4333 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
4335         PR tree-optimization/85989
4336         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4337         variable.
4338         (backprop::intersect_uses): Check it when deciding whether this
4339         is a backedge reference.
4340         (backprop::process_block): Add each phi to m_visited_phis
4341         after visiting it, then clear it at the end.
4343 2018-06-01  Richard Biener  <rguenther@suse.de>
4345         * tree-vectorizer.h (vect_dr_stmt): New function.
4346         (vect_get_load_cost): Adjust.
4347         (vect_get_store_cost): Likewise.
4348         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4349         Use vect_dr_stmt instead of DR_SMTT.
4350         (vect_record_base_alignments): Likewise.
4351         (vect_calculate_target_alignment): Likewise.
4352         (vect_compute_data_ref_alignment): Likewise and make static.
4353         (vect_update_misalignment_for_peel): Likewise.
4354         (vect_verify_datarefs_alignment): Likewise.
4355         (vector_alignment_reachable_p): Likewise.
4356         (vect_get_data_access_cost): Likewise.  Pass down
4357         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4358         (vect_get_peeling_costs_all_drs): Likewise.
4359         (vect_peeling_hash_get_lowest_cost): Likewise.
4360         (vect_enhance_data_refs_alignment): Likewise.
4361         (vect_find_same_alignment_drs): Likewise.
4362         (vect_analyze_data_refs_alignment): Likewise.
4363         (vect_analyze_group_access_1): Likewise.
4364         (vect_analyze_group_access): Likewise.
4365         (vect_analyze_data_ref_access): Likewise.
4366         (vect_analyze_data_ref_accesses): Likewise.
4367         (vect_vfa_segment_size): Likewise.
4368         (vect_small_gap_p): Likewise.
4369         (vectorizable_with_step_bound_p): Likewise.
4370         (vect_prune_runtime_alias_test_list): Likewise.
4371         (vect_analyze_data_refs): Likewise.
4372         (vect_supportable_dr_alignment): Likewise.
4373         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4374         (vect_gen_prolog_loop_niters): Likewise.
4375         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4376         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4377         modify DR_STMT.
4378         (vect_recog_mask_conversion_pattern): Likewise.
4379         (vect_try_gather_scatter_pattern): Likewise.
4380         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4381         to vect_get_store_cost.
4382         (vect_get_store_cost): Get stmt_info instead of DR.
4383         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4384         (vect_get_load_cost): Get stmt_info instead of DR.
4386 2018-06-01  Richard Biener  <rguenther@suse.de>
4388         PR middle-end/86017
4389         * gimple-fold.c (var_decl_component_p): Also allow offsetted
4390         vars wrapped in MEM_REFs.
4392 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
4394         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4395         Fix subreg tests so that we only return a choice between
4396         GENERAL_REGS and FP_REGS if the original classes included both.
4398 2018-06-01  Richard Biener  <rguenther@suse.de>
4400         PR ipa/85960
4401         * tree-ssa-structalias.c (get_function_part_constraint):
4402         Handle NULL fi->decl.
4403         (find_func_aliases_for_call): Properly handle indirect
4404         fi from direct call.
4405         (find_func_clobbers): Likewise.
4406         (ipa_pta_execute): Likewise.
4407         (create_variable_info_for): For functions that are ifunc_resolver
4408         resolve to a varinfo that contains the result of the resolver call.
4409         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4410         aliases.
4412 2018-05-31  Michael Collison  <michael.collison@arm.com>
4414         * config/aarch64/aarch64.md:
4415         (*fix_to_zero_extenddfdi2): New pattern.
4416         * gcc.target/aarch64/fix_extend1.c: New testcase.
4418 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
4420         PR middle-end/78809
4421         PR middle-end/83026
4422         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4423         and BUILT_IN_STRNCMP_EQ.
4424         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4425         BUILT_IN_STRNCMP_EQ.
4426         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4427         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4428         (gimple_fold_builtin): Likewise.
4429         * tree-ssa-strlen.c (compute_string_length): New function.
4430         (determine_min_obsize): New function.
4431         (handle_builtin_string_cmp): New function to handle calls to
4432         string compare functions.
4433         (strlen_optimize_stmt): Add handling to builtin string compare
4434         calls.
4435         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4436         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4437         * tree.c (build_common_builtin_nodes): Add new defines of
4438         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4440 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
4442         PR target/85984
4443         * bb-reorder.c (pass_partition_blocks::gate): Return false for
4444         functions with naked attribute.
4446 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
4448         * config/i386/sse.md (avx_vec_concat<mode>):
4449         Substitute concat_tg_mode mode attribute with xtg_mode.
4450         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4451         (concat_tg_mode): Remove mode attribute.
4453 2018-05-31  Martin Sebor  <msebor@redhat.com>
4455         PR c/82063
4456         * calls.c (alloc_max_size): Correct a logic error/typo.
4457         Treat excessive arguments as infinite.  Warn for invalid arguments.
4458         * doc/invoke.texi (-Walloc-size-larger-than): Update.
4460 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
4462         PR target/85829
4463         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4464         and movx for Haswell.
4466 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
4467             Cesar Philippidis  <cesar@codesourcery.com>
4469         PR middle-end/85879
4470         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4471         when emitting error on private/firstprivate reductions.
4472         * omp-low.c (lower_omp_target): Avoid reference-type processing
4473         on pointers for firstprivate clause.
4475 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4477         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4478         (st1x2): Likewise.
4479         (st1x3): Likewise.
4480         * config/aarch64/aarch64-simd.md
4481         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4482         (aarch64_ld1_x3_<mode>): Likewise
4483         (aarch64_st1x2<VALLDIF:mode>): Likewise
4484         (aarch64_st1_x2_<mode>): Likewise
4485         (aarch64_st1x3<VALLDIF:mode>): Likewise
4486         (aarch64_st1_x3_<mode>): Likewise
4487         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4488         (vld1_s8_x3): Likewise.
4489         (vld1_u16_x3): Likewise.
4490         (vld1_s16_x3): Likewise.
4491         (vld1_u32_x3): Likewise.
4492         (vld1_s32_x3): Likewise.
4493         (vld1_u64_x3): Likewise.
4494         (vld1_s64_x3): Likewise.
4495         (vld1_f16_x3): Likewise.
4496         (vld1_f32_x3): Likewise.
4497         (vld1_f64_x3): Likewise.
4498         (vld1_p8_x3): Likewise.
4499         (vld1_p16_x3): Likewise.
4500         (vld1_p64_x3): Likewise.
4501         (vld1q_u8_x3): Likewise.
4502         (vld1q_s8_x3): Likewise.
4503         (vld1q_u16_x3): Likewise.
4504         (vld1q_s16_x3): Likewise.
4505         (vld1q_u32_x3): Likewise.
4506         (vld1q_s32_x3): Likewise.
4507         (vld1q_u64_x3): Likewise.
4508         (vld1q_s64_x3): Likewise.
4509         (vld1q_f16_x3): Likewise.
4510         (vld1q_f32_x3): Likewise.
4511         (vld1q_f64_x3): Likewise.
4512         (vld1q_p8_x3): Likewise.
4513         (vld1q_p16_x3): Likewise.
4514         (vld1q_p64_x3): Likewise.
4515         (vst1_s64_x2): Likewise.
4516         (vst1_u64_x2): Likewise.
4517         (vst1_f64_x2): Likewise.
4518         (vst1_s8_x2): Likewise.
4519         (vst1_p8_x2): Likewise.
4520         (vst1_s16_x2): Likewise.
4521         (vst1_p16_x2): Likewise.
4522         (vst1_s32_x2): Likewise.
4523         (vst1_u8_x2): Likewise.
4524         (vst1_u16_x2): Likewise.
4525         (vst1_u32_x2): Likewise.
4526         (vst1_f16_x2): Likewise.
4527         (vst1_f32_x2): Likewise.
4528         (vst1_p64_x2): Likewise.
4529         (vst1q_s8_x2): Likewise.
4530         (vst1q_p8_x2): Likewise.
4531         (vst1q_s16_x2): Likewise.
4532         (vst1q_p16_x2): Likewise.
4533         (vst1q_s32_x2): Likewise.
4534         (vst1q_s64_x2): Likewise.
4535         (vst1q_u8_x2): Likewise.
4536         (vst1q_u16_x2): Likewise.
4537         (vst1q_u32_x2): Likewise.
4538         (vst1q_u64_x2): Likewise.
4539         (vst1q_f16_x2): Likewise.
4540         (vst1q_f32_x2): Likewise.
4541         (vst1q_f64_x2): Likewise.
4542         (vst1q_p64_x2): Likewise.
4543         (vst1_s64_x3): Likewise.
4544         (vst1_u64_x3): Likewise.
4545         (vst1_f64_x3): Likewise.
4546         (vst1_s8_x3): Likewise.
4547         (vst1_p8_x3): Likewise.
4548         (vst1_s16_x3): Likewise.
4549         (vst1_p16_x3): Likewise.
4550         (vst1_s32_x3): Likewise.
4551         (vst1_u8_x3): Likewise.
4552         (vst1_u16_x3): Likewise.
4553         (vst1_u32_x3): Likewise.
4554         (vst1_f16_x3): Likewise.
4555         (vst1_f32_x3): Likewise.
4556         (vst1_p64_x3): Likewise.
4557         (vst1q_s8_x3): Likewise.
4558         (vst1q_p8_x3): Likewise.
4559         (vst1q_s16_x3): Likewise.
4560         (vst1q_p16_x3): Likewise.
4561         (vst1q_s32_x3): Likewise.
4562         (vst1q_s64_x3): Likewise.
4563         (vst1q_u8_x3): Likewise.
4564         (vst1q_u16_x3): Likewise.
4565         (vst1q_u32_x3): Likewise.
4566         (vst1q_u64_x3): Likewise.
4567         (vst1q_f16_x3): Likewise.
4568         (vst1q_f32_x3): Likewise.
4569         (vst1q_f64_x3): Likewise.
4570         (vst1q_p64_x3): Likewise.
4572 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4574         * config/msp430/msp430.c (msp430_output_labelref): Prepend
4575         user_label_prefix to name.
4577         * tree-core.h: Update comment about the format of NAME string
4578         passed to handler in attribute_spec.
4580         * config/msp430/msp430.md: Remove erroneous subreg expression from
4581         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4582         zero_extend{q,h}isi2.
4584 2018-05-30  Borislav Petkov  <bp@suse.de>
4586         * doc/extend.texi: Document some architecture specific
4587         constraints and sort entries.
4589 2018-05-30  Martin Sebor  <msebor@redhat.com>
4591         PR middle-end/85369
4592         * builtins.c (expand_builtin_stpcpy_1): New function.
4593         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4594         only if the former succeeds.
4596 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4598         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4599         in saphira.
4601 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
4603         * doc/invoke.texi (-flinker-output): Document
4605 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
4607         * passes.c (ipa_write_summaries): Only modify statements if body
4608         is in memory.
4609         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4610         incrementally linking.
4611         (ipa_passes): Likewise.
4612         * lto-cgraph.c (lto_output_node): When incrementally linking do not
4613         pass down resolution info.
4614         * common.opt (flag_incremental_link): Update info.
4615         * gcc.c (plugin specs): Turn flinker-output=* to
4616         -plugin-opt=-linker-output-known
4617         * toplev.c (compile_file): Also cut compilation when doing incremental
4618         link.
4619         * flag-types. (enum lto_partition_model): Add
4620         LTO_LINKER_OUTPUT_NOLTOREL.
4621         (invoke.texi): Add -flinker-output docs.
4622         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4623         link same way as WPA; do not stream in dead initializers.
4625         * dwarf2out.c (dwarf2out_die_ref_for_decl,
4626         darf2out_register_external_decl): Support incremental link.
4628 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
4630         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4632 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
4634         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4635         it down to simple_object_copy_lto_debug_sections.
4636         (run_gcc): Determine incremental LTO link time and configure
4637         lto1 into non-wpa mode, disable renaming of debug sections.
4639 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4641         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4642         descriptions of various incorrectly documented functions.
4644 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4646         Revert:
4647         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4648         address check not strict.
4650 2018-05-30  Richard Biener  <rguenther@suse.de>
4652         PR tree-optimization/85964
4653         * tracer.c (better_p): Drop initialized count check, we only
4654         call the function with initialized counts now.
4655         (find_best_successor): Do find a best edge if one
4656         has uninitialized count.
4657         (find_best_predecessor): Likewise.  Do BB frequency check only
4658         if count is initialized.
4660 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
4662         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4663         (aarch64_ldrstr_offset_compare): New.
4664         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4665         load/store orderings.
4666         (aarch64_gen_adjusted_ldpstp): Likewise.
4668 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
4670         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4671         Check for subset of GENERAL_REGS and FP_REGS.
4672         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4673         r=w alternative.
4675 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
4677         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4678         and wi::to_poly_offset.  Add the current offset and then check
4679         whether the sum fits, rather than using an unchecked addition of
4680         a checked term.  Check for a shwi rather than a uhwi.
4681         * expr.c (get_bit_range): Use tree_to_poly_uint64.
4682         (store_constructor): Use poly_int_tree_p.
4683         (expand_expr_real_1): Likewise.
4684         * function.c (assign_temp): Likewise.
4685         * fold-const.c (const_binop): Use poly_int_tree_p and
4686         wi::to_poly_offset.
4687         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
4688         division.
4689         * ipa-icf-gimple.c (func_checker::compare_operand): Use
4690         to_poly_offset for MEM offsets.
4691         * ipa-icf.c (sem_variable::equals): Likewise.
4692         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4693         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4694         wi::to_poly_offset for BIT_FIELD_REF offsets.
4695         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4696         wi::to_poly_offset.
4697         * var-tracking.c (emit_note_insn_var_location): Use
4698         tree_to_poly_uint64.
4700 2018-05-29  Jim Wilson  <jimw@sifive.com>
4702         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4704 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
4706         PR target/85950
4707         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4708         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4709         sequence.
4710         (sse4_1_round<mode>2): Use nonimmediate_operand
4711         for operand 1 predicate.
4713 2018-05-29  Martin Sebor  <msebor@redhat.com>
4714             Richard Biener  <rguenther@suse.de>
4716         PR testsuite/85888
4717         * calls.c (get_size_range): Call determine_value_range instead
4718         of get_value_range..
4719         * tree-vrp.h (determine_value_range): Declared new function.
4720         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4722 2018-05-29  Richard Biener  <rguenther@suse.de>
4724         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4725         sure to use non-pattern stmts for get_earlier_stmt arguments.
4726         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4727         called on pattern stmts.
4728         (get_later_stmt): Likewise.
4730 2018-05-29  Martin Liska  <mliska@suse.cz>
4732         PR gcov-profile/85759
4733         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4734         env variables.
4736 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
4738         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4739         VEC_UNPACK_*_EXPR.
4740         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4741         VEC_PACK_*_EXPR.
4743         PR target/85918
4744         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4745         VEC_PACK_FLOAT_EXPR): New tree codes.
4746         * tree-pretty-print.c (op_code_prio): Handle
4747         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4748         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4749         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4750         * tree-inline.c (estimate_operator_cost): Likewise.
4751         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4752         * fold-const.c (const_binop): Likewise.
4753         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4754         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4755         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4756         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4757         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4758         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4759         * expr.c (expand_expr_real_2): Likewise.
4760         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4761         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4762         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4763         optabs.
4764         * optabs.c (expand_widen_pattern_expr): For
4765         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4766         sign from result type rather than operand's type.
4767         (expand_binop_directly): For vec_packu_float_optab and
4768         vec_packs_float_optab allow result type to be different from operand's
4769         type.
4770         * optabs-tree.c (optab_for_tree_code): Handle
4771         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4772         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
4773         * tree-vect-generic.c (expand_vector_operations_1):  Handle
4774         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4775         VEC_PACK_FLOAT_EXPR.
4776         * tree-vect-stmts.c (supportable_widening_operation): Handle
4777         FIX_TRUNC_EXPR.
4778         (supportable_narrowing_operation): Handle FLOAT_EXPR.
4779         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4780         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4781         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
4782         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4783         mode attributes.
4784         (vec_pack<floatprefix>_float_<mode>): New expander.
4785         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4786         attributes.
4787         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4788         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4789         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4790         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4791         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4792         Document.
4793         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4794         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4795         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4796         VEC_PACK_FLOAT_EXPR): Document.
4798 2018-05-29  Richard Biener  <rguenther@suse.de>
4800         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4801         member.
4802         (stmt_vec_info_vec): Make pointer.
4803         (init_stmt_vec_info_vec): Remove.
4804         (free_stmt_vec_info_vec): Likewise.
4805         (set_stmt_vec_info_vec): New function.
4806         (free_stmt_vec_infos): Likewise.
4807         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4808         (set_vinfo_for_stmt): Likewise.
4809         (get_earlier_stmt): Likewise.
4810         (get_later_stmt): Likewise.
4811         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4812         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4813         (vec_info::~vec_info): Free stmt_vec_infos.
4814         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4815         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4816         (pass_slp_vectorize::execute): Likewise.
4817         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4818         (free_stmt_vec_info_vec): Likewise.
4819         (set_stmt_vec_info_vec): New function.
4820         (free_stmt_vec_infos): Likewise.
4821         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4822         the global stmt_vec_info_vec.
4823         * tree-parloops.c (gather_scalar_reductions): Use
4824         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4825         vector.
4827 2018-05-29  Richard Biener  <rguenther@suse.de>
4829         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4831 2018-05-29  Martin Liska  <mliska@suse.cz>
4832             David Malcolm  <dmalcolm@redhat.com>
4834         * vec.c (test_reverse): New.
4835         (vec_c_tests): Add new test.
4836         * vec.h (vl_ptr>::reverse): New function.
4838 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
4840         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4842         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4843         and later.
4845 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4847         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4849 2018-05-28  Richard Biener  <rguenther@suse.de>
4851         PR tree-optimization/85933
4852         * tree-vect-data-refs.c (vect_record_base_alignments): Only
4853         look at stmts marked as vectorizable.
4855 2018-05-28  Richard Biener  <rguenther@suse.de>
4857         PR tree-optimization/85934
4858         * tree-vect-generic.c (expand_vector_operations_1): Hoist
4859         vector boolean check before scalar optimization.
4861 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
4863         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4864         for armv5te.
4866 2018-05-28  Mark Wielaard  <mark@klomp.org>
4868         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4869         if it is an expression containing a minus sign.
4871 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
4873         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4875 2018-05-27  Paul Koning  <ni1d@arrl.net>
4877         * config/pdp11/pdp11.md (truncsihi2): Remove.
4879 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
4880             Chung-Ju Wu  <jasonwucj@gmail.com>
4882         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4883         implementation.
4884         (unaligned_store_dw): Ditto.
4885         * config/nds32/nds32-memory-manipulation.c
4886         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4887         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4888         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4889         (emit_setmem_word_loop): Rename to ...
4890         (emit_setmem_doubleword_loop): ... this.
4891         (nds32_gen_dup_4_byte_to_word_value): New function.
4892         (nds32_gen_dup_8_byte_to_double_word_value): New function.
4893         (nds32_expand_setmem_loop): Refine implementation.
4894         (nds32_expand_setmem_loop_v3m): Ditto.
4895         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4896         pattern.
4898 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
4900         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4902 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
4904         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4905         (nds32_init_machine_status): Initialize machine->attr_naked_p and
4906         machine->attr_no_prologue_p.
4907         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4908         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4909         (nds32_expand_epilogue): Consider attr_naked_p.
4910         (nds32_expand_epilogue_v3pop): Likewise.
4911         (nds32_can_use_return_insn): Likewise.
4912         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4913         attr_no_prologue_p fields.
4914         * config/nds32/nds32.opt (mret-in-naked-func): New option.
4916 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
4918         PR target/85918
4919         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4920         attributes.
4921         * config/i386/sse.md
4922         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4923         Rename to ...
4924         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4925         ... this.
4926         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4927         Rename to ...
4928         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4929         ... this.
4930         (*<floatsuffix>floatv2div2sf2): Rename to ...
4931         (*float<floatunssuffix>v2div2sf2): ... this.
4932         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4933         (float<floatunssuffix>v2div2sf2_mask): ... this.
4934         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4935         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4936         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4937         to ...
4938         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4939         ... this.
4940         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4941         Rename to ...
4942         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4943         ... this.
4944         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4945         Rename to ...
4946         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4947         ... this.
4948         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4949         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4950         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4951         gen_ufix_truncv8dfv8si2.
4952         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4953         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4954         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4955         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4956         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4957         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4958         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4959         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4961 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
4963         PR target/85900
4964         PR target/85345
4965         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4967 2018-05-25  Jim Wilson  <jimw@sifive.com>
4969         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4970         * config/riscv/riscv.c (struct machine_function): Add
4971         interrupt_handler_p and attribute_checked_p fields.
4972         (riscv_attribute_table): Add interrupt.
4973         (riscv_interrupt_type_p): New.
4974         (riscv_save_reg_p): Save extra regs for interrupt handler.
4975         (riscv_use_save_libcall): Return false  for interrupt handler.
4976         (riscv_first_stack_step): Add forward declaration.
4977         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
4978         for interrupt handler with large frame.  Use it for saved reg list.
4979         (riscv_expand_prologue): Move flag_stack_usage_info support to
4980         eliminate duplication.
4981         (riscv_expand_epilogue): Generate mret for interrupt handler.
4982         (riscv_epilogue_uses): New.
4983         (riscv_can_use_return_insn): Return false for interrupt handler.
4984         (riscv_function_ok_for_sibcall): Likewise.
4985         (riscv_set_current_function): Add interrupt handler support.
4986         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4987         * config/riscv/riscv.md (UNSPECV_MRET): New.
4988         (GP_REGNUM): New.
4989         (riscv_frflags, riscv_fsflags): Use tab after opcode.
4990         (riscv_mret): New.
4991         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4993 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4995         PR tree-optimization/85712
4996         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4997         this candidate has already been replaced in-situ by a copy.
4999 2018-05-25  Jason Merrill  <jason@redhat.com>
5001         PR c++/80485 - inline function non-zero address.
5002         * symtab.c (nonzero_address): Check DECL_COMDAT.
5004 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
5006         PR target/83628
5007         * config/alpha/alpha.md (ashlsi3): New insn pattern.
5008         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
5009         extension of SImode operation.  Use const123_operand predicate.
5010         (*saddsi_1): Remove.
5011         (*saddl_se_1): Ditto.
5012         (*ssubsi_1): Ditto.
5013         (*ssubl_se_1): Ditto.
5014         * config/alpha/predicates.md (const123_operand): New predicate.
5015         * config/alpha/constraints.md (P): Use IN_RANGE.
5017 2018-05-25  Richard Biener  <rguenther@suse.de>
5019         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5020         defaulted to true.
5021         (ref_maybe_used_by_stmt_p): Likewise.
5022         (stmt_may_clobber_ref_p): Likewise.
5023         (stmt_may_clobber_ref_p_1): Likewise.
5024         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5025         and pass it along.
5026         (ref_maybe_used_by_stmt_p): Likewise.
5027         (stmt_may_clobber_ref_p): Likewise.
5028         (stmt_may_clobber_ref_p_1): Likewise.
5029         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5030         the alias oracle to disambiguate DRs with stmts DR analysis
5031         couldn't handle.
5032         (vect_analyze_data_refs): Do not give up on not analyzable
5033         DRs for BB vectorization.  Remove code truncating the dataref
5034         vector.
5036 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
5038         PR target/85832
5039         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5040         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
5041         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5043 2018-05-25  Richard Biener  <rguenther@suse.de>
5045         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5046         function, combining stmt data ref gathering and fatal analysis
5047         parts.
5048         (vect_analyze_data_refs): Remove now redudnant code and simplify.
5049         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5050         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5051         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5052         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5054 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
5056         PR tree-optimization/85720
5057         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5058         SCC if all partitions are builtins.
5059         (version_loop_by_alias_check): New parameter.  Generate cancelable
5060         runtime alias check if all partitions are builtins.
5061         (distribute_loop): Update call to above function.
5063 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
5065         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5066         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5067         (parm_default_def_partition_arg): Ditto.
5068         (set_parm_default_def_partition): Ditto.
5069         (get_parm_default_def_partitions): Ditto and make it static.
5070         (get_undefined_value_partitions): Ditto and make it static.
5071         (remove_ssa_form): Refactor call to init_var_map here.
5072         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5073         computation for loop region.
5074         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5075         (register_default_def): Delete.
5076         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5077         (parm_default_def_partition_arg): Ditto.
5078         (set_parm_default_def_partition): Ditto.
5079         (get_parm_default_def_partitions): Ditto and make it static.
5080         (get_undefined_value_partitions): Ditto and make it static.
5081         (coalesce_with_default, coalesce_with_default): Update comment.
5082         (create_coalesce_list_for_region): New func factored out from
5083         create_outofssa_var_map.
5084         (populate_coalesce_list_for_outofssa): New func factored out from
5085         create_outofssa_var_map and coalesce_ssa_name.
5086         (create_outofssa_var_map): Delete.
5087         (coalesce_ssa_name): Refactor to support live range computation.
5088         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5089         (get_parm_default_def_partitions): Delete.
5090         (get_undefined_value_partitions): Ditto.
5091         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5092         computation for loop region.
5093         (new_tree_live_info, loe_visit_block): Ditto.
5094         (live_worklist, set_var_live_on_entry): Ditto.
5095         (calculate_live_on_exit, verify_live_on_entry): Ditto.
5096         * tree-ssa-live.h (struct _var_map): New fields.
5097         (init_var_map): Change decl.
5098         (region_contains_p): New.
5100 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
5102         * tree-ssa-live.h (live_merge_and_clear): Delete.
5104 2018-05-25  Richard Biener  <rguenther@suse.de>
5106         PR c++/85912
5107         * tree-dump.c (dequeue_and_dump): Remove access to removed
5108         operand 2 of a SWITCH_EXPR.
5110 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5112         * doc/sourcebuild.texi (vect_double_cond_arith): Include
5113         multiplication and division.
5114         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5115         (cond_udiv@var{m}, cond_umod@var{m}): Document.
5116         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5117         (cond_udiv_optab, cond_umod_optab): New optabs.
5118         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5119         (IFN_COND_RDIV): New internal functions.
5120         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5121         TRUNC_MOD_EXPR and RDIV_EXPR.
5122         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5123         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5124         New unspecs.
5125         (SVE_INT_BINARY): Include mult.
5126         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5127         (optab, sve_int_op): Handle mult.
5128         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5129         UNSPEC_COND_DIV.
5130         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5131         for SVE_INT_BINARY_SD.
5133 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5135         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5136         (optab, sve_int_op): Handle div and udiv.
5137         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5138         for SVE_INT_BINARY_SD.
5139         (*<optab><mode>3): New insn for the same.
5141 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5143         * tree-vect-patterns.c: Include predict.h.
5144         (vect_recog_divmod_pattern): Restrict check for division support
5145         to when optimizing for size.
5147 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5149         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5150         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5151         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5152         (gimple_match_op::set_op): Likewise.
5153         (gimple_resimplify4): Declare.
5154         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5155         (expr::gen_transform): Likewise.
5156         (decision_tree::gen): Generate a simplification routine for 4 operands.
5157         * gimple-match-head.c (gimple_simplify): Add an overload for
5158         4 operands.  In the top-level function, handle up to 4 call
5159         arguments and call gimple_resimplify4.
5160         (gimple_resimplify4): New function.
5161         (build_call_internal): Pass a fourth operand.
5162         (maybe_push_to_seq): Likewise.
5163         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5164         Fold VEC_COND_EXPRs of an operation and a default value into
5165         an IFN_COND_* function if possible.
5166         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5167         New unspecs.
5168         (SVE_COND_FP_BINARY): Include them.
5169         (optab, sve_fp_op): Handle them.
5170         (SVE_INT_BINARY_REV): New code iterator.
5171         (SVE_COND_FP_BINARY_REV): New int iterator.
5172         (commutative): New int attribute.
5173         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5174         Declare.
5175         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5176         function.
5177         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5178         (*cond_<optab><mode>): New patterns for reversed operands.
5180 2018-05-25  Richard Biener  <rguenther@suse.de>
5182         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5183         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5184         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5185         (STMT_VINFO_GROUPED_ACCESS): Adjust.
5186         * tree-vect-data-refs.c (everywhere): Adjust users.
5187         * tree-vect-loop.c (everywhere): Likewise.
5188         * tree-vect-slp.c (everywhere): Likewise.
5189         * tree-vect-stmts.c (everywhere): Likewise.
5190         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5192 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5194         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5195         Rename to...
5196         (gcc_cv_as_section_exclude): ... this.
5197         Try Solaris as #exclude syntax.
5198         * configure: Regenerate.
5199         * config.in: Regenerate.
5200         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5201         SECTION_EXCLUDE.
5202         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5203         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5205         * varasm.c (default_elf_asm_named_section): Don't check if
5206         HAVE_GAS_SECTION_EXCLUDE is defined.
5208 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5210         * doc/md.texi: Update the documentation of the cond_* optabs
5211         to mention the new final operand.  Fix GET_MODE_NUNITS call.
5212         Describe the scalar case too.
5213         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5214         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5215         instead of 2.
5216         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5217         (get_conditional_internal_fn): Update comment.
5218         * tree-vect-loop.c (vectorizable_reduction): Pass the original
5219         accumulator value as a final argument to conditional functions.
5220         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5221         a define_expand and add an "else" operand.  Assert for now that
5222         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
5223         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5224         (*cond_<optab><mode>): New patterns.
5225         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5226         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5227         (UNSPEC_COND_EOR): Delete.
5228         (optab): Remove associated mappings.
5229         (SVE_INT_BINARY): New code iterator.
5230         (sve_int_op): Remove int attribute and add "minus" to the code
5231         attribute.
5232         (SVE_COND_INT_OP): Delete.
5233         (SVE_COND_FP_OP): Rename to...
5234         (SVE_COND_FP_BINARY): ...this.
5236 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
5238         * optabs.c (can_reuse_operands_p): New function.
5239         (maybe_legitimize_operands): Try to reuse the results for
5240         earlier operands.
5242 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
5244         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5245         Add {q} suffix to insn mnemonic.
5247 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5249         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5250         (msp430_warn_func_return): New.
5252 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
5254         * fold-const.c (tree_nonzero_bits): New function.
5255         * fold-const.h (tree_nonzero_bits): Likewise.
5256         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5257         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5259 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
5261         PR target/85900
5262         PR target/85345
5263         * varasm.c (assemble_alias): Check ifunc_resolver only on
5264         FUNCTION_DECL.
5266 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
5268         PR target/85903
5269         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5270         when memory input operand is handled.
5272 2018-05-24  Luis Machado  <luis.machado@linaro.org>
5274         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5275         global.
5276         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5278 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
5280         * match.pd: Delay FMA folds until after vectorization.
5282 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5284         PR target/83009
5285         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5286         address check not strict.
5288 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
5290         * gimple-match.h (gimple_match_op): New class.
5291         (mprts_hook): Replace parameters with a gimple_match_op *.
5292         (maybe_build_generic_op): Likewise.
5293         (gimple_simplified_result_is_gimple_val): Replace parameters with
5294         a const gimple_match_op *.
5295         (gimple_simplify): Replace code_helper * and tree * parameters with
5296         a gimple_match_op * parameter.
5297         (gimple_resimplify1): Replace code_helper *, tree and tree *
5298         parameters with a gimple_match_op * parameter.
5299         (gimple_resimplify2): Likewise.
5300         (gimple_resimplify3): Likewise.
5301         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5302         parameters with a gimple_match_op * parameter.
5303         * gimple-match-head.c (gimple_simplify): Change prototypes of
5304         auto-generated functions to take a gimple_match_op * instead of
5305         separate code_helper * and tree * parameters.  Make the same
5306         change in the top-level overload and update calls to the
5307         gimple_resimplify routines.  Update calls to the auto-generated
5308         functions and to maybe_push_res_to_seq in the publicly-facing
5309         operation-specific gimple_simplify overloads.
5310         (gimple_match_op::MAX_NUM_OPS): Define.
5311         (gimple_resimplify1): Replace rcode and ops with a single res_op
5312         parameter.  Update call to gimple_simplify.
5313         (gimple_resimplify2): Likewise.
5314         (gimple_resimplify3): Likewise.
5315         (mprts_hook): Replace parameters with a gimple_match_op *.
5316         (maybe_build_generic_op): Likewise.
5317         (build_call_internal): Replace type, nargs and ops with
5318         a gimple_match_op *.
5319         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5320         with a single gimple_match_op *.  Update calls to mprts_hook,
5321         build_call_internal and gimple_simplified_result_is_gimple_val.
5322         Factor out code that is common to the tree_code and combined_fn cases.
5323         * genmatch.c (expr::gen_transform): Replace tem_code and
5324         tem_ops with a gimple_match_op called tem_op.  Update calls
5325         to the gimple_resimplify functions and maybe_push_res_to_seq.
5326         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5327         res_ops.  Update call to the gimple_resimplify functions.
5328         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5329         (decision_tree::gen): Make the functions take a gimple_match_op *
5330         called res_op instead of separate res_code and res_ops parameters.
5331         Update call accordingly.
5332         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5333         and ops with a single res_op parameter.  Update calls to
5334         maybe_build_generic_op and maybe_push_res_to_seq.
5335         (fold_stmt_1): Update calls to gimple_simplify and
5336         replace_stmt_with_simplification.
5337         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5338         and gimple_simplified_result_is_gimple_val.
5339         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5340         gimple_simplify.
5341         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5342         with a gimple_match_op *.
5343         (vn_nary_build_or_lookup): Likewise.  Update call to
5344         vn_nary_build_or_lookup_1.
5345         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5346         gimple_match_op *.  Update calls to the gimple_resimplify routines
5347         and to gimple_simplified_result_is_gimple_val.
5348         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5349         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5350         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5351         (visit_nary_op): Likewise.
5352         (visit_reference_op_load): Likewise.
5354 2018-05-23  Luis Machado  <luis.machado@linaro.org>
5356         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5357         modifier for printing the step amount.
5359 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
5361         PR target/78849
5362         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5363         types.
5365 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
5367         * doc/sourcebuild.texi (Endianness): New subsubsection.
5369 2018-05-23  Luis Machado  <luis.machado@linaro.org>
5371         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5372         <prefetch_dynamic_strides>: New const bool field.
5373         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5374         prefetch_dynamic_strides.
5375         (exynosm1_prefetch_tune): Likewise.
5376         (thunderxt88_prefetch_tune): Likewise.
5377         (thunderx_prefetch_tune): Likewise.
5378         (thunderx2t99_prefetch_tune): Likewise.
5379         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5380         false.
5381         (aarch64_override_options_internal): Update to set
5382         PARAM_PREFETCH_DYNAMIC_STRIDES.
5383         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5384         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5385         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5386         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5387         prefetch-dynamic-strides setting.
5389 2018-05-23  Luis Machado  <luis.machado@linaro.org>
5391         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5392         <minimum_stride>: New const int field.
5393         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5394         minimum_stride field defaulting to -1.
5395         (exynosm1_prefetch_tune): Likewise.
5396         (thunderxt88_prefetch_tune): Likewise.
5397         (thunderx_prefetch_tune): Likewise.
5398         (thunderx2t99_prefetch_tune): Likewise.
5399         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5400         <default_opt_level>: Set to 3.
5401         (aarch64_override_options_internal): Update to set
5402         PARAM_PREFETCH_MINIMUM_STRIDE.
5403         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5404         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5405         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5406         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5407         stride is constant and is below the minimum stride threshold.
5409 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5411         * config/arm/arm-cpus.in (mode26): Delete.
5412         (armv4): Delete mode26 reference.
5413         * config/arm/arm.c (arm_configure_build_target): Delete use of
5414         isa_bit_mode26.
5416 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
5418         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5419         New insn pattern.
5420         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5421         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
5422         for non-SSE modes.
5423         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
5424         (floatunsdidf2): Ditto.
5426 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
5428         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5429         (fixuns_trunc<mode>si2_avx512f): Ditto.
5430         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5431         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5432         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5434 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
5436         PR rtl-optimization/79985
5437         * df-scan.c (df_insn_refs_collect): Remove special case for
5438         global registers and asm statements.
5440 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
5442         * extend.texi (Global Register Variables): Rewrite the bullet list.
5443         Note that the register is available for allocation. Note that access
5444         via inline asm must use constraints. Add note about async-signal
5445         handlers. Remove paragraph about automagic register selection.
5447 2018-05-23  Richard Biener  <rguenther@suse.de>
5449         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5450         of fixed offset from memset VN.
5452 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
5454         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5455         first_interp field.
5456         (alloc_cand_and_find_basis): Initialize first_interp field.
5457         (slsr_process_mul): Modify first_interp field.
5458         (slsr_process_add): Likewise.
5459         (slsr_process_cast): Modify first_interp field for each new
5460         interpretation.
5461         (slsr_process_copy): Likewise.
5462         (dump_candidate): Dump first_interp field.
5463         (replace_mult_candidate): Process all interpretations, not just
5464         subsequent ones.
5465         (replace_rhs_if_not_dup): Likewise.
5466         (replace_one_candidate): Likewise.
5468 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
5470         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5471         Add new boolean.
5472         (aarch64_needs_frame_chain): New function.
5473         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5475 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
5477         PR target/84882
5478         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5479         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5480         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5481         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5482         as true for strict-align.
5483         (aarch64_can_inline_p): Perform checks even when callee has no
5484         attributes to check for strict alignment.
5485         * doc/extend.texi (AArch64 Function Attributes): Document
5486         no-strict-align.
5487         * doc/invoke.texi: (AArch64 Options): Likewise.
5489 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
5491         PR tree-optimization/85853
5492         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5493         the handling of the root of the node to...
5494         (vect_slp_analyze_node_operations_1): ...this new function,
5495         and run the whole thing with the child nodes' def types
5496         set according to their SLP node's def type.
5498 2018-05-23  Richard Biener  <rguenther@suse.de>
5500         PR middle-end/85874
5501         * tree-data-ref.c (create_runtime_alias_checks): Defer
5502         and ignore overflow warnings.
5504 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
5506         PR tree-optimization/85822
5507         * tree-vrp.c (is_masked_range_test): Fix handling of negative
5508         constants.
5510 2018-05-23  Richard Biener  <rguenther@suse.de>
5512         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5513         memset constants via native_interpret_expr.
5515 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
5517         PR target/85345
5518         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5519         attribute.
5520         (cgraph_node::create_alias): Likewise.
5521         (cgraph_node::get_availability): Check ifunc_resolver instead
5522         of looking up ifunc attribute.
5523         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5524         * varasm.c (do_assemble_alias): Likewise.
5525         (assemble_alias): Likewise.
5526         (default_binds_local_p_3): Likewise.
5527         * cgraph.h (cgraph_node): Add ifunc_resolver.
5528         (cgraph_node::only_called_directly_or_aliased_p): Return false
5529         for IFUNC resolver.
5530         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5531         attribute.
5532         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5533         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5534         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5535         instead of looking up ifunc attribute.
5537 2018-05-22  Luis Machado  <luis.machado@linaro.org>
5539         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5541 2018-05-22  Martin Sebor  <msebor@redhat.com>
5543         PR middle-end/85359
5544         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5545         only when expasion succeeds.
5546         (expand_builtin_strcmp): Same.
5547         (expand_builtin_strncmp): Same.
5549 2018-05-22  Martin Sebor  <msebor@redhat.com>
5551         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5553 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
5554             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5556         * config/aarch64/aarch64-ldpstp.md: Replace uses of
5557         aarch64_mem_pair_operand with memory_operand and delete operand swapping
5558         code.
5559         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5560         Add check for legitimate_address.
5561         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5562         (aarch64_swap_ldrstr_operands): New.
5563         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5564         Define prototype.
5566 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
5567             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5569         * config/aarch64/aarch64.md: New patterns to generate stp
5570         and ldp.
5571         (store_pair_sw, store_pair_dw): New patterns to generate stp for
5572         single words and double words.
5573         (load_pair_sw, load_pair_dw): Likewise.
5574         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5575         Delete.
5576         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5577         Delete.
5578         * config/aarch64/aarch64-ldpstp.md: Modify peephole
5579         for different mode ldpstp and add peephole for merged zero stores.
5580         Likewise for loads.
5581         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5582         Add size check.
5583         (aarch64_gen_store_pair): Rename calls to match new patterns.
5584         (aarch64_gen_load_pair): Rename calls to match new patterns.
5585         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5586         (load_pair<DREG:mode><DREG2:mode>): ... This.
5587         (store_pair<mode>): Rename to...
5588         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5589         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5590         New mode iterators.
5591         (V_INT_EQUIV): Handle SImode.
5592         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5593         New predicate.
5595 2018-05-22  Martin Sebor  <msebor@redhat.com>
5597         PR c/85623
5598         * calls.c (maybe_warn_nonstring_arg): Use string length to set
5599         or ajust the presumed bound on an operation to avoid unnecessary
5600         warnings.
5602 2018-05-22  Martin Sebor  <msebor@redhat.com>
5604         PR tree-optimization/85826
5605         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5606         assuming that a DECL necesarily has a constant size.
5608 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
5610         PR middle-end/85862
5611         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5613 2018-05-22  Richard Biener  <rguenther@suse.de>
5615         PR tree-optimization/85834
5616         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5617         non-constant and non-zero memset arguments.
5619 2018-05-22  Martin Liska  <mliska@suse.cz>
5621         PR ipa/85607
5622         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5624 2018-05-22  Richard Biener  <rguenther@suse.de>
5626         PR tree-optimization/85863
5627         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5628         comparisons when vectype is specified.
5629         (vectorizable_condition): Do not specify vectype for
5630         vect_is_simple_cond when SLP vectorizing.
5632 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
5634         PR target/85657
5635         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5636         define __ibm128 as long double.
5637         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5638         as a distinct type when IEEE 128-bit support is enabled.
5639         (init_float128_ieee): Fix up conversions between IFmode and IEEE
5640         128-bit types to use the correct functions.
5641         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5642         convert between 128-bit floating point types that have different
5643         modes but the same representation, instead of using gen_lowpart to
5644         makean alias.
5645         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5646         KFmode.
5647         (IFKF_reg): New attributes to give the register constraints for
5648         IFmode and KFmode.
5649         (extend<mode>tf2_internal): New insns to mark an explicit
5650         conversion between 128-bit floating point types that have a
5651         different mode but share the same representation.
5653 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
5655         PR tree-optimization/85814
5656         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5657         a null return from get_strinfo when unsharing the next
5658         strinfo in the chain.
5660 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
5662         PR gcc/84923
5663         * varasm.c (weak_finish): Clean up weak_decls.
5665 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5667         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5668         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5669         UNSPEC_UADALP values.
5670         * config/aarch64/iterators.md (ABAL): New int iterator.
5671         (ABDL2): Likewise.
5672         (ADALP): Likewise.
5673         (sur): Add mappings for the above.
5674         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5675         New define_insn.
5676         (aarch64_<sur>abal<mode>_4): Likewise.
5677         (aarch64_<sur>adalp<mode>_3): Likewise.
5678         (<sur>sadv16qi): New define_expand.
5680 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
5682         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5683         (*movdf_internal): Ditto.
5684         (*rcpsf2_sse): Ditto.
5685         (*rsqrtsf2_sse): Ditto.
5686         (*sqrt<mode>2_sse): Ditto.
5688 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
5690         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5691         eor3q<mode>4.
5692         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5693         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5694         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5695         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5696         vbcaxq_s64): New.
5697         * config/aarch64/arm_neon.h: Likewise.
5698         * config/aarch64/iterators.md (VQ_I): New.
5700 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
5702         * config.gcc: Add arc/t-multilib-linux to tmake_file for
5703         arc*-*-linux*.
5704         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5705         MULTILIB_DIRNAMES
5707 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
5709         * config/nds32/constraints.md (S): New constraint.
5710         * config/nds32/nds32.md (call_internal): Use constraint S.
5711         (call_value_internal): Likewise.
5712         (sibcall_internal): Likewise.
5713         (sibcall_value_internal): Likewise.
5715 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
5716             Chung-Ju Wu  <jasonwucj@gmail.com>
5718         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5719         into consideration.
5721 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
5722             Chung-Ju Wu  <jasonwucj@gmail.com>
5724         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5725         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5726         (nds32_rtx_costs_impl): Simplify.
5727         (nds32_address_cost_impl): Simplify.
5728         (nds32_init_rtx_costs): New function.
5729         (nds32_rtx_costs_speed_prefer): Likewise.
5730         (nds32_rtx_costs_size_prefer): Likewise.
5731         (nds32_address_cost_speed_prefer): Likewise.
5732         (nds32_address_cost_speed_fwprop): Likewise.
5733         (nds32_address_cost_size_prefer): Likewise.
5734         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5735         * config/nds32/nds32.c (nds32_option_override): Use
5736         nds32_init_rtx_costs function.
5738 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
5740         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5741         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5742         (TARGET_PIPELINE_N8): Likewise.
5743         (TARGET_PIPELINE_N10): Likewise.
5744         (TARGET_PIPELINE_N13): Likewise.
5745         (TARGET_PIPELINE_GRAYWOLF): Likewise.
5747 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
5749         * config/nds32/nds32-fpu.md: Update copyright year.
5751 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5753         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5755 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5757         * config/nds32/nds32.c
5758         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5759         * config/nds32/nds32.opt (minline-asm-r15): New option.
5761 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5763         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5764         MASK_HW_ABS.
5765         * config/nds32/nds32.md (abssi2): New pattern.
5767 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
5769         * config/i386/i386.md (rex64namesuffix): New mode attribute.
5770         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5771         Merge insn pattern from sse_cvtsi2ss<round_name> and
5772         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5773         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5774         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5775         using SWI48 mode iterator.
5776         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5777         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5778         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5779         pattern from sse_cvttss2si<round_saeonly_name>
5780         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5781         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5782         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5783         using SWI48 mode iterator.
5784         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5785         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5786         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5787         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5788         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5789         using SWI48 mode iterator.
5790         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5791         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5792         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5793         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5794         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5795         SWI48 mode iterator.
5796         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5797         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5798         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5799         pattern from sse_cvttsd2si<round_saeonly_name>
5800         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5802 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5804         * config/nds32/nds32-md-auxiliary.c
5805         (nds32_valid_smw_lwm_base_p): Refine.
5806         (nds32_output_smw_single_word): Refine.
5807         (nds32_output_smw_double_word): New.
5808         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5810 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5812         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5813         (nds32_output_stack_pop): Refine.
5814         (nds32_expand_unaligned_load): Refine.
5815         (nds32_expand_unaligned_store): Refine.
5817 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
5818             Chung-Ju Wu  <jasonwucj@gmail.com>
5820         * config/nds32/constants.md: Add TP_REGNUM constant.
5821         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5822         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5823         UNSPEC_ADD32.
5824         * config/nds32/nds32-doubleword.md: Consider flag_pic.
5825         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5826         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5827         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5828         and PIC code generation.
5829         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5830         code generation.
5831         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5832         optimization.
5833         * config/nds32/nds32.md: Support TLS and PIC.
5834         * config/nds32/nds32.c: Support TLS and PIC.
5835         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5836         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5837         predicate.
5839 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
5841         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5842         mode with E_ prefix.
5844 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
5845             Chung-Ju Wu  <jasonwucj@gmail.com>
5847         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5848         * config/nds32/nds32-md-auxiliary.c
5849         (symbolic_reference_mentioned_p): New.
5850         (nds32_legitimize_ict_address): New.
5851         (nds32_expand_ict_move): New.
5852         (nds32_indirect_call_referenced_p): New.
5853         (nds32_symbol_binds_local_p): Delete.
5854         (nds32_long_call_p): Modify.
5855         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5856         * config/nds32/nds32-protos.h
5857         (symbolic_reference_mentioned_p): Declare.
5858         (nds32_legitimize_ict_address): Declare.
5859         (nds32_expand_ict_move): Declare.
5860         (nds32_indirect_call_referenced_p): Declare.
5861         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5862         (nds32_relax_group): Use nds32_ict_const_p as condition.
5863         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5864         (nds32_asm_file_start): Output ict_model directive in asm code.
5865         (nds32_legitimate_address_p): Consider indirect call.
5866         (nds32_print_operand): Consider indirect call.
5867         (nds32_print_operand_address): Consider indirect call.
5868         (nds32_insert_attributes): Handle "indirect_call" attribute.
5869         (TARGET_LEGITIMATE_ADDRESS_P): Define.
5870         (TARGET_LEGITIMATE_CONSTANT_P): Define.
5871         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5872         (TARGET_DELEGITIMIZE_ADDRESS): Define.
5873         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5874         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5875         (TARGET_ICT_MODEL_SMALL): Define.
5876         (TARGET_ICT_MODEL_LARGE): Define.
5877         * config/nds32/nds32.md (movsi): Consider ict model.
5878         (call, call_value): Consider ict model.
5879         (sibcall, sibcall_value): Consider ict model.
5880         * config/nds32/nds32.opt (mict-model): New option.
5881         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5882         model.
5884 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
5885             Monk Chiang  <sh.chiang04@gmail.com>
5886             Jim Wilson <jimw@sifive.com>
5888         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5889         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
5890         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5891         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5892         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
5893         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5894         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5895         compute save_libcall_adjustment properly.
5896         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5897         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5898         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5899         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5900         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5901         (ABI_SPEC): Handle mabi=ilp32e.
5902         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5903         (RVE): Add RVE mask.
5904         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5905         <-march>: Add rv32e as an example.
5907 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
5909         PR c++/82899
5910         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5911         (intra_create_variable_infos): Handle C++ constructors.
5913 2018-05-18  Martin Liska  <mliska@suse.cz>
5915         * passes.def: Remove a redundant pass.
5917 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
5919         PR bootstrap/85838
5920         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5922 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5924         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5925         (ARMv4): Update.
5926         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5927         (ARMv6m): Update.
5928         (armv2, armv2a, armv3, armv3m): Delete architectures.
5929         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5930         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5931         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5932         Delete cpus.
5933         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5934         (*mulsidi3adddi): Likewise.
5935         (mulsidi3): Likewise.
5936         (*mulsidi3_nov6): Likewise.
5937         (umulsidi3): Likewise.
5938         (umulsidi3_nov6): Likewise.
5939         (umaddsidi4): Likewise.
5940         (*umulsidi3adddi): Likewise.
5941         (smulsi3_highpart): Likewise.
5942         (*smulsi3_highpart_nov6): Likewise.
5943         (umulsi3_highpart): Likewise.
5944         (*umulsi3_highpart_nov6): Likewise.
5945         * config/arm/arm.h (arm_arch3m): Delete.
5946         * config/arm/arm.c (arm_arch3m): Delete.
5947         (arm_option_override_internal): Update armv3-related comment.
5948         (arm_configure_build_target): Delete use of isa_bit_mode32.
5949         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5950         (arm_rtx_costs_internal): Delete check of arm_arch3m.
5951         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5952         (mulsa3): Likewise.
5953         (mulusa3): Likewise.
5954         * config/arm/arm-protos.h (arm_arch3m): Delete.
5955         * config/arm/arm-tables.opt: Regenerate.
5956         * config/arm/arm-tune.md: Likewise.
5957         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5958         deleted architectures.
5960 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5962         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5963         (armv5t, armv5te): New features.
5964         (ARMv5, ARMv5e): Delete fgroups.
5965         (ARMv5t, ARMv5te): Adjust for above changes.
5966         (ARMv6m): Likewise.
5967         (armv5, armv5e): Delete arches.
5968         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5969         arm_arch5.
5970         (*call_reg_arm): Likewise.
5971         (*call_value_reg_armv5): Likewise.
5972         (*call_value_reg_arm): Likewise.
5973         (*call_symbol): Likewise.
5974         (*call_value_symbol): Likewise.
5975         (*sibcall_insn): Likewise.
5976         (*sibcall_value_insn): Likewise.
5977         (clzsi2): Likewise.
5978         (prefetch): Likewise.
5979         (define_split and define_peephole2 dependent on arm_arch5):
5980         Likewise.
5981         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5982         arm_arch5e.
5983         (TARGET_ARM_QBIT): Likewise.
5984         (TARGET_DSP_MULTIPLY): Likewise.
5985         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5986         (arm_arch5, arm_arch5e): Delete.
5987         (arm_arch5t, arm_arch5te): Declare.
5988         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5989         (arm_arch5t): Declare.
5990         (arm_option_reconfigure_globals): Update for the above.
5991         (arm_options_perform_arch_sanity_checks): Update comment, replace
5992         use of arm_arch5 with arm_arch5t.
5993         (use_return_insn): Likewise.
5994         (arm_emit_call_insn): Likewise.
5995         (output_return_instruction): Likewise.
5996         (arm_final_prescan_insn): Likewise.
5997         (arm_coproc_builtin_available): Likewise.
5998         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5999         arm_arch5e with arm_arch5t and arm_arch5te.
6000         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6001         (arm_arch5t, arm_arch5te): Declare.
6002         * config/arm/arm-tables.opt: Regenerate.
6003         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6004         * config/arm/t-multilib: Likewise.
6005         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6006         instead of arm_arch5.
6007         (*call_reg_thumb1): Likewise.
6008         (*call_value_reg_thumb1_v5): Likewise.
6009         (*call_value_reg_thumb1): Likewise.
6010         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6011         unreachable path.
6012         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6014 2018-05-18  Martin Liska  <mliska@suse.cz>
6016         PR gcov-profile/84846
6017         * doc/gcov.texi: Document -t option of gcov tool.
6019 2018-05-18  Martin Liska  <mliska@suse.cz>
6021         PR gcov-profile/84846
6022         * gcov.c (print_usage): Add new -t option.
6023         (process_args): Handle the option.
6024         (generate_results): Use stdout as output when requested by
6025         the option.
6027 2018-05-18  Martin Liska  <mliska@suse.cz>
6029         PR gcov-profile/84846
6030         * coverage.c (coverage_init): Write PWD to .gcno file.
6031         * doc/gcov.texi: Document how working directory is printed.
6032         * gcov-dump.c (dump_gcov_file): Print PWD.
6033         * gcov.c (output_intermediate_file): Likewise.
6034         (read_graph_file): Read PWD string.
6035         (output_lines): Print PWD.
6037 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6039         PR middle-end/85817
6040         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6041         for retval and return false if all args to phi are zero.
6043 2018-05-18  Richard Biener  <rguenther@suse.de>
6045         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6046         method.
6047         (evrp_dom_walker::before_dom_children): Call it.
6049 2018-05-18  Richard Biener  <rguenther@suse.de>
6051         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6052         results when processing array refs with variable index.
6054 2018-05-18  Toon Moene  <toon@moene.org>
6056         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6057         directly after that of -floop-interchange. Indicate that both
6058         options are enabled by default when specifying -O3.
6060 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6062         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6063         iterator.  Delete separate integer-mode vec_set<mode> expander.
6064         (aarch64_simd_vec_setv2di): Delete.
6065         (vec_setv2di): Delete.
6066         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6067         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
6068         the "w, r" alternative.
6070 2018-05-18  Martin Liska  <mliska@suse.cz>
6072         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6073         * tree-pass.h (make_pass_lower_switch_O0): New function.
6074         * tree-switch-conversion.c (node_has_low_bound): Remove.
6075         (node_has_high_bound): Likewise.
6076         (node_is_bounded): Likewise.
6077         (class pass_lower_switch): Make it a template type and create
6078         two instances.
6079         (pass_lower_switch::execute): Add template argument.
6080         (make_pass_lower_switch): New function.
6081         (make_pass_lower_switch_O0): New function.
6082         (do_jump_if_equal): Remove.
6083         (emit_case_nodes): Simplify to just handle all 3 cases and leave
6084         all the hard work to tree optimization passes.
6086 2018-05-18  Martin Liska  <mliska@suse.cz>
6088         * dbgcnt.c (limit_low): Renamed from limit.
6089         (limit_high): New variable.
6090         (dbg_cnt_is_enabled): Check for upper limit.
6091         (dbg_cnt): Adjust dumping.
6092         (dbg_cnt_set_limit_by_index): Add new argument for high
6093         value.
6094         (dbg_cnt_set_limit_by_name): Likewise.
6095         (dbg_cnt_process_single_pair): Parse new format.
6096         (dbg_cnt_process_opt): Use strtok.
6097         (dbg_cnt_list_all_counters): Remove 'value' and add
6098         'limit_high'.
6099         * doc/invoke.texi: Document changes.
6101 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
6103         * doc/sourcebuild.texi (scalar_all_fma): Document.
6104         * tree.def (FMA_EXPR): Delete.
6105         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6106         * internal-fn.c (ternary_direct): New macro.
6107         (expand_ternary_optab_fn): Likewise.
6108         (direct_ternary_optab_supported_p): Likewise.
6109         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6110         * builtins.c (fold_builtin_fma): Delete.
6111         (fold_builtin_3): Don't call it.
6112         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6113         * expr.c (expand_expr_real_2): Likewise.
6114         * fold-const.c (operand_equal_p): Likewise.
6115         (fold_ternary_loc): Likewise.
6116         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6117         * gimple.c (DEFTREECODE): Likewise.
6118         * gimplify.c (gimplify_expr): Likewise.
6119         * optabs-tree.c (optab_for_tree_code): Likewise.
6120         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6121         * tree-eh.c (operation_could_trap_p): Likewise.
6122         (stmt_could_throw_1_p): Likewise.
6123         * tree-inline.c (estimate_operator_cost): Likewise.
6124         * tree-pretty-print.c (dump_generic_node): Likewise.
6125         (op_code_prio): Likewise.
6126         * tree-ssa-loop-im.c (stmt_cost): Likewise.
6127         * tree-ssa-operands.c (get_expr_operands): Likewise.
6128         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6129         * fold-const-call.h (fold_fma): Delete.
6130         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6131         CFN_FNMA and CFN_FNMS.
6132         (fold_fma): Delete.
6133         * genmatch.c (combined_fn): New enum.
6134         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6135         (commutative_op): New function.
6136         (commutate): Use it.  Handle more than 2 operands.
6137         (dt_operand::gen_gimple_expr): Use commutative_op.
6138         (parser::parse_expr): Allow :c to be used with non-binary
6139         operators if the commutative operand is known.
6140         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6141         CFN_FMS, CFN_FNMA and CFN_FNMS.
6142         (backprop::process_assign_use): Remove FMA_EXPR handling.
6143         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6144         (gen_hsa_fma): New function.
6145         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6146         IFN_FNMA and IFN_FNMS.
6147         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6148         * gimple-fold.h (follow_all_ssa_edges): Declare.
6149         * gimple-fold.c (follow_all_ssa_edges): New function.
6150         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6151         gimple_build interface and use follow_all_ssa_edges to fold the result.
6152         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6153         instead of checking for optabs directly.
6154         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6155         rather than FMA_EXPRs.
6156         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6157         call to IFN_FMA instead of an FMA_EXPR.
6159 2018-05-17  Jim Wilson  <jimw@sifive.com>
6161         * expr.c (do_tablejump): When converting index to Pmode, if we have a
6162         sign extended promoted subreg, and the range does not have the sign bit
6163         set, then do a sign extend.
6165         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6166         test, check for sign extended subreg and/or constant operands, and
6167         do a sign extend in that case.
6169 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
6171         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6172         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6173         Add untyped.
6174         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6175         Change logics_shift_reg to logics_shift_imm.
6176         (thunderx2t99_fp_loadpair_basic): Delete.
6177         (thunderx2t99_fp_storepair_basic): Delete.
6178         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6179         (thunderx2t99_asimd_polynomial): Delete.
6180         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6181         and neon_fp_mul_d_scalar_q.
6182         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6183         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6184         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6185         (thunderx2t99_asimd_lut): Add missing tbl types.
6186         (thunderx2t99_asimd_ext): Delete.
6187         (thunderx2t99_asimd_load1_1_mult): Delete.
6188         (thunderx2t99_asimd_load1_2_mult): Delete.
6189         (thunderx2t99_asimd_load1_ldp): New.
6190         (thunderx2t99_asimd_load1): New.
6191         (thunderx2t99_asimd_load2): Add missing *load2* types.
6192         (thunderx2t99_asimd_load3): New.
6193         (thunderx2t99_asimd_load4): New.
6194         (thunderx2t99_asimd_store1_1_mult): Delete.
6195         (thunderx2t99_asimd_store1_2_mult): Delete.
6196         (thunderx2t99_asimd_store2_mult): Delete.
6197         (thunderx2t99_asimd_store2_onelane): Delete.
6198         (thunderx2t99_asimd_store_stp): New.
6199         (thunderx2t99_asimd_store1): New.
6200         (thunderx2t99_asimd_store2): New.
6201         (thunderx2t99_asimd_store3): New.
6202         (thunderx2t99_asimd_store4): New.
6204 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
6206         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6207         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
6209 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
6210             Segher Boessenkool  <segher@kernel.crashing.org>
6212         PR target/85698
6213         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6214         operand.
6216 2018-05-17  Richard Biener  <rguenther@suse.de>
6218         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6219         for pruning loop and prune defs feeding only already visited PHIs.
6221 2018-05-17  Richard Biener  <rguenther@suse.de>
6223         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6225 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
6226             Richard Biener  <rguenther@suse.de>
6228         PR tree-optimization/85793
6229         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6230         for VMAT_ELEMENTWISE.
6232 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
6234         * internal-fn.h (lookup_internal_fn): Declare
6235         * internal-fn.c (lookup_internal_fn): New function.
6236         * gimple.c (gimple_build_call_from_tree): Handle calls to
6237         internal functions.
6238         * gimple-pretty-print.c (dump_gimple_call): Print "." before
6239         internal function names.
6240         * tree-pretty-print.c (dump_generic_node): Likewise.
6241         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6243 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
6245         * gimple-fold.h (gimple_build): Make the function forms take
6246         combined_fn rather than built_in_function.
6247         (gimple_simplify): Likewise.
6248         * gimple-match-head.c (gimple_simplify): Likewise.
6249         * gimple-fold.c (gimple_build): Likewise.
6250         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6251         rather than gimple_build_call_internal.
6252         (get_initial_defs_for_reduction): Likewise.
6253         (vect_create_epilog_for_reduction): Likewise.
6254         (vectorizable_live_operation): Likewise.
6256 2018-05-17  Martin Liska  <mliska@suse.cz>
6258         * gimple-ssa-sprintf.c (format_directive): Do not use
6259         space in between 'G_' and '('.
6261 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
6263         PR target/85323
6264         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6265         even if the mask is not all ones.
6267         PR target/85323
6268         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6269         vector.
6270         (ix86_gimple_fold_builtin): Likewise.
6272         PR target/85323
6273         * config/i386/i386.c: Include tree-vector-builder.h.
6274         (ix86_vector_shift_count): New function.
6275         (ix86_fold_builtin): Fold shift builtins by scalar count.
6276         (ix86_gimple_fold_builtin): Likewise.
6278         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6279         _mm512_setzero): New intrinsics.
6281 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
6282             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6284         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6285         code generation for cases where splatting a value is not useful.
6286         * simplify-rtx.c (simplify_ternary_operation): Simplify
6287         vec_merge across a vec_duplicate and a paradoxical subreg forming
6288         a vector mode to a vec_concat.
6290 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
6292         * config.gcc: Support "goldmont-plus".
6293         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6294         "goldmont-plus".
6295         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6296         PROCESSOR_GOLDMONT_PLUS.
6297         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6298         (processor_target_table): Add "goldmont-plus".
6299         (PTA_GOLDMONT_PLUS): Define.
6300         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6301         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6302         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6303         (fold_builtin_cpu): Add "goldmont-plus".
6304         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6305         (ix86_option_override_internal): Add "goldmont-plus".
6306         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6307         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6308         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6309         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6311 2018-05-17  Richard Biener  <rguenther@suse.de>
6313         PR tree-optimization/85757
6314         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6315         remove defs that only feed that PHI from further processing.
6317 2018-05-16  Jim Wilson  <jimw@sifive.com>
6319         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6320         asterisk to name.
6321         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6323 2018-05-16  Mark Wielaard  <mark@klomp.org>
6325         * dwarf2out.c (count_index_strings): New function.
6326         (output_indirect_strings): Call count_index_strings and generate
6327         header for dwarf_version >= 5.
6329 2018-05-16  Mark Wielaard  <mark@klomp.org>
6331         * dwarf2out.c (dwarf_FORM): New function.
6332         (set_indirect_string): Use dwarf_FORM.
6333         (reset_indirect_string): Likewise.
6334         (size_of_die): Likewise.
6335         (value_format): Likewise.
6336         (output_die): Likewise.
6337         (add_skeleton_AT_string): Likewise.
6338         (output_macinfo_op): Likewise.
6339         (index_string): Likewise.
6340         (output_index_string_offset): Likewise.
6341         (output_index_string): Likewise.
6342         (count_index_strings): Likewise.
6344 2018-05-16  Carl Love  <cel@us.ibm.com>
6346         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6347         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6349 2018-05-16  Martin Jambor  <mjambor@suse.cz>
6351         * ipa-prop.c (ipa_free_all_edge_args): Remove.
6352         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6354 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
6356         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6357         (fnma<mode>4): Likewise.
6358         (fms<mode>4): Likewise.
6359         (fnms<mode>4): Likewise.
6360         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6361         (aarch64_fnma<mode>4): Likewise.
6362         (aarch64_fms<mode>4): Likewise.
6363         (aarch64_fnms<mode>4): Likewise.
6364         (aarch64_fnmadd<mode>4): Likewise.
6366 2018-05-16  Jason Merrill  <jason@redhat.com>
6368         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
6370 2018-05-16  Richard Biener  <rguenther@suse.de>
6372         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6373         (dump_stmt_cost): Declare.
6374         (add_stmt_cost): Dump cost we add.
6375         (add_stmt_costs): New function.
6376         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6377         No longer exported.
6378         (vect_analyze_stmt): Adjust prototype.
6379         (vectorizable_condition): Likewise.
6380         (vectorizable_live_operation): Likewise.
6381         (vectorizable_reduction): Likewise.
6382         (vectorizable_induction): Likewise.
6383         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6384         cost vector to pass to vectorizable_ and record afterwards.
6385         (vect_model_reduction_cost): Take cost vector argument and adjust.
6386         (vect_model_induction_cost): Likewise.
6387         (vectorizable_reduction): Likewise.
6388         (vectorizable_induction): Likewise.
6389         (vectorizable_live_operation): Likewise.
6390         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6391         SLP_TREE_NUMBER_OF_VEC_STMTS.
6392         (vect_analyze_slp_cost_1): Remove.
6393         (vect_analyze_slp_cost): Likewise.
6394         (vect_slp_analyze_node_operations): Take visited args and
6395         a target cost vector.  Avoid processing already visited stmt sets.
6396         (vect_slp_analyze_operations): Use a local cost vector to gather
6397         costs and register those of non-discarded instances.
6398         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6399         (vect_schedule_slp_instance): Remove copying of
6400         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
6401         zero.
6402         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6403         adding cost.  Record cost entry location.
6404         (vect_prologue_cost_for_slp_op): Function to compute cost of
6405         a constant or invariant generated for SLP vect in the prologue,
6406         split out from vect_analyze_slp_cost_1.
6407         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
6408         (vect_model_promotion_demotion_cost): Likewise.
6409         (vect_model_store_cost): Likewise, make static.
6410         (vect_model_load_cost): Likewise.
6411         (vectorizable_bswap): Add cost vector arg and adjust.
6412         (vectorizable_call): Likewise.
6413         (vectorizable_simd_clone_call): Likewise.
6414         (vectorizable_conversion): Likewise.
6415         (vectorizable_assignment): Likewise.
6416         (vectorizable_shift): Likewise.
6417         (vectorizable_operation): Likewise.
6418         (vectorizable_store): Likewise.
6419         (vectorizable_load): Likewise.
6420         (vectorizable_condition): Likewise.
6421         (vectorizable_comparison): Likewise.
6422         (can_vectorize_live_stmts): Likewise.
6423         (vect_analyze_stmt): Likewise.
6424         (vect_transform_stmt): Adjust calls to vectorizable_*.
6425         * tree-vectorizer.c: Include gimple-pretty-print.h.
6426         (dump_stmt_cost): New function.
6428 2018-05-16  Richard Biener  <rguenther@suse.de>
6430         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6431         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6432         * tree-ssa-dse.c: Include tree-ssa-loop.h.
6433         (check_name): New callback.
6434         (dse_classify_store): Track cycles via a visited bitmap of PHI
6435         defs and simplify handling of in-loop and across loop dead stores
6436         and properly fail for loop-variant refs.  Handle byte-tracking with
6437         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6438         limiting the walk.
6440 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
6442         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6443         (vect_get_mask_type_for_stmt): Likewise.
6444         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6445         split out from...
6446         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
6447         to determine the statement's vector type and the vector type that
6448         should be used for calculating nunits.  Deal with cases in which
6449         the type has to be deferred.
6450         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6451         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6452         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6453         (vect_determine_vf_for_stmt): New functions, split out from...
6454         (vect_determine_vectorization_factor): ...here.
6455         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6456         (vect_get_mask_type_for_stmt): New functions, split out from
6457         vect_determine_vectorization_factor.
6459 2018-05-16  Richard Biener  <rguenther@suse.de>
6461         * tree-cfg.c (verify_gimple_assign_ternary): Properly
6462         verify the [VEC_]COND_EXPR embedded comparison.
6464 2018-05-15  Martin Sebor  <msebor@redhat.com>
6466         PR tree-optimization/85753
6467         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6468         RECORD_TYPE in addition to ARRAY_TYPE.
6470 2018-05-15  Martin Sebor  <msebor@redhat.com>
6472         PR middle-end/85643
6473         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6475 2018-05-15  Richard Biener  <rguenther@suse.de>
6477         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6478         add by_clobber_p one.  Change algorithm to collect all defs
6479         representing uses we need to walk and try reducing them to
6480         a single one before failing.
6481         (dse_dom_walker::dse_optimize_stmt): Adjust.
6483 2018-05-13  Mark Wielaard  <mark@klomp.org>
6485         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6486         (size_of_loc_descr): Likewise.
6487         (output_loc_operands): Likewise.
6488         (output_loc_operands_raw): Likewise.
6489         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6490         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6491         (hash_loc_operands): Likewise.
6492         (compare_loc_operands): Likewise.
6494 2018-05-14  Mark Wielaard  <mark@klomp.org>
6496         * dwarf2out.c (count_index_addrs): New function.
6497         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6499 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6501         PR tree-optimization/83648
6502         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6503         return value as malloc candidate.
6505 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6507         PR ipa/85734
6508         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6509         param as true in call to suggest_attribute.
6511 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
6513         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6514         -mreadonly-in-sdata.
6516 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6518         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6519         New pattern.
6520         (aarch64_crypto_aesd_fused): Likewise.
6522 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
6524         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6525         (movsi_aarch64): Likewise.
6526         (load_pairsi): Likewise.
6527         (load_pairdi): Likewise.
6528         (store_pairsi): Likewise.
6529         (store_pairdi): Likewise.
6530         (load_pairsf): Likewise.
6531         (load_pairdf): Likewise.
6532         (store_pairsf): Likewise.
6533         (store_pairdf): Likewise.
6534         (zero_extend): Likewise.
6535         (trunc): Swap alternatives.
6536         (fcvt_target): Add '?' to prefer w over r.
6538 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
6540         PR target/85756
6541         * config/i386/i386.md: Disallow non-commutative arithmetics in
6542         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6543         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
6544         in the peephole2 before it.
6546 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
6548         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6549         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6550         (ix86_handle_option): Handle -mcldemote.
6551         * config.gcc: New header.
6552         * config/i386/cldemoteintrin.h: New file.
6553         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6554         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6555         -mcldemote.
6556         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6557         OPTION_MASK_ISA_CLDEMOTE.
6558         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6559         (ix86_valid_target_attribute_inner_p): Ditto.
6560         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6561         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6562         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6563         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6564         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6565         (cldemote): New.
6566         * config/i386/i386.opt: Add -mcldemote.
6567         * config/i386/x86intrin.h: New header.
6568         * doc/invoke.texi: Add -mcldemote.
6570 2018-05-14  Richard Biener  <rguenther@suse.de>
6572         * doc/match-and-simplify.texi: Adjust :s documentation.
6574 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
6576         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6577         intended memcpy size.
6578         (REORDER_45): Likewise.
6580 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
6582         * sort.cc: New file.
6583         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6584         * vec.c (qsort_chk): Use gcc_qsort.
6585         * Makefile.in (OBJS-libcommon): Add sort.o.
6586         (build/sort.o): New target.  Use it...
6587         (BUILD_RTL): ... here, and...
6588         (build/gencfn-macros): ... here, and...
6589         (build/genmatch): ... here.
6591 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
6592             Chung-Ju Wu  <jasonwucj@gmail.com>
6594         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6595         * config/nds32/nds32-graywolf.md: New file.
6596         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6597         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6598         pipeline.
6599         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6600         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6601         * config/nds32/nds32.md (pipeline_model): Add graywolf.
6602         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6603         * config/nds32/pipelines.md: Include n15 settings.
6605 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
6606             Chung-Ju Wu  <jasonwucj@gmail.com>
6608         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6609         * config/nds32/nds32-n13.md: New file.
6610         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6611         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6612         pipeline.
6613         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6614         * config/nds32/nds32.md (pipeline_model): Add n13.
6615         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6616         * config/nds32/pipelines.md: Include n13 settings.
6618 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
6619             Chung-Ju Wu  <jasonwucj@gmail.com>
6621         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6622         * config/nds32/nds32-n10.md: New file.
6623         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6624         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6625         pipeline.
6626         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6627         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6628         * config/nds32/nds32.md (pipeline_model): Add n10.
6629         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6630         * config/nds32/pipelines.md: Include n10 settings.
6632 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
6633             Kito Cheng  <kito.cheng@gmail.com>
6634             Chung-Ju Wu  <jasonwucj@gmail.com>
6636         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6637         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6638         Add enum values for DSP extension instructions.
6639         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6640         New constraints.
6641         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6642         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6643         New code iterators.
6644         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6645         * config/nds32/nds32-dspext.md: New file for DSP implementation.
6646         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6647         * config/nds32/nds32-intrinsic.md: Likewise.
6648         * config/nds32/nds32_intrinsic.h: Likewise.
6649         * config/nds32/nds32-md-auxiliary.c: Likewise.
6650         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6651         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6652         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6653         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6654         * config/nds32/nds32-protos.h: New declarations for DSP extension.
6655         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6656         TYPE_DMAC in switch statement.
6657         * config/nds32/nds32.c: New checking and implementation for DSP
6658         extension instructions.
6659         * config/nds32/nds32.h: Likewise.
6660         * config/nds32/nds32.md: Likewise.
6661         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6662         * config/nds32/predicates.md: Implement new predicates for DSP
6663         extension.
6665 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
6667         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6668         Reformat alternatives and attributes so it is easier to identify
6669         which constraints/attributes go with which instruction.
6670         (mov<mode>_hardfloat32, FMOVE64): Likewise.
6671         (mov<mode>_softfloat32, FMOVE64): Likewise.
6672         (mov<mode>_hardfloat64, FMOVE64): Likewise.
6673         (mov<mode>_softfloat64, FMOVE64): Likewise.
6675 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6677         * doc/extend.texi (PowerPC Built-in Functions): Rename this
6678         subsection.
6679         (Basic PowerPC Built-in Functions): The new name of the
6680         subsection previously known as "PowerPC Built-in Functions".
6681         (Basic PowerPC Built-in Functions Available on all Configurations):
6682         New subsubsection.
6683         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6684         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6685         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6686         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6688 2018-05-11  Martin Jambor  <mjambor@suse.cz>
6690         PR ipa/85655
6691         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6692         single const.
6694 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
6696         PR target/85733
6697         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6699 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
6701         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6702         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6703         (ix86_handle_option): Handle -mwaitpkg.
6704         * config.gcc: New header.
6705         * config/i386/cpuid.h (bit_WAITPKG): New bit.
6706         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6707         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6708         function type.
6709         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6710         OPTION_MASK_ISA_WAITPKG.
6711         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6712         (ix86_option_override_internal): Add PTA_WAITPKG.
6713         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6714         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6715         IX86_BUILTIN_TPAUSE.
6716         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6717         __builtin_ia32_umwait and __builtin_ia32_tpause.
6718         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6719         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6720         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6721         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6722         UNSPECV_TPAUSE): New.
6723         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6724         * config/i386/i386.opt: Add -mwaitpkg.
6725         * config/i386/waitpkgintrin.h: New file.
6726         * config/i386/x86intrin.h: New header.
6727         * doc/invoke.texi: Add -mwaitpkg.
6729 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
6731         PR target/85606
6732         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6733         equivalent.
6734         (cortex-m0): Use armv6s-m isa.
6735         (cortex-m0plus): Likewise.
6736         (cortex-m1): Likewise.
6737         (cortex-m0.small-multiply): Likewise.
6738         (cortex-m0plus.small-multiply): Likewise.
6739         (cortex-m1.small-multiply): Likewise.
6741 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6742             Jakub Jelinek  <jakub@redhat.com>
6744         PR tree-optimization/85692
6745         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6746         source permute as well.
6748 2018-05-11  Martin Liska  <mliska@suse.cz>
6750         PR sanitizer/85556
6751         * doc/extend.texi: Document LLVM style format for no_sanitize
6752         attribute.
6754 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
6756         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6757         mode_supports_vsx_dform_quad to mode_supports_dq_form.
6758         (mode_supports_vsx_dform_quad): Likewise.
6759         (mode_supports_vmx_dform): Move these functions to be next to the
6760         other mode_supports functions.
6761         (mode_supports_dq_form): Likewise.
6762         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6763         mode_supports_dq_form.
6764         (reg_offset_addressing_ok_p): Likewise.
6765         (offsettable_ok_by_alignment): Likewise.
6766         (rs6000_legitimate_offset_address_p): Likewise.
6767         (legitimate_lo_sum_address_p): Likewise.
6768         (rs6000_legitimize_address): Likewise.
6769         (rs6000_legitimize_reload_address): Likewise.
6770         (rs6000_secondary_reload_inner): Likewise.
6771         (rs6000_preferred_reload_class): Likewise.
6772         (rs6000_output_move_128bit): Likewise.
6774 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
6776         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6777         Generate SImode target register for null target.
6778         <case IX86_BUILTIN_XGETBV>: Ditto.
6779         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6780         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6782 2018-05-10  Carl Love  <cel@us.ibm.com>
6784         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6785         dcbtt and dcbtstt if operands[2] is 0.
6787 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
6789         PR target/85693
6790         * config/i386/sse.md (usadv64qi): New expander.
6792 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
6794         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6795         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6796         -maltivec=be support.
6797         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6798         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6799         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6800         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6801         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6802         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6803         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6804         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6805         altivec_vsumsws): Adjust.
6806         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6807         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6808         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6809         support.
6810         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6811         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6812         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6813         (altivec_lve<VI_char>x): Delete expand.
6814         (*altivec_lve<VI_char>x_internal): Rename to...
6815         (altivec_lve<VI_char>x): ... this.
6816         (altivec_lvxl_<mode>): Delete expand.
6817         (*altivec_lvxl_<mode>_internal): Rename to ...
6818         (altivec_lvxl_<mode>): ... this.
6819         (altivec_stvxl_<mode>): Delete expand.
6820         (*altivec_stvxl_<mode>_internal): Rename to ...
6821         (altivec_stvxl_<mode>): ... this.
6822         (altivec_stve<VI_char>x): Delete expand.
6823         (*altivec_stve<VI_char>x_internal): Rename to ...
6824         (altivec_stve<VI_char>x): ... this.
6825         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6826         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6827         reduc_plus_scal_<mode>): Adjust.
6828         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6829         comment.
6830         (rs6000_cpu_cpp_builtins): Adjust.
6831         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6832         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6833         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6834         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6835         -maltivec=be support.
6836         (rs6000_split_vec_extract_var): Adjust.
6837         (rs6000_split_v4si_init): Adjust.
6838         (swap_selector_for_mode): Delete.
6839         (altivec_expand_lvx_be, altivec_expand_stvx_be,
6840         altivec_expand_stvex_be): Delete.
6841         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6842         -maltivec=be support.
6843         (rs6000_gimple_fold_builtin): Ditto.
6844         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6845         Adjust.
6846         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6847         (TARGET_DIRECT_MOVE_64BIT): Adjust.
6848         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6849         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6850         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6851         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6852         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6853         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6854         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6855         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6856         anonymous split): Adjust.
6857         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6858         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6860 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
6862         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6863         when --with-gxx-include-dir is also specified.
6864         * configure: Regenerate.
6866 2018-05-09  Jim Wilson  <jimw@sifive.com>
6868         PR target/84797
6869         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6870         * config/riscv/t-withmultilib: New.
6871         * config/riscv/withmultilib.h: New.
6872         * doc/install.texi: Document RISC-V --with-multilib-list support.
6874 2018-05-09  Richard Biener  <rguenther@suse.de>
6876         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6877         vector.
6878         (vect_bb_vectorization_profitable_p): Adjust.  Compute
6879         actual scalar cost using the cost vector and the add_stmt_cost
6880         machinery.
6882 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
6884         PR rtl-optimization/85645
6885         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6886         in the REG_CFA_REGISTER note for LR, don't leave it empty.
6888 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
6890         PR rtl-optimization/85645
6891         * shrink-wrap.c (spread_components): Return a boolean saying if
6892         anything was changed.
6893         (try_shrink_wrapping_separate): Iterate spread_components until
6894         nothing changes anymore.
6896 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
6898         PR rtl-optimization/85645
6899         * regrename.c (build_def_use): Also kill the chains that include the
6900         destination of a REG_CFA_REGISTER note.
6902 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
6904         PR rtl-optimization/85645
6905         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6906         insn that has a REG_CFA_REGISTER note.
6908 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
6910         * cfgexpand.c (expand_clobber): New function.
6911         (expand_gimple_stmt_1): Use it.
6912         * tree-vect-stmts.c (vect_clobber_variable): New function,
6913         split out from...
6914         (vectorizable_simd_clone_call): ...here.
6915         (vectorizable_store): Emit a clobber either side of an
6916         IFN_STORE_LANES sequence.
6917         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6919 2018-05-09  Tom de Vries  <tom@codesourcery.com>
6921         PR target/85626
6922         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6923         (define_insn "trap_if_false"): Add exit after trap.
6925 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
6927         PR rtl-optimization/85638
6928         * bb-reorder.c: Include common/common-target.h.
6929         (create_forwarder_block): New function extracted from...
6930         (fix_up_crossing_landing_pad): ...here.  Rename into...
6931         (dw2_fix_up_crossing_landing_pad): ...this.
6932         (sjlj_fix_up_crossing_landing_pad): New function.
6933         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6934         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6935         from both partitions and exit the loop after one iteration.
6937 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6939         Revert:
6940         * doc/extend.texi (PowerPC Built-in Functions): Rename this
6941         subsection.
6942         (Basic PowerPC Built-in Functions): The new name of the
6943         subsection previously known as "PowerPC Built-in Functions".
6944         (Basic PowerPC Built-in Functions Available on all Configurations):
6945         New subsubsection.
6946         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6947         subsubsection.
6948         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6949         subsubsection.
6950         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6951         subsubsection.
6952         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6953         subsubsection.
6955 2018-05-08  Jim Wilson  <jimw@sifive.com>
6957         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6958         (LD_EMUL_SUFFIX): New.
6959         (LINK_SPEC): Use it.
6961 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6963         * doc/extend.texi (PowerPC Built-in Functions): Rename this
6964         subsection.
6965         (Basic PowerPC Built-in Functions): The new name of the
6966         subsection previously known as "PowerPC Built-in Functions".
6967         (Basic PowerPC Built-in Functions Available on all Configurations):
6968         New subsubsection.
6969         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6970         subsubsection.
6971         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6972         subsubsection.
6973         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6974         subsubsection.
6975         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6976         subsubsection.
6978 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
6980         PR target/85683
6981         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6982         after cmpelim optimization.
6984 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
6986         * config.gcc: Support "goldmont".
6987         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6988         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6989         PROCESSOR_GOLDMONT.
6990         * config/i386/i386.c (m_GOLDMONT): Define.
6991         (processor_target_table): Add "goldmont".
6992         (PTA_GOLDMONT): Define.
6993         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6994         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6995         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6996         (fold_builtin_cpu): Add "goldmont".
6997         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6998         (ix86_option_override_internal): Add "goldmont".
6999         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7000         (processor_type): Add PROCESSOR_GOLDMONT.
7001         * config/i386/i386.md: Add CPU "glm".
7002         * config/i386/glm.md: New file.
7003         * config/i386/x86-tune.def: Add m_GOLDMONT.
7004         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7006 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
7008         PR target/85572
7009         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7010         E_V4DImode.
7011         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7012         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
7013         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
7015         PR target/85317
7016         * config/i386/i386.c (ix86_fold_builtin): Handle
7017         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7019         PR target/85480
7020         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7021         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7023 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
7025         PR target/85658
7026         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7027         (check_arch): Likewise.
7028         (check_fpu): Return the result rather than printing it.
7029         (end arch): Fix operator precedence.
7030         (end cpu): Likewise.
7031         (END): Print the result from check_fpu.
7033 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
7034             Alan Hayward  <alan.hayward@arm.com>
7035             David Sherwood  <david.sherwood@arm.com>
7037         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7038         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7039         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7040         (*fcmuo<mode>_and): New patterns.
7042 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
7044         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7045         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7046         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7047         (cmp_op, sve_imm_con): New code attributes.
7048         (SVE_COND_INT_CMP, imm_con): Delete.
7049         (cmp_op): Remove above unspecs from int attribute.
7050         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7051         to...
7052         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
7053         comparison-specific unspecs.
7054         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7055         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7056         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7057         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7058         (*vec_fcm<cmp_op><mode>): Rename to...
7059         (*fcm<cmp_op><mode>): ...this and adjust likewise.
7060         (*vec_fcmuo<mode>): Rename to...
7061         (*fcmuo<mode>): ...this and adjust likewise.
7062         (*pred_fcm<cmp_op><mode>): New pattern.
7063         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7064         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7065         functions.
7066         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7067         and UNORDERED.
7068         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7069         (aarch64_emit_sve_predicated_cond): New function.
7070         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7071         (aarch64_emit_unspec_cond_or): Replace with...
7072         (aarch64_emit_sve_or_conds): ...this new function.  Use
7073         aarch64_emit_sve_ptrue_op for the individual comparisons and
7074         aarch64_emit_binop to OR them together.
7075         (aarch64_emit_inverted_unspec_cond): Replace with...
7076         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
7077         aarch64_emit_sve_ptrue_op for the comparison and
7078         aarch64_emit_unop to invert the result.
7079         (aarch64_expand_sve_vec_cmp_float): Update after the above
7080         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
7082 2018-05-07  Nathan Sidwell  <nathan@acm.org>
7084         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7085         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7086         (Backwards Compatibility): Likewise.
7088 2018-05-07  Luis Machado  <luis.machado@linaro.org>
7090         PR bootstrap/85681
7091         Revert:
7092         2018-05-07  Luis Machado  <luis.machado@linaro.org>
7094         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7095         <prefetch_dynamic_strides>: New const bool field.
7096         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7097         prefetch_dynamic_strides.
7098         (exynosm1_prefetch_tune): Likewise.
7099         (thunderxt88_prefetch_tune): Likewise.
7100         (thunderx_prefetch_tune): Likewise.
7101         (thunderx2t99_prefetch_tune): Likewise.
7102         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7103         to false.
7104         (aarch64_override_options_internal): Update to set
7105         PARAM_PREFETCH_DYNAMIC_STRIDES.
7106         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7107         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7108         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7109         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7110         prefetch-dynamic-strides setting.
7112         2018-05-07  Luis Machado  <luis.machado@linaro.org>
7114         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7115         <minimum_stride>: New const int field.
7116         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7117         minimum_stride field.
7118         (exynosm1_prefetch_tune): Likewise.
7119         (thunderxt88_prefetch_tune): Likewise.
7120         (thunderx_prefetch_tune): Likewise.
7121         (thunderx2t99_prefetch_tune): Likewise.
7122         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7123         (aarch64_override_options_internal): Update to set
7124         PARAM_PREFETCH_MINIMUM_STRIDE.
7125         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7126         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7127         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7128         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7129         stride is constant and is below the minimum stride threshold.
7131 2018-05-07  Luis Machado  <luis.machado@linaro.org>
7133         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7134         to 512.
7136 2018-05-07  Luis Machado  <luis.machado@linaro.org>
7138         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7139         <prefetch_dynamic_strides>: New const bool field.
7140         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7141         prefetch_dynamic_strides.
7142         (exynosm1_prefetch_tune): Likewise.
7143         (thunderxt88_prefetch_tune): Likewise.
7144         (thunderx_prefetch_tune): Likewise.
7145         (thunderx2t99_prefetch_tune): Likewise.
7146         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7147         to false.
7148         (aarch64_override_options_internal): Update to set
7149         PARAM_PREFETCH_DYNAMIC_STRIDES.
7150         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7151         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7152         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7153         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7154         prefetch-dynamic-strides setting.
7156 2018-05-07  Luis Machado  <luis.machado@linaro.org>
7158         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7159         <minimum_stride>: New const int field.
7160         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7161         minimum_stride field.
7162         (exynosm1_prefetch_tune): Likewise.
7163         (thunderxt88_prefetch_tune): Likewise.
7164         (thunderx_prefetch_tune): Likewise.
7165         (thunderx2t99_prefetch_tune): Likewise.
7166         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7167         (aarch64_override_options_internal): Update to set
7168         PARAM_PREFETCH_MINIMUM_STRIDE.
7169         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7170         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7171         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7172         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7173         stride is constant and is below the minimum stride threshold.
7175 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
7177         PR c++/85659
7178         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7179         the type is addressable.  Don't force op into register if it has
7180         BLKmode.
7182 2018-05-05  Roland McGrath  <mcgrathr@google.com>
7184         PR other/77609
7185         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7186         any section for which we don't know a specific type it should have,
7187         regardless of name.  Previously this was done only for the exact
7188         names ".init_array", ".fini_array", and ".preinit_array".
7189         (default_elf_asm_named_section): Add comment about
7190         relationship with default_section_type_flags and SECTION_NOTYPE.
7191         (get_section): Don't consider it a type conflict if one side has
7192         SECTION_NOTYPE and the other doesn't, as long as neither has the
7193         SECTION_BSS et al used in the default_section_type_flags logic.
7195 2018-05-05  Tom de Vries  <tom@codesourcery.com>
7197         PR target/85653
7198         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7199         (workaround_barsyncs): New function.
7200         (nvptx_reorg): Use workaround_barsyncs.
7201         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7202         (define_expand "nvptx_membar_cta"): New define_expand.
7203         (define_insn "*nvptx_membar_cta"): New insn.
7205 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
7207         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7208         To improve optimization opportunities.
7209         * builtin-types.def: The new needed builtin types for the above.
7211 2018-05-04  Richard Biener  <rguenther@suse.de>
7213         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7214         * gimple-ssa-store-merging.c
7215         (imm_store_chain_info::output_merged_store): Remove redundant create,
7216         release split_store vector contents on failure.
7217         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7218         scalar stmt vector on cache hit.
7220 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
7222         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7223         Xilinx FP support.
7224         * config.gcc (powerpc-xilinx-eabi*): Remove.
7225         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7226         support.
7227         (fusion_addis_mem_combo_load): Ditto.
7228         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7229         FP support.
7230         (rs6000_cpu_cpp_builtins): Ditto.
7231         * config/rs6000/rs6000-linux.c
7232         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7233         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7234         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7235         support.
7236         (rs6000_setup_reg_addr_masks): Ditto.
7237         (rs6000_init_hard_regno_mode_ok): Ditto.
7238         (rs6000_option_override_internal): Ditto.
7239         (legitimate_lo_sum_address_p): Ditto.
7240         (rs6000_legitimize_address): Ditto.
7241         (rs6000_legitimize_reload_address): Ditto.
7242         (rs6000_legitimate_address_p): Ditto.
7243         (abi_v4_pass_in_fpr): Ditto.
7244         (setup_incoming_varargs): Ditto.
7245         (rs6000_gimplify_va_arg): Ditto.
7246         (rs6000_split_multireg_move): Ditto.
7247         (rs6000_savres_strategy): Ditto.
7248         (rs6000_emit_prologue_components): Ditto.
7249         (rs6000_emit_epilogue_components): Ditto.
7250         (rs6000_emit_prologue): Ditto.
7251         (rs6000_emit_epilogue): Ditto.
7252         (rs6000_elf_file_end): Ditto.
7253         (rs6000_function_value): Ditto.
7254         (rs6000_libcall_value): Ditto.
7255         * config/rs6000/rs6000.h: Ditto.
7256         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7257         (TARGET_MINMAX): ... this.  New.
7258         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7259         * config/rs6000/rs6000.md: Remove Xilinx FP support.
7260         (*movsi_internal1_single): Delete.
7261         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7262         mfpu=, mxilinx-fpu): Delete.
7263         * config/rs6000/singlefp.h: Delete.
7264         * config/rs6000/sysv4.h: Remove Xilinx FP support.
7265         * config/rs6000/t-rs6000: Ditto.
7266         * config/rs6000/t-xilinx: Delete.
7267         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7268         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7269         (VStype_simple): Delete.
7270         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7271         * config/rs6000/xfpu.h: Delete.
7272         * config/rs6000/xfpu.md: Delete.
7273         * config/rs6000/xilinx.h: Delete.
7274         * config/rs6000/xilinx.opt: Delete.
7275         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7276         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7278 2018-05-04  Tom de Vries  <tom@codesourcery.com>
7280         PR libgomp/85639
7281         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7282         if ignore == 0.
7284 2018-05-04  Richard Biener  <rguenther@suse.de>
7286         PR middle-end/85627
7287         * tree-complex.c (update_complex_assignment): We are always in SSA form.
7288         (expand_complex_div_wide): Likewise.
7289         (expand_complex_operations_1): Likewise.
7290         (expand_complex_libcall): Preserve EH info of the original stmt.
7291         (tree_lower_complex): Handle removed blocks.
7292         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7293         on complex multiplication and division libcall builtins.
7295 2018-05-04  Richard Biener  <rguenther@suse.de>
7297         PR middle-end/85574
7298         * fold-const.c (negate_expr_p): Restrict negation of operand
7299         zero of a division to when we know that can happen without
7300         overflow.
7301         (fold_negate_expr_1): Likewise.
7303 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
7305         PR libstdc++/85466
7306         * real.h (real_nextafter): Declare.
7307         * real.c (real_nextafter): New function.
7308         * fold-const-call.c (fold_const_nextafter): New function.
7309         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7310         CASE_CFN_NEXTTOWARD.
7311         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7312         even when arg1_mode is different from arg0_mode.
7314 2018-05-03  Nathan Sidwell  <nathan@acm.org>
7316         * doc/extend.texi (Deprecated Features): Remove
7317         -ffriend-injection.
7318         (Backwards Compatibility): Likewise.
7319         * doc/invoke.texi (C++ Language Options): Likewise.
7320         (C++ Dialect Options): Likewise.
7322 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
7324         PR target/85530
7325         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7326         _mm512_mask_mullox_epi64): New intrinsics.
7328 2018-05-03  Tom de Vries  <tom@codesourcery.com>
7330         PR testsuite/85106
7331         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7332         dump files): Add offload-tree.
7334 2018-05-03  Richard Biener  <rguenther@suse.de>
7336         PR tree-optimization/85615
7337         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7338         to loops not nested in BBs loop father to avoid creating multi-entry
7339         loops.
7341 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7343         PR tree-optimization/70291
7344         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7345         arguments.  Change return type to tree.  Emit libcall as a new
7346         statement rather than replacing existing one when inplace_p is true.
7347         (expand_complex_multiplication_components): New function.
7348         (expand_complex_multiplication): Expand floating-point complex
7349         multiplication using the above.
7350         (expand_complex_division): Rename inner_type parameter to type.
7351         Update expand_complex_libcall call-site.
7352         (expand_complex_operations_1): Update expand_complex_multiplication
7353         and expand_complex_division call-sites.
7355 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
7357         PR target/85582
7358         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7359         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7360         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7361         the highest significant bit of the shift count mask is clear.  In
7362         check whether and[sq]i3 is needed verify that all significant bits
7363         of the shift count other than the highest are set.
7365 2018-05-02  Tom de Vries  <tom@codesourcery.com>
7367         PR libgomp/82428
7368         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7369         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7370         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7371         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7372         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7373         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7374         __builtin_goacc_parlevel_size.
7376 2018-05-02  Richard Biener  <rguenther@suse.de>
7378         PR tree-optimization/85597
7379         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7380         do not use split vect_get_vec_defs call but call vect_get_slp_defs
7381         directly.
7383 2018-05-02  Tom de Vries  <tom@codesourcery.com>
7385         PR testsuite/85106
7386         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7387         dump files): Add ltrans-tree.
7389 2018-05-02  Tom de Vries  <tom@codesourcery.com>
7391         PR testsuite/85106
7392         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7393         dump files): Add wpa-ipa.
7395 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
7397         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
7398         powerpc*-*-linux*paired* target.
7399         * config/rs6000/750cl.h: Delete.
7400         * config/rs6000/paired.h: Delete.
7401         * config/rs6000/paired.md: Delete.
7402         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7403         float support.
7404         * config/rs6000/rs6000-builtin.def: Remove paired float support.
7405         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7406         comment.  Remove paired float support.
7407         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7408         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7409         VECTOR_PAIRED.
7410         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7411         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7412         declarations.
7413         * config/rs6000/rs6000.c: Remove paired float support.
7414         (paired_expand_vector_init, paired_expand_vector_move,
7415         paired_emit_vector_compare, paired_emit_vector_cond_expr,
7416         (paired_expand_lv_builtin, paired_expand_stv_builtin,
7417         paired_expand_builtin, paired_expand_predicate_builtin,
7418         paired_init_builtins): Delete.
7419         * config/rs6000/rs6000.h: Remove paired float support.
7420         * config/rs6000/rs6000.md: Remove paired float support.
7421         (move_from_CR_ov_bit): Delete.
7422         * config/rs6000/rs6000.opt (mpaired): Delete.
7423         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7424         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7426 2018-05-02  Richard Biener  <rguenther@suse.de>
7428         PR middle-end/85567
7429         * gimplify.c (gimplify_save_expr): When in SSA form allow
7430         SAVE_EXPRs to compute to SSA vars.
7432 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
7434         PR target/85582
7435         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7436         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7437         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7438         clobber operands[2], instead use a new pseudo.  Formatting fixes.
7440 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
7442         PR tree-optimization/85586
7443         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7444         exit early for statements in the same group if the accesses are
7445         not strided.
7447 2018-05-02  Tom de Vries  <tom@codesourcery.com>
7449         PR lto/85451
7450         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7451         error message.
7453 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
7455         PR tree-optimization/85143
7456         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7458 2018-05-01  Tom de Vries  <tom@codesourcery.com>
7460         PR lto/85451
7461         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7462         not found" error message.
7464 2018-05-01  Tom de Vries  <tom@codesourcery.com>
7466         PR other/83786
7467         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7468         * vec.c (test_ordered_remove_if): New function.
7469         (vec_c_tests): Call test_ordered_remove_if.
7470         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7471         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7472         * tree-vect-patterns.c (vect_pattern_recog_1): Use
7473         VEC_ORDERED_REMOVE_IF.
7475 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7477         PR tree-optimization/82665
7478         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7479         pointer subtraction where arguments come from a memchr call.
7481 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
7483         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7484         --push-state --as-needed and --pop-state instead of --as-needed and
7485         --no-as-needed if ld supports it.
7486         * configure: Regenerated.
7488         PR web/85578
7489         * doc/install.texi2html: Replace _002d with - and _002a with * in
7490         generated html files using sed.
7492 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
7494         PR c++/85523
7495         * gcc-rich-location.c (blank_line_before_p): New function.
7496         (use_new_line): New function.
7497         (gcc_rich_location::add_fixit_insert_formatted): New function.
7498         * gcc-rich-location.h
7499         (gcc_rich_location::add_fixit_insert_formatted): New function.
7501 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
7503         * selftest.c (assert_streq): Rename "expected" and "actual" to
7504         "val1" and "val2".  Extend NULL-handling to cover both inputs
7505         symmetrically, while still requiring both to be non-NULL for a pass.
7506         * selftest.h (assert_streq): Rename "expected" and "actual" to
7507         "val1" and "val2".
7508         (ASSERT_EQ): Likewise.
7509         (ASSERT_EQ_AT): Likewise.
7510         (ASSERT_KNOWN_EQ): Likewise.
7511         (ASSERT_KNOWN_EQ_AT): Likewise.
7512         (ASSERT_NE): Likewise.
7513         (ASSERT_MAYBE_NE): Likewise.
7514         (ASSERT_MAYBE_NE_AT): Likewise.
7515         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
7516         the assertion to pass.
7517         (ASSERT_STREQ_AT): Likewise.
7519 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
7521         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7522         interaction with -pie.
7524 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
7526         * selftest.h: Fix alphabetization of per-source-file selftest
7527         declarations.
7529 2018-04-30  Jason Merrill  <jason@redhat.com>
7531         PR c++/61982 - dead stores to destroyed objects.
7532         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7533         of clobber.
7535 2018-04-30  Jason Merrill  <jason@redhat.com>
7537         * tree.c (build_clobber): New.
7538         * tree.h: Declare it.
7539         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7541 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
7543         * diagnostic-show-locus.c (layout::layout): Update for
7544         location_get_source_line returning a char_span.
7545         (struct char_span): Move to input.h.
7546         (struct correction): Update for fields in char_span becoming
7547         private.
7548         (struct source_line): Update for location_get_source_line
7549         returning a char_span.
7550         (layout::print_line): Likewise.
7551         * edit-context.c (edited_file::print_content): Likewise.
7552         (edited_file::print_diff_hunk): Likewise.
7553         (edited_file::print_run_of_changed_lines): Likewise.
7554         (edited_file::get_num_lines): Likewise.
7555         (edited_line::edited_line): Likewise.
7556         * final.c (asm_show_source): Likewise.
7557         * input.c (location_get_source_line): Convert return type
7558         from const char * to char_span, losing the final "line_len"
7559         param.
7560         (dump_location_info): Update for the above.
7561         (get_substring_ranges_for_loc): Likewise.  Use a char_span
7562         when handling the literal within the line.
7563         (test_reading_source_line): Update for location_get_source_line
7564         returning a char_span.
7565         * input.h (class char_span): Move here from
7566         diagnostic-show-locus.c, converting from a struct to a class.
7567         Make data members private.
7568         (char_span::operator bool): New.
7569         (char_span::length): New.
7570         (char_span::get_buffer): New.
7571         (char_span::operator[]): New.
7572         (char_span::subspan): Make const.
7573         (char_span::xstrdup): New.
7574         (location_get_source_line): Convert return type from const char *
7575         to char_span, losing the final "line_size" param.
7577 2018-04-30  Jan Hubicka  <jh@suse.cz>
7579         * lto-wrapper.c (ltrans_priorities): New static var.
7580         (cmp_priority): New.
7581         (run_gcc): Read priorities and if doing parallel build order
7582         the Makefile by them.
7584 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
7586         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7588 2018-04-30  Richard Biener  <rguenther@suse.de>
7590         * tree-cfg.c (verify_address): Remove base argument, add
7591         flag whether to check TREE_ADDRESSABLE and do that.
7592         (verify_expr): Remove.
7593         (verify_types_in_gimple_reference): Add pieces from verify_expr.
7594         (verify_gimple_assign_single): Likewise.
7595         (verify_gimple_switch): Likewise.
7596         (verify_expr_location_1): Dereference tp once.  Add (disabled)
7597         piece from verify_expr.
7598         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7600 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
7602         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7604 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
7606         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7607         (small_data_pattern): Likewise.
7608         (arc_rewrite_small_data): Likewise.
7609         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7610         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7611         (get_symbol_alignment): New function.
7612         (legitimate_small_data_address_p): Likewise.
7613         (legitimate_scaled_address): Update, call
7614         legitimate_small_data_address_p.
7615         (output_sdata): New static variable.
7616         (arc_print_operand): Update how we handle small data operands.
7617         (arc_print_operand_address): Likewise.
7618         (arc_legitimate_address_p): Update, use
7619         legitimate_small_data_address_p.
7620         (arc_rewrite_small_data_p): Remove.
7621         (arc_rewrite_small_data_1): Likewise.
7622         (arc_rewrite_small_data): Likewise.
7623         (small_data_pattern): Likewise.
7624         (compact_sda_memory_operand): Update to use
7625         legitimate_small_data_address_p and get_symbol_alignment.
7626         (prepare_move_operands): Don't rewite sdata pattern.
7627         (prepare_extend_operands): Remove.
7628         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7629         pattern.
7630         (zero_extendqisi2): Likewise.
7631         (zero_extendhisi2): Likewise.
7632         (extendqihi2): Likewise.
7633         (extendqisi2): Likewise.
7634         (extendhisi2): Likewise.
7635         (addsi3): Likewise.
7636         (subsi3): Likewise.
7637         (andsi3): Likewise.
7638         * config/arc/constraints.md (Usd): Change it to memory constraint.
7640 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
7642         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7643         as source of std instructions.
7644         * config/arc/arc.md (movsi_insn): Update pattern predicate to
7645         allow 6-bit constants as source for store instructions.
7646         (movdi_insn): Update instruction pattern to allow 6-bit constants
7647         as source for store instructions.
7649 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
7651         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7653 2018-04-30  Nathan Sidwell  <nathan@acm.org>
7654             Sandra Loosemore <sandra@codesourcery.com>
7656         * dumpfile.c (dump_open): Allow '-' for stdout.
7657         * doc/invoke.texi (Developer Options): Document dump filename
7658         determination early.  Document stdin/stdout selection.
7660 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
7662         Microblaze Target: PIC data text relative
7664         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7665         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7666         Add declaration.
7667         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7668         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7669         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7670         New addressing mode for data-text relative position indepenedent code.
7671         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7672         'ADDRESS_SYMBOLIC_TXT_REL'.
7673         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7674         (microblaze_legitimate_pic_operand): Exclude function calls from
7675         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7676         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7677         addresses cases.
7678         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7679         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7680         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7681         for 'address + offset'.
7682         (microblaze_expand_prologue): Add new function prologue call for
7683         'r20' assignation.
7684         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7685         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7686         table in case of TARGET_PIC_DATA_TEXT_REL.
7687         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7688         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7689         Add new macros 'UNSPEC_TEXT',
7690         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7691         + exclude function calls from 'UNSPEC_PLT' in case of data text
7692         relative mode.
7693         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7694         new target hook for generating address diff vector tables in case of
7695         flag_pic.
7696         * doc/tm.texi : Regenerate.
7697         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7698         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7699         of addr diff vector generation.
7700         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7701         target hook definition.
7702         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7703         Add default function for generate_pic_addr_diff_vec -> flag_pic.
7704         * doc/invoke.texi (Add new pic option): Add new microblaze pic
7705         option for data text relative.
7707 2018-04-30  Richard Biener  <rguenther@suse.de>
7709         * tree-chrec.h (evolution_function_is_constant_p): Remove
7710         redundant check.
7711         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7713 2018-04-30  Richard Biener  <rguenther@suse.de>
7715         PR bootstrap/85571
7716         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7718 2018-04-30  Richard Biener  <rguenther@suse.de>
7720         PR tree-optimization/28364
7721         PR tree-optimization/85275
7722         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7723         copying first exit test.
7725 2018-04-28  Mark Wielaard  <mark@klomp.org>
7727         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7728         dwarf_version >= 5.
7729         (dwarf_AT): Handle DW_AT_addr_base.
7730         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7732 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
7734         PR target/84431
7735         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7736         (*ashl<dwi>3_doubleword_mask_1): Ditto.
7737         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7738         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7740 2018-04-28  Richard Biener  <rguenther@suse.de>
7742         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7743         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7744         to reflect use.  Only add interesting stmts.
7746 2018-04-27  Martin Jambor  <mjambor@suse.cz>
7748         PR ipa/85549
7749         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7750         the jump function allows for passing through aggregate values.
7752 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
7754         * input.h (in_system_header_at): Convert from macro to inline
7755         function.
7756         (from_macro_expansion_at): Likewise.
7757         (from_macro_definition_at): Likewise.
7759 2018-04-27  Jeff Law  <law@redhat.com>
7761         * config.gcc: Mark tile* targets as deprecated/obsolete.
7763 2018-04-27  Richard Biener  <rguenther@suse.de>
7765         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7766         fix for ILP32.
7768 2018-04-27  Richard Biener  <rguenther@suse.de>
7770         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7772 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
7774         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7775         with Yd constraint. Set "preferred_for_speed" attribute from
7776         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7777         with Yd constraint.
7778         (*movdi_internal): Ditto.
7779         (movti_interunit splitters): Remove
7780         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7781         (movdi_interunit splitters): Ditto.
7782         * config/i386/constraints.md (Ye): Remove.
7783         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7785 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7787         PR target/85512
7788         * config/aarch64/constraints.md (Usg): Limit to 31.
7789         (Usj): Limit to 63.
7791 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
7793         PR tree-optimization/85529
7794         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7795         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7796         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7797         zero extension or masking of the MSB bit.
7798         (optimize_range_tests): Add FIRST_BB argument, pass it through
7799         to optimize_range_tests_var_bound.
7800         (maybe_optimize_range_tests, reassociate_bb): Adjust
7801         optimize_range_tests callers.
7803 2018-04-26  Richard Biener  <rguenther@suse.de>
7804             Jakub Jelinek  <jakub@redhat.com>
7806         * cgraph.h (symbol_table): Just declare debug method here.
7807         * symtab.c (symbol_table::debug): Define.
7809 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
7811         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7813 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
7815         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7816         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7817         (*movdi_internal): Substitute Yi and Yj constraint with x
7818         and Ym and Yn constraint with y constraint.  Update "isa"
7819         attribute and set "preferred_for_speed" attribute from
7820         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7821         (*movsi_internal): Ditto.
7822         (*movdf_internal): Ditto.
7823         (*movsf_internal): Ditto.
7824         (*zero_extendsidi2): Ditto.
7825         * config/i386/sse.md (vec_set<mode>_0): Ditto.
7826         (sse2_loadld): Ditto.
7827         (*vec_extract<ssevecmodelower>_0): Ditto.
7828         (*vec_extractv4si_0_zext_sse4): Ditto.
7829         (vec_concatv2di): Ditto.
7830         (*vec_dup<mode>): Ditto.
7831         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7832         * config/i386/constraints.md (Yi): Remove.
7833         (Yj): Remove.
7834         (Ym): Remove.
7835         (Yn): Remove.
7837 2018-04-26  Nathan Sidwell  <nathan@acm.org>
7839         * dumpfile.c (dump_open): New.
7840         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7841         (dump_finish): Detect stdio/stderr by value not name.
7843 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
7845         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7847 2018-04-26  Tom de Vries  <tom@codesourcery.com>
7849         PR target/84952
7850         * config/nvptx/nvptx.c (verify_neutering_jumps)
7851         (verify_neutering_labels): New function
7852         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7854 2018-04-26  Tom de Vries  <tom@codesourcery.com>
7856         PR target/84025
7857         * config/nvptx/nvptx.c (needs_neutering_p): New function.
7858         (nvptx_single): Use needs_neutering_p to skip over insns that do not
7859         need neutering.
7861 2018-04-26  Richard Biener <rguenther@suse.de>
7862             Tom de Vries  <tom@codesourcery.com>
7864         PR lto/85422
7865         * lto-streamer-out.c (output_function): Fixup loops if required to match
7866         discovery done in the reader.
7868 2018-04-26  Richard Biener  <rguenther@suse.de>
7870         PR tree-optimization/85116
7871         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7872         have a loop exit from the single latch predecessor.  Remove
7873         case of header with just condition.
7874         (ch_base::copy_headers): Exclude infinite loops from any
7875         processing.
7876         (pass_ch::execute): Record exits.
7878 2018-04-26  Richard Biener  <rguenther@suse.de>
7880         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7881         prologue cost vector and pass it to vect_get_load_cost.
7882         (vect_get_peeling_costs_all_drs): Likewise.
7883         (vect_peeling_hash_get_lowest_cost): Likewise.
7884         (vect_enhance_data_refs_alignment): Likewise.
7886 2018-04-26  Richard Biener  <rguenther@suse.de>
7888         PR middle-end/85450
7889         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7890         checking of integer<->pointer conversions.
7891         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7892         sign-/zero-extending pointer types.
7893         (expand_omp_for_static_chunk): Likewise.
7895 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
7896             Jean Lee  <xiaoyur347@gmail.com>
7898         * config/mips/mips.c (mips_asan_shadow_offset): New function.
7899         (TARGET_ASAN_SHADOW_OFFSET): Define.
7900         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7901         true for -fsanitize=address.
7903 2018-04-25  Mark Wielaard  <mark@klomp.org>
7905         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7906         shorter ones.
7908 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
7910         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7911         than "alu", remove explicit "memory" and "imm_disp" attributes.
7912         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7914         PR middle-end/85414
7915         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7916         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7917         gen_lowpart_no_emit.
7919 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
7921         PR target/85473
7922         * config/i386/i386.c (ix86_expand_builtin): Change memory
7923         operand to XI, extend p0 to Pmode.
7924         * config/i386/i386.md: Change unspec volatile and operand
7925         1 mode to XI, change operand 0 mode to P.
7927 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
7929         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7930         GET_MODE_MASK before any checking.
7931         (nds32_can_use_bset_p): Likewise.
7932         (nds32_can_use_btgl_p): Likewise.
7934 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
7936         * config/nds32/nds32-doubleword.md: New define_split pattern for
7937         illegal register number.
7939 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
7941         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7943 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
7945         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7947 2018-04-25  Richard Biener  <rguenther@suse.de>
7949         * lto-streamer.h (LTO_major_version): Bump to 8.
7951 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
7953         * BASE-VER: Set to 9.0.0.
7955 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
7957         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7958         in __abskf2 and __powikf2.
7960 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7962         PR target/85512
7963         * config/aarch64/constraints.md (Usg, Usj): New constraints.
7964         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7965         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7966         Use the above on operand 2.  Reindent.
7967         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7969 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
7971         PR target/85485
7972         * common/config/i386/i386-common.c (ix86_handle_option): Don't
7973         handle OPT_mcet.
7974         * config/i386/i386.opt (mcet): Removed.
7975         * doc/install.texi: Remove -mcet documentation.
7976         * doc/invoke.texi: Likewise.
7978 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
7980         PR target/85485
7981         * doc/install.texi: Remove -mcet from bootstrap-cet.
7983 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
7985         PR target/85511
7986         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7987         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7988         if TARGET_64BIT.
7990         PR target/85503
7991         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7992         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7993         containing a CONST_VECTOR.
7995 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
7997         * doc/install.texi: Update newlib dependency for nvptx.
7999 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
8001         PR target/85508
8002         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8003         instead of INTVAL when shifting x left.
8005 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
8007         PR tree-optimization/85478
8008         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8009         vect_grouped_store_supported for single element vectors.
8011 2018-04-24  Richard Biener  <rguenther@suse.de>
8013         PR target/85491
8014         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8015         load cost increase to the case of non-constant step.
8017 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
8019         PR target/84828
8020         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8021         destination if any_malformed_asm.
8023 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
8025         PR middle-end/85496
8026         * expr.c (store_field): In the bitfield case, if the value comes from
8027         a function call and is returned in registers by means of a PARALLEL,
8028         do not change the mode of the temporary unless BLKmode and VOIDmode.
8030 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
8032         PR rtl-optimization/85423
8033         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8034         dependencies to debug insns when the previous insn is non-debug.
8036 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
8038         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8039         enums into a single definition.
8040         (fls): Fix predicates and printing.
8041         (seti): Likewise.
8043 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
8045         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8046         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8047         and short u6 immediate.
8048         (check_if_valid_sleep_operand): Remove.
8049         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8051 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8053         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8054         flag_always_save_lp condition.
8055         * config/nds32/nds32.opt (malways-save-lp): New option.
8057 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
8059         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8060         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8061         * config/nds32/nds32.h
8062         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8063         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8065 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
8067         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8068         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8070 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
8071             Chung-Ju Wu  <jasonwucj@gmail.com>
8073         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8074         Declare.
8075         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8076         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8078 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8080         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8082 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8084         * config/nds32/nds32-protos.h (nds32_data_alignment,
8085         nds32_local_alignment): Declare.
8086         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8087         nds32_local_alignment): New functions.
8088         (TARGET_CONSTANT_ALIGNMENT): Define.
8089         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8091 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8093         * config/nds32/nds32.c
8094         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8095         (TARGET_MODES_TIEABLE_P): Likewise.
8097 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8099         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8100         level Ofast and Og.
8102 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
8103             Chung-Ju Wu  <jasonwucj@gmail.com>
8105         * config/nds32/constants.md (unspec_volatile_element): Add enum values
8106         for unaligned access.
8107         * config/nds32/nds32-intrinsic.c: Implementation of expanding
8108         unaligned access.
8109         * config/nds32/nds32-intrinsic.md: Likewise.
8110         * config/nds32/nds32_intrinsic.h: Likewise.
8111         * config/nds32/nds32.h (nds32_builtins): Likewise.
8112         * config/nds32/nds32.opt (munaligned-access): New option.
8113         * config/nds32/nds32.c (nds32_asm_file_start): Display
8114         flag_unaligned_access status.
8116 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
8118         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8119         -mno-relax is present.
8120         * config/riscv/linux.h (LINK_SPEC): Ditto.
8122 2018-04-20  Martin Sebor  <msebor@redhat.com>
8124         PR c/85365
8125         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8126         for null pointers.
8127         (gimple_fold_builtin_stxcpy_chk): Same.
8128         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8130 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
8132         PR target/85456
8133         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8134         __powikf2 when long double is IEEE 128-bit.
8136 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
8138         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8139         step to make sure stack always aligned.
8141 2018-04-20  Carl Love  <cel@us.ibm.com>
8143         PR target/83402
8144         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8145         size check for arg0.
8147 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
8148             Tom de Vries  <tom@codesourcery.com>
8150         PR target/85445
8151         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8152         Emit insns for calls too.
8153         (nvptx_find_par): Always look for worker-level predecessor insn.
8154         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
8155         calls.
8156         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8157         (nvptx_process_pars): Propagate frames for calls.
8159 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
8161         PR target/85469
8162         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8163         Removed.
8164         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8165         (ix86_handle_option): Don't handle OPT_mibt.
8166         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8167         __SHSTK__.
8168         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8169         has_ibt and ibt.
8170         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8171         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8172         (ix86_target_macros): Define __CET__ with flag_cf_protection
8173         for -fcf-protection.
8174         * config/i386/i386.c (isa2_opts): Remove -mibt.
8175         * config/i386/i386.h (TARGET_IBT): Removed.
8176         (TARGET_IBT_P): Likewise.
8177         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8178         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8179         * config/i386/i386.opt (mcet): Update help message.
8180         (mshstk): Likewise.
8181         (mibt): Removed.
8182         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
8183         -mcet as an alias for -mshstk.
8185 2018-04-20  Richard Biener <rguenther@suse.de>
8187         PR middle-end/85475
8188         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8189         complexity by forcing a single use of the multiply operand.
8191 2018-04-20  Martin Jambor  <mjambor@suse.cz>
8193         ipa/85449
8194         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8195         recursion dependency to only apply to non-clones.
8197 2018-04-20  Martin Jambor  <mjambor@suse.cz>
8199         ipa/85447
8200         * ipa-cp.c (create_specialized_node): Check that clones of
8201         self-recursive edges exist during IPA-CP.
8203 2018-04-19  Toon Moene  <toon@moene.org>
8205         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8206         by -O3.
8208 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
8210         PR tree-optimization/85467
8211         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8212         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
8213         VECTOR_CST element to type.
8215 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
8217         PR target/85397
8218         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8219         * config/i386/i386.md (builtin_setjmp_setup): Removed.
8220         (builtin_longjmp): Likewise.
8221         (save_stack_nonlocal): New pattern.
8222         (restore_stack_nonlocal): Likewise.
8224 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
8226         PR target/85404
8227         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8228         Replace ASM_OUTPUT_LABEL with fprintf.
8230 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
8232         PR target/85417
8233         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8234         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8235         * config/i386/i386-c.c (ix86_target_macros_internal): Also
8236         define __IBT__ and __SHSTK__ for -fcf-protection.
8237         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8238         TARGET_IBT.
8239         (ix86_trampoline_init): Likewise.
8240         (x86_output_mi_thunk): Likewise.
8241         (ix86_notrack_prefixed_insn_p): Likewise.
8242         (ix86_option_override_internal): Don't disallow -fcf-protection.
8243         * config/i386/i386.md (rdssp<mode>): Also enable for
8244         -fcf-protection.
8245         (incssp<mode>): Likewise.
8246         (nop_endbr): Likewise.
8247         * config/i386/i386.opt (mcet): Change help message to built-in
8248         functions only.
8249         (mibt): Likewise.
8250         (mshstk): Likewise.
8251         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8252         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
8253         enable CET built-in functions.
8255 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
8257         * common/config/i386/i386-common.c
8258         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8259         OPTION_MASK_ISA_MOVDIRI_UNSET,
8260         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8261         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8262         * config.gcc (movdirintrin.h): New header.
8263         * config/i386/cpuid.h (bit_MOVDIRI,
8264         bit_MOVDIR64B): New bits.
8265         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8266         and -mmvodir64b.
8267         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8268         (VOID, PVOID, PCVOID)): New function types.
8269         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8270         __builtin_ia32_directstoreu_u64,
8271         __builtin_ia32_movdir64b): New builtins.
8272         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8273         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8274         and -mmovdiri.
8275         (ix86_valid_target_attribute_inner_p): Ditto.
8276         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8277         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8278         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8279         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8280         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8281         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8282         (movdiri<mode>, movdir64b_<mode>): New.
8283         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8284         * config/i386/immintrin.h: Include movdirintrin.h.
8285         * config/i386/movdirintrin.h: New file.
8286         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8288 2018-04-19  Richard Biener  <rguenther@suse.de>
8290         PR middle-end/85455
8291         * cfg.c (clear_bb_flags): When loop state says we have
8292         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8294 2018-04-19  Richard Biener  <rguenther@suse.de>
8296         PR tree-optimization/84737
8297         * tree-vect-data-refs.c (vect_copy_ref_info): New function
8298         copying restrict info.
8299         (vect_setup_realignment): Use it.
8300         * tree-vectorizer.h (vect_copy_ref_info): Declare.
8301         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8302         the first DR to all generated stores.
8303         (vectorizable_load): Likewise for loads.
8305 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
8307         PR tree-optimization/85446
8308         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8309         the integral and pointer types to have the same precision.
8311         * doc/install.texi: Document --disable-cet being the default and
8312         --enable-cet=auto.
8314 2018-04-18  Martin Liska  <mliska@suse.cz>
8316         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8317         style.
8319 2018-04-18  Martin Liska  <mliska@suse.cz>
8321         Revert
8322         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
8324         PR ipa/83983
8325         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8326         arguments if they are comparable.
8328 2018-04-18  Martin Liska  <mliska@suse.cz>
8330         Revert
8331         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
8333         PR lto/84805
8334         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8335         incomplete types.
8337 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
8339         PR target/85388
8340         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8341         ENDBR after calling __morestack.
8343 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
8345         PR jit/85384
8346         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8347         by using gcc_base_ver to generate a gcc_driver_version, and use
8348         it when generating GCC_DRIVER_NAME.
8349         * configure: Regenerate.
8351 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
8353         PR target/81084
8354         * config.gcc: Obsolete powerpc*-*-*spe*.
8356 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
8358         PR debug/84637
8359         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8360         (stabstr_D): Change type of unum from unsigned int to
8361         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
8362         type.
8364 2018-04-17  Jim Wilson  <jimw@sifive.com>
8366         PR 84856
8367         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8368         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8369         Set arg_pointer_offset after using pretend_args_size.
8371 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
8373         PR rtl-optimization/85431
8374         * dse.c (record_store): Ignore zero width stores.
8376         PR sanitizer/85230
8377         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
8378         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8379         __builtin_stack_restore rather than after it.
8380         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8381         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8382         argument instead of virtual_dynamic_stack_rtx.
8384 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8386         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8387         New prototype.
8388         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8389         Add note to error message to explain internal mapping of overloaded
8390         built-in function name to non-overloaded built-in function name.
8391         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8392         function.
8394 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
8396         PR target/85424
8397         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8398         where the inputs overlap with the output.
8400 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
8402         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8403         (=v, v) alternative and explicit "memory" attribute.
8404         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
8405         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8406         attributes.
8407         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8408         "sselog1" type instead of "sselog".
8409         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8410         "sselog".  Remove explicit "memory" attribute.
8411         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8412         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8413         attributes.
8414         (vec_extract_hi_v32hi): Merge all alternatives into one, use
8415         "sselog1" type instead of "sselog".  Remove explicit "memory"
8416         attribute.
8417         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8418         use "sselog1" type instead of "sselog".  Remove explicit "memory"
8419         attribute.
8420         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8421         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8422         attributes.
8423         (vec_extract_hi_v64qi): Merge all alternatives into one, use
8424         "sselog1" type instead of "sselog".  Remove explicit "memory"
8425         attribute.
8426         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8427         use "sselog1" type instead of "sselog".  Remove explicit "memory"
8428         attribute.
8430         PR target/85430
8431         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8433         PR middle-end/85414
8434         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8435         on a SUBREG.
8437 2018-04-17  Martin Jambor  <mjambor@suse.cz>
8439         PR ipa/85421
8440         * ipa-cp.c (create_specialized_node): Call
8441         expand_all_artificial_thunks if necessary.
8443 2018-04-17  Martin Liska  <mliska@suse.cz>
8445         PR lto/85405
8446         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8447         in message, remote space in between '_G' and '('.
8449 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
8451         PR target/85281
8452         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8453         avx512f_vmcmp<mode>3<round_saeonly_name>,
8454         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8455         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8456         avx512f_rndscale<mode><round_saeonly_name>,
8457         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8458         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8459         Use %<iptr>2 instead of %2 for -masm=intel.
8460         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8461         avx512f_vcvttss2usi<round_saeonly_name>,
8462         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8463         -masm=intel.
8464         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8465         avx512f_vcvttsd2usi<round_saeonly_name>,
8466         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8467         Use %q1 instead of %1 for -masm=intel.
8468         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8469         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8470         of %3 for -masm=intel.
8471         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8472         -masm=intel.
8473         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8474         -masm=intel.
8475         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8476         -masm=intel.
8477         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8478         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8479         %g1.
8480         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8481         -masm=intel.
8482         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8483         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8484         %g1 and one with %0 and %1.
8485         (avx512er_vmrcp28<mode><round_saeonly_name>,
8486         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8487         %1 for -masm=intel.
8488         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8489         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8490         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8491         of %0 and %{%4%} for -masm=intel.
8492         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8493         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8494         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8495         order of %0 and %{%5%}%{z%} for -masm=intel.
8497 2018-04-17  Jan Hubicka  <jh@suse.cz>
8499         PR lto/85405
8500         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8502 2018-04-17  Martin Liska  <mliska@suse.cz>
8504         PR ipa/85329
8505         * multiple_target.c (create_dispatcher_calls): Set apostrophes
8506         for target_clone error message.  Make default implementation
8507         clone to be a local declaration.
8508         (separate_attrs): Add new argument and check for an empty
8509         string.
8510         (expand_target_clones): Handle it.
8511         (ipa_target_clone): Make redirection just for target_clones
8512         functions.
8514 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
8515             Tom de Vries  <tom@codesourcery.com>
8517         PR middle-end/84955
8518         * omp-expand.c (expand_oacc_for): Add dummy false branch for
8519         tiled basic blocks without omp continue statements.
8521 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
8523         PR target/83660
8524         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8525         vec_extract expression as having side effects to make sure it gets
8526         a cleanup point.
8528 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
8530         PR target/85403
8531         * config/i386/i386.c (get_builtin_code_for_version): Check
8532         error_mark_node.
8534 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
8536         PR target/84331
8537         * gcc/config.gcc: Support "skylake".
8538         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8539         PROCESSOR_SKYLAKE.
8540         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8541         (processor_target_table): Add "skylake".
8542         (ix86_option_override_internal): Add "skylake".
8543         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8544         PROCESSOR_CANNONLAKE.
8545         (get_builtin_code_for_version): Fix priority for
8546         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8547         PROCESSOR_SKYLAKE-AVX512.
8548         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8549         (processor_type): Add PROCESSOR_SKYLAKE.
8551 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
8552             Jason Merrill  <jason@redhat.com>
8554         PR c++/85112
8555         * convert.c (convert_to_integer_1): Use direct recursion for
8556         enumeral types and types with a precision less than the number
8557         of bits in their mode.
8559 2018-04-16  Julia Koval  <julia.koval@intel.com>
8561         PR target/84413
8562         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8563         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8565 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
8567         PR target/85293
8568         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8569         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8570         and -mno-direct-move.
8572 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
8574         PR target/83402
8575         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8576         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8577         Ensure negative shifts result in {0}.
8579 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
8581         PR rtl-optimization/79916
8582         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8583         regs (if any) to define how to gnerate SD moves when LRA is in
8584         progress.
8586 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
8588         PR rtl-optimization/85393
8589         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8590         * except.c (expand_dw2_landing_pad_for_region): Make static.
8591         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8592         a label and unconditional jump to old_bb, rather than
8593         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8594         basic block.
8596         PR rtl-optimization/85376
8597         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8598         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8599         instead of a specific value.
8601 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
8602             Bin Cheng  <bin.cheng@arm.com>
8604         PR tree-optimization/82965
8605         PR tree-optimization/83991
8606         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8607         by_profile_only parameter.
8608         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8609         information if the loop was predicted to iterate too many times.
8610         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8612 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
8614         PR lto/71991
8615         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8616         always inline.
8618 2018-04-13  Martin Liska  <mliska@suse.cz>
8619             Jakub Jelinek  <jakub@redhat.com>
8621         PR middle-end/81657
8622         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8623         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8624         * builtins.c (expand_builtin_memory_copy_args): Use
8625         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8626         handle dest_addr == pc_rtx.
8628 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
8630         PR target/85291
8631         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8632         asked to not generate direct moves.
8633         (fix_trunc<mode>si2_stfiwx): Similar.
8634         (fix_trunc<mode>si2_internal): Similar.
8636 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
8638         PR debug/83157
8639         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8640         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8641         lookup if dest in some wider mode is known to be const0_rtx and
8642         if so, record permanent equivalence for it to be ZERO_EXTEND of
8643         the narrower mode destination.
8645 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
8647         * lto-streamer-out.c (output_function): Revert 259346.
8648         * omp-expand.c (expand_oacc_for): Likewise.
8650 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
8652         PR rtl-optimization/85354
8653         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8654         * sel-sched.c (sel_global_init): ... here.
8656 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
8658         PR target/85238
8659         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8660         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8661         mode for PE-COFF targets.
8662         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8663         (i386_pe_asm_lto_end): Likewise.
8664         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8665         (TARGET_ASM_LTO_END): Likewise.
8666         * config/i386/winnt.c (saved_debug_info_level): New static variable.
8667         (i386_pe_asm_lto_start): New function.
8668         (i386_pe_asm_lto_end): Likewise.
8670 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
8671             Richard Biener  <rguenther@suse.de>
8673         PR middle-end/84955
8674         * lto-streamer-out.c (output_function): Fix CFG loop state before
8675         streaming out.
8676         * omp-expand.c (expand_oacc_for): Handle calls to internal
8677         functions like regular functions.
8679 2018-04-12  Richard Biener  <rguenther@suse.de>
8681         PR lto/85371
8682         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8683         for the early LTO debug to properly generate references to it
8684         during DIE emission.  Do not re-use that for the skeleton for
8685         split-dwarf.
8686         (dwarf2out_early_finish): Likewise.
8688 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
8690         PR target/85328
8691         * config/i386/sse.md
8692         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8693         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8694         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8695         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8696         and output is a reg, avoid creating invalid lowpart subreg, but
8697         instead split into a 512-bit move.  Don't split if not AVX512VL,
8698         input is xmm16+ reg and output is a mem.
8699         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8700         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8701         xmm16+ reg and output is a mem.
8703 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8705         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8706         also for flag_dwarf2_cfi_asm.
8708 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
8710         PR rtl-optimization/85342
8711         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8712         a bool scalar var inside of the loop instead.  Don't try to update
8713         recog_data.operand after failed apply_change_group.
8715 2018-04-12  Tom de Vries  <tom@codesourcery.com>
8717         PR target/85296
8718         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8719         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
8720         array with flexible array member as array without given dimension.
8721         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8722         argument for undefined param to true.
8724 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
8726         PR target/85321
8727         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8728         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8729         from PowerPC section.
8730         * config/rs6000/sysv4.opt (mcall-): Improve help text.
8731         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8732         help text that is too long.
8733         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8734         help text that is too long.
8735         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8736         help text that is too long.
8738 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
8740         * config/alpha/alpha.md (stack_probe_internal): Rename
8741         from "probe_stack".  Update all callers.
8743 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
8745         PR rtl-optimization/84566
8746         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8747         sched_macro_fuse_insns.
8749 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
8751         PR target/84301
8752         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8753         (compute_block_dependences): ... from here.
8755 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
8757         PR tree-optimization/85331
8758         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8759         from int to HOST_WIDE_INT.
8761 2018-04-11  Martin Jambor  <mjambor@suse.cz>
8763         PR ipa/84149
8764         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8765         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8766         not the same as the source val.
8767         (cgraph_edge_brings_value_p): New parameter.
8768         (gather_edges_for_value): Pass destination value to
8769         cgraph_edge_brings_value_p.
8770         (perhaps_add_new_callers): Likewise.
8771         (get_info_about_necessary_edges): Likewise and exclude values brought
8772         only by self-recursive edges.
8773         (create_specialized_node): Redirect only clones of self-calling edges.
8774         (+self_recursive_pass_through_p): New function.
8775         (find_more_scalar_values_for_callers_subset): Use it.
8776         (find_aggregate_values_for_callers_subset): Likewise.
8777         (known_aggs_to_agg_replacement_list): Removed.
8778         (decide_whether_version_node): Re-calculate known constants for all
8779         remaining context clones.
8781 2018-04-11  Richard Biener  <rguenther@suse.de>
8783         PR lto/85339
8784         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8785         from early DWARF output.
8786         (dwarf2out_early_finish): Output line info unconditionally into
8787         early DWARF and add reference to it.
8789 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
8791         PR target/85281
8792         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8793         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8794         other than V2DFmode using iptr mode attribute.
8795         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8797 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
8799         PR rtl-optimization/84659
8800         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8802 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
8804         PR debug/85302
8805         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8806         SIZEP is NULL.
8807         (output_loc_list): Pass address of a dummy size variable even in the
8808         locview handling loop.
8809         (index_location_lists): Add comment on why skip_loc_list_entry can't
8810         call size_of_locs.
8812 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8814         PR target/85261
8815         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8816         into register.
8818 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
8820         PR target/85321
8821         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8822         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8823         and -mstring-compare-inline-limit.
8825 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
8827         PR target/85287
8828         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8829         for stack clash protection in a register whenever we need it to be in
8830         a register.
8832 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
8834         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8835         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8837 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
8839         PR target/85321
8840         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8841         the help text.
8842         (mlong-double-): Ditto.
8843         * config/rs6000/sysv4.opt (msdata=): Ditto.
8844         (mtls-size=): Ditto.
8846 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8848         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8849         erroneous entries for
8850         "vector int vec_ldl (int, long int *)", and
8851         "vector unsigned int vec_ldl (int, unsigned long int *)".
8852         Add comments and entries for
8853         "vector bool char vec_ldl (int, bool char *)",
8854         "vector bool short vec_ldl (int, bool short *)",
8855         "vector bool int vec_ldl (int, bool int *)",
8856         "vector bool long long vec_ldl (int, bool long long *)",
8857         "vector pixel vec_ldl (int, pixel *)",
8858         "vector long long vec_ldl (int, long long *)",
8859         "vector unsigned long long vec_ldl (int, unsigned long long *)".
8860         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8861         type tree bool_long_long_type_node and correct definition of
8862         bool_V2DI_type_node to make reference to this new type tree.
8863         (rs6000_mangle_type): Replace erroneous reference to
8864         bool_long_type_node with bool_long_long_type_node.
8865         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8866         comments to emphasize sign distinctions for char and int types and
8867         replace RS6000_BTI_bool_long constant with
8868         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
8869         use of RS6000_BTI_pixel.
8870         (bool_long_type_node): Remove this macro definition.
8871         (bool_long_long_type_node): New macro definition
8873 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
8875         PR rtl-optimization/85300
8876         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8877         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8878         simplify_unary_operation fails.
8880 2018-04-10  Martin Liska  <mliska@suse.cz>
8882         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8883         cgraph_edge and ipa_ref.
8885 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
8887         PR target/85177
8888         PR target/85255
8889         * config/i386/sse.md
8890         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8891         computation of the VEC_MERGE selector from mask.
8892         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8893         Fix decoding of the VEC_MERGE selector into mask.
8895 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
8897         PR tree-optimization/85286
8898         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8900 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
8902         * final.c (final_1): Set insn_last_address as well as
8903         insn_current_address.
8905 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8907         PR target/85173
8908         * explow.c (emit_stack_probe): Call validize_mem on memory location
8909         before passing it to gen_probe_stack.  Create address operand and
8910         legitimize it for the probe_stack_address case.
8912 2018-04-09  Jan Hubicka  <jh@suse.cz>
8914         PR lto/85078
8915         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8916         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8917         * tree.c (free_lang_data_in_type): Fix handling of binfos;
8918         walk basetypes.
8919         (free_lang_data): Rebuild type inheritance graph.
8921 2018-04-09  Martin Sebor  <msebor@redhat.com>
8923         * invoke.texi (-finline-small-functions): Mention other optimization
8924         options.
8925         (-findirect-inlining, -fpartial-inlining): Same.
8926         (-finline-functions-called-once): Same.
8927         (-freorder-blocks-and-partition): Same.
8929 2018-04-09  Jan Hubicka  <jh@suse.cz>
8931         PR rtl/84058
8932         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8933         jumps; choose last target that matches the criteria (i.e.
8934         no partition changes for non-crossing jumps).
8935         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8936         support for redirecting crossing jumps to non-crossing.
8938 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
8940         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8941         also for naked functions.
8943 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
8945         * config/arc/arc.md (add_shift): New pattern.
8946         (add_shift2): Likewise.
8947         (sub_shift): Likewise.
8948         (sub_shift_cmp0_noout): Likewise.
8949         (compare_si_ashiftsi): Likewise.
8950         (xbfu_cmp0_noout): New combine pattern.
8951         (xbfu_cmp0"): Likewise.
8952         (movsi_set_cc_insn): Place the predicable variant first.
8953         (commutative_binary_cmp0_noout): Remove clobber.
8954         (commutative_binary_cmp0): New pattern.
8955         (noncommutative_binary_cmp0): Likewise.
8956         (noncommutative_binary_cmp0_noout): Likewise.
8957         (noncommutative_binary_comparison_result_used): Removed.
8958         (rsub_cmp0): New pattern.
8959         (rsub_cmp0_noout): Likewise.
8960         (extzvsi): Changed, keep only meaningful variants.
8961         (SQH, SEZ): New iterators.
8962         (SQH_postfix): New mode attribute.
8963         (SEZ_prefix): New code attribute.
8964         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8965         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8966         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8967         of numerical value.
8968         (noncommutative_operator): Check the availability of barrel
8969         shifter option.
8971 2018-04-09  Richard Biener  <rguenther@suse.de>
8973         PR tree-optimization/85284
8974         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8975         Only use the niter constraining form of simple_iv when the exit
8976         is always executed.
8978 2018-04-09  Tom de Vries  <tom@codesourcery.com>
8980         PR target/84041
8981         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8982         (define_expand "*memory_barrier"): New define_expand.
8983         (define_insn "memory_barrier"): New insn.
8985 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
8987         PR rtl-optimization/80463
8988         PR rtl-optimization/83972
8989         PR rtl-optimization/83480
8991         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8992         correct producer for the insn.
8993         (tidy_control_flow): Fixup seqnos in case of debug insns.
8995 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
8997         PR rtl-optimization/83913
8999         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9000         different sched-times when merging exprs.
9002 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
9004         PR rtl-optimization/83962
9006         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9007         tidy_fallthru_edge and tidy_control_flow.
9009 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
9011         PR rtl-optimization/83530
9013         * sel-sched.c (force_next_insn): New global variable.
9014         (remove_insn_for_debug): When force_next_insn is true, also leave only
9015         next insn in the ready list.
9016         (sel_sched_region): When the region wasn't scheduled, make another pass
9017         over it with force_next_insn set to 1.
9019 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
9021         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9022         into tm_file.
9023         * config/nds32/constants.md (unspec_volatile_element): Add enum values
9024         for interrupt control.
9025         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9026         functions for interrupt control.
9027         * config/nds32/nds32-intrinsic.md: Likewise.
9028         * config/nds32/nds32_intrinsic.h: Likewise.
9029         * config/nds32/nds32.h (nds32_builtins): Likewise.
9031 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
9033         * config/nds32/nds32.c (nds32_init_machine_status,
9034         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9035         strict_aligned_p field.
9036         (nds32_expand_to_rtl_hook): New function.
9037         (TARGET_EXPAND_TO_RTL_HOOK): Define.
9038         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9040 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
9041             Chung-Ju Wu  <jasonwucj@gmail.com>
9043         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9044         * config/nds32/nds32-n7.md: New file.
9045         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9046         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9047         pipeline.
9048         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9049         * config/nds32/nds32.md (pipeline_model): Add n7.
9050         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9051         * config/nds32/pipelines.md: Include n7 settings.
9053 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
9054             Chung-Ju Wu  <jasonwucj@gmail.com>
9056         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9057         * config/nds32/nds32-e8.md: New file.
9058         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9059         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9060         pipeline.
9061         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9062         * config/nds32/nds32.md (pipeline_model): Add e8.
9063         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9064         * config/nds32/pipelines.md: Include e8 settings.
9066 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
9067             Chung-Ju Wu  <jasonwucj@gmail.com>
9069         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9070         * config/nds32/nds32-n8.md: New file.
9071         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9072         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9073         pipeline.
9074         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9075         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9076         * config/nds32/nds32.md (pipeline_model): Add n8.
9077         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9078         * config/nds32/pipelines.md: Include n8 settings.
9080 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
9081             Chung-Ju Wu  <jasonwucj@gmail.com>
9083         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9084         * config/nds32/nds32-n9-2r1w.md: New file.
9085         * config/nds32/nds32-n9-3r2w.md: New file.
9086         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9087         nds32_register_ports): New or modify for cpu n9.
9088         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9089         pipeline.
9090         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9091         * config/nds32/nds32-utils.c: New file.
9092         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9093         TARGET_MUL_SLOW): Define.
9094         * config/nds32/nds32.md (pipeline_model): New attribute.
9095         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9096         New options that support cpu n9.
9097         * config/nds32/pipelines.md: Include n9 settings.
9098         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9100 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
9102         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9103         information if necessary.
9104         (output_cond_branch_compare_zero): Likewise.
9105         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9106         (nds32_target_alignment): Refine for alignment.
9107         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9108         (FUNCTION_BOUNDARY): Modify.
9109         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9110         align case.
9111         * config/nds32/nds32.opt (malways-align, malign-functions): New.
9113 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
9115         * config/nds32/constants.md (unspec_volatile_element): Add values for
9116         TLB operation and data prefetch.
9117         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9118         functions for TLB operation and data prefetch.
9119         * config/nds32/nds32-intrinsic.md: Likewise.
9120         * config/nds32/nds32_intrinsic.h: Likewise.
9121         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9122         (nds32_print_operand): Likewise.
9123         * config/nds32/nds32.h (nds32_builtins): Likewise.
9125 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
9126         Andrew Pinski <pinsika@gcc.gnu.org>
9128         PR middle-end/82976
9129         * match.pd: Use constant_boolean_node of correct type instead of
9130         boolean_true_node or boolean_false_node for simplifying
9131         pointer comparisons to zero.
9133 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
9135         PR tree-optimization/80021
9136         * tree.c (verify_type_variant): Make error call in verify_variant_match
9137         translatable and remove final full stop.
9139 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
9141         * config/nds32/constants.md (unspec_volatile_element): Add
9142         UNSPEC_VOLATILE_EH_RETURN.
9143         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9144         nds32_output_stack_pop): Support dwarf exception handling process.
9145         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9146         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9147         exception handling process.
9148         (nds32_compute_stack_frame): Likewise.
9149         (nds32_return_addr_rtx): Likewise.
9150         (nds32_initial_elimination_offset): Likewise.
9151         (nds32_expand_prologue): Likewise.
9152         (nds32_expand_epilogue): Likewise.
9153         (nds32_dynamic_chain_address): New function.
9154         * config/nds32/nds32.h (machine_function): Add fields for dwarf
9155         exception handling.
9156         (DYNAMIC_CHAIN_ADDRESS): Define.
9157         (EH_RETURN_DATA_REGNO): Define.
9158         (EH_RETURN_STACKADJ_RTX): Define.
9159         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9160         patterns for dwarf exception handling.
9162 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
9164         * config/nds32/nds32.h: Clean up obsolete macros.
9166 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
9168         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9169         Add enum values for particular instructions.
9170         * config/nds32/nds32-intrinsic.c: Implementation of expanding
9171         particular intrinsic functions.
9172         * config/nds32/nds32-intrinsic.md: Likewise.
9173         * config/nds32/nds32_intrinsic.h: Likewise.
9174         * config/nds32/nds32.h (nds32_builtins): Likewise.
9175         * config/nds32/nds32.md (type): Add pbsad and pbsada.
9176         (btst, ave): New patterns for particular instructions.
9178 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
9180         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9181         Add enum values for atomic load/store and memory sync.
9182         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9183         and memory sync.
9184         * config/nds32/nds32-intrinsic.md: Likewise.
9185         * config/nds32/nds32_intrinsic.h: Likewise.
9186         * config/nds32/nds32.h (nds32_builtins): Likewise.
9188 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
9190         PR tree-optimization/85257
9191         * fold-const.c (native_encode_vector): If not all elts could fit
9192         and off is -1, return 0 rather than offset.
9193         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9194         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9195         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
9196         adjust buffer in native_interpret_expr call.
9198 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
9200         * config/nds32/constants.md (unspec_volatile_element): Add cache
9201         control enum values.
9202         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9203         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9204         * config/nds32/nds32.c (nds32_cctl_names): New.
9205         (nds32_print_operand): Handle cache control register names.
9206         * config/nds32/nds32.h (nds32_builtins): New enum values.
9207         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9208         macros.
9209         * config/nds32/nds32.md (type): Add mmu.
9210         * config/nds32/pipelines.md (simple_insn): Add mmu.
9212 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
9214         * config/nds32/nds32.md (type): Remove call.
9215         * config/nds32/pipelines.md (simple_insn): Likewise.
9217 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
9219         * config/nds32/constants.md (unspec_volatile_element): Add
9220         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9221         UNSPEC_VOLATILE_FMFCFG.
9222         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9223         description for fmfcfg and fmfcsr.
9224         (bdesc_1arg): Add fmtcsr.
9225         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9226         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9227         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9228         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9229         unspec_fmfcfg): New patterns.
9230         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9231         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9232         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9233         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9234         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9235         __nds32__fmfcfg): Define.
9237 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
9239         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9240         intrinsic register names.
9241         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9242         intrinsic register enum values and macros.
9244 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
9246         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9247         for load/store addressing form.
9248         (nds32_print_operand_address): Likewise.
9250 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
9252         PR target/85196
9253         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9254         based on LABEL_REF.  Remove useless assertion.
9255         (pic_address_needs_scratch): Fix formatting.
9256         (sparc_legitimize_pic_address): Minor tweaks.
9257         (sparc_delegitimize_address): Adjust assertion accordingly.
9258         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9259         into symbolic_operand.
9260         (movsi_high_pic_label_ref): Likewise.
9261         (movsi_lo_sum_pic_label_ref): Likewise.
9262         (movdi_pic_label_ref): Likewise.
9263         (movdi_high_pic_label_ref): Likewise.
9264         (movdi_lo_sum_pic_label_ref): Likewise.
9266 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
9268         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9269         custom LIB_SPEC setup.
9271 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
9272             Kito Cheng  <kito.cheng@gmail.com>
9274         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9275         * config/riscv/freebsd.h: New.
9277 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
9279         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9280         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9281         file.
9283 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
9284             Kito Cheng  <kito.cheng@gmail.com>
9286         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9287         nds32_output_call, nds32_symbol_binds_local_p): New functions.
9288         * config/nds32/nds32-protos.h (nds32_output_call,
9289         nds32_output_return): Declare.
9290         * config/nds32/nds32.md: Refine all the call and return patterns.
9292 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
9294         PR debug/85252
9295         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9296         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9298         PR rtl-optimization/84872
9299         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9300         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9301         EDGE_CROSSING edge.
9303 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
9305         * expr.c (copy_blkmode_to_reg): Revert 254862.
9306         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9308 2018-04-06  Richard Biener  <rguenther@suse.de>
9310         PR middle-end/85244
9311         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9312         after seeing a component reference with an adjacent field.  Treat
9313         refs to arrays at struct end of external decls similar to
9314         refs to unconstrained commons.
9316 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
9318         PR sanitizer/85213
9319         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9320         look through SAVE_EXPRs with non-side-effects argument.  Adjust
9321         recursive calls.
9322         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9323         save_p here.
9325 2018-04-06  Richard Biener  <rguenther@suse.de>
9327         PR middle-end/85180
9328         * alias.c (find_base_term): New wrapper around find_base_term
9329         unwinding CSELIB_VAL_PTR changes.
9330         (find_base_term): Do not restore CSELIB_VAL_PTR during the
9331         recursion.
9333 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9335         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9336         instructions.
9337         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9338         constant definitions.
9339         ("nop"): lr 0,0 -> nopr r0
9340         ("nop_lr0", "nop_lr1"): New insn definitions.
9342 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
9344         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9345         NDS32_V3PUSH_AVAILABLE_P macro.
9347 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
9348             Chung-Ju Wu  <jasonwucj@gmail.com>
9350         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9351         (nds32*-*-*): Add float and fpu_config into supported_defaults.
9352         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9353         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9354         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9355         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9356         * config/nds32/constraints.md: New constraints and checking for hard
9357         float configuration.
9358         * config/nds32/iterators.md: New mode iterator and attribute for hard
9359         float configuration.
9360         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9361         patterns.
9362         * config/nds32/nds32-fpu.md: New file.
9363         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9364         deal with hard float code generation.
9365         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9366         ARCH_V3S.
9367         (abi_type, float_reg_number): New enum type.
9368         * config/nds32/nds32-predicates.c: New predicates for hard float.
9369         * config/nds32/nds32-protos.h: Declare functions for hard float.
9370         * config/nds32/nds32.c: Implementation for hard float configuration.
9371         * config/nds32/nds32.h: Definitions for hard float configuration.
9372         * config/nds32/nds32.md: Include hard float machine description and
9373         modify patterns for hard float configuration.
9374         * config/nds32/nds32.opt: New options for hard float configuration.
9375         * config/nds32/predicates.md: New predicates for hard float
9376         configuration.
9378 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
9380         * common/config/nds32/nds32-common.c
9381         (nds32_option_optimization_table): Enable -mreleax-hint by default.
9383 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
9385         PR middle-end/85195
9386         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9387         CONSTRUCTOR_ELT (ctor, ...)->value.
9389 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
9391         PR target/85193
9392         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9394 2018-04-05  Tom de Vries  <tom@codesourcery.com>
9396         PR target/85204
9397         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9398         cond jump.
9400 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
9401             Kito Cheng  <kito.cheng@gmail.com>
9403         * config/nds32/constraints.md (U33): Fine-tune checking condition.
9404         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9405         * config/nds32/nds32.h (nds32_16bit_address_type): Add
9406         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9408 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
9409             Kito Cheng  <kito.cheng@gmail.com>
9411         * config/nds32/constraints.md (Ufe): New memory constraint.
9412         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9413         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9414         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9415         operands.
9416         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9417         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9419 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9421         * config/nds32/nds32.md: Use optimize_size in the condition for
9422         alu-shift instructions.
9424 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9426         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9428 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9430         * config/nds32/nds32.md (negsi2): Refine pattern.
9432 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
9433             Chung-Ju Wu  <jasonwucj@gmail.com>
9435         * config/nds32/iterators.md (shift_rotate): New code iterator.
9436         (shift): New code attribute.
9437         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9438         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9439         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9440         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9441         bit-wise operations.
9442         (andsi3, *andsi3): Ditto.
9443         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9444         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9445         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9446         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9447         nds32_ior_operand, nds32_xor_operand): New predicates.
9449 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9451         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9452         (addsi3, subsi3): ... this.
9454 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9456         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9458 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9460         * config/nds32/nds32.md: Adjust indention.
9462 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
9464         * config/nds32/nds32.md (feature): New attribute.
9466 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
9468         * config/nds32/nds32.md (subtype): New attribute.
9470 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
9472         PR target/85203
9473         * config/arm/arm-builtins.c (arm_expand_builtin): Change
9474         expansion to perform a bitwise AND of the argument followed by a
9475         boolean negation of the result.
9477 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
9479         PR rtl-optimization/84878
9480         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9481         the basic block.  Assert the use reference is not artificial and that
9482         it has an associated insn.
9484 2018-04-04  Michael Matz  <matz@suse.de>
9486         * builtins.c (compute_objsize): Pass correct operand
9487         to array_at_struct_end_p.
9489 2018-04-04  Richard Biener  <rguenther@suse.de>
9491         PR lto/85176
9492         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9493         from contexts for DINFO_LEVEL_TERSE and below.
9495 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
9497         * config/nds32/nds32-doubleword.md (move_<mode>): Require
9498         resiter_operand condition.
9499         * config/nds32/nds32.md (*move<mode>): Ditto.
9501 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
9502             Monk Chiang  <sh.chiang04@gmail.com>
9504         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9506 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
9508         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9510 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
9511             Kito Cheng  <kito.cheng@gmail.com>
9513         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9514         nds32_cond_code_str, output_cond_branch,
9515         output_cond_branch_compare_zero, nds32_expand_cbranch,
9516         nds32_expand_cstore, nds32_expand_movcc,
9517         nds32_output_cbranchsi4_equality_zero,
9518         nds32_output_cbranchsi4_equality_reg,
9519         nds32_output_cbranchsi4_equality_reg_or_const_int,
9520         nds32_output_cbranchsi4_greater_less_zero: New functions.
9521         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9522         nds32_expand_cstore, nds32_expand_movcc,
9523         nds32_output_cbranchsi4_equality_zero,
9524         nds32_output_cbranchsi4_equality_reg,
9525         nds32_output_cbranchsi4_equality_reg_or_const_int,
9526         nds32_output_cbranchsi4_greater_less_zero): Declare.
9527         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9528         nds32_rimm11s_operand): New predicates.
9529         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9530         * config/nds32/nds32.md: Rewrite all the branch and conditional move
9531         patterns.
9533 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
9535         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9536         * config/nds32/nds32.md: Ditto.
9537         * config/nds32/pipelines.md: Ditto.
9539 2018-04-04  Richard Biener  <rguenther@suse.de>
9541         PR tree-optimization/85168
9542         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9543         propagating abnormals.
9545 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
9547         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9549 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
9550             Kito Cheng  <kito.cheng@gmail.com>
9552         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9553         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9554         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9555         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9556         * config/nds32/nds32.md (sibcall_internal): New.
9557         (sibcall_register): Remove.
9558         (sibcall_immediate): Remove.
9559         (sibcall_value_internal): New.
9560         (sibcall_value_register): Remove.
9561         (sibcall_value_immediate): Remove.
9562         * config/nds32/predicates.md (nds32_general_register_operand): New.
9563         (nds32_call_address_operand): New.
9565 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
9567         PR rtl-optimization/85167
9568         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9569         bb_defs if *split_p, instead preinitialize it to NULL.
9571         PR tree-optimization/85156
9572         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9573         evaluating the argument multiple times.
9575 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
9577         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9578         than vector.
9579         (_mm_cvtpd_ps): Likewise.
9580         (_mm_cvttpd_epi32): Likewise.
9581         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9582         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9583         vector, pixel, and bool following altivec.h include.
9585 2018-04-03  Martin Sebor  <msebor@redhat.com>
9587         * doc/extend.texi (Common Function Attributes): Clarify.
9588         (const attribute): Likewise.
9589         (pure attribute): Likewise.
9591 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
9593         PR target/85169
9594         * config/i386/i386.c (ix86_expand_vector_set): Use
9595         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
9597 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
9599         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9600         instructions when changing rounding bits to preserve precision bits
9601         in the x87 control word.
9603 2018-04-03  Martin Liska  <mliska@suse.cz>
9605         PR tree-optimization/82491
9606         * rtl.h (strip_offset_and_add): Replace += suboffset with
9607         poly_uint64 () + suboffset.
9609 2018-03-29  Martin Liska  <mliska@suse.cz>
9610             Martin Jambor  <mjambor@suse.cz>
9612         PR ipa/84947
9613         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9614         param_type is not an integral or pointer type.
9616 2018-04-03  Richard Biener  <rguenther@suse.de>
9618         * sese.h (recompute_all_dominators): Remove.
9620 2018-04-02  Martin Sebor  <msebor@redhat.com>
9622         * doc/invoke.texi (-Wrestrict): Fix typos.
9624 2018-04-02  Jim Wilson  <jimw@sifive.com>
9626         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9627         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9628         (<optab>di3, <optab>si3_extend): Likewise.
9629         (<optab>si3_mask, <optab>si3_mask_1): New.
9630         (<optab>di3_mask, <optab>di3_mask_1): New.
9631         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9632         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9633         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9635 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
9637         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9638         example.
9640 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
9642         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9643         (nds32_canonicalize_comparison): New function.
9645 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
9646             Kito Cheng  <kito.cheng@gmail.com>
9647             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
9649         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9650         * config/nds32/constants.md (unspec_volatile_element): Add
9651         UNSPEC_VOLATILE_RELAX_GROUP.
9652         * config/nds32/nds32-relax-opt.c: New file.
9653         * config/nds32/nds32-predicates.c
9654         (nds32_symbol_load_store_p): New function.
9655         * config/nds32/nds32-protos.h
9656         (nds32_symbol_load_store_p): Declare function.
9657         (make_pass_nds32_relax_opt): Declare new rtl pass function.
9658         * config/nds32/nds32.c
9659         (nds32_register_pass): New function to register pass.
9660         (nds32_register_passes): New function to register passes.
9661         * config/nds32/nds32.md (relax_group): New pattern.
9662         * config/nds32/nds32.opt (mrelax-hint): New option.
9663         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9665 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
9667         * config/nds32/t-nds32: Modify files dependency.
9669 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
9671         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9672         (PROFILE_HOOK): Define its implementation.
9674 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
9676         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9677         type and 32-bit size.
9679 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
9681         PR middle-end/85090
9682         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9683         (V_128_256): New mode iterator.
9684         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9685         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9686         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9687         of V.
9688         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9689         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9691 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
9693         PR target/83315
9694         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9695         NaN inputs correctly.
9697 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
9699         PR target/80546
9700         * config/rs6000/vsx.md (??r): New mode attribute.
9701         (*vsx_mov<mode>_64bit): Use it.
9702         (*vsx_mov<mode>_32bit): Likewise.
9704 2018-03-30  Martin Sebor  <msebor@redhat.com>
9706         PR tree-optimization/84818
9707         * builtins.c (check_access): Use warning_n.
9709 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9711         PR target/83822
9712         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9713         condition.
9714         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9715         condition.
9717 2018-03-30  Julia Koval  <julia.koval@intel.com>
9719         PR target/84413
9720         * x86-tune.def (movx, partial_reg_dependency): Enable for
9721         m_SKYLAKE_AVX512.
9723 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
9725         PR inline-asm/84985
9726         * lra-constraints.c (process_alt_operands): Move setting
9727         this_alternative_matches below.
9729 2018-03-29  Martin Liska  <mliska@suse.cz>
9731         PR lto/84995.
9732         * doc/invoke.texi: Document how LTO works with debug info.
9733         Describe auto-load support of binutils.  Mention 'x86-64'
9734         as valid option value of -march option.
9736 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
9738         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9740         PR c/85094
9741         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9742         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9743         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9744         checking.
9746 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
9748         PR target/84912
9749         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9750         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9751         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9752         for RS6000_BTM_POWERPC64.
9753         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9754         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9755         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9756         definition.
9757         (DIVDE): Use it.
9758         (DIVDEU): Likewise.
9760 2018-03-28 Carl Love  <cel@us.ibm.com>
9762         Revert
9763         2017-09-27  Carl Love  <cel@us.ibm.com>
9765         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9766         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9767         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9768         fctiw instruction.
9770 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9772         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9773         instead of __vector bool.
9774         (_mm_max_pu8): Likewise.
9775         (_mm_min_pi16): Likewise.
9777 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
9779         PR target/84912
9780         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9781         (DIVWEUO): Likewise.
9782         (DIVDEO): Likewise.
9783         (DIVDEUO): Likewise.
9784         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9785         DIVWEUO and DIVDEUO.
9786         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9787         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9788         (div_extend): Likewise.
9789         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9790         builtin function.
9791         (__builtin_divweuo): Likewise.
9792         (__builtin_divdeo): Likewise.
9793         (__builtin_divdeuo): Likewise.
9795 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
9797         PR target/85095
9798         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9799         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9801         PR tree-optimization/82004
9802         * gimple-match-head.c (optimize_pow_to_exp): New function.
9803         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9804         Don't fold to exp if optimize_pow_to_exp is false.
9806 2018-03-28  Martin Liska  <mliska@suse.cz>
9808         PR other/84819
9809         * calls.c (initialize_argument_information): Fix trailing space.
9810         * common.opt: Fix typo and provide better explanation for
9811         -fsanitize-coverage option.
9812         * config/i386/i386.opt: Fix typo.
9814 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
9815             Martin Liska  <mliska@suse.cz>
9817         PR sanitizer/85081
9818         * gimplify.c (asan_poison_variable): Don't do the check for
9819         gimplify_omp_ctxp here.
9820         (gimplify_decl_expr): Do it here.
9821         (gimplify_target_expr): Likewise.
9823 2018-03-28  Martin Liska  <mliska@suse.cz>
9825         PR target/84988
9826         * config/i386/i386.c (ix86_function_arg_advance): Do not call
9827         chkp_type_bounds_count if MPX is not enabled.
9829 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
9831         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9833 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
9835         PR target/84914
9836         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9837         function to create the function decl for complex long double
9838         multiply and divide for -mabi=ieeelongdouble.
9839         (init_float128_ieee): Call it.
9841 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
9843         PR target/85044
9844         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9845         -fcf-protection=branch -mibt.
9846         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9848 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9850         PR target/81863
9851         * config/arm/arm.c (arm_valid_symbolic_address): Handle
9852         arm_word_relocations.
9854 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
9856         PR target/85056
9857         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9858         extern array declarations.
9860 2018-03-27  Richard Biener  <rguenther@suse.de>
9862         PR middle-end/84067
9863         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9864         explicit single_use checks.
9866 2018-03-27  Richard Biener  <rguenther@suse.de>
9868         PR tree-optimization/85082
9869         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9870         Valueize the VUSE.
9872 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9874         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9875         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9876         Turn on fasynchronous-unwind-tables and funwind-tables.
9878 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
9880         PR target/85073
9881         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9882         (*bmi_blsr_<mode>_ccz): Ditto.
9884 2018-03-26  Tom de Vries  <tom@codesourcery.com>
9886         PR tree-optimization/85063
9887         * omp-general.c (offloading_function_p): New function.  Factor out
9888         of ...
9889         * omp-offload.c (pass_omp_target_link::gate): ... here.
9890         * omp-general.h (offloading_function_p): Declare.
9891         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9892         with attribute omp declare target for offloading functions.
9894 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
9896         PR tree-optimization/84005
9897         * tree-data-ref.h (get_base_for_alignment): Declare.
9898         * tree-data-ref.c (get_base_for_alignment_1): New function.
9899         (get_base_for_alignment): Likewise.
9900         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9901         get_base_for_alignment to find a suitable base object, instead
9902         of always using drb->base_address.
9904 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
9906         PR inline-asm/85022
9907         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9908         known size by default.
9910 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
9912         PR inline-asm/85030
9913         * lra-constraints.c (process_alt_operands): Don't match BLKmode
9914         and non BLKmode operands.
9916 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9918         PR target/85026
9919         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9920         Clean up attributes.
9922 2018-03-23  Richard Biener  <rguenther@suse.de>
9924         PR debug/85020
9925         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9926         we are going to emit early debug for LTO.
9928 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
9930         PR inline-asm/85034
9931         * function.c (match_asm_constraints_1): Don't optimize if input
9932         doesn't satisfy general_operand predicate for output's mode.
9934         PR inline-asm/85022
9935         * alias.c (write_dependence_p): Don't require for x_canonicalized
9936         non-VOIDmode if x has VOIDmode.
9938         PR sanitizer/85029
9939         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9940         just don't try to optimize it rather than assert it never happens.
9942 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9944         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9945         macro expansions for definition of ST_INTERNAL_<mode> and
9946         LD_INTERNAL_<mode> builtins.
9947         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9948         Remove prototype.
9949         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9950         function.
9951         (altivec_expand_st_builtin): Likewise.
9952         (altivec_expand_builtin): Remove calls to deleted functions.
9953         (rs6000_address_for_altivec): Delete this function.
9954         * config/rs6000/vector.md: Remove expands for
9955         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9957 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
9959         PR target/84826
9960         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9961         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9962         re-computing once computed.
9963         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9964         (arm_init_machine_status): Initialize
9965         machine->static_chain_stack_bytes.
9967 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9969         PR target/84760
9970         * doc/extend.texi: Add four new prototypes for vec_ld.
9971         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9972         definitions for more logical presentation.
9973         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9974         entries for V1TI variants of __builtin_altivec_ld builtin.
9975         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9976         handling of V1TI variant of LVX icode pattern.
9977         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9978         (rs6000_gimple_fold_builtin): Likewise.
9979         (altivec_init_builtins): Add code to define
9980         __builtin_altivec_lvx_v1ti function.
9982 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
9984         PR inline-asm/84941
9985         * function.c (match_asm_constraints_1): Don't do the optimization
9986         if input isn't a REG, SUBREG, MEM or constant.
9988 2018-03-22  Tom de Vries  <tom@codesourcery.com>
9990         PR tree-optimization/84956
9991         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9992         bb_has_abnormal_pred.
9994 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
9996         PR sanitizer/85018
9997         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9998         DECL_INITIAL (decl) to decl at the end.
9999         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10000         adjust the comment.
10002 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
10004         * doc/extend.texi (__builtin_tgmath): Document when complex
10005         integer types are treated as _Complex _Float64.
10007 2018-03-21  Tom de Vries  <tom@codesourcery.com>
10009         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10011 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
10013         PR tree-optimization/84960
10014         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10015         if it is ENTRY block, move them into single succ of ENTRY in that case.
10017 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
10019         PR tree-optimization/84811
10020         * poly-int.h (poly_span_traits): Remove the T3 parameter and
10021         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10022         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10023         (known_subrange_p): Update accordingly.  Cast each value involved
10024         in the size comparison, rather than casting the result of the
10025         subtraction.
10027 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
10029         PR tree-optimization/84982
10030         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10031         by flipping the least significant bit rather than all bits from
10032         bitpos to bitpos + bitsize - 1.
10034 2018-03-21  Nathan Sidwell  <nathan@acm.org>
10036         * doc/extend.texi (Deprecated Features): Remove mention of
10037         long-deleted deprecations.
10039 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10041         PR jit/84288
10042         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10043         * configure: Regenerate.
10045 2018-03-21  Tom de Vries  <tom@codesourcery.com>
10047         PR tree-optimization/83126
10048         * tree-parloops.c (num_phis): New function.
10049         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10051 2018-03-21  Nathan Sidwell  <nathan@acm.org>
10053         * doc/extend.texi (Deprecated Features): Update deprecated flags,
10054         mention anon-struct/union members and trailing attributes.
10056 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
10058         PR tree-optimization/84969
10059         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10060         builtin memset partitions if they set different rhs values.
10062 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
10064         PR rtl-optimization/84989
10065         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10066         VEC_DUPLICATE with scalar result mode.
10068 2018-03-21  Martin Liska  <mliska@suse.cz>
10070         PR ipa/84963
10071         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10072         not intended return statement.
10074 2018-03-21  Martin Liska  <mliska@suse.cz>
10076         PR target/84988
10077         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10078         (chkp_find_bound_slots_1): Limit number of iterations.
10080 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
10082         PR target/84838
10083         * Minor grammar fixes for x86 options.
10085 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
10087         PR debug/84875
10088         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10089         holding REG_CFA_RESTORE notes, instead turn them into a USE.
10091 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
10093         PR target/83789
10094         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10095         (altivec_lvx_<mode>_1op): Likewise.
10096         (altivec_stvx_<mode>_2op): Likewise.
10097         (altivec_stvx_<mode>_1op): Likewise.
10098         (altivec_lvx_<VM2:mode>): New define_expand.
10099         (altivec_stvx_<VM2:mode>): Likewise.
10100         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10101         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10102         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10103         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10104         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10105         (rs6000_gen_lvx): Likewise.
10106         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10107         (altivec_expand_stv_builtin): Likewise.
10108         (altivec_expand_builtin): Likewise.
10109         * config/rs6000/vector.md: Likewise.
10111 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10113         PR target/82518
10114         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10115         BYTES_BIG_ENDIAN.
10117 2018-03-20  Richard Biener  <rguenther@suse.de>
10119         PR target/84986
10120         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10121         sign-conversions as zero, fall back to standard scalar_stmt
10122         cost for the rest.
10124 2018-03-20  Martin Liska  <mliska@suse.cz>
10126         PR ipa/84825
10127         * predict.c (rebuild_frequencies): Handle case when we have
10128         PROFILE_ABSENT, but flag_guess_branch_prob is false.
10130 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
10132         PR target/84990
10133         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10134         flag_section_anchors.
10135         * varasm.c (use_blocks_for_decl_p): Remove hack for
10136         dw2_force_const_mem.
10138         PR target/84845
10139         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10140         to ...
10141         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
10142         be created, use lowpart_subreg of operands[0] rather than operands[0]
10143         itself.
10144         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10145         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10146         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10147         and n constraint instead of aarch64_shift_imm_di and Usd.
10148         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10149         (*aarch64_<optab>_reg_minus<mode>3): ... this.
10151 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
10153         PR target/82989
10154         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10155         to favor GPR over NEON registers.
10156         (<shift>di3_neon): Likewise.
10158 2018-03-20  Tom de Vries  <tom@codesourcery.com>
10160         PR target/84952
10161         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10162         (nvptx_process_pars): Emit bar.sync asap and alap.
10164 2018-03-20  Tom de Vries  <tom@codesourcery.com>
10166         PR target/84954
10167         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10168         seen_label if seen_label is already set.
10170 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
10172         PR target/84945
10173         * config/i386/i386.c (fold_builtin_cpu): For features above 31
10174         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10175         Use 1U instead of 1.  Formatting fixes.
10177         PR c/84953
10178         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10179         instead of TREE_TYPE (s1) for the return value.
10181 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
10183         PR tree-optimization/84946
10184         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10185         bitsize + bitsize in poly_uint64 rather than poly_int64.
10187         PR sanitizer/78651
10188         * dwarf2asm.c: Include fold-const.c.
10189         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10190         of decl rather than decl itself.
10192         PR rtl-optimization/84643
10193         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10195 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
10197         PR sanitizer/78651
10198         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10199         calling assemble_variable.
10201 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
10203         PR target/81647
10204         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10205         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10207 2018-03-19  Jim Wilson  <jimw@sifive.com>
10209         PR bootstrap/84856
10210         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10211         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10212         (riscv_first_stack_step): Likewise.
10213         (riscv_option_override): Use STACK_BOUNDARY instead of
10214         MIN_STACK_BOUNDARY.
10215         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10216         MIN_STACK_BOUNDARY.
10217         (BIGGEST_ALIGNMENT): Set to 128.
10218         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10219         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10220         STACK_BOUNDARY.
10222 2018-03-19  Richard Biener  <rguenther@suse.de>
10224         PR tree-optimization/84933
10225         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10226         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10228 2018-03-19  Richard Biener  <rguenther@suse.de>
10230         PR tree-optimization/84859
10231         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10232         (cond_if_else_store_replacement): Perform sinking operation on
10233         single-store BBs regardless of MAX_STORES_TO_SINK setting.
10234         Generalize what a BB with a single eligible store is.
10236 2018-03-19  Richard Biener  <rguenther@suse.de>
10238         PR tree-optimization/84929
10239         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10240         chrec_is_positive against non-chrec arg.
10242 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
10244         PR target/84711
10245         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10247 2018-03-18  Martin Liska  <mliska@suse.cz>
10249         PR rtl-optimization/84635
10250         * regrename.c (build_def_use): Use matches_mode only when
10251         matches >= 0.
10253 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
10255         PR tree-optimization/84913
10256         * tree-vect-loop.c (vectorizable_reduction): Don't try to
10257         vectorize chains of COND_EXPRs.
10259 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
10261         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10263 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
10265         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10267 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
10269         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10271 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
10272             Kito Cheng  <kito.cheng@gmail.com>
10274         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10275         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10276         (nds32_adjust_reg_alloc_order): New function.
10277         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10279 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
10281         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10282         nds32_print_operand, nds32_print_operand_address): Use
10283         HOST_WIDE_INT_PRINT_DEC instead.
10285 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
10287         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10289 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
10291         PR target/84902
10292         * config/i386/i386.c (initial_ix86_tune_features,
10293         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10294         unsigned long long.
10295         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10296         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10297         rather than 1u << ix86_tune.  Formatting fix.
10298         (ix86_option_override_internal): Change ix86_arch_mask from
10299         unsigned int to unsigned HOST_WIDE_INT, initialize to
10300         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10301         (ix86_function_specific_restore): Likewise.
10303 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
10305         PR target/84899
10306         * postreload.c (reload_combine_recognize_pattern): Perform
10307         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10308         truncate_int_for_mode the result for the destination's mode.
10310         PR c/84909
10311         * hsa-gen.c (mem_type_for_type): Fix comment typo.
10312         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10313         Likewise.
10314         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10315         Likewise.
10317 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
10319         PR target/84876
10320         * lra-assigns.c (lra_split_hard_reg_for): Don't use
10321         regno_allocno_class_array and sorted_pseudos.
10322         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10323         insns where regno is used.
10325 2018-03-16  Martin Liska  <mliska@suse.cz>
10327         PR ipa/84833
10328         * multiple_target.c (create_dispatcher_calls): Redirect
10329         reference in the symbol table.
10331 2018-03-16  Martin Liska  <mliska@suse.cz>
10333         PR ipa/84722
10334         * multiple_target.c (create_dispatcher_calls): Redirect also
10335         an alias.
10337 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
10339         PR c++/79937
10340         PR c++/82410
10341         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10342         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10343         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10345 2018-03-16  Julia Koval  <julia.koval@intel.com>
10347         * doc/invoke.texi (Skylake Server): Add CLWB.
10348         Cannonlake): Remove CLWB.
10350 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
10352         PR tree-optimization/84841
10353         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10354         1 << 3.
10355         (FLOAT_ONE_CONST_TYPE): Define.
10356         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10357         (sort_by_operand_rank): Put entries with higher constant_type last
10358         rather than first to match comments.
10360 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
10362         * config/nios2/nios2.md (movsi_internal): Fix thinko in
10363         split predicate.
10365 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
10367         PR c++/79085
10368         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10369         check and use address of target always.
10371 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
10373         PR target/84574
10374         * config/i386/i386.c (indirect_thunk_needed): Update comments.
10375         (indirect_thunk_bnd_needed): Likewise.
10376         (indirect_thunks_used): Likewise.
10377         (indirect_thunks_bnd_used): Likewise.
10378         (indirect_return_needed): New.
10379         (indirect_return_bnd_needed): Likewise.
10380         (output_indirect_thunk_function): Add a bool argument for
10381         function return.
10382         (output_indirect_thunk_function): Don't generate alias for
10383         function return thunk.
10384         (ix86_code_end): Call output_indirect_thunk_function to generate
10385         function return thunks.
10386         (ix86_output_function_return): Set indirect_return_bnd_needed
10387         and indirect_return_needed instead of indirect_thunk_bnd_needed
10388         and indirect_thunk_needed.
10390 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
10392         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10393         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10394         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10396 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
10397             Paul Hua <paul.hua.gm@gmail.com>
10399         PR c/84852
10400         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10402 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
10404         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10405         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10406         resp. SFmode cases.
10408 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
10410         PR target/84711
10411         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10412         instead of GET_MODE_SIZE when comparing Units.
10414 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
10416         PR target/68256
10417         * varasm.c (hash_section): Return an unchangeble hash value
10418         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10419         Return !aarch64_can_use_per_function_literal_pools_p ().
10421 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
10423         PR target/84860
10424         * optabs.c (emit_conditional_move): Pass address of cmode's copy
10425         rather than address of cmode as last argument to prepare_cmp_insn.
10427 2018-03-15  Julia Koval  <julia.koval@intel.com>
10429         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10430         F_AVX512VNNI, F_AVX512BITALG): New.
10432 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
10434         PR target/83451
10435         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
10436         insn for floating-point loads and stores.
10438 2018-03-14  Carl Love  <cel@us.ibm.com>
10440         * config/rs6000/rs6000-c.c: Add macro definitions for
10441         ALTIVEC_BUILTIN_VEC_PERMXOR.
10442         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10443         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10444         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10445         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10446         UNSPEC_VPERMXOR.
10447         * config/doc/extend.texi: Add prototypes for vec_permxor.
10449 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
10451         PR c/84852
10452         * diagnostic-show-locus.c (class layout_point): Convert m_line
10453         from int to linenum_type.
10454         (line_span::comparator): Use linenum "compare" function when
10455         comparing line numbers.
10456         (test_line_span): New function.
10457         (layout_range::contains_point): Convert param "row" from int to
10458         linenum_type.
10459         (layout_range::intersects_line_p): Likewise.
10460         (layout::will_show_line_p): Likewise.
10461         (layout::print_source_line): Likewise.
10462         (layout::should_print_annotation_line_p): Likewise.
10463         (layout::print_annotation_line): Likewise.
10464         (layout::print_leading_fixits): Likewise.
10465         (layout::annotation_line_showed_range_p): Likewise.
10466         (struct line_corrections): Likewise for field m_row.
10467         (line_corrections::line_corrections): Likewise for param "row".
10468         (layout::print_trailing_fixits): Likewise.
10469         (layout::get_state_at_point): Likewise.
10470         (layout::get_x_bound_for_row): Likewise.
10471         (layout::print_line): Likewise.
10472         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10473         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10474         * input.c (selftest::test_linenum_comparisons): New function.
10475         (selftest::input_c_tests): Call it.
10476         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10477         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10478         * selftest.h (ASSERT_GT): New macro.
10479         (ASSERT_GT_AT): New macro.
10480         (ASSERT_LT): New macro.
10481         (ASSERT_LT_AT): New macro.
10483 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
10485         PR rtl-optimization/84780
10486         * combine.c (distribute_links): Don't make a link based on pc_rtx.
10488 2018-03-14  Martin Liska  <mliska@suse.cz>
10490         * tree.c (record_node_allocation_statistics): Use
10491         get_stats_node_kind.
10492         (get_stats_node_kind): New function extracted from
10493         record_node_allocation_statistics.
10494         (free_node): Use get_stats_node_kind.
10496 2018-03-14  Richard Biener  <rguenther@suse.de>
10498         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10499         that the value-set of ANTIC_IN doesn't grow.
10501         Revert
10502         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10503         member.
10504         (BB_VISITED_WITH_VISITED_SUCCS): New define.
10505         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10507 2018-03-14  Julia Koval  <julia.koval@intel.com>
10509         * config.gcc (icelake-client, icelake-server): New.
10510         (icelake): Remove.
10511         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10512         (initial_ix86_arch_features): Ditto.
10513         (PTA_SKYLAKE): Add SGX.
10514         (PTA_ICELAKE): Remove.
10515         (PTA_ICELAKE_CLIENT): New.
10516         (PTA_ICELAKE_SERVER): New.
10517         (ix86_option_override_internal): Split up icelake on icelake client and
10518         icelake server.
10519         (get_builtin_code_for_version): Ditto.
10520         (fold_builtin_cpu): Ditto.
10521         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10522         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10523         * config/i386/i386.h (processor_type): Ditto.
10524         * doc/invoke.texi: Ditto.
10526 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
10528         PR sanitizer/83392
10529         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10530         INTEGER_CST offset, add it together with bitpos / 8 and
10531         sign extend based on POINTER_SIZE.
10533         PR target/84844
10534         Revert
10535         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
10537         PR target/78090
10538         * config/i386/constraints.md (Yc): New register constraint.
10539         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10540         Use Yc constraint for alternative 2 of operand 0.  Remove
10541         preferred_for_speed attribute.
10543 2018-03-14  Richard Biener  <rguenther@suse.de>
10545         PR tree-optimization/84830
10546         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10547         with the old one to avoid oscillations.
10549 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
10551         PR target/83712
10552         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10553         pseudos.
10554         (assign_by_spills): Return a flag of reload assignment failure.
10555         Do not process the reload assignment failures.  Do not spill other
10556         reload pseudos if they has the same reg class.  Update n if
10557         necessary.
10558         (lra_assign): Add a return arg.  Set up from the result of
10559         assign_by_spills call.
10560         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10561         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
10562         usage_insns if it is not NULL.
10563         (spill_hard_reg_in_range): New function.
10564         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10565         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10566         function prototypes.
10567         (lra_assign): Change prototype.
10568         * lra.c (lra): Add code to deal with fails by splitting hard reg
10569         live ranges.
10571 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
10573         * config/riscv/riscv.opt (mrelax): New option.
10574         * config/riscv/riscv.c (riscv_file_start): Emit ".option
10575         "norelax" when riscv_mrelax is disabled.
10576         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10578 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
10580         PR target/84743
10581         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10582         reassociation for int modes.
10584 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
10586         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10587         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10588         for big-endian.
10589         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10590         * config/aarch64/aarch64-sve.md
10591         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10592         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10593         (*extend<mode><Vwide>2): Rename to...
10594         (aarch64_sve_extend<mode><Vwide>2): ...this.
10595         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10596         renaming the old pattern to...
10597         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
10598         unsigned packs.
10599         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10600         define_expand, renaming the old pattern to...
10601         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10602         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10603         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10604         account when deciding which SVE instruction the optab should use.
10605         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10607 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
10609         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10610         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10611         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10612         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10613         (tlsdesc_small_<mode>): Turn a define_expand and use
10614         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
10615         (tlsdesc_small_advsimd_<mode>): ...this.
10616         (tlsdesc_small_sve_<mode>): New pattern.
10618 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
10620         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10621         (UNSPEC_UMUL_HIGHPART): New constants.
10622         (MUL_HIGHPART): New int iteraor.
10623         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10624         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10625         define_expand.
10626         (*<su>mul<mode>3_highpart): New define_insn.
10628 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
10630         PR lto/84805
10631         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10632         incomplete types.
10634 2018-03-13  Martin Liska  <mliska@suse.cz>
10636         PR ipa/84658.
10637         * (sem_item_optimizer::sem_item_optimizer): Initialize new
10638         vector.
10639         (sem_item_optimizer::~sem_item_optimizer): Release it.
10640         (sem_item_optimizer::merge_classes): Register variable aliases.
10641         (sem_item_optimizer::fixup_pt_set): New function.
10642         (sem_item_optimizer::fixup_points_to_sets): Likewise.
10643         * ipa-icf.h: Declare new variables and functions.
10645 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
10647         PR middle-end/84834
10648         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10649         integer_pow2p@2 and test integer_pow2p in condition.
10650         (A < 0 ? C : 0): Similarly for @1.
10652         PR middle-end/84831
10653         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10654         characters starting at p contain '\0' character, don't look beyond
10655         that.
10657         PR target/84827
10658         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10659         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10661         PR target/84828
10662         * reg-stack.c (change_stack): Change update_end var from int to
10663         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10664         also call set_block_for_insn on the newly added insns and rescan.
10666         PR target/84786
10667         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10668         on the last operand.
10670         PR c++/84704
10671         * tree.c (stabilize_reference_1): Return save_expr (e) for
10672         STATEMENT_LIST even if it doesn't have side-effects.
10674 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
10676         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10678 2018-03-12  Renlin Li  <renlin.li@arm.com>
10680         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10681         aarch64_output_scalar_simd_mov_immediate.
10683 2018-03-12  Martin Sebor  <msebor@redhat.com>
10685         PR tree-optimization/83456
10686         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10687         for perfectly overlapping calls to memcpy.
10688         (gimple_fold_builtin_memory_chk): Same.
10689         (gimple_fold_builtin_strcpy): Handle no-warning.
10690         (gimple_fold_builtin_stxcpy_chk): Same.
10691         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10693 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
10695         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10696         parameter.  Use it for SFmode.
10697         (rs6000_function_arg_advance_1): Adjust.
10698         (rs6000_function_arg): Adjust.
10699         (rs6000_gimplify_va_arg): Pass false for that new parameter.
10701 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
10703         PR rtl-optimization/84169
10704         PR rtl-optimization/84780
10705         * combine.c (can_combine_p): Check for a 2-insn combination whether
10706         the destination register is used between the two insns, too.
10708 2018-03-12  Richard Biener  <rguenther@suse.de>
10710         PR tree-optimization/84803
10711         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10712         for refs DR analysis didn't process.
10714 2018-03-12  Richard Biener  <rguenther@suse.de>
10716         PR tree-optimization/84777
10717         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10718         force-vectorize loops ignore whether we are optimizing for size.
10720 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10722         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10723         (TARGET_MD_ASM_ADJUST): Define.
10725 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
10726             Kito Cheng  <kito.cheng@gmail.com>
10727             Chung-Ju Wu  <jasonwucj@gmail.com>
10729         * config/nds32/nds32.c (nds32_compute_stack_frame,
10730         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10731         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10732         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10733         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10734         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10735         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10736         * config/nds32/nds32.md (prologue, epilogue): Use macro
10737         NDS32_V3PUSH_AVAILABLE_P to do checking.
10739 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
10741         PR debug/58150
10742         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10743         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10744         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
10745         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
10746         addition of most attributes on !orig_type_die or the attribute not
10747         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10749 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
10750             Chung-Ju Wu  <jasonwucj@gmail.com>
10752         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10753         __NDS32_VH__ macro.
10754         * config/nds32/nds32.opt (mvh): New option.
10756 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
10757             Chung-Ju Wu  <jasonwucj@gmail.com>
10759         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10760         function.
10761         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10762         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10763         definition.
10765 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
10766             Chung-Ju Wu  <jasonwucj@gmail.com>
10768         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10769         function.
10770         * config/nds32/nds32-multiple.md (strlensi): New pattern.
10771         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10773 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
10774             Kito Cheng  <kito.cheng@gmail.com>
10775             Chung-Ju Wu  <jasonwucj@gmail.com>
10777         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10778         UNSPEC_FFMISM and UNSPEC_FLMISM.
10779         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10780         for ffb, ffmism and flmism.
10781         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10782         (unspec_ffmism): Ditto.
10783         (unspec_flmism): Ditto.
10784         (nds32_expand_builtin_impl): Check if string extension is available.
10785         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10786         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10788 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
10790         Reverting patch:
10791         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
10793         PR target/83712
10794         * lra-assigns.c (assign_by_spills): Return a flag of reload
10795         assignment failure.  Do not process the reload assignment
10796         failures.  Do not spill other reload pseudos if they has the same
10797         reg class.
10798         (lra_assign): Add a return arg.  Set up from the result of
10799         assign_by_spills call.
10800         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10801         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
10802         usage_insns if it is not NULL.
10803         (spill_hard_reg_in_range): New function.
10804         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10805         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10806         function prototypes.
10807         (lra_assign): Change prototype.
10808         * lra.c (lra): Add code to deal with fails by splitting hard reg
10809         live ranges.
10811 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
10813         PR target/84807
10814         * config/i386/i386.opt: Replace Enforcment with Enforcement.
10816 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
10818         PR debug/84620
10819         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10820         (dw_val_node): Add val_symbolic_view.
10821         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10822         (symview_upper_bound): New.
10823         (new_line_info_table): Initialize symviews_since_reset.
10824         (dwarf2out_source_line): Count symviews_since_reset and set
10825         symview_upper_bound.
10826         (dw_val_equal_p): Handle symview.
10827         (add_AT_symview): New.
10828         (print_dw_val): Handle symview.
10829         (attr_checksum, attr_checksum_ordered): Likewise.
10830         (same_dw_val_p, size_of_die): Likewise.
10831         (value_format, output_die): Likewise.
10832         (add_high_low_attributes): Use add_AT_symview for entry_view.
10833         (dwarf2out_finish): Reset symview_upper_bound, clear
10834         zero_view_p.
10836 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
10838         PR target/83969
10839         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10840         Add strict argument and use it.
10841         (rs6000_split_multireg_move): Update for new strict argument.
10842         (mem_operand_gpr): Disallow all non-offsettable addresses.
10843         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10845 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
10847         PR target/84772
10848         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10849         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10850         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10852         PR c++/84767
10853         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10854         decl, use remap_type if we want to use the type.
10856 2018-03-09  Martin Sebor  <msebor@redhat.com>
10858         PR tree-optimization/84526
10859         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10860         Remove dead code.
10861         (builtin_access::generic_overlap): Be prepared to handle non-array
10862         base objects.
10864 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
10866         PR rtl-optimization/84682
10867         * lra-constraints.c (process_address_1): Check is_address flag
10868         for address constraints.
10869         (process_alt_operands): Likewise.
10870         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10871         preprocess_constraints.
10872         * recog.h (preprocess_constraints): Add oploc parameter.
10873         Adjust callers.
10874         * recog.c (preprocess_constraints): Test address_operand for
10875         CT_ADDRESS constraints.
10877 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
10879         PR target/83712
10880         * lra-assigns.c (assign_by_spills): Return a flag of reload
10881         assignment failure.  Do not process the reload assignment
10882         failures.  Do not spill other reload pseudos if they has the same
10883         reg class.
10884         (lra_assign): Add a return arg.  Set up from the result of
10885         assign_by_spills call.
10886         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10887         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
10888         usage_insns if it is not NULL.
10889         (spill_hard_reg_in_range): New function.
10890         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10891         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10892         function prototypes.
10893         (lra_assign): Change prototype.
10894         * lra.c (lra): Add code to deal with fails by splitting hard reg
10895         live ranges.
10897 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10899         PR target/83193
10900         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10901         Accept complain bool parameter.  Only emit errors if it is true.
10902         (arm_parse_cpu_option_name): Likewise.
10903         (arm_target_thumb_only): Adjust callers of the above.
10904         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10905         prototype to take a default true bool parameter.
10906         (arm_parse_arch_option_name): Likewise.
10908 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
10909             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
10911         PR jit/64089
10912         PR jit/84288
10913         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10914         * configure: Regenerate.
10915         * configure.ac ("linker --version-script option"): New.
10916         ("linker soname option"): New.
10918 2018-03-09  Richard Biener  <rguenther@suse.de>
10920         PR tree-optimization/84775
10921         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10922         immediate uses of predicate stmts and mark them modified.
10924         Revert
10925         PR tree-optimization/84178
10926         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10927         to caller.
10928         (version_loop_for_if_conversion): Delay update_ssa call.
10929         (tree_if_conversion): Delay update_ssa until after predicate
10930         insertion.
10932 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
10934         PR target/84763
10935         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10936         when the function accesses prior frames.
10938 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
10940         PR debug/84456
10941         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10942         gen_llsym, otherwise call maybe_gen_llsym.
10944         PR inline-asm/84742
10945         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10946         has ',' character inside of it.
10948 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10950         PR target/84748
10951         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10952         as clobbering CC_REGNUM.
10954 2018-03-08  Richard Biener  <rguenther@suse.de>
10956         PR middle-end/84552
10957         * tree-scalar-evolution.c: Include tree-into-ssa.h.
10958         (follow_copies_to_constant): Do not follow SSA names registered
10959         for update.
10961 2018-03-08  Richard Biener  <rguenther@suse.de>
10963         PR tree-optimization/84178
10964         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10965         to caller.
10966         (version_loop_for_if_conversion): Delay update_ssa call.
10967         (tree_if_conversion): Delay update_ssa until after predicate
10968         insertion.
10970 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
10972         PR tree-optimization/84178
10973         * tree-if-conv.c (release_bb_predicate): Remove the
10974         the assertion that the stmts have NULL use_ops.
10975         Discard the statements, asserting that they haven't
10976         yet been added to a BB.
10978 2018-03-08  Richard Biener  <rguenther@suse.de>
10980         PR tree-optimization/84746
10981         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10982         (phi_translate): Pass in destination ANTIC_OUT set.
10983         (phi_translate_1): Likewise.  For a simplified result lookup
10984         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10985         (phi_translate_set): Adjust.
10986         (do_pre_regular_insertion): Likewise.
10987         (do_pre_partial_partial_insertion): Likewise.
10989 2018-03-08  Martin Liska  <mliska@suse.cz>
10991         PR gcov-profile/84735
10992         * doc/gcov.texi: Document usage of profile files.
10993         * gcov-io.h: Document changes in the format.
10995 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
10997         PR debug/84404
10998         PR debug/84408
10999         * dwarf2out.c (struct dw_line_info_table): Update comments for
11000         view == -1.
11001         (FORCE_RESET_NEXT_VIEW): New.
11002         (FORCE_RESETTING_VIEW_P): New.
11003         (RESETTING_VIEW_P): Check for -1 too.
11004         (ZERO_VIEW_P): Likewise.
11005         (new_line_info_table): Force-reset next view.
11006         (dwarf2out_begin_function): Likewise.
11007         (dwarf2out_source_line): Simplify zero_view_p initialization.
11008         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11009         view directly.  Omit view when omitting .loc at line 0.
11011 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
11013         PR tree-optimization/84740
11014         * tree-switch-conversion.c (process_switch): Call build_constructors
11015         only if info.phi_count is non-zero.
11017         PR tree-optimization/84739
11018         * tree-tailcall.c (find_tail_calls): Check call arguments against
11019         DECL_ARGUMENTS (current_function_decl) rather than
11020         DECL_ARGUMENTS (func) when checking for tail recursion.
11022 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
11024         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11025         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11026         Volker Reichelt's entry and add entries for people that perform
11027         GCC fuzzy testing and report numerous bugs.
11029 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
11031         PR target/82411
11032         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11033         readonly data in sdata, if that is disabled.
11034         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11035         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11036         -mreadonly-in-sdata option.
11038 2018-03-07  Martin Sebor  <msebor@redhat.com>
11040         PR tree-optimization/84468
11041         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11042         basic block when looking for nul assignment.
11044 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
11046         PR target/84277
11047         * except.h (output_function_exception_table): Adjust prototype.
11048         * except.c (output_function_exception_table): Remove FNNAME parameter
11049         and add SECTION parameter.  Ouput one part of the table at a time.
11050         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11051         the first part of the exception table and emit unwind directives.
11052         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11053         (i386_pe_seh_cold_init): Likewise.
11054         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11055         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11056         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11057         (ix86_output_call_insn): Emit a nop in one more case for SEH.
11058         * config/i386/winnt.c: Include except.h.
11059         (struct seh_frame_state): Add reg_offset, after_prologue and
11060         in_cold_section fields.
11061         (i386_pe_seh_end_prologue): Set seh->after_prologue.
11062         (i386_pe_seh_cold_init): New function.
11063         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11064         to seh->in_cold_section.
11065         (seh_emit_push): Record the offset of the push.
11066         (seh_emit_save): Record the offet of the save.
11067         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11068         Test seh->after_prologue to disregard the epilogue.
11069         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11070         (i386_pe_end_cold_function): New function.
11072 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
11074         PR fortran/84565
11075         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11076         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11078         PR c++/84704
11079         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11080         on tmp_var.
11081         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11082         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11084         PR middle-end/84723
11085         * multiple_target.c: Include tree-inline.h and intl.h.
11086         (expand_target_clones): Diagnose and fail if node->definition and
11087         !tree_versionable_function_p (node->decl).
11089 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
11091         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11092         sprint_ul.
11093         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11094         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11095         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11097 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
11099         PR target/84710
11100         * combine.c (try_combine): Use reg_or_subregno instead of handling
11101         just paradoxical SUBREGs and REGs.
11103 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
11105         * config/arc/arc.c (arc_finalize_pic): Remove function.
11106         (arc_must_save_register): We use single base PIC register, remove
11107         checks to save/restore the PIC register.
11108         (arc_expand_prologue): Likewise.
11109         * config/arc/arc-protos.h (arc_set_default_type_attributes):
11110         Remove.
11111         (arc_verify_short): Likewise.
11112         (arc_attr_type): Likewise.
11113         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11114         (walk_stores): Likewise.
11115         (arc_address_cost): Make it static.
11116         (arc_verify_short): Likewise.
11117         (branch_dest): Likewise.
11118         (arc_attr_type): Likewise.
11119         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11120         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11121         (arc_final_prescan_insn): Remove inserting the nops due to
11122         hardware hazards.  It is done in reorg step.
11123         (insn_length_variant_t): Remove.
11124         (insn_length_parameters_t): Likewise.
11125         (arc_insn_length_parameters): Likewise.
11126         (arc_get_insn_variants): Likewise.
11127         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11129 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
11131         PR inline-asm/84683
11132         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11133         assertion failure.
11135         PR tree-optimization/84687
11136         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11137         on new_node->decl.
11138         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11140 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11142         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11143         Rename to ppc_speculation_barrier.
11144         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11145         __builtin_ppc_speculation_barrier.
11147 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
11149         PR target/84700
11150         * combine.c (combine_simplify_rtx): Don't try to simplify if
11151         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11152         are equal to x.
11154 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
11156         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11157         to 32 bytes when compiling for POWER9.
11159 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
11161         PR target/84564
11162         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11163         regparm >= 3 with no arg reg available also for calls with
11164         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
11166         PR target/84524
11167         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11168         orig,vex.
11169         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
11171 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
11173         PR target/84264
11174         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11176 2018-03-05  Richard Biener  <rguenther@suse.de>
11178         PR tree-optimization/84486
11179         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11180         When inserting a __builtin_assume_aligned call set the LHS
11181         SSA name alignment info accordingly.
11183 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
11185         PR tree-optimization/84114
11186         * config/aarch64/aarch64.c (aarch64_reassociation_width)
11187         Avoid reassociation of FLOAT_MODE addition.
11189 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
11191         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11192         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11193         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11194         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11195         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11196         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11197         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11198         and -mwbnoinvd.
11199         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11200         __builtin_ia32_wbinvd): New builtins.
11201         (SPECIAL_ARGS2): New.
11202         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11203         (SPECIAL_ARGS2): New.
11204         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11205         (ix86_valid_target_attribute_inner_p): Ditto.
11206         (ix86_init_mmx_sse_builtins): Add special_args2.
11207         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11208         TARGET_WBNOINVD_P): New.
11209         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11210         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11211         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11212         * config/i386/immintrin.h (_wbinvd): New intrinsic.
11213         * config/i386/pconfigintrin.h: New file.
11214         * config/i386/wbnoinvdintrin.h: Ditto.
11215         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11216         wbnoinvdintrin.h.
11217         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11219 2018-03-05  Richard Biener  <rguenther@suse.de>
11221         PR tree-optimization/84670
11222         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11223         member.
11224         (BB_VISITED_WITH_VISITED_SUCCS): New define.
11225         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11226         (compute_antic_aux): Only assert the number of values in ANTIC_IN
11227         doesn't grow if all successors (recursively) were visited at least
11228         once.
11230 2018-03-05  Richard Biener  <rguenther@suse.de>
11232         PR tree-optimization/84650
11233         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11234         if executed in the loop pipeline.
11236 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
11238         * doc/configfiles.texi (Configuration Files): Move info about
11239         conditionalizing $target-protos.h to...
11240         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
11241         differs from $target-protos.h.
11243 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
11244             Chung-Ju Wu  <jasonwucj@gmail.com>
11246         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11247         * config/nds32/nds32-multiple.md (setmemsi): Define.
11248         * config/nds32/nds32-memory-manipulation.c
11249         (nds32_gen_dup_4_byte_to_word_value): New.
11250         (emit_setmem_word_loop): New.
11251         (emit_setmem_byte_loop): New.
11252         (nds32_expand_setmem_loop): New.
11253         (nds32_expand_setmem_loop_v3m): New.
11254         (nds32_expand_setmem_unroll): New.
11255         (nds32_expand_setmem): New.
11257 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
11258             Chung-Ju Wu  <jasonwucj@gmail.com>
11260         * config/nds32/nds32-memory-manipulation.c
11261         (nds32_emit_load_store): New.
11262         (nds32_emit_post_inc_load_store): New.
11263         (nds32_emit_mem_move): New.
11264         (nds32_emit_mem_move_block): New.
11265         (nds32_expand_movmemsi_loop_unknown_size): New.
11266         (nds32_expand_movmemsi_loop_known_size): New.
11267         (nds32_expand_movmemsi_loop): New.
11268         (nds32_expand_movmemsi_unroll): New.
11269         (nds32_expand_movmemqi): Rename ...
11270         (nds32_expand_movmemsi): ... to this.
11271         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11272         (movmemsi): ... to this.
11273         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11274         (nds32_expand_movmemsi): ... to this.
11276 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
11277             Monk Chiang  <sh.chiang04@gmail.com>
11278             Chung-Ju Wu  <jasonwucj@gmail.com>
11280         * config/nds32/nds32-protos.h
11281         (nds32_expand_load_multiple): New arguments.
11282         (nds32_expand_store_multiple): Ditto.
11283         (nds32_valid_multiple_load_store): Rename ...
11284         (nds32_valid_multiple_load_store_p): ... to this.
11285         * config/nds32/nds32-memory-manipulation.c
11286         (nds32_expand_load_multiple): Refine implementation.
11287         (nds32_expand_store_multiple): Ditto.
11288         * config/nds32/nds32-multiple.md
11289         (load_multiple): Update nds32_expand_load_multiple interface.
11290         (store_multiple): Update nds32_expand_store_multiple interface.
11291         * config/nds32/nds32-predicates.c
11292         (nds32_valid_multiple_load_store): Rename ...
11293         (nds32_valid_multiple_load_store_p): ... to this and refine
11294         implementation.
11295         * config/nds32/predicates.md
11296         (nds32_load_multiple_and_update_address_operation): New predicate.
11297         (nds32_store_multiple_and_update_address_operation): New predicate.
11299 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
11300             Chung-Ju Wu  <jasonwucj@gmail.com>
11302         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11303         (combo): New attribute.
11304         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11306 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
11308         * config/nds32/nds32.opt: Change -mcmodel= default value.
11310 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
11311             Monk Chiang  <sh.chiang04@gmail.com>
11312             Chung-Ju Wu  <jasonwucj@gmail.com>
11314         * config/nds32/constants.md (unspec_element): New enum.
11315         * config/nds32/constraints.md (Umw): New constraint.
11316         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11317         * config/nds32/nds32-intrinsic.md: Likewise.
11318         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11319         (nds32_valid_smw_lwm_base_p): New.
11320         (nds32_output_smw_single_word): New.
11321         (nds32_output_lmw_single_word): New.
11322         (nds32_expand_unaligned_load): New.
11323         (nds32_expand_unaligned_store): New.
11324         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11325         (nds32_output_smw_single_word): Declare.
11326         (nds32_output_lmw_single_word): Declare.
11327         (nds32_expand_unaligned_load): Declare.
11328         (nds32_expand_unaligned_store): Declare.
11329         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11330         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11331         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11332         NDS32_BUILTIN_UASTORE_DW.
11333         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11334         predicate.
11336 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
11337             Kito Cheng  <kito.cheng@gmail.com>
11338             Chung-Ju Wu  <jasonwucj@gmail.com>
11340         * config/nds32/nds32-intrinsic.c
11341         (nds32_expand_builtin_null_ftype_reg): Delete.
11342         (nds32_expand_builtin_reg_ftype_imm): Ditto.
11343         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11344         (nds32_read_argument): New.
11345         (nds32_legitimize_target): Ditto.
11346         (nds32_legitimize_argument): Ditto.
11347         (nds32_check_constant_argument): Ditto.
11348         (nds32_expand_unop_builtin): Ditto.
11349         (nds32_expand_unopimm_builtin): Ditto.
11350         (nds32_expand_binop_builtin): Ditto.
11351         (nds32_builtin_decl_impl): Ditto.
11352         (builtin_description): Ditto.
11353         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11354         (nds32_init_builtins_impl): Ditto.
11355         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11356         (nds32_builtin_decl): New.
11357         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11358         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11360 2018-03-02  Jeff Law  <law@redhat.com>
11362         * reorg.c (stop_search_p): Handle DEBUG_INSN.
11363         (redundant_insn, fill_simple_delay_slots): Likewise.
11364         (fill_slots_from_thread): Likewise.
11365         * resource.c (mark_referenced_resources): Likewise.
11366         (mark_set_resources, find_dead_or_set_registers): Likewise.
11368 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
11370         * substring-locations.h (format_warning_va): Formatting fix for
11371         ATTRIBUTE_GCC_DIAG.
11372         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11373         argument.
11374         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11375         * substring-locations.c: Include intl.h.
11376         (format_warning_va): Turned into small wrapper around
11377         format_warning_n_va, renamed to ...
11378         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11379         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11380         use ngettext.
11381         (format_warning_at_substring_n): New function.
11382         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11383         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
11384         format_warning_at_substring with just a shorter name instead of
11385         const function pointer.
11386         (fmtwarn_n): New function.
11387         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11388         appropriate, get rid of all the fmtstr temporaries, move conditionals
11389         with G_() wrapped string literals directly into fmtwarn arguments,
11390         cast dir.len to (int), formatting fixes.
11392 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
11394         * doc/invoke.texi: Remove "Cilk Plus" references.
11396 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
11397             Richard Biener  <rguenther@suse.de>
11399         PR ipa/84628
11400         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11401         for error or warning attributes if CALL_FROM_THUNK_P is set.
11402         Formatting fixes.
11404 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
11406         PR target/56540
11407         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11408         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11410         PR target/56540
11411         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11412         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11414         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11415         instead of -1U in last predictors element's probability member.
11417 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
11419         PR ipa/83983
11420         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11421         arguments if they are comparable.
11423 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
11425         PR tree-optimization/84634
11426         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11427         masks and masked_loop_p with a single loop_masks, making sure it's
11428         null for bb vectorization.
11430 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
11432         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11433         (vect_analyze_data_ref_access): Use loop->safe_len rather than
11434         loop->force_vectorize to check whether there is no alias.
11436 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
11438         PR target/84614
11439         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11440         prototypes.
11441         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11442         comments.
11443         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11444         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11445         instead of a loop around prev_real_insn.
11446         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11447         prev_real_insn.
11449         PR inline-asm/84625
11450         * config/i386/i386.c (ix86_print_operand): Use conditional
11451         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11452         zero vector.
11454 2018-03-02  Richard Biener  <rguenther@suse.de>
11456         PR tree-optimization/84427
11457         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11458         (bitmap_set_subtract_values): Rewrite to handle multiple
11459         exprs per value.
11460         (clean): Likewise.
11461         (prune_clobbered_mems): Likewise.
11462         (phi_translate): Take edge instead of pred/phiblock.
11463         (phi_translate_1): Likewise.
11464         (phi_translate_set): Likewise.  Insert all translated
11465         exprs for a value into the set, keeping possibly multiple
11466         expressions per value.
11467         (compute_antic_aux): Adjust for phi_translate changes.
11468         When intersecting union the expressions and prune those
11469         not in the final value set, keeping possibly multiple
11470         expressions per value.  Do not use value-insertion
11471         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11472         all expressions.  Add verification that the value-sets
11473         only shrink during iteration.
11474         (compute_partial_antic_aux): Adjust for the phi_translate changes.
11475         (do_pre_regular_insertion): Likewise.
11476         (do_pre_partial_partial_insertion): Likewise.
11478 2018-03-02  Richard Biener  <rguenther@suse.de>
11480         PR target/82005
11481         * config/darwin.c (saved_debug_info_level): New static global.
11482         (darwin_asm_lto_start): Disable debug info generation for LTO out.
11483         (darwin_asm_lto_end): Restore debug info generation settings.
11485 2018-03-01  Martin Liska  <mliska@suse.cz>
11487         PR sanitizer/82484
11488         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11489         volatile arguments.
11491 2018-03-01  Richard Biener  <rguenther@suse.de>
11493         PR debug/84645
11494         * dwarf2out.c (gen_variable_die): Properly handle late VLA
11495         type annotation with LTO when debug was disabled at compile-time.
11497 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
11499         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11500         XINT with INTVAL.
11501         (mips_final_postscan_insn): Likewise.
11503 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
11505         PR rtl-optimization/84528
11506         * alias.c (init_alias_target): Add commentary.
11507         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11508         a unique base value if the frame pointer is not eliminated
11509         to the stack pointer.
11511 2018-03-01  Tom de Vries  <tom@codesourcery.com>
11513         PR rtl-optimization/83327
11514         * lra-int.h (hard_regs_spilled_into): Declare.
11515         * lra.c (hard_regs_spilled_into): Define.
11516         (init_reg_info): Init hard_regs_spilled_into.
11517         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11518         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11519         (process_bb_lives): Handle hard_regs_spilled_into.
11520         (lra_create_live_ranges_1): Before doing liveness propagation, clear
11521         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11523 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
11525         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11526         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11527         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11528         * config/rs6000/aix72.h: New file.
11530 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
11532         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11533         instead of warning_at with conditional singular and plural messages
11534         where possible.
11536         PR target/52991
11537         * stor-layout.c (update_alignment_for_field): For
11538         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11539         && !DECL_PACKED (field), do the alignment update, just use
11540         only desired_align instead of MAX (type_align, desired_align)
11541         as the alignment.
11542         (place_field): Don't do known_align < desired_align handling
11543         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11544         is non-NULL, instead do it after rli->prev_field handling and
11545         only if not within a bitfield word.  For DECL_PACKED (field)
11546         use type_align of BITS_PER_UNIT.
11548 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
11550         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11551         superfluous parentheses and trailing spaces.
11553 2018-02-28  Richard Biener  <rguenther@suse.de>
11555         PR tree-optimization/84584
11556         * graphite-scop-detection.c (scop_detection::add_scop): Discard
11557         SCoPs with fake exit edge.
11559 2018-02-28  Martin Liska  <mliska@suse.cz>
11561         PR testsuite/84597
11562         * timevar.c (timer::print): Fix format to properly print 100%
11563         values.
11565 2018-02-28  Richard Biener  <rguenther@suse.de>
11567         PR middle-end/84607
11568         * genmatch.c (capture_info::walk_match): Do not mark
11569         captured expressions without operands as expr_p given
11570         they act more like predicates and should be subject to
11571         "lost tail" side-effect preserving.
11573 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
11575         PR rtl-optimization/81611
11576         * auto-inc-dec.c (attempt_change): Move dead note from
11577         mem_insn if it's the next use of regno
11578         (find_address): Take address use of reg holding
11579         non-incremented value.  Add parm to limit search to the named
11580         reg only.
11581         (merge_in_block): Attempt to use a mem insn that is the next
11582         use of the original regno.
11584 2018-02-27  Martin Sebor  <msebor@redhat.com>
11586         PR c++/83871
11587         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11588         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11590 2018-02-27  Martin Sebor  <msebor@redhat.com>
11592         PR translation/84207
11593         * diagnostic-core.h (warning_n, error_n, inform_n): Change
11594         n argument to unsigned HOST_WIDE_INT.
11595         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11596         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
11597         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11598         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11600 2018-02-27  Richard Biener  <rguenther@suse.de>
11602         PR tree-optimization/84512
11603         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11604         Do not use the estimate returned from record_stmt_cost for
11605         the scalar iteration cost but sum properly using add_stmt_cost.
11607 2018-02-27  Richard Biener  <rguenther@suse.de>
11609         PR tree-optimization/84466
11610         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11611         Adjust last change to less strictly validate use operands.
11613 2018-02-27  Martin Liska  <mliska@suse.cz>
11615         PR gcov-profile/84548
11616         * gcov.c (process_file): Allow partial overlap and consider it
11617         also as group functions.
11618         (output_lines): Properly calculate range of lines for a group.
11620 2018-02-27  Martin Liska  <mliska@suse.cz>
11622         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11623         'ggc' suffixes.  Change first column width.
11624         (timer::print): Fix formatting of the column.
11626 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
11628         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11629         preserve inline entry blocks for the sake of debug inline
11630         entry point markers alone.
11631         (remove_unused_locals): Suggest in comments a better place to
11632         force the preservation of inline entry blocks that are
11633         otherwise unused, but do not preserve them.
11635 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
11637         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11639 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
11641         PR target/84039
11642         * config/i386/constraints.md (Bs): Replace
11643         ix86_indirect_branch_register with
11644         TARGET_INDIRECT_BRANCH_REGISTER.
11645         (Bw): Likewise.
11646         * config/i386/i386.md (indirect_jump): Likewise.
11647         (tablejump): Likewise.
11648         (*sibcall_memory): Likewise.
11649         (*sibcall_value_memory): Likewise.
11650         Peepholes of indirect call and jump via memory: Likewise.
11651         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11652         (*sibcall_value_GOT_32): Likewise.
11653         * config/i386/predicates.md (indirect_branch_operand): Likewise.
11654         (GOT_memory_operand): Likewise.
11655         (call_insn_operand): Likewise.
11656         (sibcall_insn_operand): Likewise.
11657         (GOT32_symbol_operand): Likewise.
11658         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11660 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
11662         PR rtl-optimization/83496
11663         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11664         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
11665         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11666         redundant insn, if any.
11667         (relax_delay_slots): Likewise.
11668         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11670 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
11672         PR tree-optimization/83965
11673         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11674         that grouped statements are part of a reduction chain.  Return
11675         true if the statement is not marked as a reduction itself but
11676         is part of a group.
11677         (vect_recog_dot_prod_pattern): Don't check whether the statement
11678         is part of a group here.
11679         (vect_recog_sad_pattern): Likewise.
11680         (vect_recog_widen_sum_pattern): Likewise.
11682 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
11684         PR debug/84545
11685         * final.c (rest_of_clean_state): Also look for calls inside sequences.
11687 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
11689         PR target/84530
11690         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11691         the bool argument.
11692         (ix86_output_indirect_function_return): New prototype.
11693         (ix86_split_simple_return_pop_internal): Likewise.
11694         * config/i386/i386.c (indirect_return_via_cx): New.
11695         (indirect_return_via_cx_bnd): Likewise.
11696         (indirect_thunk_name): Handle return va CX_REG.
11697         (output_indirect_thunk_function): Create alias for
11698         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11699         (ix86_output_indirect_jmp): Remove the bool argument.
11700         (ix86_output_indirect_function_return): New function.
11701         (ix86_split_simple_return_pop_internal): Likewise.
11702         * config/i386/i386.md (*indirect_jump): Don't pass false
11703         to ix86_output_indirect_jmp.
11704         (*tablejump_1): Likewise.
11705         (simple_return_pop_internal): Change it to define_insn_and_split.
11706         Call ix86_split_simple_return_pop_internal to split it for
11707         -mfunction-return=.
11708         (simple_return_indirect_internal): Call
11709         ix86_output_indirect_function_return instead of
11710         ix86_output_indirect_jmp.
11712 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
11714         PR bootstrap/84405
11715         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11716         memset and value initialization afterwards.
11718 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
11720         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11722 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
11724         PR target/84521
11725         * common/config/aarch64/aarch64-common.c
11726         (aarch_option_optimization_table[]): Switch
11727         off fomit-frame-pointer
11729 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
11730             Chung-Ju Wu  <jasonwucj@gmail.com>
11732         * config/nds32/nds32-multiple.md (load_multiple): Disallow
11733         volatile memory.
11734         (store_multiple): Ditto.
11736 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
11738         * config.gcc: Add --with-cpu support for nds32 target.
11739         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11740         * config/nds32/nds32.opt: Add -mcpu= option.
11742 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
11744         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11745         isel=yes): Warn for these deprecated options.
11747 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
11749         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11750         ISA_2_5_MASKS_EMBEDDED.
11752 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
11754         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11755         p->max as pointers rather than using iterative_hash_expr.
11757 2018-02-23  Carl Love  <cel@us.ibm.com>
11759         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11760         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11761         BU_P8V_OVERLOAD_2.
11762         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11763         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11764         P8V_BUILTIN_VEC_VUNSIGNED2.
11766 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
11768         PR target/81572
11769         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11770         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11771         LRA_UNKNOWN_ALT.
11772         * lra-constraints.c (curr_insn_transform): Set up
11773         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
11774         LRA_UNKNOWN_ALT.
11775         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11776         * lra-eliminations.c (spill_pseudos): Ditto.
11777         (process_insn_for_elimination): Ditto.
11778         * lra-lives.c (reg_early_clobber_p): Use the new macros.
11779         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11780         LRA_NON_CLOBBERED_ALT.
11782 2018-02-22  Martin Sebor  <msebor@redhat.com>
11784         PR tree-optimization/84480
11785         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11786         to maybe_diag_stxncpy_trunc.  Call it.
11787         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11788         from gimple_fold_builtin_strcpy.  Print inlining stack.
11789         (handle_builtin_stxncpy): Print inlining stack.
11790         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11792 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
11794         PR target/84176
11795         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11796         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11797         and -fcheck-pointer-bounds are used together.
11798         (indirect_thunk_prefix): New enum.
11799         (indirect_thunk_need_prefix): New function.
11800         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
11801         "_nt" instead of "_bnd" for NOTRACK prefix.
11802         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11803         (output_indirect_thunk_function): Likewise.
11804         (): Likewise.
11805         (ix86_code_end): Update output_indirect_thunk_function calls.
11806         (ix86_output_indirect_branch_via_reg): Replace
11807         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11808         (ix86_output_indirect_branch_via_push): Likewise.
11809         (ix86_output_function_return): Likewise.
11810         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11811         incompatible with -fcf-protection=branch and
11812         -fcheck-pointer-bounds.
11814 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
11816         PR target/83335
11817         * config/aarch64/aarch64.c (aarch64_print_address_internal):
11818         Change gcc_assert call to output_operand_lossage.
11820 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
11822         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11824 2018-02-22  DJ Delorie  <dj@redhat.com>
11825             Sebastian Perta  <sebastian.perta@renesas.com>
11826             Oleg Endo  <olegendo@gcc.gnu.org>
11828         * config/rx/rx.c (rx_rtx_costs): New function.
11829         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11831 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
11833         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11835 2018-02-22  Martin Liska  <mliska@suse.cz>
11837         PR driver/83193
11838         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11839         Add "native" as a possible value.
11841 2018-02-22  Martin Liska  <mliska@suse.cz>
11843         PR driver/83193
11844         * config/i386/i386.c (ix86_option_override_internal):
11845         Add "native" as a possible value for -march and -mtune.
11847 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
11849         PR target/84502
11850         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11851         to all type variants.
11853         PR tree-optimization/84503
11854         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11855         width as info->bitpos + info->bitsize - start.
11856         (merged_store_group::merge_overlapping): Simplify width computation.
11857         (check_no_overlap): New function.
11858         (imm_store_chain_info::try_coalesce_bswap): Compute expected
11859         start + width and last_order of the group, fail if check_no_overlap
11860         fails.
11861         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11862         to group if check_no_overlap fails.
11864 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
11866         * config/rs6000/altivec.md: Delete contraint arguments to
11867         define_expand, define_split, and define_peephole2, and in
11868         define_insn_and_split if always unused.
11869         * config/rs6000/darwin.md: Ditto.
11870         * config/rs6000/dfp.md: Ditto.
11871         * config/rs6000/rs6000.md: Ditto.
11872         * config/rs6000/sync.md: Ditto.
11873         * config/rs6000/vector.md: Ditto.
11874         * config/rs6000/vsx.md: Ditto.
11876 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
11878         * config/rs6000/altivec.md: Write output control strings as braced
11879         blocks instead of double-quoted strings.
11880         * config/rs6000/darwin.md: Ditto.
11881         * config/rs6000/rs6000.md: Ditto.
11882         * config/rs6000/vector.md: Ditto.
11883         * config/rs6000/vsx.md: Ditto.
11885 2018-02-21  Jason Merrill  <jason@redhat.com>
11887         PR c++/84314 - ICE with templates and fastcall attribute.
11888         * attribs.c (build_type_attribute_qual_variant): Remove assert.
11890 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
11892         * ipa-cp.c (determine_versionability): Fix comment typos.
11894 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
11896         PR c/84229
11897         * ipa-cp.c (determine_versionability): Do not version functions caling
11898         va_arg_pack.
11900 2018-02-21  Martin Liska  <mliska@suse.cz>
11902         PR driver/83193
11903         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11904         Add "native" as a possible value.
11905         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
11906         the macro when native cpu detection is available.
11908 2018-02-21  Martin Liska  <mliska@suse.cz>
11910         PR driver/83193
11911         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11912         Add "native" as a possible value.
11913         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11914         when native cpu detection is available.
11916 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
11917             Martin Sebor  <msebor@redhat.com>
11919         PR tree-optimization/84478
11920         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11921         false.
11922         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11923         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11924         support which is conservatively correct, for 2 only stay conservative
11925         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
11926         argument to the 2 argument get_range_strlen, adjust 6 arg
11927         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11928         false.
11929         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11930         (gimple_fold_builtin_strlen): Pass true as last argument to
11931         get_range_strlen.
11933 2018-02-20  Martin Sebor  <msebor@redhat.com>
11935         PR middle-end/84095
11936         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11937         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
11938         (builtin_memref::builtin_memref): Factor out parts into
11939         set_base_and_offset and call it.
11941 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
11943         PR middle-end/84406
11944         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11945         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11946         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
11947         search at the associated MODE_INT.
11949 2018-02-20  Jeff Law  <law@redhat.com>
11951         PR middle-end/82123
11952         PR tree-optimization/81592
11953         PR middle-end/79257
11954         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11955         for range data rather than using global data.
11956         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11957         range data rather than using global data.
11958         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11959         pass it to children as needed.
11960         (struct directive::fmtresult): Similarly.
11961         (struct directive::set_width): Similarly.
11962         (struct directive::set_precision): Similarly.
11963         (format_integer, format_directive, parse_directive): Similarly.
11964         (format_none): Accept unnamed vr_values parameter.
11965         (format_percent, format_floating, format_character): Similarly.
11966         (format_string, format_plain): Similarly.
11967         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11968         the EVRP range analyzer for range data rather than using global data.
11969         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11970         gimple-ssa-evrp-analyze.h
11971         (class sprintf_dom_walker): Add after_dom_children member function.
11972         Add evrp_range_analyzer member.
11973         (sprintf_dom_walker::before_dom_children): Call into the EVRP
11974         range analyzer as needed.
11975         (sprintf_dom_walker::after_dom_children): New member function.
11976         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11977         if not optimizing.
11978         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11979         (evrp_range_analyzer::pop_to_marker): Likewise.
11981 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
11983         PR tree-optimization/84419
11984         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11985         with the required type if its current type is compatible but
11986         different.
11988 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
11990         PR middle-end/82004
11991         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11992         after vectorization.
11994 2018-02-20  Martin Liska  <mliska@suse.cz>
11996         PR driver/83193
11997         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11998         possible values if we don't have a hint.
12000 2018-02-20  Martin Liska  <mliska@suse.cz>
12002         PR c/84310
12003         PR target/79747
12004         * final.c (shorten_branches): Build align_tab array with one
12005         more element.
12006         * opts.c (finish_options): Add alignment option limit check.
12007         (MAX_CODE_ALIGN): Likewise.
12008         (MAX_CODE_ALIGN_VALUE): Likewise.
12009         * doc/invoke.texi: Document maximum allowed option value for
12010         all -falign-* options.
12012 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
12014         PR target/84146
12015         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12016         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12017         * var-tracking.c (emit_note_insn_var_location): Remove all references
12018         to NOTE_INSN_CALL_ARG_LOCATION.
12019         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12020         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12021         Use copy_rtx_if_shared.
12022         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12023         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12024         (dwarf2out_var_location): Remove handling of
12025         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12026         on call_insn.
12027         * final.c (final_scan_insn): Remove all references to
12028         NOTE_INSN_CALL_ARG_LOCATION.
12029         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
12030         before dumping final insns.
12031         * except.c (emit_note_eh_region_end): Remove all references to
12032         NOTE_INSN_CALL_ARG_LOCATION.
12033         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12034         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12035         * config/arc/arc.c (hwloop_optimize): Likewise.
12036         * config/arm/arm.c (create_fix_barrier): Likewise.
12037         * config/s390/s390.c (s390_chunkify_start): Likewise.
12038         * config/sh/sh.c (find_barrier): Likewise.
12039         * config/i386/i386.c (rest_of_insert_endbranch,
12040         ix86_seh_fixup_eh_fallthru): Likewise.
12041         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12042         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12043         * config/frv/frv.c (frv_function_prologue): Likewise.
12044         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
12045         reg note.
12046         (note_outside_basic_block_p): Remove all references to
12047         NOTE_INSN_CALL_ARG_LOCATION.
12048         * gengtype.c (adjust_field_rtx_def): Likewise.
12049         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12050         Likewise.
12051         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12052         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12054         PR c++/84444
12055         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12056         is ADDR_EXPR.
12058         PR tree-optimization/84452
12059         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12060         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12061         is NULL.
12063 2018-02-19  Martin Liska  <mliska@suse.cz>
12065         PR sanitizer/82183
12066         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12068 2018-02-19  Martin Liska  <mliska@suse.cz>
12069             Richard Sandiford  <richard.sandiford@linaro.org>
12071         PR tree-optimization/82491
12072         * gimple-fold.c (get_base_constructor): Make earlier bail out
12073         to prevent ubsan.
12075 2018-02-19  Carl Love  <cel@us.ibm.com>
12077         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12078         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12079         BU_P8V_OVERLOAD_1.
12080         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12081         P8V_BUILTIN_VEC_NEG.
12083 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
12085         * config/rl78/rl78.md (movdf): New define expand.
12087 2018-02-19  Martin Liska  <mliska@suse.cz>
12089         PR other/80589
12090         * doc/invoke.texi: Fix typo.
12091         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12093 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
12095         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12096         handle rs6000_single_float and rs6000_double_float specially for
12097         e500 family CPUs.
12099 2018-02-16  Jeff Law  <law@redhat.com>
12101         * config/rx/rx.c (add_pop_cfi_notes): New function.;
12102         (pop_regs): Use it.
12104 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
12106         PR ipa/84425
12107         * ipa-inline.c (inline_small_functions): Fix a typo.
12109 2018-02-16  Nathan Sidwell  <nathan@acm.org>
12111         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12113 2018-02-16  Carl Love  <cel@us.ibm.com>
12115         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12116         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12117         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12118         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12119         expansion to P8V_BUILTIN_VEC_FLOAT2.
12121 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
12123         PR rtl-optimization/70023
12124         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12125         src_regno into account.
12127 2018-02-16  Carl Love  <cel@us.ibm.com>
12129         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12130         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12131         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12132         * config/rs6000/rs6000.c: Remove case statements for
12133         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12134         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12135         and P9V_BUILTIN_VEC_VINSERT4B.
12136         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12137         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12138         * config/rs6000/vsx.md:
12139         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12140         vec_insert4b.
12142 2018-02-16  Carl Love  <cel@us.ibm.com>
12144         * config/rs6000/altivec.h: Add builtin names vec_extract4b
12145         vec_insert4b.
12146         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12147         definitions.
12148         * config/rs6000/rs6000-c.c: Add the definitions for
12149         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12150         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12151         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12152         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
12153         definition for insert4b and define insn *insert3b_internal.
12154         * doc/extend.texi: Add documentation for vec_extract4b.
12156 2018-02-16  Nathan Sidwell  <nathan@acm.org>
12158         * doc/extend.texi (Backwards Compatibility): Mention friend
12159         injection.  Note for-scope is deprecated.
12160         * doc/invoke.texi (-ffriend-injection): Deprecate.
12162 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
12164         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12165         that moved to I2, also allow destinations that are a paradoxical
12166         subreg (instead of a normal reg).
12168 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
12170         PR target/83831
12171         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12172         to QImode.
12174 2018-02-16  Richard Biener  <rguenther@suse.de>
12176         PR tree-optimization/84037
12177         PR tree-optimization/84016
12178         PR target/82862
12179         * config/i386/i386.c (ix86_builtin_vectorization_cost):
12180         Adjust vec_construct for the fact we need additional higher latency
12181         128bit inserts for AVX256 and AVX512 vector builds.
12182         (ix86_add_stmt_cost): Scale vector construction cost for
12183         elementwise loads.
12185 2018-02-16  Richard Biener  <rguenther@suse.de>
12187         PR tree-optimization/84417
12188         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12189         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12190         (non_rewritable_lvalue_p): Likewise, use poly-ints.
12192 2018-02-16  Martin Liska  <mliska@suse.cz>
12194         PR sanitizer/84307
12195         * internal-fn.def (ASAN_CHECK): Set proper flags.
12196         (ASAN_MARK): Likewise.
12198 2018-02-16  Julia Koval  <julia.koval@intel.com>
12200         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12201         from PTA_CANNONLAKE.
12203 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
12205         PR target/84272
12206         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12207         Use ++iter rather than iter++ for std::list iterators.
12208         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
12209         defer deleting them until all nodes in the forest are processed.  Do
12210         free even leaf nodes.  Change to_process into auto_vec.
12212         PR bootstrap/84405
12213         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12214         * vec.h (vec_default_construct): Use memset instead of placement new
12215         if BROKEN_VALUE_INITIALIZATION is defined.
12216         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12217         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12218         is defined.
12220         PR rtl-optimization/83723
12221         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12222         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
12223         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
12224         recursive calls.
12225         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12226         callers.
12227         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12229 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
12231         PR rtl-optimization/81443
12232         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12233         from inner REGs to paradoxical SUBREGs.
12235 2018-02-16  Richard Biener  <rguenther@suse.de>
12237         PR tree-optimization/84399
12238         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12239         For operands we can analyze at their definition make sure we can
12240         analyze them at each use as well.
12242 2018-02-16  Richard Biener  <rguenther@suse.de>
12244         PR tree-optimization/84190
12245         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12246         volatile accesses if the decl isn't volatile.
12248 2018-02-15  Jason Merrill  <jason@redhat.com>
12250         PR c++/84314 - ICE with templates and fastcall attribute.
12251         * attribs.c (build_type_attribute_qual_variant): Don't clobber
12252         TYPE_CANONICAL on an existing type.
12254 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
12256         PR tree-optimization/84383
12257         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12258         dstoff nor call operand_equal_p if dstbase is NULL.
12260         PR tree-optimization/84334
12261         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12262         also a CONSTANT_CLASS_P, punt.
12264 2018-02-14  Jim Wilson  <jimw@sifive.com>
12266         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12267         first SMALL_OPERAND check.  New local min_second_step.  Move assert
12268         to where locals are set.  Add TARGET_RVC support.
12269         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12271 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
12273         * doc/invoke.texi: Correct -Wformat-overflow code sample.
12275 2018-02-14  Martin Sebor  <msebor@redhat.com>
12277         PR tree-optimization/83698
12278         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12279         arrays constrain the offset range to their bounds.
12280         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12281         (builtin_access::overlap): Avoid setting the size of overlap if it's
12282         already been set.
12283         (maybe_diag_overlap): Also consider arrays when deciding what values
12284         of offsets to include in diagnostics.
12286 2018-02-14  Martin Sebor  <msebor@redhat.com>
12288         PR c/84108
12289         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12290         that correspond to the kind of a declaration.
12292 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
12294         PR target/83984
12295         * config/pa/pa.md: Load address of PIC label using the linkage table
12296         if the label is nonlocal.
12298 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12300         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12301         warning message if user requests -maltivec=be.
12302         * doc/invoke.texi: Document deprecation of -maltivec=be.
12304 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
12306         PR target/84220
12307         * config/rs6000/rs6000-c.c: Update definitions for
12308         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12309         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12311 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
12313         PR target/84239
12314         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12315         add _get_ssp intrinsics. Remove argument from
12316         __builtin_ia32_rdssp[d|q].
12317         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12318         * config/i386/i386-builtin.def: Remove argument from
12319         __builtin_ia32_rdssp[d|q].
12320         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12321         ix86_expand_special_args_builtin for _rdssp[d|q].
12322         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12323         Clear register before usage.
12324         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12325         Add documentation for new _get_ssp and _inc_ssp intrinsics.
12327 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
12329         PR tree-optimization/84357
12330         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12331         operand 1 of an ARRAY_REF too.
12333 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
12335         PR target/83831
12336         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12337         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12338         declarations.
12339         (set_of_reg): New struct.
12340         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12341         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12342         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12343         functions.
12344         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12345         Split into bitclr, bitset, bitinvert patterns if appropriate.
12346         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12347         use rx_fuse_in_memory_bitop.
12348         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12349         to named insn, correct maximum insn length.
12351 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
12353         PR target/79242
12354         * machmode.def: Define a complex mode for PARTIAL_INT.
12355         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12356         MODE_PARTIAL_INT.
12357         * doc/rtl.texi: Document CSPImode.
12358         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12359         handling.
12360         (msp430_hard_regno_nregs_with_padding): Likewise.
12362 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
12364         PR target/84279
12365         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12367 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
12369         PR rtl-optimization/84169
12370         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
12371         we generated a parallel as new i3 and we split that to new i2 and i3
12372         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
12373         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
12374         those to i2, not i1.  Partially rewrite this scan code.
12376 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
12378         PR c/82210
12379         * stor-layout.c (place_field): For variable length fields, adjust
12380         offset_align afterwards not just based on the field's alignment,
12381         but also on the size.
12383         PR middle-end/84309
12384         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12385         of exps and logs in the use_exp2 case.
12387 2018-02-13  Jeff Law  <law@redhat.com>
12389         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12390         entry for "vector".
12392         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12393         ARGS as unused.
12395 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
12397         PR debug/84342
12398         PR debug/84319
12399         * common.opt (gas-loc-support, gas-locview-support): New.
12400         (ginline-points, ginternal-reset-location-views): New.
12401         * doc/invoke.texi: Document them.  Use @itemx where intended.
12402         (gvariable-location-views): Adjust.
12403         * target.def (reset_location_view): New.
12404         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12405         (TARGET_RESET_LOCATION_VIEW): New.
12406         * doc/tm.texi: Rebuilt.
12407         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12408         (dwarf2out_default_as_locview_support): New.
12409         (output_asm_line_debug_info): Use option variables.
12410         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12411         (output_loc_list): Likewise.
12412         (add_high_low_attributes): Check option variables.
12413         Don't output entry view attribute in strict mode.
12414         (gen_inlined_subroutine_die): Check option variables.
12415         (dwarf2out_inline_entry): Likewise.
12416         (init_sections_and_labels): Likewise.
12417         (dwarf2out_early_finish): Likewise.
12418         (maybe_reset_location_view): New, from...
12419         (dwarf2out_var_location): ... here.  Call it.
12420         * debug.h (dwarf2out_default_as_loc_support): Declare.
12421         (dwarf2out_default_as_locview_support): Declare.
12422         * hooks.c (hook_int_rtx_insn_0): New.
12423         * hooks.h (hook_int_rtx_insn_0): Declare.
12424         * toplev.c (process_options): Take -gas-loc-support and
12425         -gas-locview-support from dwarf2out.  Enable
12426         -gvariable-location-views by default only with locview
12427         assembler support.  Enable -ginternal-reset-location-views by
12428         default only if the target defines the corresponding hook.
12429         Enable -ginline-points by default if location views are
12430         enabled; force it disabled if statement frontiers are
12431         disabled.
12432         * tree-inline.c (expand_call_inline): Check option variables.
12433         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12435 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
12437         PR tree-optimization/84321
12438         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12439         handling.  Also check whether the anti-range contains any values
12440         that satisfy the mask; switch to a VR_RANGE if not.
12442 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
12444         PR sanitizer/84340
12445         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12447 2018-02-13  Martin Jambor  <mjambor@suse.cz>
12449         PR c++/83990
12450         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12451         of call statements, also set location of a load to a temporary.
12453 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
12455         * config/rl78/rl78.c (add_vector_labels): New function.
12456         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12457         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12458         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12459         which checks that no arguments are passed.
12460         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12461         * doc/extend.texi: Documentation for the new attribute.
12463 2018-02-13  Andreas Schwab  <schwab@suse.de>
12465         * config/riscv/linux.h (CPP_SPEC): Define.
12467 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
12469         PR target/84335
12470         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12471         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12472         OPTION_MASK_ISA_AES as first argument to def_builtin_const
12473         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12474         instead of OPTION_MASK_ISA_PCLMUL as first argument to
12475         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12476         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12477         temporarily for AES and PCLMUL builtins.
12479         PR tree-optimization/84339
12480         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12481         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12482         Formatting fixes.
12484         PR middle-end/84309
12485         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12486         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12487         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12488         inline function.
12489         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12490         inline function.
12491         * omp-simd-clone.h: New file.
12492         * omp-simd-clone.c: Include omp-simd-clone.h.
12493         (expand_simd_clones): No longer static.
12494         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12495         cgraph.h and omp-simd-clone.h.
12496         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12497         (vect_recog_widen_shift_pattern): Formatting fix.
12498         (vect_pattern_recog_1): Don't check optab for calls.
12500         PR target/84336
12501         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12502         operands[2] into a REG before using gen_lowpart on it.
12504 2018-02-12  Jeff Law  <law@redhat.com>
12506         PR target/83760
12507         * config/sh/sh.c (find_barrier): Consider a sibling call
12508         a barrier as well.
12510         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12511         successfully back substituting a reg.
12513 2018-02-12  Richard Biener  <rguenther@suse.de>
12515         PR tree-optimization/84037
12516         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12517         parameter, move visited init to caller.
12518         (vect_slp_analyze_operations): Separate cost from validity
12519         check, initialize visited once for all instances.
12520         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12521         for all instances.
12522         * tree-vect-stmts.c (vect_model_simple_cost): Make early
12523         out an assert.
12524         (vect_model_promotion_demotion_cost): Likewise.
12525         (vectorizable_bswap): Guard cost modeling with !slp_node
12526         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12527         SLP stmts.
12528         (vectorizable_call): Likewise.
12529         (vectorizable_conversion): Likewise.
12530         (vectorizable_assignment): Likewise.
12531         (vectorizable_shift): Likewise.
12532         (vectorizable_operation): Likewise.
12533         (vectorizable_store): Likewise.
12534         (vectorizable_load): Likewise.
12535         (vectorizable_condition): Likewise.
12536         (vectorizable_comparison): Likewise.
12538 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
12540         PR sanitizer/84307
12541         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12542         (ASAN_MARK): Fix fnspec to account for return value, change pointer
12543         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12545 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
12547         PR middle-end/83665
12548         * params.def (inline-min-speedup): Increase from 8 to 15.
12549         (max-inline-insns-auto): Decrease from 40 to 30.
12550         * ipa-split.c (consider_split): Add some buffer for function to
12551         be considered inlining candidate.
12552         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12553         default values.
12555 2018-02-12  Richard Biener  <rguenther@suse.de>
12557         PR tree-optimization/84037
12558         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12559         matched stmts if we cannot swap the non-matched ones.
12561 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
12563         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12564         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12565         _mm_maskz_scalef_round_ss): New intrinsics.
12566         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12567         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12568         __builtin_ia32_scalefss_round): Remove.
12569         (__builtin_ia32_scalefsd_mask_round,
12570         __builtin_ia32_scalefss_mask_round): New intrinsics.
12571         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12572         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12573         ((match_operand:VF_128 2 "<round_nimm_predicate>"
12574         "<round_constraint>")): Changed to ...
12575         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12576         "<round_scalar_constraint>")): ... this.
12577         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12578         %0, %1, %2<round_op3>}"): Changed to ...
12579         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12580         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12581         %2<round_scalar_mask_op3>}"): ... this.
12582         * config/i386/subst.md (round_scalar_nimm_predicate): New.
12584 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
12586         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12587         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12588         (_mm_maskz_sqrt_round_ss): New intrinsics.
12589         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12590         (__builtin_ia32_sqrtsd_mask_round)
12591         (__builtin_ia32_sqrtss_mask_round): New builtins.
12592         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12593         (__builtin_ia32_sqrtss_round): Remove.
12594         (__builtin_ia32_sqrtsd_mask_round)
12595         (__builtin_ia32_sqrtss_mask_round): New builtins.
12596         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12597         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12598         ((match_operand:VF_128 1 "vector_operand"
12599         "xBm,<round_constraint>")): Changed to ...
12600         ((match_operand:VF_128 1 "vector_operand"
12601         "xBm,<round_scalar_constraint>")): ... this.
12602         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12603         %0, %2, %<iptr>1<round_op3>}): Changed to ...
12604         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12605         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12606         %<iptr>1<round_scalar_mask_op3>}): ... this.
12607         ((set_attr "prefix" "<round_prefix>")): Changed to ...
12608         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12610 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
12612         PR target/84266
12613         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12614         Cast vec_cmpeq result to correct type.
12615         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12616         Cast vec_cmpgt result to correct type.
12618 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
12620         * final.c (final_scan_insn_1): Renamed from...
12621         (final_scan_insn): ... this.  New wrapper, to recover
12622         seen from the outermost call in recursive ones.
12623         * config/sparc/sparc.c (output_return): Drop seen from call.
12624         (output_sibcall): Likewise.
12625         * config/visium/visium.c (output_branch): Likewise.
12627 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
12629         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12630         function label.
12632 2018-02-10  Alan Modra  <amodra@gmail.com>
12634         PR target/84300
12635         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12636         Specify LR as an input.
12638 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
12640         PR sanitizer/83987
12641         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12642         remove_member_access_dummy_vars): New functions.
12643         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12644         lower_omp_1, execute_lower_omp): Use them.
12646         PR rtl-optimization/84308
12647         * shrink-wrap.c (spread_components): Release todo vector.
12649 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
12651         PR rtl-optimization/57193
12652         * ira-color.c (struct allocno_color_data): Add member
12653         conflict_allocno_hard_prefs.
12654         (update_conflict_allocno_hard_prefs): New.
12655         (bucket_allocno_compare_func): Add a preference based on
12656         conflict_allocno_hard_prefs.
12657         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12658         (color_allocnos): Remove a dead code.  Initiate
12659         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
12661 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
12663         PR target/84226
12664         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12665         constraint from =wa to wa.  Avoid a subreg on the output operand,
12666         instead use a pseudo and subreg it in a move.
12667         (p9_xxbrd_<mode>): Changed to ...
12668         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12669         (p9_xxbrd_v2df): New expander.
12670         (p9_xxbrw_<mode>): Changed to ...
12671         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12672         (p9_xxbrw_v4sf): New expander.
12674 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
12676         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12678 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
12680         PR target/83926
12681         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12682         multiply in 32-bit mode.
12683         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12684         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12685         mode.
12687 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
12689         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12690         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12691         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12692         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12694 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
12696         PR lto/84213
12697         * dwarf2out.c (is_trivial_indirect_ref): New function.
12698         (dwarf2out_late_global_decl): Do not generate a location
12699         attribute for variables that have a non-trivial DECL_VALUE_EXPR
12700         and that are not defined in the current unit.
12702 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
12704         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12705         instead of a libcall for UNORDERED.
12707 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
12709         PR target/82641
12710         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12711         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12713 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
12715         PR target/PR84295
12716         * config/s390/s390.c (s390_set_current_function): Invoke
12717         s390_indirect_branch_settings also if fndecl didn't change.
12719 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
12721         * config/rs6000/rs6000.md (blockage): Set length to zero.
12723 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
12725         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12727 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
12729         PR sanitizer/84285
12730         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12731         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12732         -static-lib*san.
12734         PR debug/84252
12735         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12736         PARALLEL incoming that failed vt_get_decl_and_offset check.
12738         PR middle-end/84237
12739         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12740         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12741         TREE_READONLY bit.
12742         (get_variable_section): For decls in named .bss* sections pass true as
12743         second argument to bss_initializer_p.
12745 2018-02-09  Marek Polacek  <polacek@redhat.com>
12746             Jakub Jelinek  <jakub@redhat.com>
12748         PR c++/83659
12749         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12750         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
12751         Sync some changes from cxx_fold_indirect_ref.
12753 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
12755         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12756         markers.
12757         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12758         (BLOCK_INLINE_ENTRY_LABEL): New.
12759         (dwarf2out_var_location): Disregard inline entry markers.
12760         (inline_entry_data): New struct.
12761         (inline_entry_data_hasher): New hashtable type.
12762         (inline_entry_data_hasher::hash): New.
12763         (inline_entry_data_hasher::equal): New.
12764         (inline_entry_data_table): New variable.
12765         (add_high_low_attributes): Add DW_AT_entry_pc and
12766         DW_AT_GNU_entry_view attributes if a pending entry is found
12767         in inline_entry_data_table.  Add old entry_pc attribute only
12768         if debug nonbinding markers are disabled.
12769         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12770         markers are enabled.
12771         (block_within_block_p, dwarf2out_inline_entry): New.
12772         (dwarf2out_finish): Check that no entries remained in
12773         inline_entry_data_table.
12774         * final.c (reemit_insn_block_notes): Handle inline entry notes.
12775         (final_scan_insn, notice_source_line): Likewise.
12776         (rest_of_clean_state): Skip inline entry markers.
12777         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12778         markers.
12779         * gimple.c (gimple_build_debug_inline_entry): New.
12780         * gimple.h (enum gimple_debug_subcode): Add
12781         GIMPLE_DEBUG_INLINE_ENTRY.
12782         (gimple_build_debug_inline_entry): Declare.
12783         (gimple_debug_inline_entry_p): New.
12784         (gimple_debug_nonbind_marker_p): Adjust.
12785         * insn-notes.def (INLINE_ENTRY): New.
12786         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12787         inline entry marker notes.
12788         (print_insn): Likewise.
12789         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12790         (INSN_DEBUG_MARKER_KIND): Likewise.
12791         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12792         * tree-inline.c (expand_call_inline): Build and insert
12793         debug_inline_entry stmt.
12794         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12795         inline entry blocks early, if nonbind markers are enabled.
12796         (dump_scope_block): Dump fragment info.
12797         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12798         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12799         (gimple_build_debug_inline_entry): New.
12800         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12801         Enable/disable inline entry points too.
12802         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12803         (DEBUG_INSN): Describe inline entry markers.
12805         * common.opt (gvariable-location-views): New.
12806         (gvariable-location-views=incompat5): New.
12807         * config.in: Rebuilt.
12808         * configure: Rebuilt.
12809         * configure.ac: Test assembler for view support.
12810         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12811         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12812         * dwarf2out.c (var_loc_view): New typedef.
12813         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12814         (dwarf2out_locviews_in_attribute): New.
12815         (dwarf2out_locviews_in_loclist): New.
12816         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12817         (enum dw_line_info_opcode): Add LI_adv_address.
12818         (struct dw_line_info_table): Add view.
12819         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12820         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12821         (zero_view_p): New variable.
12822         (ZERO_VIEW_P): New macro.
12823         (output_asm_line_debug_info): New.
12824         (struct var_loc_node): Add view.
12825         (add_AT_view_list, AT_loc_list): New.
12826         (add_var_loc_to_decl): Add view param.  Test it against last.
12827         (new_loc_list): Add view params.  Record them.
12828         (AT_loc_list_ptr): Handle loc and view lists.
12829         (view_list_to_loc_list_val_node): New.
12830         (print_dw_val): Handle dw_val_class_view_list.
12831         (size_of_die): Likewise.
12832         (value_format): Likewise.
12833         (loc_list_has_views): New.
12834         (gen_llsym): Set vl_symbol too.
12835         (maybe_gen_llsym, skip_loc_list_entry): New.
12836         (dwarf2out_maybe_output_loclist_view_pair): New.
12837         (output_loc_list): Output view list or entries too.
12838         (output_view_list_offset): New.
12839         (output_die): Handle dw_val_class_view_list.
12840         (output_dwarf_version): New.
12841         (output_compilation_unit_header): Use it.
12842         (output_skeleton_debug_sections): Likewise.
12843         (output_rnglists, output_line_info): Likewise.
12844         (output_pubnames, output_aranges): Update version comments.
12845         (output_one_line_info_table): Output view numbers in asm comments.
12846         (dw_loc_list): Determine current endview, pass it to new_loc_list.
12847         Call maybe_gen_llsym.
12848         (loc_list_from_tree_1): Adjust.
12849         (add_AT_location_description): Create view list attribute if
12850         needed, check it's absent otherwise.
12851         (convert_cfa_to_fb_loc_list): Adjust.
12852         (maybe_emit_file): Call output_asm_line_debug_info for test.
12853         (dwarf2out_var_location): Reset views as needed.  Precompute
12854         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
12855         attribute.  Set view.
12856         (new_line_info_table): Reset next view.
12857         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12858         (dwarf2out_source_line): Likewise.  Output view resets and labels to
12859         the assembler, or select appropriate line info opcodes.
12860         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12861         (optimize_string_length): Catch it.  Adjust.
12862         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
12863         dw_val_class_view_list, and remove it if no longer needed.
12864         (hash_loc_list): Hash view numbers.
12865         (loc_list_hasher::equal): Compare them.
12866         (optimize_location_lists): Check whether a view list symbol is
12867         needed, and whether the locview attribute is present, and
12868         whether they match.  Remove the locview attribute if no longer
12869         needed.
12870         (index_location_lists): Call skip_loc_list_entry for test.
12871         (dwarf2out_finish): Call output_asm_line_debug_info for test.
12872         Use output_dwarf_version.
12873         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12874         (struct dw_val_node): Add val_view_list.
12875         * final.c (SEEN_NEXT_VIEW): New.
12876         (set_next_view_needed): New.
12877         (clear_next_view_needed): New.
12878         (maybe_output_next_view): New.
12879         (final_start_function): Rename to...
12880         (final_start_function_1): ... this.  Take pointer to FIRST,
12881         add SEEN parameter.  Emit param bindings in the initial view.
12882         (final_start_function): Reintroduce SEEN-less interface.
12883         (final): Rename to...
12884         (final_1): ... this.  Take SEEN parameter.  Output final pending
12885         next view at the end.
12886         (final): Reintroduce seen-less interface.
12887         (final_scan_insn): Output pending next view before switching
12888         sections or ending a block.  Mark the next view as needed when
12889         outputting variable locations.  Notify debug backend of section
12890         changes, and of location view changes.
12891         (rest_of_handle_final): Adjust.
12892         * toplev.c (process_options): Autodetect value for debug variable
12893         location views option.  Warn on incompat5 without -gdwarf-5.
12894         * doc/invoke.texi (gvariable-location-views): New.
12895         (gvariable-location-views=incompat5): New.
12896         (gno-variable-location-views): New.
12898 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
12900         PR tree-optimization/84136
12901         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12902         that the result of find_edge is non-NULL.
12904 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
12906         PR target/83008
12907         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12908         storing integer register in SImode.  Fix cost of 256 and 512
12909         byte aligned SSE register store.
12911 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
12913         * config/i386/i386.c (ix86_multiplication_cost): Fix
12914         multiplication cost for TARGET_AVX512DQ.
12916 2018-02-08  Marek Polacek  <polacek@redhat.com>
12918         PR tree-optimization/84238
12919         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12920         get_range_strlen.
12922 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
12924         PR tree-optimization/84265
12925         * tree-vect-stmts.c (vectorizable_store): Don't treat
12926         VMAT_CONTIGUOUS accesses as grouped.
12927         (vectorizable_load): Likewise.
12929 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
12931         PR tree-optimization/81635
12932         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12933         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12934         (test_round_for_mask): New functions.
12935         (wide_int_cc_tests): Call test_round_for_mask.
12936         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12937         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12938         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12939         range returned by get_range_info.
12941 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
12943         PR ipa/81360
12944         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12945         * symtab.c: Include builtins.h
12946         (symtab_node::output_to_lto_symbol_table_p): Move here
12947         from lto-streamer-out.c:output_symbol_p.
12948         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12949         (output_symbol_p): Move all logic to symtab.c
12950         (produce_symtab): Update.
12952 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
12954         * config/s390/s390-opts.h (enum indirect_branch): Define.
12955         * config/s390/s390-protos.h (s390_return_addr_from_memory)
12956         (s390_indirect_branch_via_thunk)
12957         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12958         (enum s390_indirect_branch_type): Define.
12959         * config/s390/s390.c (struct s390_frame_layout, struct
12960         machine_function): Remove.
12961         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12962         (indirect_branch_table_label_no, indirect_branch_table_name):
12963         Define variables.
12964         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12965         (enum s390_indirect_branch_option): Define.
12966         (s390_return_addr_from_memory): New function.
12967         (s390_handle_string_attribute): New function.
12968         (s390_attribute_table): Add new attribute handler.
12969         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12970         (s390_indirect_branch_via_thunk): New function.
12971         (s390_indirect_branch_via_inline_thunk): New function.
12972         (s390_function_ok_for_sibcall): When jumping via thunk disallow
12973         sibling call optimization for non z10 compiles.
12974         (s390_emit_call): Force indirect branch target to be a single
12975         register.  Add r1 clobber for non-z10 compiles.
12976         (s390_emit_epilogue): Emit return jump via return_use expander.
12977         (s390_reorg): Handle JUMP_INSNs as execute targets.
12978         (s390_option_override_internal): Perform validity checks for the
12979         new command line options.
12980         (s390_indirect_branch_attrvalue): New function.
12981         (s390_indirect_branch_settings): New function.
12982         (s390_set_current_function): Invoke s390_indirect_branch_settings.
12983         (s390_output_indirect_thunk_function):  New function.
12984         (s390_code_end): Implement target hook.
12985         (s390_case_values_threshold): Implement target hook.
12986         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12987         macros.
12988         * config/s390/s390.h (struct s390_frame_layout)
12989         (struct machine_function): Move here from s390.c.
12990         (TARGET_INDIRECT_BRANCH_NOBP_RET)
12991         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12992         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12993         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12994         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12995         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12996         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12997         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12998         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12999         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13000         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13001         (mnemonic attribute): Add values which aren't recognized
13002         automatically.
13003         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13004         pattern for branch conversion.  Fix mnemonic attribute.
13005         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13006         indirect branch via thunk if requested.
13007         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13008         ("*indirect_jump"): Disable for branch conversion using out of
13009         line thunks.
13010         ("indirect_jump_via_thunk<mode>_z10")
13011         ("indirect_jump_via_thunk<mode>")
13012         ("indirect_jump_via_inlinethunk<mode>_z10")
13013         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13014         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13015         ("casesi_jump_via_inlinethunk<mode>_z10")
13016         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13017         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13018         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13019         ("*indirect2_jump"): Disable for branch conversion.
13020         ("casesi_jump"): Turn into expander and expand patterns for branch
13021         conversion.
13022         ("return_use"): New expander.
13023         ("*return"): Emit return via thunk and rename it to ...
13024         ("*return<mode>"): ... this one.
13025         * config/s390/s390.opt: Add new options and and enum for the
13026         option values.
13028 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
13030         * lra-constraints.c (match_reload): Unconditionally use
13031         gen_lowpart_SUBREG, rather than selecting between that
13032         and equivalent gen_rtx_SUBREG code.
13034 2018-02-08  Richard Biener  <rguenther@suse.de>
13036         PR tree-optimization/84233
13037         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13038         changed flag instead of boguously re-using phi_inserted.
13040 2018-02-08  Martin Jambor  <mjambor@suse.cz>
13042         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13043         static local variables.
13045 2018-02-08  Richard Biener  <rguenther@suse.de>
13047         PR tree-optimization/84278
13048         * tree-vect-stmts.c (vectorizable_store): When looking for
13049         smaller vector types to perform grouped strided loads/stores
13050         make sure the mode is supported by the target.
13051         (vectorizable_load): Likewise.
13053 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
13055         * config/aarch64/aarch64.c (aarch64_components_for_bb):
13056         Increase LDP/STP opportunities by adding adjacent callee-saves.
13058 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
13060         PR rtl-optimization/84068
13061         PR rtl-optimization/83459
13062         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13064 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
13066         PR tree-optimization/84224
13067         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13068         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13069         non-zero arguments.
13071 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
13073         PR target/84113
13074         * config/rs6000/altivec.md (*restore_world): Remove LR use.
13075         * config/rs6000/predicates.md (restore_world_operation): Adjust op
13076         count, remove one USE.
13078 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
13080         * doc/install.texi (Configuration): Document the
13081         --with-long-double-format={ibm,ieee} PowerPC configuration
13082         options.
13084         PR target/84154
13085         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13086         Convert from define_expand to be define_insn_and_split.  Rework
13087         float/double/_Float128 conversions to QI/HI/SImode to work with
13088         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
13089         conversions to QI/HImode types did a store and then a load to
13090         truncate the value.  For conversions to VSX registers, don't split
13091         the insn, instead emit the code directly.  Use the code iterator
13092         any_fix to combine signed and unsigned conversions.
13093         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13094         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13095         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13096         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13097         (fix_<mode>di2_hw): Likewise.
13098         (fixuns_<mode>di2_hw): Likewise.
13099         (fix_<mode>si2_hw): Likewise.
13100         (fixuns_<mode>si2_hw): Likewise.
13101         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13102         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13103         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13104         fix<uns>_trunc<SFDF:mode>si2_p8.
13105         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13106         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13107         (fix<uns>_<mode>_mem): Likewise.
13108         (fctiw<u>z_<mode>_mem): Likewise.
13109         (fix<uns>_<mode>_mem): Likewise.
13110         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13111         the register allocator from doing a direct move to the GPRs to do
13112         a store, and instead use the ISA 3.0 store byte/half-word from
13113         vector register instruction.  For IEEE 128-bit floating point,
13114         also optimize stores of 32-bit ints.
13115         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13117 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
13119         * genextract.c (push_pathstr_operand): New function to support
13120         [a-zA-Z].
13121         (walk_rtx): Call push_pathstr_operand.
13122         (print_path): Support [a-zA-Z].
13124 2018-02-07  Richard Biener  <rguenther@suse.de>
13126         PR tree-optimization/84037
13127         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13128         (cse_and_gimplify_to_preheader): Declare.
13129         (vect_get_place_in_interleaving_chain): Likewise.
13130         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13131         ivexpr_map.
13132         (_loop_vec_info::~_loop_vec_info): Delete it.
13133         (cse_and_gimplify_to_preheader): New function.
13134         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13135         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13136         (vectorizable_load): Likewise.  For grouped stores always base
13137         the IV on the first element.
13138         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13139         condition before gimplifying.
13141 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
13143         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13144         *DIV_EXPR and *MOD_EXPR.
13146 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
13148         PR target/84248
13149         * config/i386/i386.c (ix86_option_override_internal): Mask out
13150         the CF_SET bit when checking -fcf-protection.
13152 2018-02-07  Tom de Vries  <tom@codesourcery.com>
13154         PR libgomp/84217
13155         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13156         enough.
13158 2018-02-07  Richard Biener  <rguenther@suse.de>
13160         PR tree-optimization/84204
13161         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13162         this place.
13164         PR tree-optimization/84205
13165         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13166         special-case isl_ast_op_zdiv_r.
13168         PR tree-optimization/84223
13169         * graphite-scop-detection.c (gather_bbs::before_dom_children):
13170         Only add conditions from within the region.
13171         (gather_bbs::after_dom_children): Adjust.
13173 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
13175         PR target/84209
13176         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13177         * config/avr/avr.md: Only post-reload split REG-REG moves if
13178         either register is GENERAL_REG_P.
13180 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
13182         PR tree-optimization/84235
13183         * tree-ssa-scopedtables.c
13184         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13185         if the subtraction is performed in floating point type where NaNs are
13186         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13187         build 1.  Formatting fix.
13189 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
13191         PR target/84146
13192         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13193         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13194         and skip it regardless of bb boundaries.  Use CALL_P macro,
13195         don't test INSN_P (insn) together with CALL_P or JUMP_P check
13196         unnecessarily, formatting fix.
13198 2018-02-06  Michael Collison  <michael.collison@arm.com>
13200         * config/arm/thumb2.md:
13201         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13202         (*thumb_mov_notscc): Ditto.
13204 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
13206         PR target/84154
13207         * config/rs6000/rs6000.md (su code attribute): Use "u" for
13208         unsigned_fix, not "s".
13210 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13212         * configure.ac (gcc_fn_eh_frame_ro): New function.
13213         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13214         correct .eh_frame permissions.
13215         * configure: Regenerate.
13217 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
13219         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13220         irrelevant options.
13222 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13224         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13225         Display warning message for -mno-speculate-indirect-jumps.
13227 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
13229         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13230         Undocumented.
13231         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13233 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
13235         PR tree-optimization/84225
13236         * tree-eh.c (find_trapping_overflow): Only call
13237         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13239 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
13241         PR target/84145
13242         * config/i386/i386.c: Reimplement the check of possible options
13243         -mibt/-mshstk conbination. Change error messages.
13244         * doc/invoke.texi: Fix a typo: remove extra '='.
13246 2018-02-06  Marek Polacek  <polacek@redhat.com>
13248         PR tree-optimization/84228
13249         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13251 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
13253         PR target/82641
13254         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13255         emitted arch directives.
13256         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13257         __ARM_FEATURE_COPROC before changing architectures.
13259 2018-02-06  Richard Biener  <rguenther@suse.de>
13261         * config/i386/i386.c (print_reg): Fix typo.
13262         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13264 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
13266         * configure: Regenerate.
13268 2018-02-05  Martin Sebor  <msebor@redhat.com>
13270         PR tree-optimization/83369
13271         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13272         inlining context.
13274 2018-02-05  Martin Liska  <mliska@suse.cz>
13276         * doc/invoke.texi: Cherry-pick upstream r323995.
13278 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
13280         * ira.c (ira_init_register_move_cost): Adjust comment.
13282 2018-02-05  Martin Liska  <mliska@suse.cz>
13284         PR gcov-profile/84137
13285         * doc/gcov.texi: Fix typo in documentation.
13287 2018-02-05  Martin Liska  <mliska@suse.cz>
13289         PR gcov-profile/83879
13290         * doc/gcov.texi: Document necessity of --dynamic-list-data when
13291         using dlopen functionality.
13293 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
13295         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13296         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13297         _mm_maskz_range_ss, _mm_mask_range_round_ss,
13298         _mm_maskz_range_round_ss): New intrinsics.
13299         (__builtin_ia32_rangesd128_round)
13300         (__builtin_ia32_rangess128_round): Remove.
13301         (__builtin_ia32_rangesd128_mask_round,
13302         __builtin_ia32_rangess128_mask_round): New builtins.
13303         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13304         __builtin_ia32_rangess128_round): Remove.
13305         (__builtin_ia32_rangesd128_mask_round,
13306         __builtin_ia32_rangess128_mask_round): New builtins.
13307         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13308         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13309         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13310         "<round_saeonly_constraint>")): Changed to ...
13311         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13312         "<round_saeonly_scalar_constraint>")): ... this.
13313         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13314         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13315         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13316         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13317         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13319 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
13321         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13322         options.
13323         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13324         Remove all values except native, 8540 and 8548.
13326 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
13328         * config/i386/i386.c (ix86_output_function_return): Pass
13329         INVALID_REGNUM, instead of -1, as invalid register number to
13330         indirect_thunk_name and output_indirect_thunk.
13332 2018-02-02  Julia Koval  <julia.koval@intel.com>
13334         * config.gcc: Add -march=icelake.
13335         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13336         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13337         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13338         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13339         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13340         (processor_target_table): Add icelake.
13341         (ix86_option_override_internal): Handle new PTAs.
13342         (get_builtin_code_for_version): Handle icelake.
13343         (M_INTEL_COREI7_ICELAKE): New.
13344         (fold_builtin_cpu): Handle icelake.
13345         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13346         * doc/invoke.texi: Add -march=icelake.
13348 2018-02-02  Julia Koval  <julia.koval@intel.com>
13350         * config/i386/i386.c (ix86_option_override_internal): Change flags type
13351         to wide_int_bitmask.
13352         * wide-int-bitmask.h: New.
13354 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
13356         PR target/84066
13357         * config/i386/i386.md: Replace Pmode with word_mode in
13358         builtin_setjmp_setup and builtin_longjmp to support x32.
13360 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
13362         PR target/56010
13363         PR target/83743
13364         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13365         #include "opts.h".
13366         (rs6000_supported_cpu_names): New static variable.
13367         (linux_cpu_translation_table): Likewise.
13368         (elf_platform) <cpu>: Define new static variable and use it.
13369         Translate kernel AT_PLATFORM name to canonical name if needed.
13370         Error if platform name is unknown.
13372 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
13374         PR target/84089
13375         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13377 2018-02-01  Jeff Law  <law@redhat.com>
13379         PR target/84128
13380         * config/i386/i386.c (release_scratch_register_on_entry): Add new
13381         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
13382         the scratch if RELEASE_VIA_POP is false.
13383         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13384         If we have to save a temporary register, decrement SIZE appropriately.
13385         Pass new arguments to release_scratch_register_on_entry.
13386         (ix86_adjust_stack_and_probe): Likewise.
13387         (ix86_emit_probe_stack_range): Pass new arguments to
13388         release_scratch_register_on_entry.
13390 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
13392         PR rtl-optimization/84157
13393         * combine.c (change_zero_ext): Use REG_P predicate in
13394         front of HARD_REGISTER_P predicate.
13396 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
13398         * config/avr/avr.c (avr_option_override): Move disabling of
13399         -fdelete-null-pointer-checks to...
13400         * common/config/avr/avr-common.c (avr_option_optimization_table):
13401         ...here.
13403 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13405         PR tree-optimization/81635
13406         * tree-data-ref.c (split_constant_offset_1): For types that
13407         wrap on overflow, try to use range info to prove that wrapping
13408         cannot occur.
13410 2018-02-01  Renlin Li  <renlin.li@arm.com>
13412         PR target/83370
13413         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13414         TAILCALL_ADDR_REGS.
13415         (aarch64_register_move_cost): Likewise.
13416         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13417         TAILCALL_ADDR_REGS.
13418         (REG_CLASS_NAMES): Likewise.
13419         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13420         TAILCALL_ADDR_REGS. Remove IP registers.
13421         * config/aarch64/aarch64.md (Ucs): Update register constraint.
13423 2018-02-01  Richard Biener  <rguenther@suse.de>
13425         * domwalk.h (dom_walker::dom_walker): Add additional constructor
13426         for specifying RPO order and allow NULL for that.
13427         * domwalk.c (dom_walker::dom_walker): Likewise.
13428         (dom_walker::walk): Handle NULL RPO order.
13429         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13430         in RPO order.
13431         (rewrite_update_dom_walker): Likewise.
13432         (mark_def_dom_walker): Likewise.
13434 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13436         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13437         (aarch64_maybe_expand_sve_subreg_move): Declare.
13438         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13439         * config/aarch64/predicates.md (aarch64_any_register_operand): New
13440         predicate.
13441         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13442         that are semantically a reverse operation.
13443         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13444         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13445         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13446         functions.
13447         (aarch64_can_change_mode_class): For big-endian, forbid changes
13448         between two SVE modes if they have different element sizes.
13450 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13452         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13453         the TImode handling for big-endian targets.
13455 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13457         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13458         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
13459         not just bytes.
13460         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13461         Remove BSWAP handing for big-endian targets and use the form of
13462         LD1RQ appropariate for the mode.
13464 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13466         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13467         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13468         duplicated element.
13470 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
13472         PR tearget/83845
13473         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13474         check for operands that need to go through aarch64_sve_reload_be.
13476 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
13478         PR tree-optimization/81661
13479         PR tree-optimization/84117
13480         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13481         * tree-eh.c: Include gimplify.h.
13482         (find_trapping_overflow, replace_trapping_overflow,
13483         rewrite_to_non_trapping_overflow): New functions.
13484         * tree-vect-loop.c: Include tree-eh.h.
13485         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13486         * tree-data-ref.c: Include tree-eh.h.
13487         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13489 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
13491         PR rtl-optimization/84123
13492         * combine.c (change_zero_ext): Check if hard register satisfies
13493         can_change_dest_mode before calling gen_lowpart_SUBREG.
13495 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
13497         PR target/82444
13498         * ira.c (ira_init_register_move_cost): Remove assert.
13500 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
13502         PR rtl-optimization/84071
13503         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13504         * doc/tm.texi: Regenerate.
13506 2018-01-31  Richard Biener  <rguenther@suse.de>
13508         PR tree-optimization/84132
13509         * tree-data-ref.c (analyze_miv_subscript): Properly
13510         check whether evolution_function_is_affine_multivariate_p
13511         before calling gcd_of_steps_may_divide_p.
13513 2018-01-31  Julia Koval  <julia.koval@intel.com>
13515         PR target/83618
13516         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13517         * config/i386/i386.md (rdpid_rex64) New.
13518         (rdpid): Make 32bit only.
13520 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
13522         PR lto/84105
13523         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13524         an IDENTIFIER_NODE for FUNCTION_TYPE's.
13526 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
13528         Revert
13529         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
13531         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13533 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
13535         PR rtl-optimization/84071
13536         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13537         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13539 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
13541         * config/arc/arc.c (arc_handle_aux_attribute): New function.
13542         (arc_attribute_table): Add 'aux' attribute.
13543         (arc_in_small_data_p): Consider aux like variables.
13544         (arc_is_aux_reg_p): New function.
13545         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13546         (arc_get_aux_arg): New function.
13547         (prepare_move_operands): Handle aux-register access.
13548         (arc_handle_aux_attribute): New function.
13549         * doc/extend.texi (ARC Variable attributes): Add subsection.
13551 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
13553         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13554         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13555         (arc_attribute_table): Add 'uncached' attribute.
13556         (arc_print_operand): Print '.di' flag for uncached memory
13557         accesses.
13558         (arc_in_small_data_p): Do not consider for small data the uncached
13559         types.
13560         (arc_is_uncached_mem_p): New function.
13561         * config/arc/predicates.md (compact_store_memory_operand): Check
13562         for uncached memory accesses.
13563         (nonvol_nonimm_operand): Likewise.
13564         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13566 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
13568         PR c/84100
13569         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13570         falign-loops=): Add Optimization flag.
13572 2018-01-30  Jeff Law  <law@redhat.com>
13574         PR target/84064
13575         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13576         INT_REGISTERS_SAVED.  Check it prior to calling
13577         get_scratch_register_on_entry.
13578         (ix86_adjust_stack_and_probe): Similarly.
13579         (ix86_emit_probe_stack_range): Similarly.
13580         (ix86_expand_prologue): Corresponding changes.
13582 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13584         PR target/40411
13585         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13586         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13588 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
13590         PR target/84112
13591         * lra-constraints.c (curr_insn_transform): Process AND in the
13592         address.
13594 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
13596         PR rtl-optimization/83986
13597         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13598         dependence against last_pending_memory_flush in addition to
13599         pending_jump_insns.
13601 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
13603         PR tree-optimization/81611
13604         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13605         copies.
13607 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13609         PR target/83758
13610         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13611         a reg rtx.
13613 2018-01-30  Richard Biener  <rguenther@suse.de>
13614             Jakub Jelinek  <jakub@redhat.com>
13616         PR tree-optimization/84111
13617         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13618         inner loops added during recursion, as they don't have up-to-date
13619         SSA form.
13621 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
13623         PR ipa/81360
13624         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13625         (can_inline_edge_by_limits_p): ... here.
13626         (can_early_inline_edge_p, check_callers,
13627         update_caller_keys, update_callee_keys, recursive_inlining,
13628         add_new_edges_to_heap, speculation_useful_p,
13629         inline_small_functions,
13630         inline_small_functions, flatten_function,
13631         inline_to_all_callers_1): Update.
13633 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
13635         * profile-count.c (profile_count::combine_with_ipa_count): Handle
13636         zeros correctly.
13638 2018-01-30  Richard Biener  <rguenther@suse.de>
13640         PR tree-optimization/83008
13641         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13642         invariant and constant vector uses in stmts when they need
13643         more than one stmt.
13645 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13647         PR bootstrap/84017
13648         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13649         * configure: Regenerate.
13651 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
13653         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13654         pattern.
13655         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13656         Use gen_rtx_REG rather than gen_lowpart.
13658 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
13660         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13661         rather than 0 when creating partial subregs.
13663 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
13665         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13666         of usage.
13668 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
13670         PR target/81550
13671         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13672         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13673         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13674         flags.  This restores the settings used before the 2017-07-24.
13675         Turning off pre increment/decrement/modify allows IVOPTS to
13676         optimize DF/SF loops where the index is an int.
13678 2018-01-29  Richard Biener  <rguenther@suse.de>
13679             Kelvin Nilsen  <kelvin@gcc.gnu.org>
13681         PR bootstrap/80867
13682         * tree-vect-stmts.c (vectorizable_call): Don't call
13683         targetm.vectorize_builtin_md_vectorized_function if callee is
13684         NULL.
13686 2018-01-22  Carl Love  <cel@us.ibm.com>
13688         * doc/extend.tex: Fix typo in second arg in
13689         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13691 2018-01-29  Richard Biener  <rguenther@suse.de>
13693         PR tree-optimization/84086
13694         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13695         (flush_ssaname_freelist): When SSA names were released reset
13696         the SCEV hash table.
13698 2018-01-29  Richard Biener  <rguenther@suse.de>
13700         PR tree-optimization/84057
13701         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13702         removed paths when removing edges.
13704 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
13706         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13707         -mfunction-return=@var{choice}.
13709 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13711         PR diagnostic/84034
13712         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13713         Handle CR like TAB.
13714         (layout::print_source_line): Likewise.
13715         (test_get_line_width_without_trailing_whitespace): Add test cases.
13717 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
13719         PR middle-end/84040
13720         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13721         debug insns.
13723 2018-01-26  Jim Wilson  <jimw@sifive.com>
13725         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13727         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13728         specified.
13730 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13732         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13733         and CMP + SUB-immediate -> SUBS.
13735 2018-01-26  Martin Sebor  <msebor@redhat.com>
13737         PR tree-optimization/83896
13738         * tree-ssa-strlen.c (get_string_len): Rename...
13739         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
13740         Avoid assuming length is constant.
13741         (handle_char_store): Use HOST_WIDE_INT for string length.
13743 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
13745         PR target/81763
13746         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13747         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13749 2018-01-26  Richard Biener  <rguenther@suse.de>
13751         PR rtl-optimization/84003
13752         * dse.c (record_store): Only record redundant stores when
13753         the earlier store aliases at least all accesses the later one does.
13755 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
13757         PR rtl-optimization/83985
13758         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13759         REG_CFA_RESTORE insns.
13760         (delete_unmarked_insns): Don't ignore separate shrink wrapping
13761         REG_CFA_RESTORE insns here.
13763         PR c/83989
13764         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13765         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13767 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
13769         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13770         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13771         (arc_init): Likewise.
13772         (arc_override_options): Likewise.
13773         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13774         value.
13775         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13776         support.
13777         * config/arc/arc.h (TARGET_DBNZ): Define.
13778         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13779         properly set the tune attribute.
13780         (dbnz): Use TARGET_DBNZ guard.
13781         * config/arc/arc.opt (mtune): Add core3 option.
13783 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
13785         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13786         recognize new pic like addresses.
13787         (arc_delegitimize_address): Clean up.
13789 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
13791         * config/arc/arc-arches.def: Option mrf16 valid for all
13792         architectures.
13793         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13794         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13795         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13796         * config/arc/arc-tables.opt: Regenerate.
13797         * config/arc/arc.c (arc_conditional_register_usage): Handle
13798         reduced register file case.
13799         (arc_file_start): Set must have build attributes.
13800         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13801         mrf16 option value.
13802         * config/arc/arc.opt (mrf16): Add new option.
13803         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13804         * config/arc/genmultilib.awk: Handle new mrf16 option.
13805         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13806         * config/arc/t-multilib: Regenerate.
13807         * doc/invoke.texi (ARC Options): Document mrf16 option.
13809 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
13811         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13812         * config/arc/arc.c (arc_handle_secure_attribute): New function.
13813         (arc_attribute_table): Add 'secure_call' attribute.
13814         (arc_print_operand): Print secure call operand.
13815         (arc_function_ok_for_sibcall): Don't optimize tail calls when
13816         secure.
13817         (arc_is_secure_call_p): New function.  * config/arc/arc.md
13818         (call_i): Add support for sjli instruction.
13819         (call_value_i): Likewise.
13820         * config/arc/constraints.md (Csc): New constraint.
13822 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
13823             John Eric Martin  <John.Martin@emmicro-us.com>
13825         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13826         * config/arc/arc.c (_arc_jli_section): New struct.
13827         (arc_jli_section): New type.
13828         (rc_jli_sections): New static variable.
13829         (arc_handle_jli_attribute): New function.
13830         (arc_attribute_table): Add jli_always and jli_fixed attribute.
13831         (arc_file_end): New function.
13832         (TARGET_ASM_FILE_END): Define.
13833         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13834         (arc_add_jli_section): New function.
13835         (jli_call_scan): Likewise.
13836         (arc_reorg): Call jli_call_scan.
13837         (arc_output_addsi): Remove 'S' from printing asm operand.
13838         (arc_is_jli_call_p): New function.
13839         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13840         operand.
13841         (movhi_insn): Likewise.
13842         (movsi_insn): Likewise.
13843         (movsi_set_cc_insn): Likewise.
13844         (loadqi_update): Likewise.
13845         (load_zeroextendqisi_update): Likewise.
13846         (load_signextendqisi_update): Likewise.
13847         (loadhi_update): Likewise.
13848         (load_zeroextendhisi_update): Likewise.
13849         (load_signextendhisi_update): Likewise.
13850         (loadsi_update): Likewise.
13851         (loadsf_update): Likewise.
13852         (movsicc_insn): Likewise.
13853         (bset_insn): Likewise.
13854         (bxor_insn): Likewise.
13855         (bclr_insn): Likewise.
13856         (bmsk_insn): Likewise.
13857         (bicsi3_insn): Likewise.
13858         (cmpsi_cc_c_insn): Likewise.
13859         (movsi_ne): Likewise.
13860         (movsi_cond_exec): Likewise.
13861         (clrsbsi2): Likewise.
13862         (norm_f): Likewise.
13863         (normw): Likewise.
13864         (swap): Likewise.
13865         (divaw): Likewise.
13866         (flag): Likewise.
13867         (sr): Likewise.
13868         (kflag): Likewise.
13869         (ffs): Likewise.
13870         (ffs_f): Likewise.
13871         (fls): Likewise.
13872         (call_i): Remove 'S' asm letter, add jli instruction.
13873         (call_value_i): Likewise.
13874         * config/arc/arc.op (mjli-always): New option.
13875         * config/arc/constraints.md (Cji): New constraint.
13876         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13877         operand.
13878         (subsf3_fpx): Likewise.
13879         (mulsf3_fpx): Likewise.
13880         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13881         asm operand.
13882         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13883         function attrbutes.
13884         * doc/invoke.texi (ARC): Document mjli-always option.
13886 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
13888         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13889         avoid addition with 0 and use incw and decw where possible.
13891 2018-01-26  Richard Biener  <rguenther@suse.de>
13893         PR tree-optimization/81082
13894         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13895         association if it requires casting to unsigned.
13896         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13897         from fold_plusminus_mult_expr to catch important cases late when
13898         range info is available.
13900 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13902         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13903         * configure.ac (hidden_linkonce): New test.
13904         * configure: Regenerate.
13905         * config.in: Regenerate.
13907 2018-01-26  Julia Koval  <julia.koval@intel.com>
13909         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13910         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13911         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13912         _mm_mask_bitshuffle_epi64_mask): Fix type.
13913         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13914         USI_FTYPE_V4DI_V4DI_USI): Remove.
13915         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13916         __builtin_ia32_vpshufbitqmb256_mask,
13917         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13918         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13919         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13921 2018-01-26  Alan Modra  <amodra@gmail.com>
13923         PR target/84033
13924         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13925         UNSPEC_VBPERMQ.  Sort other unspecs.
13927 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
13929         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13931 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
13933         PR middle-end/83055
13934         * predict.c (drop_profile): Do not push/pop cfun; update also
13935         node->count.
13936         (handle_missing_profiles): Fix logic looking for zero profiles.
13938 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
13940         PR middle-end/83977
13941         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13942         on functions with #pragma omp declare simd or functions with simd
13943         attribute.
13944         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13945         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13946         Remove trailing \n from warning_at calls.
13948 2018-01-25  Tom de Vries  <tom@codesourcery.com>
13950         PR target/84028
13951         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13952         for neutered workers.
13954 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
13956         PR target/68467
13957         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13958         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13960 2018-01-24  Jeff Law  <law@redhat.com>
13962         PR target/83994
13963         * i386.c (get_probe_interval): Move to earlier point.
13964         (ix86_compute_frame_layout): If -fstack-clash-protection and
13965         the frame is larger than the probe interval, then use pushes
13966         to save registers rather than reg->mem moves.
13967         (ix86_expand_prologue): Remove conditional for int_registers_saved
13968         assertion.
13970 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
13972         PR target/84014
13973         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13974         min/max for never referenced object.
13976 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
13978         PR middle-end/83977
13979         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13980         here.
13981         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13982         attributes from DECL_ATTRIBUTES (decl) without affecting
13983         DECL_ATTRIBUTES (current_function_decl).
13984         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13985         functions with non-NULL DECL_ABSTRACT_ORIGIN.
13987 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
13989         PR tree-optimization/83979
13990         * fold-const.c (fold_comparison): Use constant_boolean_node
13991         instead of boolean_{true,false}_node.
13993 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
13995         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13996         with zero counts.
13998 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14000         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14001         Simplify the clause that sets the length attribute.
14002         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14003         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14004         clause that sets the length attribute.
14005         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14007 2018-01-24  Tom de Vries  <tom@codesourcery.com>
14009         PR target/83589
14010         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14011         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14012         Add strict parameter.
14013         (prevent_branch_around_nothing): Insert dummy insn between branch to
14014         label and label with no ptx insn inbetween.
14015         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14017 2018-01-24  Tom de Vries  <tom@codesourcery.com>
14019         PR target/81352
14020         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14021         for neutered threads in warp.
14022         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14024 2018-01-24  Richard Biener  <rguenther@suse.de>
14026         PR tree-optimization/83176
14027         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14028         operands.
14030 2018-01-24  Richard Biener  <rguenther@suse.de>
14032         PR tree-optimization/82819
14033         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14034         code generating pluses that are no-ops in the target precision.
14036 2018-01-24  Richard Biener  <rguenther@suse.de>
14038         PR middle-end/84000
14039         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14041 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
14043         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14044         to merge probabilities.
14045         * predict.c (probably_never_executed): Also mark as cold functions
14046         with global 0 profile and guessed local profile.
14047         * profile-count.c (profile_probability::combine_with_count): New
14048         member function.
14049         * profile-count.h (profile_probability::operator*,
14050         profile_probability::operator*=, profile_probability::operator/,
14051         profile_probability::operator/=): Reduce precision to adjusted
14052         and set value to guessed on contradictory divisions.
14053         (profile_probability::combine_with_freq): Remove.
14054         (profile_probability::combine_wiht_count): Declare.
14055         (profile_count::force_nonzero):: Set to adjusted.
14056         (profile_count::probability_in):: Set quality to adjusted.
14057         * tree-ssa-tail-merge.c (replace_block_by): Use
14058         combine_with_count.
14060 2018-01-23  Andrew Waterman  <andrew@sifive.com>
14061             Jim Wilson  <jimw@sifive.com>
14063         * config/riscv/riscv.c (riscv_stack_boundary): New.
14064         (riscv_option_override): Set riscv_stack_boundary.  Handle
14065         riscv_preferred_stack_boundary_arg.
14066         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14067         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14068         (STACK_BOUNDARY): Set to riscv_stack_boundary.
14069         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14070         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14071         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14073 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
14075         PR target/83905
14076         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14077         of struct ix86_frame.
14078         (ix86_expand_epilogue): Likewise.  Add a local variable for
14079         the reg_save_offset field in struct ix86_frame.
14081 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
14083         PR tree-optimization/82604
14084         * tree-loop-distribution.c (enum partition_kind): New enum item
14085         PKIND_PARTIAL_MEMSET.
14086         (partition_builtin_p): Support above new enum item.
14087         (generate_code_for_partition): Ditto.
14088         (compute_access_range): Differentiate cases that equality can be
14089         proven at all loops, the innermost loops or no loops.
14090         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14091         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
14092         (finalize_partitions, distribute_loop): Don't fuse partition of
14093         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14094         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14095         parloop is enabled.
14097 2018-01-23  Martin Liska  <mliska@suse.cz>
14099         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14100         order to ignore the predictor.
14101         (PRED_POLYMORPHIC_CALL): Likewise.
14102         (PRED_RECURSIVE_CALL): Likewise.
14104 2018-01-23  Martin Liska  <mliska@suse.cz>
14106         * tree-profile.c (tree_profiling): Print function header to
14107         aware reader which function we are working on.
14108         * value-prof.c (gimple_find_values_to_profile): Do not print
14109         not interesting value histograms.
14111 2018-01-23  Martin Liska  <mliska@suse.cz>
14113         * profile-count.h (enum profile_quality): Add
14114         profile_uninitialized as the first value. Do not number values
14115         as they are zero based.
14116         (profile_count::verify): Update sanity check.
14117         (profile_probability::verify): Likewise.
14119 2018-01-23  Nathan Sidwell  <nathan@acm.org>
14121         * doc/invoke.texi (ffor-scope): Deprecate.
14123 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
14125         PR tree-optimization/83510
14126         * domwalk.c (set_all_edges_as_executable): New function.
14127         (dom_walker::dom_walker): Convert bool param
14128         "skip_unreachable_blocks" to enum reachability.  Move setup of
14129         edge flags to set_all_edges_as_executable and only do it when
14130         reachability is REACHABLE_BLOCKS.
14131         * domwalk.h (enum dom_walker::reachability): New enum.
14132         (dom_walker::dom_walker): Convert bool param
14133         "skip_unreachable_blocks" to enum reachability.
14134         (set_all_edges_as_executable): New decl.
14135         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
14136         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14137         "reachability".
14138         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14139         but converting true to REACHABLE_BLOCKS.
14140         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14141         * tree-vrp.c
14142         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14143         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14144         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14145         REACHABLE_BLOCKS.
14146         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14147         if check_all_array_refs will be called.
14149 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
14151         * tree.c (selftest::test_location_wrappers): Add more test
14152         coverage.
14154 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
14156         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14157         (selftest::test_bit_in_range): Likewise.
14159 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
14161         PR testsuite/83888
14162         * doc/sourcebuild.texi (vect_float): Say that the selector
14163         only describes the situation when -funsafe-math-optimizations is on.
14164         (vect_float_strict): Document.
14166 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
14168         PR tree-optimization/83965
14169         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14170         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14171         instead of checking only for a reduction.
14172         (vect_recog_widen_sum_pattern): Likewise.
14174 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
14176         * predict.c (probably_never_executed): Only use precise profile info.
14177         (compute_function_frequency): Skip after inlining hack since we now
14178         have quality checking.
14180 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
14182         * profile-count.h (profile_probability::very_unlikely,
14183         profile_probability::unlikely, profile_probability::even): Set
14184         precision to guessed.
14186 2018-01-23  Richard Biener  <rguenther@suse.de>
14188         PR tree-optimization/83963
14189         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14190         Properly terminate dominator walk when crossing the exit edge not
14191         when visiting its source block.
14193 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
14195         PR c++/83918
14196         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14197         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14199 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
14201         PR tree-optimization/83957
14202         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
14203         semicolon after for body surrounded by braces.
14205         PR tree-optimization/83081
14206         * profile-count.h (profile_probability::split): New method.
14207         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14208         Use profile_probability::split.
14209         (do_compare_rtx_and_jump): Fix adjustment of probabilities
14210         when splitting a single conditional jump into 2.
14212 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
14214         PR tree-optimization/69452
14215         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14216         decl.
14218 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14220         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14221         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14222         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14224 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14226         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14227         declaration.
14228         * config/rl78/rl78.md (movdi): New define_expand.
14229         * config/rl78/rl78.c (rl78_split_movdi): New function.
14231 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
14233         PR target/83862
14234         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14235         no longer used.
14236         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14237         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14238         128-bit to produce an UNSPEC move to get the double word with the
14239         signbit and then a shift directly to do signbit.
14240         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14241         implementation with a new version that just does either a direct
14242         move or a regular move.  Move memory interface to separate insns.
14243         Move insns so they are next to the expander.
14244         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14245         with signbit move.  Split big and little endian case.
14246         (signbit<mode>2_dm_mem_le): Likewise.
14247         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14248         (signbit<mode>2_dm2): Likewise.
14250 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14252         * config/rl78/rl78.md (anddi3): New define_expand.
14254 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14256         * config/rl78/rl78.md (umindi3): New define_expand.
14258 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14260         * config/rl78/rl78.md (smindi3): New define_expand.
14262 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14264         * config/rl78/rl78.md (smaxdi3): New define_expand.
14266 2018-01-22  Carl Love  <cel@us.ibm.com>
14268         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14269         LVX_V1TI): Add macro expansion.
14270         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14271         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14272         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14273         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14274         Change check to determine if the instruction is a byte reversing
14275         entry.  Fix typo in comment.
14276         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14277         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14278         Add def_builtin calls for new builtins.
14279         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14280         Add define_insn expansion.
14282 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14284         * config/rl78/rl78.md (umaxdi3): New define_expand.
14286 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
14288         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14289         for non-QImode registers.
14291 2018-01-22  Richard Biener  <rguenther@suse.de>
14293         PR tree-optimization/83963
14294         * graphite-scop-detection.c (scop_detection::get_sese): Delay
14295         including the loop exit block.
14296         (scop_detection::merge_sese): Likewise.
14297         (scop_detection::add_scop): Do it here instead.
14299 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14301         * doc/sourcebuild.texi (arm_softfloat): Document.
14303 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
14305         PR gcc/77734
14306         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14307         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14308         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14310 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14311             David Edelsohn  <dje.gcc@gmail.com>
14313         PR target/83946
14314         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14315         Change "crset eq" to "crset 2".
14316         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14317         (*call_indirect_aix<mode>_nospec): Likewise.
14318         (*call_value_indirect_aix<mode>_nospec): Likewise.
14319         (*call_indirect_elfv2<mode>_nospec): Likewise.
14320         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14321         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14322         change assembly output from . to $.
14323         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14324         (indirect_jump<mode>_nospec): Change assembly output from . to $.
14325         (*tablejump<mode>_internal1_nospec): Likewise.
14327 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
14329         PR target/80870
14330         * config/sh/sh_optimize_sett_clrt.cc:
14331         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14333 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
14335         PR tree-optimization/83940
14336         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14337         offset_dt to vect_constant_def rather than vect_unknown_def_type.
14338         (vect_check_load_store_mask): Add a mask_dt_out parameter and
14339         use it to pass back the definition type.
14340         (vect_check_store_rhs): Likewise rhs_dt_out.
14341         (vect_build_gather_load_calls): Add a mask_dt argument and use
14342         it instead of a call to vect_is_simple_use.
14343         (vectorizable_store): Update calls to vect_check_load_store_mask
14344         and vect_check_store_rhs.  Use the dt returned by the latter instead
14345         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
14346         instead of calls to vect_is_simple_use.  Pass the scalar rather
14347         than the vector operand to vect_is_simple_use when handling
14348         second and subsequent copies of an rhs value.
14349         (vectorizable_load): Update calls to vect_check_load_store_mask
14350         and vect_build_gather_load_calls.  Use the cached mask_dt and
14351         gs_info.offset_dt instead of calls to vect_is_simple_use.
14353 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
14355         PR middle-end/83945
14356         * tree-emutls.c: Include gimplify.h.
14357         (lower_emutls_2): New function.
14358         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14359         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14360         it before further processing.
14362         PR target/83930
14363         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14364         UINTVAL (trueop1) instead of INTVAL (op1).
14366 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
14368         PR debug/81570
14369         PR debug/83728
14370         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14371         INCOMING_FRAME_SP_OFFSET if not defined.
14372         (scan_trace): Add ENTRY argument.  If true and
14373         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14374         emit a note to adjust the CFA offset.
14375         (create_cfi_notes): Adjust scan_trace callers.
14376         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14377         INCOMING_FRAME_SP_OFFSET in the CIE.
14378         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14379         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14380         Likewise.
14381         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14382         * doc/tm.texi: Regenerated.
14384 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14386         PR rtl-optimization/83147
14387         * lra-constraints.c (remove_inheritance_pseudos): Use
14388         lra_substitute_pseudo_within_insn.
14390 2018-01-19  Tom de Vries  <tom@codesourcery.com>
14391             Cesar Philippidis  <cesar@codesourcery.com>
14393         PR target/83920
14394         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14396 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
14398         PR target/83790
14399         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14400         spaces for function labels.
14402 2018-01-19  Martin Liska  <mliska@suse.cz>
14404         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14405         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14406         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14407         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14408         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14409         (PRED_CONST_RETURN): Change from 69 to 65.
14410         (PRED_NULL_RETURN): Change from 91 to 71.
14411         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14412         (PRED_LOOP_GUARD): Change from 66 to 73.
14414 2018-01-19  Martin Liska  <mliska@suse.cz>
14416         * predict.c (predict_insn_def): Add new assert.
14417         (struct branch_predictor): Change type to signed integer.
14418         (test_prediction_value_range): Amend test to cover
14419         PROB_UNINITIALIZED.
14420         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14421         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14422         (PRED_LOOP_ITERATIONS_MAX): Likewise.
14423         (PRED_LOOP_IV_COMPARE): Likewise.
14424         * predict.h (PROB_UNINITIALIZED): Define new constant.
14426 2018-01-19  Martin Liska  <mliska@suse.cz>
14428         * predict.c (dump_prediction): Add new format for
14429         analyze_brprob.py script which is enabled with -details
14430         suboption.
14431         * profile-count.h (precise_p): New function.
14433 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
14435         PR tree-optimization/83922
14436         * tree-vect-loop.c (vect_verify_full_masking): Return false if
14437         there are no statements that need masking.
14438         (vect_active_double_reduction_p): New function.
14439         (vect_analyze_loop_operations): Use it when handling phis that
14440         are not in the loop header.
14442 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
14444         PR tree-optimization/83914
14445         * tree-vect-loop.c (vectorizable_induction): Don't convert
14446         init_expr or apply the peeling adjustment for inductions
14447         that are nested within the vectorized loop.
14449 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14451         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14452         instead of NEG.
14454 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
14456         PR sanitizer/81715
14457         PR testsuite/83882
14458         * function.h (gimplify_parameters): Add gimple_seq * argument.
14459         * function.c: Include gimple.h and options.h.
14460         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14461         for the added local temporaries if needed.
14462         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14463         if there are any parameter cleanups, wrap whole body into a
14464         try/finally with the cleanups.
14466 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
14468         PR target/82964
14469         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14470         Use GET_MODE_CLASS for scalar floating point.
14472 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
14474         PR ipa/82256
14475         patch by PaX Team
14476         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14477         Fix call of call_cgraph_insertion_hooks.
14479 2018-01-18  Martin Sebor  <msebor@redhat.com>
14481         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14483 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
14485         PR ipa/83619
14486         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14487         frequencies.
14489 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
14491         PR other/70268
14492         * common.opt: (-ffile-prefix-map): New option.
14493         * opts.c (common_handle_option): Defer it.
14494         * opts-global.c (handle_common_deferred_options): Handle it.
14495         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14496         * file-prefix-map.h: New file.
14497         (remap_debug_filename, add_debug_prefix_map): ...here.
14498         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14499         * final.c (debug_prefix_map, add_debug_prefix_map
14500         remap_debug_filename): Move to...
14501         * file-prefix-map.c: New file.
14502         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14503         generalize, get rid of alloca(), use strrchr() instead of strchr().
14504         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14505         Implement in terms of add_prefix_map().
14506         (remap_macro_filename, remap_debug_filename): Implement in term of
14507         remap_filename().
14508         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14509         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14510         * dbxout.c: Include file-prefix-map.h.
14511         * varasm.c: Likewise.
14512         * vmsdbgout.c: Likewise.
14513         * xcoffout.c: Likewise.
14514         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14515         * doc/cppopts.texi (-fmacro-prefix-map): Document.
14516         * doc/invoke.texi (-ffile-prefix-map): Document.
14517         (-fdebug-prefix-map): Update description.
14519 2018-01-18  Martin Liska  <mliska@suse.cz>
14521         * config/i386/i386.c (indirect_thunk_name): Document that also
14522         lfence is emitted.
14523         (output_indirect_thunk): Document why both instructions
14524         (pause and lfence) are generated.
14526 2018-01-18  Richard Biener  <rguenther@suse.de>
14528         PR tree-optimization/83887
14529         * graphite-scop-detection.c
14530         (scop_detection::get_nearest_dom_with_single_entry): Remove.
14531         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14532         (scop_detection::merge_sese): Re-implement with a flood-fill
14533         algorithm that properly finds a SESE region if it exists.
14535 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
14537         PR c/61240
14538         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14539         pointer_diff optimizations use view_convert instead of convert.
14541 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14543         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14544         Generate different code for -mno-speculate-indirect-jumps.
14545         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14546         (*call_indirect_aix<mode>): Disable for
14547         -mno-speculate-indirect-jumps.
14548         (*call_indirect_aix<mode>_nospec): New define_insn.
14549         (*call_value_indirect_aix<mode>): Disable for
14550         -mno-speculate-indirect-jumps.
14551         (*call_value_indirect_aix<mode>_nospec): New define_insn.
14552         (*sibcall_nonlocal_sysv<mode>): Generate different code for
14553         -mno-speculate-indirect-jumps.
14554         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14556 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
14558         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14559         long double type, set the flags for noting the default long double
14560         type, even if we don't pass or return a long double type.
14562 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
14564         PR ipa/83051
14565         * ipa-inline.c (flatten_function): Do not overwrite final inlining
14566         failure.
14568 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
14570         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14571         support for merge[hl].
14572         (fold_mergehl_helper): New helper function.
14573         (tree-vector-builder.h): New #include for tree_vector_builder usage.
14574         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14575         (altivec_vmrglw_direct): Add xxmrglw insn.
14577 2018-01-17  Andrew Waterman  <andrew@sifive.com>
14579         * config/riscv/riscv.c (riscv_conditional_register_usage): If
14580         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14582 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
14584         PR lto/83121
14585         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14586         call the lto_location_cache before reading the
14587         DECL_SOURCE_LOCATION of the types.
14589 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
14590             Richard Sandiford  <richard.sandiford@linaro.org>
14592         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14593         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14594         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14595         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14596         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14597         Add declaration.
14598         * config/aarch64/constraints.md (aarch64_movti_operand):
14599         Limit immediates.
14600         * config/aarch64/predicates.md (Uti): Add new constraint.
14602 2018-01-17 Carl Love  <cel@us.ibm.com>
14604         * config/rs6000/vsx.md (define_expand xl_len_r,
14605         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14606         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14607         lxvll.
14608         (define_expand, define_insn): Move the shift left from  the
14609         define_insn to the define_expand for lxvl and stxvl instructions.
14610         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14611         and XL_LEN_R definitions to PURE.
14613 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
14615         * config/i386/i386.c (indirect_thunk_name): Declare regno
14616         as unsigned int.  Compare regno with INVALID_REGNUM.
14617         (output_indirect_thunk): Ditto.
14618         (output_indirect_thunk_function): Ditto.
14619         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
14620         in the call to output_indirect_thunk_function.
14622 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
14624         PR middle-end/83884
14625         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14626         rather than the size of inner_type to determine the stack slot size
14627         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14629 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
14631         PR target/83546
14632         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14633         to PTA_SILVERMONT.
14635 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
14637         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14638         endian Linux systems to optionally enable multilibs for selecting
14639         the long double type if the user configured an explicit type.
14640         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14641         have no long double multilibs if not defined.
14642         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14643         warn if the user used -mabi={ieee,ibm}longdouble and we built
14644         multilibs for long double.
14645         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14646         appropriate multilib option.
14647         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14648         multilib options.
14649         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14650         for building long double multilibs.
14651         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14653 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
14655         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14656         copies.
14658         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14659         64 bits.
14660         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14661         128 bits.
14663         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14664         variables.
14666         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14667         return value.
14669 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
14671         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14672         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14674 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14676         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14677         different rtl trees depending on TARGET_64BIT.
14678         (rs6000_gen_lvx): Likewise.
14680 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
14682         * config/visium/visium.md (nop): Tweak comment.
14683         (hazard_nop): Likewise.
14685 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14687         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14688         -mspeculate-indirect-jumps.
14689         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14690         for -mno-speculate-indirect-jumps.
14691         (*call_indirect_elfv2<mode>_nospec): New define_insn.
14692         (*call_value_indirect_elfv2<mode>): Disable for
14693         -mno-speculate-indirect-jumps.
14694         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14695         (indirect_jump): Emit different RTL for
14696         -mno-speculate-indirect-jumps.
14697         (*indirect_jump<mode>): Disable for
14698         -mno-speculate-indirect-jumps.
14699         (*indirect_jump<mode>_nospec): New define_insn.
14700         (tablejump): Emit different RTL for
14701         -mno-speculate-indirect-jumps.
14702         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14703         (tablejumpsi_nospec): New define_expand.
14704         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14705         (tablejumpdi_nospec): New define_expand.
14706         (*tablejump<mode>_internal1): Disable for
14707         -mno-speculate-indirect-jumps.
14708         (*tablejump<mode>_internal1_nospec): New define_insn.
14709         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14710         option.
14712 2018-01-16  Artyom Skrobov tyomitch@gmail.com
14714         * caller-save.c (insert_save): Drop unnecessary parameter.  All
14715         callers updated.
14717 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
14718             Richard Biener  <rguenth@suse.de>
14720         PR libgomp/83590
14721         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14722         return early, inline manually is_gimple_sizepos.  Make sure if we
14723         call gimplify_expr we don't end up with a gimple constant.
14724         * tree.c (variably_modified_type_p): Don't return true for
14725         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
14726         * gimplify.h (is_gimple_sizepos): Remove.
14728 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
14730         PR tree-optimization/83857
14731         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14732         vectorizable_live_operation for pure SLP statements.
14733         (vectorizable_live_operation): Handle PHIs.
14735 2018-01-16  Richard Biener  <rguenther@suse.de>
14737         PR tree-optimization/83867
14738         * tree-vect-stmts.c (vect_transform_stmt): Precompute
14739         nested_in_vect_loop_p since the scalar stmt may get invalidated.
14741 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
14743         PR c/83844
14744         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14745         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14746         If off is not INTEGER_CST, issue a may not be aligned warning
14747         rather than isn't aligned.  Use isn%'t rather than isn't.
14748         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14749         into MULT_EXPR.
14750         <case MULT_EXPR>: Improve the case when bottom and one of the
14751         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14752         operand, in that case check if the other operand is multiple of
14753         bottom divided by the INTEGER_CST operand.
14755 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
14757         PR target/83858
14758         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14759         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14760         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14761         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14762         * config/pa/pa.c (pa_function_arg_advance): Likewise.
14763         (pa_function_arg, pa_arg_partial_bytes): Likewise.
14764         (pa_function_arg_size): New function.
14766 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
14768         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14769         in a separate statement.
14771 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
14773         PR tree-optimization/83847
14774         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14775         group gathers and scatters.
14777 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
14779         PR rtl-optimization/86620
14780         * params.def (max-sched-ready-insns): Bump minimum value to 1.
14782         PR rtl-optimization/83213
14783         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14784         to last if both are JUMP_INSNs.
14786         PR tree-optimization/83843
14787         * gimple-ssa-store-merging.c
14788         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14789         store_immediate_info for bswap/nop orig_stores.
14791 2018-01-15  Andrew Waterman  <andrew@sifive.com>
14793         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14794         !TARGET_MUL.
14795         <UDIV>: Increase cost if !TARGET_DIV.
14797 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
14799         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14800         (define_attr "cr_logical_3op"): New.
14801         (cceq_ior_compare): Adjust.
14802         (cceq_ior_compare_complement): Adjust.
14803         (*cceq_rev_compare): Adjust.
14804         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14805         (is_cracked_insn): Adjust.
14806         (insn_must_be_first_in_group): Adjust.
14807         * config/rs6000/40x.md: Adjust.
14808         * config/rs6000/440.md: Adjust.
14809         * config/rs6000/476.md: Adjust.
14810         * config/rs6000/601.md: Adjust.
14811         * config/rs6000/603.md: Adjust.
14812         * config/rs6000/6xx.md: Adjust.
14813         * config/rs6000/7450.md: Adjust.
14814         * config/rs6000/7xx.md: Adjust.
14815         * config/rs6000/8540.md: Adjust.
14816         * config/rs6000/cell.md: Adjust.
14817         * config/rs6000/e300c2c3.md: Adjust.
14818         * config/rs6000/e500mc.md: Adjust.
14819         * config/rs6000/e500mc64.md: Adjust.
14820         * config/rs6000/e5500.md: Adjust.
14821         * config/rs6000/e6500.md: Adjust.
14822         * config/rs6000/mpc.md: Adjust.
14823         * config/rs6000/power4.md: Adjust.
14824         * config/rs6000/power5.md: Adjust.
14825         * config/rs6000/power6.md: Adjust.
14826         * config/rs6000/power7.md: Adjust.
14827         * config/rs6000/power8.md: Adjust.
14828         * config/rs6000/power9.md: Adjust.
14829         * config/rs6000/rs64.md: Adjust.
14830         * config/rs6000/titan.md: Adjust.
14832 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
14834         * config/i386/predicates.md (indirect_branch_operand): Rewrite
14835         ix86_indirect_branch_register logic.
14837 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
14839         * config/i386/constraints.md (Bs): Update
14840         ix86_indirect_branch_register check.  Don't check
14841         ix86_indirect_branch_register with GOT_memory_operand.
14842         (Bw): Likewise.
14843         * config/i386/predicates.md (GOT_memory_operand): Don't check
14844         ix86_indirect_branch_register here.
14845         (GOT32_symbol_operand): Likewise.
14847 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
14849         * config/i386/predicates.md (constant_call_address_operand):
14850         Rewrite ix86_indirect_branch_register logic.
14851         (sibcall_insn_operand): Likewise.
14853 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
14855         * config/i386/constraints.md (Bs): Replace
14856         ix86_indirect_branch_thunk_register with
14857         ix86_indirect_branch_register.
14858         (Bw): Likewise.
14859         * config/i386/i386.md (indirect_jump): Likewise.
14860         (tablejump): Likewise.
14861         (*sibcall_memory): Likewise.
14862         (*sibcall_value_memory): Likewise.
14863         Peepholes of indirect call and jump via memory: Likewise.
14864         * config/i386/i386.opt: Likewise.
14865         * config/i386/predicates.md (indirect_branch_operand): Likewise.
14866         (GOT_memory_operand): Likewise.
14867         (call_insn_operand): Likewise.
14868         (sibcall_insn_operand): Likewise.
14869         (GOT32_symbol_operand): Likewise.
14871 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
14873         PR middle-end/83837
14874         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14875         type rather than type addr's type points to.
14876         (expand_omp_atomic_mutex): Likewise.
14877         (expand_omp_atomic): Likewise.
14879 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
14881         PR target/83839
14882         * config/i386/i386.c (output_indirect_thunk_function): Use
14883         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14884         for  __x86_return_thunk.
14886 2018-01-15  Richard Biener  <rguenther@suse.de>
14888         PR middle-end/83850
14889         * expmed.c (extract_bit_field_1): Fix typo.
14891 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14893         PR target/83687
14894         * config/arm/iterators.md (VF): New mode iterator.
14895         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14896         Remove integer-related logic from pattern.
14897         (neon_vabd<mode>_3): Likewise.
14899 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
14901         PR middle-end/82694
14902         * common.opt (fstrict-overflow): No longer an alias.
14903         (fwrapv-pointer): New option.
14904         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14905         also for pointer types based on flag_wrapv_pointer.
14906         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14907         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14908         opts->x_flag_wrapv got set.
14909         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14910         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14911         POINTER_TYPE_OVERFLOW_UNDEFINED.
14912         * match.pd: Likewise in address comparison pattern.
14913         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14915 2018-01-15  Richard Biener  <rguenther@suse.de>
14917         PR lto/83804
14918         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14919         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
14920         Reset type names to their identifier if their TYPE_DECL doesn't
14921         have linkage (and thus is used for ODR and devirt).
14922         (save_debug_info_for_decl): Remove.
14923         (save_debug_info_for_type): Likewise.
14924         (add_tree_to_fld_list): Adjust.
14925         * tree-pretty-print.c (dump_generic_node): Make dumping of
14926         type names more robust.
14928 2018-01-15  Richard Biener  <rguenther@suse.de>
14930         * BASE-VER: Bump to 8.0.1.
14932 2018-01-14  Martin Sebor  <msebor@redhat.com>
14934         PR other/83508
14935         * builtins.c (check_access): Avoid warning when the no-warning bit
14936         is set.
14938 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
14940         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14941         * ira-color (allocno_hard_regs_compare): Likewise.
14943 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
14945         PR target/83013
14946         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14947         Use .pushsection/.popsection.
14949 2018-01-14  Martin Sebor  <msebor@redhat.com>
14951         PR c++/81327
14952         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14954 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
14956         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14957         entry from extra_headers.
14958         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14959         extra_headers, make the list bitwise identical to the i?86-*-* one.
14961 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
14963         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14964         -mcmodel=large with -mindirect-branch=thunk,
14965         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14966         -mfunction-return=thunk-extern.
14967         * doc/invoke.texi: Document -mcmodel=large is incompatible with
14968         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14969         -mfunction-return=thunk and -mfunction-return=thunk-extern.
14971 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
14973         * config/i386/i386.c (print_reg): Print the name of the full
14974         integer register without '%'.
14975         (ix86_print_operand): Handle 'V'.
14976         * doc/extend.texi: Document 'V' modifier.
14978 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
14980         * config/i386/constraints.md (Bs): Disallow memory operand for
14981         -mindirect-branch-register.
14982         (Bw): Likewise.
14983         * config/i386/predicates.md (indirect_branch_operand): Likewise.
14984         (GOT_memory_operand): Likewise.
14985         (call_insn_operand): Likewise.
14986         (sibcall_insn_operand): Likewise.
14987         (GOT32_symbol_operand): Likewise.
14988         * config/i386/i386.md (indirect_jump): Call convert_memory_address
14989         for -mindirect-branch-register.
14990         (tablejump): Likewise.
14991         (*sibcall_memory): Likewise.
14992         (*sibcall_value_memory): Likewise.
14993         Disallow peepholes of indirect call and jump via memory for
14994         -mindirect-branch-register.
14995         (*call_pop): Replace m with Bw.
14996         (*call_value_pop): Likewise.
14997         (*sibcall_pop_memory): Replace m with Bs.
14998         * config/i386/i386.opt (mindirect-branch-register): New option.
14999         * doc/invoke.texi: Document -mindirect-branch-register option.
15001 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
15003         * config/i386/i386-protos.h (ix86_output_function_return): New.
15004         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15005         set function_return_type.
15006         (indirect_thunk_name): Add ret_p to indicate thunk for function
15007         return.
15008         (output_indirect_thunk_function): Pass false to
15009         indirect_thunk_name.
15010         (ix86_output_indirect_branch_via_reg): Likewise.
15011         (ix86_output_indirect_branch_via_push): Likewise.
15012         (output_indirect_thunk_function): Create alias for function
15013         return thunk if regno < 0.
15014         (ix86_output_function_return): New function.
15015         (ix86_handle_fndecl_attribute): Handle function_return.
15016         (ix86_attribute_table): Add function_return.
15017         * config/i386/i386.h (machine_function): Add
15018         function_return_type.
15019         * config/i386/i386.md (simple_return_internal): Use
15020         ix86_output_function_return.
15021         (simple_return_internal_long): Likewise.
15022         * config/i386/i386.opt (mfunction-return=): New option.
15023         (indirect_branch): Mention -mfunction-return=.
15024         * doc/extend.texi: Document function_return function attribute.
15025         * doc/invoke.texi: Document -mfunction-return= option.
15027 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
15029         * config/i386/i386-opts.h (indirect_branch): New.
15030         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15031         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15032         with local indirect jump when converting indirect call and jump.
15033         (ix86_set_indirect_branch_type): New.
15034         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15035         (indirectlabelno): New.
15036         (indirect_thunk_needed): Likewise.
15037         (indirect_thunk_bnd_needed): Likewise.
15038         (indirect_thunks_used): Likewise.
15039         (indirect_thunks_bnd_used): Likewise.
15040         (INDIRECT_LABEL): Likewise.
15041         (indirect_thunk_name): Likewise.
15042         (output_indirect_thunk): Likewise.
15043         (output_indirect_thunk_function): Likewise.
15044         (ix86_output_indirect_branch_via_reg): Likewise.
15045         (ix86_output_indirect_branch_via_push): Likewise.
15046         (ix86_output_indirect_branch): Likewise.
15047         (ix86_output_indirect_jmp): Likewise.
15048         (ix86_code_end): Call output_indirect_thunk_function if needed.
15049         (ix86_output_call_insn): Call ix86_output_indirect_branch if
15050         needed.
15051         (ix86_handle_fndecl_attribute): Handle indirect_branch.
15052         (ix86_attribute_table): Add indirect_branch.
15053         * config/i386/i386.h (machine_function): Add indirect_branch_type
15054         and has_local_indirect_jump.
15055         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15056         to true.
15057         (tablejump): Likewise.
15058         (*indirect_jump): Use ix86_output_indirect_jmp.
15059         (*tablejump_1): Likewise.
15060         (simple_return_indirect_internal): Likewise.
15061         * config/i386/i386.opt (mindirect-branch=): New option.
15062         (indirect_branch): New.
15063         (keep): Likewise.
15064         (thunk): Likewise.
15065         (thunk-inline): Likewise.
15066         (thunk-extern): Likewise.
15067         * doc/extend.texi: Document indirect_branch function attribute.
15068         * doc/invoke.texi: Document -mindirect-branch= option.
15070 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
15072         PR ipa/83051
15073         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15075 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
15077         * ipa-inline.c (want_inline_small_function_p): Return false if
15078         inlining has already failed with CIF_FINAL_ERROR.
15079         (update_caller_keys): Call want_inline_small_function_p before
15080         can_inline_edge_p.
15081         (update_callee_keys): Likewise.
15083 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15085         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15086         New function.
15087         (rs6000_quadword_masked_address_p): Likewise.
15088         (quad_aligned_load_p): Likewise.
15089         (quad_aligned_store_p): Likewise.
15090         (const_load_sequence_p): Add comment to describe the outer-most loop.
15091         (mimic_memory_attributes_and_flags): New function.
15092         (rs6000_gen_stvx): Likewise.
15093         (replace_swapped_aligned_store): Likewise.
15094         (rs6000_gen_lvx): Likewise.
15095         (replace_swapped_aligned_load): Likewise.
15096         (replace_swapped_load_constant): Capitalize argument name in
15097         comment describing this function.
15098         (rs6000_analyze_swaps): Add a third pass to search for vector loads
15099         and stores that access quad-word aligned addresses and replace
15100         with stvx or lvx instructions when appropriate.
15101         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15102         New function prototype.
15103         (rs6000_quadword_masked_address_p): Likewise.
15104         (rs6000_gen_lvx): Likewise.
15105         (rs6000_gen_stvx): Likewise.
15106         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15107         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15108         when memory address is aligned.
15109         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15110         this split to select lvx instruction when memory address is aligned.
15111         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15112         instruction when memory address is aligned.
15113         (*vsx_le_perm_load_v16qi): Likewise.
15114         (four unnamed splitters): Modify to select the stvx instruction
15115         when memory is aligned.
15117 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
15119         * predict.c (determine_unlikely_bbs): Handle correctly BBs
15120         which appears in the queue multiple times.
15122 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15123             Alan Hayward  <alan.hayward@arm.com>
15124             David Sherwood  <david.sherwood@arm.com>
15126         * tree-vectorizer.h (vec_lower_bound): New structure.
15127         (_loop_vec_info): Add check_nonzero and lower_bounds.
15128         (LOOP_VINFO_CHECK_NONZERO): New macro.
15129         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15130         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15131         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15132         fields.  Make seg_len the distance travelled, not including the
15133         access size.
15134         (dr_direction_indicator): Declare.
15135         (dr_zero_step_indicator): Likewise.
15136         (dr_known_forward_stride_p): Likewise.
15137         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15138         tree-ssanames.h.
15139         (runtime_alias_check_p): Allow runtime alias checks with
15140         variable strides.
15141         (operator ==): Compare access_size and align.
15142         (prune_runtime_alias_test_list): Rework for new distinction between
15143         the access_size and seg_len.
15144         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
15145         segment lengths.
15146         (get_segment_min_max): New function.
15147         (create_intersect_range_checks): Use it.
15148         (dr_step_indicator): New function.
15149         (dr_direction_indicator): Likewise.
15150         (dr_zero_step_indicator): Likewise.
15151         (dr_known_forward_stride_p): Likewise.
15152         * tree-loop-distribution.c (data_ref_segment_size): Return
15153         DR_STEP * (niters - 1).
15154         (compute_alias_check_pairs): Update call to the dr_with_seg_len
15155         constructor.
15156         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15157         (vect_preserves_scalar_order_p): New function, split out from...
15158         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
15159         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15160         (vect_vfa_access_size): New function.
15161         (vect_vfa_align): Likewise.
15162         (vect_compile_time_alias): Take access_size_a and access_b arguments.
15163         (dump_lower_bound): New function.
15164         (vect_check_lower_bound): Likewise.
15165         (vect_small_gap_p): Likewise.
15166         (vectorizable_with_step_bound_p): Likewise.
15167         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15168         depencies if the vectorization factor is 1.  Convert the checks
15169         for nonzero steps into checks on the bounds of DR_STEP.  Try using
15170         a bunds check for variable steps if the minimum required step is
15171         relatively small. Update calls to the dr_with_seg_len
15172         constructor and to vect_compile_time_alias.
15173         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15174         function.
15175         (vect_loop_versioning): Call it.
15176         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15177         when retrying.
15178         (vect_estimate_min_profitable_iters): Account for any bounds checks.
15180 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15181             Alan Hayward  <alan.hayward@arm.com>
15182             David Sherwood  <david.sherwood@arm.com>
15184         * doc/sourcebuild.texi (vect_scatter_store): Document.
15185         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15186         optabs.
15187         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15188         Document.
15189         * genopinit.c (main): Add supports_vec_scatter_store and
15190         supports_vec_scatter_store_cached to target_optabs.
15191         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15192         IFN_MASK_SCATTER_STORE.
15193         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15194         functions.
15195         * internal-fn.h (internal_store_fn_p): Declare.
15196         (internal_fn_stored_value_index): Likewise.
15197         * internal-fn.c (scatter_store_direct): New macro.
15198         (expand_scatter_store_optab_fn): New function.
15199         (direct_scatter_store_optab_supported_p): New macro.
15200         (internal_store_fn_p): New function.
15201         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15202         IFN_MASK_SCATTER_STORE.
15203         (internal_fn_mask_index): Likewise.
15204         (internal_fn_stored_value_index): New function.
15205         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15206         for scatter stores.
15207         * optabs-query.h (supports_vec_scatter_store_p): Declare.
15208         * optabs-query.c (supports_vec_scatter_store_p): New function.
15209         * tree-vectorizer.h (vect_get_store_rhs): Declare.
15210         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15211         true for scatter stores.
15212         (vect_gather_scatter_fn_p): Handle scatter stores too.
15213         (vect_check_gather_scatter): Consider using scatter stores if
15214         supports_vec_scatter_store_p.
15215         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15216         scatter stores too.
15217         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15218         internal_fn_stored_value_index.
15219         (check_load_store_masking): Handle scatter stores too.
15220         (vect_get_store_rhs): Make public.
15221         (vectorizable_call): Use internal_store_fn_p.
15222         (vectorizable_store): Handle scatter store internal functions.
15223         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15224         when deciding whether the end of the group has been reached.
15225         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15226         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15227         (mask_scatter_store<mode>): New insns.
15229 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15230             Alan Hayward  <alan.hayward@arm.com>
15231             David Sherwood  <david.sherwood@arm.com>
15233         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15234         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15235         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15236         function.
15237         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15238         Use vect_truncate_gather_scatter_offset if we can't treat the
15239         operation as a normal gather load or scatter store.
15240         (get_group_load_store_type): Take the gather_scatter_info
15241         as argument.  Try using a gather load or scatter store for
15242         single-element groups.
15243         (get_load_store_type): Update calls to get_group_load_store_type
15244         and vect_use_strided_gather_scatters_p.
15246 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15247             Alan Hayward  <alan.hayward@arm.com>
15248             David Sherwood  <david.sherwood@arm.com>
15250         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15251         optional tree argument.
15252         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15253         null target hooks.
15254         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15255         but continue to use the current value as a fallback.
15256         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15257         to compare the updates.
15258         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15259         (get_load_store_type): Use it when handling a strided access.
15260         (vect_get_strided_load_store_ops): New function.
15261         (vect_get_data_ptr_increment): Likewise.
15262         (vectorizable_load): Handle strided gather loads.  Always pass
15263         a step to vect_create_data_ref_ptr and bump_vector_ptr.
15265 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15266             Alan Hayward  <alan.hayward@arm.com>
15267             David Sherwood  <david.sherwood@arm.com>
15269         * doc/md.texi (gather_load@var{m}): Document.
15270         (mask_gather_load@var{m}): Likewise.
15271         * genopinit.c (main): Add supports_vec_gather_load and
15272         supports_vec_gather_load_cached to target_optabs.
15273         * optabs-tree.c (init_tree_optimization_optabs): Use
15274         ggc_cleared_alloc to allocate target_optabs.
15275         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15276         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15277         functions.
15278         * internal-fn.h (internal_load_fn_p): Declare.
15279         (internal_gather_scatter_fn_p): Likewise.
15280         (internal_fn_mask_index): Likewise.
15281         (internal_gather_scatter_fn_supported_p): Likewise.
15282         * internal-fn.c (gather_load_direct): New macro.
15283         (expand_gather_load_optab_fn): New function.
15284         (direct_gather_load_optab_supported_p): New macro.
15285         (direct_internal_fn_optab): New function.
15286         (internal_load_fn_p): Likewise.
15287         (internal_gather_scatter_fn_p): Likewise.
15288         (internal_fn_mask_index): Likewise.
15289         (internal_gather_scatter_fn_supported_p): Likewise.
15290         * optabs-query.c (supports_at_least_one_mode_p): New function.
15291         (supports_vec_gather_load_p): Likewise.
15292         * optabs-query.h (supports_vec_gather_load_p): Declare.
15293         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15294         and memory_type field.
15295         (NUM_PATTERNS): Bump to 15.
15296         * tree-vect-data-refs.c: Include internal-fn.h.
15297         (vect_gather_scatter_fn_p): New function.
15298         (vect_describe_gather_scatter_call): Likewise.
15299         (vect_check_gather_scatter): Try using internal functions for
15300         gather loads.  Recognize existing calls to a gather load function.
15301         (vect_analyze_data_refs): Consider using gather loads if
15302         supports_vec_gather_load_p.
15303         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15304         (vect_get_gather_scatter_offset_type): Likewise.
15305         (vect_convert_mask_for_vectype): Likewise.
15306         (vect_add_conversion_to_patterm): Likewise.
15307         (vect_try_gather_scatter_pattern): Likewise.
15308         (vect_recog_gather_scatter_pattern): New pattern recognizer.
15309         (vect_vect_recog_func_ptrs): Add it.
15310         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15311         internal_fn_mask_index and internal_gather_scatter_fn_p.
15312         (check_load_store_masking): Take the gather_scatter_info as an
15313         argument and handle gather loads.
15314         (vect_get_gather_scatter_ops): New function.
15315         (vectorizable_call): Check internal_load_fn_p.
15316         (vectorizable_load): Likewise.  Handle gather load internal
15317         functions.
15318         (vectorizable_store): Update call to check_load_store_masking.
15319         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15320         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15321         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15322         (aarch64_gather_scale_operand_d): New predicates.
15323         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15324         (mask_gather_load<mode>): New insns.
15326 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15327             Alan Hayward  <alan.hayward@arm.com>
15328             David Sherwood  <david.sherwood@arm.com>
15330         * optabs.def (fold_left_plus_optab): New optab.
15331         * doc/md.texi (fold_left_plus_@var{m}): Document.
15332         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15333         * internal-fn.c (fold_left_direct): Define.
15334         (expand_fold_left_optab_fn): Likewise.
15335         (direct_fold_left_optab_supported_p): Likewise.
15336         * fold-const-call.c (fold_const_fold_left): New function.
15337         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15338         * tree-parloops.c (valid_reduction_p): New function.
15339         (gather_scalar_reductions): Use it.
15340         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15341         (vect_finish_replace_stmt): Declare.
15342         * tree-vect-loop.c (fold_left_reduction_fn): New function.
15343         (needs_fold_left_reduction_p): New function, split out from...
15344         (vect_is_simple_reduction): ...here.  Accept reductions that
15345         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15346         (vect_force_simple_reduction): Also store the reduction type in
15347         the assignment's STMT_VINFO_REDUC_TYPE.
15348         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15349         (merge_with_identity): New function.
15350         (vect_expand_fold_left): Likewise.
15351         (vectorize_fold_left_reduction): Likewise.
15352         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
15353         scalar phi in place for it.  Check for target support and reject
15354         cases that would reassociate the operation.  Defer the transform
15355         phase to vectorize_fold_left_reduction.
15356         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15357         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15358         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15360 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15362         * tree-if-conv.c (predicate_mem_writes): Remove redundant
15363         call to ifc_temp_var.
15365 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15366             Alan Hayward  <alan.hayward@arm.com>
15367             David Sherwood  <david.sherwood@arm.com>
15369         * target.def (legitimize_address_displacement): Take the original
15370         offset as a poly_int.
15371         * targhooks.h (default_legitimize_address_displacement): Update
15372         accordingly.
15373         * targhooks.c (default_legitimize_address_displacement): Likewise.
15374         * doc/tm.texi: Regenerate.
15375         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15376         as an argument, moving assert of ad->disp == ad->disp_term to...
15377         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
15378         Try calling targetm.legitimize_address_displacement before expanding
15379         the address rather than afterwards, and adjust for the new interface.
15380         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15381         Match the new hook interface.  Handle SVE addresses.
15382         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15383         new hook interface.
15385 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15387         * Makefile.in (OBJS): Add early-remat.o.
15388         * target.def (select_early_remat_modes): New hook.
15389         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15390         * doc/tm.texi: Regenerate.
15391         * targhooks.h (default_select_early_remat_modes): Declare.
15392         * targhooks.c (default_select_early_remat_modes): New function.
15393         * timevar.def (TV_EARLY_REMAT): New timevar.
15394         * passes.def (pass_early_remat): New pass.
15395         * tree-pass.h (make_pass_early_remat): Declare.
15396         * early-remat.c: New file.
15397         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15398         function.
15399         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15401 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15402             Alan Hayward  <alan.hayward@arm.com>
15403             David Sherwood  <david.sherwood@arm.com>
15405         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15406         vfm1 with a bound_epilog parameter.
15407         (vect_do_peeling): Update calls accordingly, and move the prologue
15408         call earlier in the function.  Treat the base bound_epilog as 0 for
15409         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
15410         this base when peeling for gaps.
15411         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15412         with fully-masked loops.
15413         (vect_estimate_min_profitable_iters): Handle the single peeled
15414         iteration in that case.
15416 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15417             Alan Hayward  <alan.hayward@arm.com>
15418             David Sherwood  <david.sherwood@arm.com>
15420         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15421         single-element interleaving even if the size is not a power of 2.
15422         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15423         accesses for single-element interleaving if the group size is
15424         not a power of 2.
15426 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15427             Alan Hayward  <alan.hayward@arm.com>
15428             David Sherwood  <david.sherwood@arm.com>
15430         * doc/md.texi (fold_extract_last_@var{m}): Document.
15431         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15432         * optabs.def (fold_extract_last_optab): New optab.
15433         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15434         * internal-fn.c (fold_extract_direct): New macro.
15435         (expand_fold_extract_optab_fn): Likewise.
15436         (direct_fold_extract_optab_supported_p): Likewise.
15437         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15438         * tree-vect-loop.c (vect_model_reduction_cost): Handle
15439         EXTRACT_LAST_REDUCTION.
15440         (get_initial_def_for_reduction): Do not create an initial vector
15441         for EXTRACT_LAST_REDUCTION reductions.
15442         (vectorizable_reduction): Leave the scalar phi in place for
15443         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
15444         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
15445         epilogue code for EXTRACT_LAST_REDUCTION and defer the
15446         transform phase to vectorizable_condition.
15447         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15448         split out from...
15449         (vect_finish_stmt_generation): ...here.
15450         (vect_finish_replace_stmt): New function.
15451         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15452         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15453         pattern.
15454         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15456 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15457             Alan Hayward  <alan.hayward@arm.com>
15458             David Sherwood  <david.sherwood@arm.com>
15460         * doc/md.texi (extract_last_@var{m}): Document.
15461         * optabs.def (extract_last_optab): New optab.
15462         * internal-fn.def (EXTRACT_LAST): New internal function.
15463         * internal-fn.c (cond_unary_direct): New macro.
15464         (expand_cond_unary_optab_fn): Likewise.
15465         (direct_cond_unary_optab_supported_p): Likewise.
15466         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15467         loops using EXTRACT_LAST.
15468         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15469         (extract_last_<mode>): ...this optab.
15470         (vec_extract<mode><Vel>): Update accordingly.
15472 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15473             Alan Hayward  <alan.hayward@arm.com>
15474             David Sherwood  <david.sherwood@arm.com>
15476         * target.def (empty_mask_is_expensive): New hook.
15477         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15478         * doc/tm.texi: Regenerate.
15479         * targhooks.h (default_empty_mask_is_expensive): Declare.
15480         * targhooks.c (default_empty_mask_is_expensive): New function.
15481         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15482         if the target says that empty masks are expensive.
15483         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15484         New function.
15485         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15487 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15488             Alan Hayward  <alan.hayward@arm.com>
15489             David Sherwood  <david.sherwood@arm.com>
15491         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15492         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15493         (vect_use_loop_mask_for_alignment_p): New function.
15494         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15495         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15496         niters_skip argument.  Make sure that the first niters_skip elements
15497         of the first iteration are inactive.
15498         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15499         Update call to vect_set_loop_masks_directly.
15500         (get_misalign_in_elems): New function, split out from...
15501         (vect_gen_prolog_loop_niters): ...here.
15502         (vect_update_init_of_dr): Take a code argument that specifies whether
15503         the adjustment should be added or subtracted.
15504         (vect_update_init_of_drs): Likewise.
15505         (vect_prepare_for_masked_peels): New function.
15506         (vect_do_peeling): Skip prologue peeling if we're using a mask
15507         instead.  Update call to vect_update_inits_of_drs.
15508         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15509         mask_skip_niters.
15510         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15511         alignment.  Do not include the number of peeled iterations in
15512         the minimum threshold in that case.
15513         (vectorizable_induction): Adjust the start value down by
15514         LOOP_VINFO_MASK_SKIP_NITERS iterations.
15515         (vect_transform_loop): Call vect_prepare_for_masked_peels.
15516         Take the number of skipped iterations into account when calculating
15517         the loop bounds.
15518         * tree-vect-stmts.c (vect_gen_while_not): New function.
15520 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15521             Alan Hayward  <alan.hayward@arm.com>
15522             David Sherwood  <david.sherwood@arm.com>
15524         * doc/sourcebuild.texi (vect_fully_masked): Document.
15525         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15526         default value to 0.
15527         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15528         split out from...
15529         (vect_analyze_loop_2): ...here. Don't check the vectorization
15530         factor against the number of loop iterations if the loop is
15531         fully-masked.
15533 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15534             Alan Hayward  <alan.hayward@arm.com>
15535             David Sherwood  <david.sherwood@arm.com>
15537         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15538         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15539         (dump_groups): Update accordingly.
15540         (iv_use::mem_type): New member variable.
15541         (address_p): New function.
15542         (record_use): Add a mem_type argument and initialize the new
15543         mem_type field.
15544         (record_group_use): Add a mem_type argument.  Use address_p.
15545         Remove obsolete null checks of base_object.  Update call to record_use.
15546         (find_interesting_uses_op): Update call to record_group_use.
15547         (find_interesting_uses_cond): Likewise.
15548         (find_interesting_uses_address): Likewise.
15549         (get_mem_type_for_internal_fn): New function.
15550         (find_address_like_use): Likewise.
15551         (find_interesting_uses_stmt): Try find_address_like_use before
15552         calling find_interesting_uses_op.
15553         (addr_offset_valid_p): Use the iv mem_type field as the type
15554         of the addressed memory.
15555         (add_autoinc_candidates): Likewise.
15556         (get_address_cost): Likewise.
15557         (split_small_address_groups_p): Use address_p.
15558         (split_address_groups): Likewise.
15559         (add_iv_candidate_for_use): Likewise.
15560         (autoinc_possible_for_pair): Likewise.
15561         (rewrite_groups): Likewise.
15562         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15563         (determine_group_iv_cost): Update after split of USE_ADDRESS.
15564         (get_alias_ptr_type_for_ptr_address): New function.
15565         (rewrite_use_address): Rewrite address uses in calls that were
15566         identified by find_address_like_use.
15568 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15569             Alan Hayward  <alan.hayward@arm.com>
15570             David Sherwood  <david.sherwood@arm.com>
15572         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15573         TARGET_MEM_REFs.
15574         * gimple-expr.h (is_gimple_addressable: Likewise.
15575         * gimple-expr.c (is_gimple_address): Likewise.
15576         * internal-fn.c (expand_call_mem_ref): New function.
15577         (expand_mask_load_optab_fn): Use it.
15578         (expand_mask_store_optab_fn): Likewise.
15580 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15581             Alan Hayward  <alan.hayward@arm.com>
15582             David Sherwood  <david.sherwood@arm.com>
15584         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15585         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15586         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15587         (cond_umax@var{mode}): Document.
15588         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15589         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15590         (cond_umin_optab, cond_umax_optab): New optabs.
15591         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15592         (COND_IOR, COND_XOR): New internal functions.
15593         * internal-fn.h (get_conditional_internal_fn): Declare.
15594         * internal-fn.c (cond_binary_direct): New macro.
15595         (expand_cond_binary_optab_fn): Likewise.
15596         (direct_cond_binary_optab_supported_p): Likewise.
15597         (get_conditional_internal_fn): New function.
15598         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15599         Cope with reduction statements that are vectorized as calls rather
15600         than assignments.
15601         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15602         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15603         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15604         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15605         (UNSPEC_COND_EOR): New unspecs.
15606         (optab): Add mappings for them.
15607         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15608         (sve_int_op, sve_fp_op): New int attributes.
15610 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15611             Alan Hayward  <alan.hayward@arm.com>
15612             David Sherwood  <david.sherwood@arm.com>
15614         * optabs.def (while_ult_optab): New optab.
15615         * doc/md.texi (while_ult@var{m}@var{n}): Document.
15616         * internal-fn.def (WHILE_ULT): New internal function.
15617         * internal-fn.h (direct_internal_fn_supported_p): New override
15618         that takes two types as argument.
15619         * internal-fn.c (while_direct): New macro.
15620         (expand_while_optab_fn): New function.
15621         (convert_optab_supported_p): Likewise.
15622         (direct_while_optab_supported_p): New macro.
15623         * wide-int.h (wi::udiv_ceil): New function.
15624         * tree-vectorizer.h (rgroup_masks): New structure.
15625         (vec_loop_masks): New typedef.
15626         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15627         and fully_masked_p.
15628         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15629         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15630         (vect_max_vf): New function.
15631         (slpeel_make_loop_iterate_ntimes): Delete.
15632         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15633         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15634         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15635         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15636         internal-fn.h, stor-layout.h and optabs-query.h.
15637         (vect_set_loop_mask): New function.
15638         (add_preheader_seq): Likewise.
15639         (add_header_seq): Likewise.
15640         (interleave_supported_p): Likewise.
15641         (vect_maybe_permute_loop_masks): Likewise.
15642         (vect_set_loop_masks_directly): Likewise.
15643         (vect_set_loop_condition_masked): Likewise.
15644         (vect_set_loop_condition_unmasked): New function, split out from
15645         slpeel_make_loop_iterate_ntimes.
15646         (slpeel_make_loop_iterate_ntimes): Rename to..
15647         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
15648         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15649         (vect_do_peeling): Update call accordingly.
15650         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15651         loops.
15652         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15653         mask_compare_type, can_fully_mask_p and fully_masked_p.
15654         (release_vec_loop_masks): New function.
15655         (_loop_vec_info): Use it to free the loop masks.
15656         (can_produce_all_loop_masks_p): New function.
15657         (vect_get_max_nscalars_per_iter): Likewise.
15658         (vect_verify_full_masking): Likewise.
15659         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15660         retries, and free the mask rgroups before retrying.  Check loop-wide
15661         reasons for disallowing fully-masked loops.  Make the final decision
15662         about whether use a fully-masked loop or not.
15663         (vect_estimate_min_profitable_iters): Do not assume that peeling
15664         for the number of iterations will be needed for fully-masked loops.
15665         (vectorizable_reduction): Disable fully-masked loops.
15666         (vectorizable_live_operation): Likewise.
15667         (vect_halve_mask_nunits): New function.
15668         (vect_double_mask_nunits): Likewise.
15669         (vect_record_loop_mask): Likewise.
15670         (vect_get_loop_mask): Likewise.
15671         (vect_transform_loop): Handle the case in which the final loop
15672         iteration might handle a partial vector.  Call vect_set_loop_condition
15673         instead of slpeel_make_loop_iterate_ntimes.
15674         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15675         (check_load_store_masking): New function.
15676         (prepare_load_store_mask): Likewise.
15677         (vectorizable_store): Handle fully-masked loops.
15678         (vectorizable_load): Likewise.
15679         (supportable_widening_operation): Use vect_halve_mask_nunits for
15680         booleans.
15681         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15682         (vect_gen_while): New function.
15683         * config/aarch64/aarch64.md (umax<mode>3): New expander.
15684         (aarch64_uqdec<mode>): New insn.
15686 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15687             Alan Hayward  <alan.hayward@arm.com>
15688             David Sherwood  <david.sherwood@arm.com>
15690         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15691         (reduc_xor_scal_optab): New optabs.
15692         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15693         (reduc_xor_scal_@var{m}): Document.
15694         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15695         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15696         internal functions.
15697         * fold-const-call.c (fold_const_call): Handle them.
15698         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15699         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15700         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15701         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15702         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15703         (UNSPEC_XORV): New unspecs.
15704         (optab): Add entries for them.
15705         (BITWISEV): New int iterator.
15706         (bit_reduc_op): New int attributes.
15708 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15709             Alan Hayward  <alan.hayward@arm.com>
15710             David Sherwood  <david.sherwood@arm.com>
15712         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15713         * internal-fn.def (VEC_SHL_INSERT): New internal function.
15714         * optabs.def (vec_shl_insert_optab): New optab.
15715         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15716         (duplicate_and_interleave): Likewise.
15717         * tree-vect-loop.c: Include internal-fn.h.
15718         (neutral_op_for_slp_reduction): New function, split out from
15719         get_initial_defs_for_reduction.
15720         (get_initial_def_for_reduction): Handle option 2 for variable-length
15721         vectors by loading the neutral value into a vector and then shifting
15722         the initial value into element 0.
15723         (get_initial_defs_for_reduction): Replace the code argument with
15724         the neutral value calculated by neutral_op_for_slp_reduction.
15725         Use gimple_build_vector for constant-length vectors.
15726         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15727         but the first group_size elements have a neutral value.
15728         Use duplicate_and_interleave otherwise.
15729         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15730         Update call to get_initial_defs_for_reduction.  Handle SLP
15731         reductions for variable-length vectors by creating one vector
15732         result for each scalar result, with the elements associated
15733         with other scalar results stubbed out with the neutral value.
15734         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15735         Require IFN_VEC_SHL_INSERT for double reductions on
15736         variable-length vectors, or SLP reductions that have
15737         a neutral value.  Require can_duplicate_and_interleave_p
15738         support for variable-length unchained SLP reductions if there
15739         is no neutral value, such as for MIN/MAX reductions.  Also require
15740         the number of vector elements to be a multiple of the number of
15741         SLP statements when doing variable-length unchained SLP reductions.
15742         Update call to vect_create_epilog_for_reduction.
15743         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15744         and remove initial values.
15745         (duplicate_and_interleave): Make public.
15746         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15747         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15749 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15750             Alan Hayward  <alan.hayward@arm.com>
15751             David Sherwood  <david.sherwood@arm.com>
15753         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15754         (can_duplicate_and_interleave_p): New function.
15755         (vect_get_and_check_slp_defs): Take the vector of statements
15756         rather than just the current one.  Remove excess parentheses.
15757         Restriction rejectinon of vect_constant_def and vect_external_def
15758         for variable-length vectors to boolean types, or types for which
15759         can_duplicate_and_interleave_p is false.
15760         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15761         (duplicate_and_interleave): New function.
15762         (vect_get_constant_vectors): Use gimple_build_vector for
15763         constant-length vectors and suitable variable-length constant
15764         vectors.  Use duplicate_and_interleave for other variable-length
15765         vectors.  Don't defer the update when inserting new statements.
15767 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15768             Alan Hayward  <alan.hayward@arm.com>
15769             David Sherwood  <david.sherwood@arm.com>
15771         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15772         min_profitable_iters doesn't go negative.
15774 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15775             Alan Hayward  <alan.hayward@arm.com>
15776             David Sherwood  <david.sherwood@arm.com>
15778         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15779         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15780         * optabs.def (vec_mask_load_lanes_optab): New optab.
15781         (vec_mask_store_lanes_optab): Likewise.
15782         * internal-fn.def (MASK_LOAD_LANES): New internal function.
15783         (MASK_STORE_LANES): Likewise.
15784         * internal-fn.c (mask_load_lanes_direct): New macro.
15785         (mask_store_lanes_direct): Likewise.
15786         (expand_mask_load_optab_fn): Handle masked operations.
15787         (expand_mask_load_lanes_optab_fn): New macro.
15788         (expand_mask_store_optab_fn): Handle masked operations.
15789         (expand_mask_store_lanes_optab_fn): New macro.
15790         (direct_mask_load_lanes_optab_supported_p): Likewise.
15791         (direct_mask_store_lanes_optab_supported_p): Likewise.
15792         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15793         parameter.
15794         (vect_load_lanes_supported): Likewise.
15795         * tree-vect-data-refs.c (strip_conversion): New function.
15796         (can_group_stmts_p): Likewise.
15797         (vect_analyze_data_ref_accesses): Use it instead of checking
15798         for a pair of assignments.
15799         (vect_store_lanes_supported): Take a masked_p parameter.
15800         (vect_load_lanes_supported): Likewise.
15801         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15802         vect_store_lanes_supported and vect_load_lanes_supported.
15803         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15804         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15805         parameter.  Don't allow gaps for masked accesses.
15806         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
15807         and vect_load_lanes_supported.
15808         (get_load_store_type): Take a masked_p parameter and update
15809         call to get_group_load_store_type.
15810         (vectorizable_store): Update call to get_load_store_type.
15811         Handle IFN_MASK_STORE_LANES.
15812         (vectorizable_load): Update call to get_load_store_type.
15813         Handle IFN_MASK_LOAD_LANES.
15815 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15816             Alan Hayward  <alan.hayward@arm.com>
15817             David Sherwood  <david.sherwood@arm.com>
15819         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15820         modes for SVE.
15821         * config/aarch64/aarch64-protos.h
15822         (aarch64_sve_struct_memory_operand_p): Declare.
15823         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15824         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15825         (VPRED, vpred): Handle SVE structure modes.
15826         * config/aarch64/constraints.md (Utx): New constraint.
15827         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15828         (aarch64_sve_struct_nonimmediate_operand): New predicates.
15829         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15830         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15831         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15832         structure modes.  Split into pieces after RA.
15833         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15834         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15835         New patterns.
15836         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15837         SVE structure modes.
15838         (aarch64_classify_address): Likewise.
15839         (sizetochar): Move earlier in file.
15840         (aarch64_print_operand): Handle SVE register lists.
15841         (aarch64_array_mode): New function.
15842         (aarch64_sve_struct_memory_operand_p): Likewise.
15843         (TARGET_ARRAY_MODE): Redefine.
15845 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15846             Alan Hayward  <alan.hayward@arm.com>
15847             David Sherwood  <david.sherwood@arm.com>
15849         * target.def (array_mode): New target hook.
15850         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15851         * doc/tm.texi: Regenerate.
15852         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15853         * hooks.c (hook_optmode_mode_uhwi_none): New function.
15854         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15855         targetm.array_mode.
15856         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
15857         type sizes.
15859 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15860             Alan Hayward  <alan.hayward@arm.com>
15861             David Sherwood  <david.sherwood@arm.com>
15863         * fold-const.c (fold_binary_loc): Check the argument types
15864         rather than the result type when testing for a vector operation.
15866 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15868         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15869         * doc/tm.texi: Regenerate.
15871 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
15872             Alan Hayward  <alan.hayward@arm.com>
15873             David Sherwood  <david.sherwood@arm.com>
15875         * doc/invoke.texi (-msve-vector-bits=): Document new option.
15876         (sve): Document new AArch64 extension.
15877         * doc/md.texi (w): Extend the description of the AArch64
15878         constraint to include SVE vectors.
15879         (Upl, Upa): Document new AArch64 predicate constraints.
15880         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15881         enum.
15882         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15883         (msve-vector-bits=): New option.
15884         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15885         SVE when these are disabled.
15886         (sve): New extension.
15887         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15888         modes.  Adjust their number of units based on aarch64_sve_vg.
15889         (MAX_BITSIZE_MODE_ANY_MODE): Define.
15890         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15891         aarch64_addr_query_type.
15892         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15893         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15894         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15895         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15896         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15897         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15898         (aarch64_simd_imm_zero_p): Delete.
15899         (aarch64_check_zero_based_sve_index_immediate): Declare.
15900         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15901         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15902         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15903         (aarch64_sve_float_mul_immediate_p): Likewise.
15904         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15905         rather than an rtx.
15906         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15907         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15908         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15909         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15910         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15911         (aarch64_regmode_natural_size): Likewise.
15912         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15913         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15914         left one place.
15915         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15916         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15917         for VG and the SVE predicate registers.
15918         (V_ALIASES): Add a "z"-prefixed alias.
15919         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15920         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15921         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15922         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15923         (REG_CLASS_NAMES): Add entries for them.
15924         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
15925         and the predicate registers.
15926         (aarch64_sve_vg): Declare.
15927         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15928         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15929         (REGMODE_NATURAL_SIZE): Define.
15930         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15931         SVE macros.
15932         * config/aarch64/aarch64.c: Include cfgrtl.h.
15933         (simd_immediate_info): Add a constructor for series vectors,
15934         and an associated step field.
15935         (aarch64_sve_vg): New variable.
15936         (aarch64_dbx_register_number): Handle VG and the predicate registers.
15937         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15938         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15939         (VEC_ANY_DATA, VEC_STRUCT): New constants.
15940         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15941         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15942         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15943         (aarch64_get_mask_mode): New functions.
15944         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15945         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15946         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
15947         predicate modes and predicate registers.  Explicitly restrict
15948         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
15949         to store a vector mode if it is recognized by
15950         aarch64_classify_vector_mode.
15951         (aarch64_regmode_natural_size): New function.
15952         (aarch64_hard_regno_caller_save_mode): Return the original mode
15953         for predicates.
15954         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15955         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15956         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15957         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15958         functions.
15959         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
15960         does not overlap dest if the function is frame-related.  Handle
15961         SVE constants.
15962         (aarch64_split_add_offset): New function.
15963         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15964         them aarch64_add_offset.
15965         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15966         and update call to aarch64_sub_sp.
15967         (aarch64_add_cfa_expression): New function.
15968         (aarch64_expand_prologue): Pass extra temporary registers to the
15969         functions above.  Handle the case in which we need to emit new
15970         DW_CFA_expressions for registers that were originally saved
15971         relative to the stack pointer, but now have to be expressed
15972         relative to the frame pointer.
15973         (aarch64_output_mi_thunk): Pass extra temporary registers to the
15974         functions above.
15975         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
15976         IP0 and IP1 values for SVE frames.
15977         (aarch64_expand_vec_series): New function.
15978         (aarch64_expand_sve_widened_duplicate): Likewise.
15979         (aarch64_expand_sve_const_vector): Likewise.
15980         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15981         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
15982         into the register, rather than emitting a SET directly.
15983         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15984         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15985         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15986         (offset_9bit_signed_scaled_p): New functions.
15987         (aarch64_replicate_bitmask_imm): New function.
15988         (aarch64_bitmask_imm): Use it.
15989         (aarch64_cannot_force_const_mem): Reject expressions involving
15990         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
15991         (aarch64_classify_index): Handle SVE indices, by requiring
15992         a plain register index with a scale that matches the element size.
15993         (aarch64_classify_address): Handle SVE addresses.  Assert that
15994         the mode of the address is VOIDmode or an integer mode.
15995         Update call to aarch64_classify_symbol.
15996         (aarch64_classify_symbolic_expression): Update call to
15997         aarch64_classify_symbol.
15998         (aarch64_const_vec_all_in_range_p): New function.
15999         (aarch64_print_vector_float_operand): Likewise.
16000         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
16001         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
16002         and the FP immediates 1.0 and 0.5.
16003         (aarch64_print_address_internal): Handle SVE addresses.
16004         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16005         (aarch64_regno_regclass): Handle predicate registers.
16006         (aarch64_secondary_reload): Handle big-endian reloads of SVE
16007         data modes.
16008         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16009         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16010         (aarch64_convert_sve_vector_bits): New function.
16011         (aarch64_override_options): Use it to handle -msve-vector-bits=.
16012         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16013         rather than an rtx.
16014         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16015         Handle SVE vector and predicate modes.  Accept VL-based constants
16016         that need only one temporary register, and VL offsets that require
16017         no temporary registers.
16018         (aarch64_conditional_register_usage): Mark the predicate registers
16019         as fixed if SVE isn't available.
16020         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16021         Return true for SVE vector and predicate modes.
16022         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16023         rather than an unsigned int.  Handle SVE modes.
16024         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
16025         SVE modes.
16026         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16027         if SVE is enabled.
16028         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16029         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16030         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16031         (aarch64_sve_float_mul_immediate_p): New functions.
16032         (aarch64_sve_valid_immediate): New function.
16033         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16034         Explicitly reject structure modes.  Check for INDEX constants.
16035         Handle PTRUE and PFALSE constants.
16036         (aarch64_check_zero_based_sve_index_immediate): New function.
16037         (aarch64_simd_imm_zero_p): Delete.
16038         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16039         vector modes.  Accept constants in the range of CNT[BHWD].
16040         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16041         ask for an Advanced SIMD mode.
16042         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16043         (aarch64_simd_vector_alignment): Handle SVE predicates.
16044         (aarch64_vectorize_preferred_vector_alignment): New function.
16045         (aarch64_simd_vector_alignment_reachable): Use it instead of
16046         the vector size.
16047         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16048         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16049         functions.
16050         (MAX_VECT_LEN): Delete.
16051         (expand_vec_perm_d): Add a vec_flags field.
16052         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16053         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16054         (aarch64_evpc_ext): Don't apply a big-endian lane correction
16055         for SVE modes.
16056         (aarch64_evpc_rev): Rename to...
16057         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
16058         (aarch64_evpc_rev_global): New function.
16059         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16060         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16061         MAX_VECT_LEN.
16062         (aarch64_evpc_sve_tbl): New function.
16063         (aarch64_expand_vec_perm_const_1): Update after rename of
16064         aarch64_evpc_rev.  Handle SVE permutes too, trying
16065         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16066         than aarch64_evpc_tbl.
16067         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16068         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16069         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16070         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16071         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16072         (aarch64_expand_sve_vcond): New functions.
16073         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16074         of aarch64_vector_mode_p.
16075         (aarch64_dwarf_poly_indeterminate_value): New function.
16076         (aarch64_compute_pressure_classes): Likewise.
16077         (aarch64_can_change_mode_class): Likewise.
16078         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16079         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16080         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16081         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16082         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16083         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16084         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16085         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16086         constraints.
16087         (Dn, Dl, Dr): Accept const as well as const_vector.
16088         (Dz): Likewise.  Compare against CONST0_RTX.
16089         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16090         of "vector" where appropriate.
16091         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16092         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16093         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16094         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16095         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16096         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16097         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16098         (v_int_equiv): Extend to SVE modes.
16099         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16100         mode attributes.
16101         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16102         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16103         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16104         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16105         (SVE_COND_FP_CMP): New int iterators.
16106         (perm_hilo): Handle the new unpack unspecs.
16107         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16108         attributes.
16109         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16110         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16111         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16112         (aarch64_equality_operator, aarch64_constant_vector_operand)
16113         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16114         (aarch64_sve_nonimmediate_operand): Likewise.
16115         (aarch64_sve_general_operand): Likewise.
16116         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16117         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16118         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16119         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16120         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16121         (aarch64_sve_float_arith_immediate): Likewise.
16122         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16123         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16124         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16125         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16126         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16127         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16128         (aarch64_sve_float_arith_operand): Likewise.
16129         (aarch64_sve_float_arith_with_sub_operand): Likewise.
16130         (aarch64_sve_float_mul_operand): Likewise.
16131         (aarch64_sve_vec_perm_operand): Likewise.
16132         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16133         (aarch64_mov_operand): Accept const_poly_int and const_vector.
16134         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16135         as well as const_vector.
16136         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16137         in file.  Use CONST0_RTX and CONSTM1_RTX.
16138         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
16139         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16140         Use aarch64_simd_imm_zero.
16141         * config/aarch64/aarch64-sve.md: New file.
16142         * config/aarch64/aarch64.md: Include it.
16143         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16144         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16145         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16146         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16147         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16148         (sve): New attribute.
16149         (enabled): Disable instructions with the sve attribute unless
16150         TARGET_SVE.
16151         (movqi, movhi): Pass CONST_POLY_INT operaneds through
16152         aarch64_expand_mov_immediate.
16153         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16154         CNT[BHSD] immediates.
16155         (movti): Split CONST_POLY_INT moves into two halves.
16156         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16157         Split additions that need a temporary here if the destination
16158         is the stack pointer.
16159         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16160         (*add<mode>3_poly_1): New instruction.
16161         (set_clobber_cc): New expander.
16163 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
16165         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16166         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
16167         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16168         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16169         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
16170         Change innermode from fixed_mode_size to machine_mode.
16171         (simplify_subreg): Update call accordingly.  Handle a constant-sized
16172         subreg of a variable-length CONST_VECTOR.
16174 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
16175             Alan Hayward  <alan.hayward@arm.com>
16176             David Sherwood  <david.sherwood@arm.com>
16178         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16179         (add_offset_to_base): New function, split out from...
16180         (create_mem_ref): ...here.  When handling a scale other than 1,
16181         check first whether the address is valid without the offset.
16182         Add it into the base if so, leaving the index and scale as-is.
16184 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
16186         PR c++/83778
16187         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16188         fold_for_warn before checking if arg2 is INTEGER_CST.
16190 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
16192         * config/rs6000/predicates.md (load_multiple_operation): Delete.
16193         (store_multiple_operation): Delete.
16194         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16195         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16196         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16197         guarded by TARGET_STRING.
16198         (rs6000_output_load_multiple): Delete.
16199         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16200         OPTION_MASK_STRING / TARGET_STRING handling.
16201         (print_operand) <'N', 'O'>: Add comment that these are unused now.
16202         (const rs6000_opt_masks) <"string">: Change mask to 0.
16203         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16204         (MASK_STRING): Delete.
16205         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16206         parts.  Simplify.
16207         (load_multiple): Delete.
16208         (*ldmsi8): Delete.
16209         (*ldmsi7): Delete.
16210         (*ldmsi6): Delete.
16211         (*ldmsi5): Delete.
16212         (*ldmsi4): Delete.
16213         (*ldmsi3): Delete.
16214         (store_multiple): Delete.
16215         (*stmsi8): Delete.
16216         (*stmsi7): Delete.
16217         (*stmsi6): Delete.
16218         (*stmsi5): Delete.
16219         (*stmsi4): Delete.
16220         (*stmsi3): Delete.
16221         (movmemsi_8reg): Delete.
16222         (corresponding unnamed define_insn): Delete.
16223         (movmemsi_6reg): Delete.
16224         (corresponding unnamed define_insn): Delete.
16225         (movmemsi_4reg): Delete.
16226         (corresponding unnamed define_insn): Delete.
16227         (movmemsi_2reg): Delete.
16228         (corresponding unnamed define_insn): Delete.
16229         (movmemsi_1reg): Delete.
16230         (corresponding unnamed define_insn): Delete.
16231         * config/rs6000/rs6000.opt (mno-string): New.
16232         (mstring): Replace by deprecation warning stub.
16233         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16235 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
16237         * regrename.c (regrename_do_replace): If replacing the same
16238         reg multiple times, try to reuse last created gen_raw_REG.
16240         PR debug/81155
16241         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16242         main to workaround a bug in GDB.
16244 2018-01-12  Tom de Vries  <tom@codesourcery.com>
16246         PR target/83737
16247         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16249 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
16251         PR rtl-optimization/80481
16252         * ira-color.c (get_cap_member): New function.
16253         (allocnos_conflict_by_live_ranges_p): Use it.
16254         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16255         (setup_slot_coalesced_allocno_live_ranges): Ditto.
16257 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
16259         PR target/83628
16260         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16261         (*saddl_se_1): Ditto.
16262         (*ssubsi_1): Ditto.
16263         (*ssubl_se_1): Ditto.
16265 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
16267         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16268         rather than wi::to_widest for DR_INITs.
16269         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16270         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16271         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16272         INTEGER_CSTs.
16273         (vect_analyze_group_access_1): Note that here.
16275 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
16277         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16278         polynomial type sizes.
16280 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
16282         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16283         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16284         (gimple_add_tmp_var): Likewise.
16286 2018-01-12  Martin Liska  <mliska@suse.cz>
16288         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16289         (gimple_alloc_sizes): Likewise.
16290         (dump_gimple_statistics): Use PRIu64 in printf format.
16291         * gimple.h: Change uint64_t to int.
16293 2018-01-12  Martin Liska  <mliska@suse.cz>
16295         * tree-core.h: Use uint64_t instead of int.
16296         * tree.c (tree_node_counts): Likewise.
16297         (tree_node_sizes): Likewise.
16298         (dump_tree_statistics): Use PRIu64 in printf format.
16300 2018-01-12  Martin Liska  <mliska@suse.cz>
16302         * Makefile.in: As qsort_chk is implemented in vec.c, add
16303         vec.o to linkage of gencfn-macros.
16304         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16305         passing the info to record_node_allocation_statistics.
16306         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16307         and pass the info.
16308         * ggc-common.c (struct ggc_usage): Add operator== and use
16309         it in operator< and compare function.
16310         * mem-stats.h (struct mem_usage): Likewise.
16311         * vec.c (struct vec_usage): Remove operator< and compare
16312         function. Can be simply inherited.
16314 2018-01-12  Martin Jambor  <mjambor@suse.cz>
16316         PR target/81616
16317         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16318         * tree-ssa-math-opts.c: Include domwalk.h.
16319         (convert_mult_to_fma_1): New function.
16320         (fma_transformation_info): New type.
16321         (fma_deferring_state): Likewise.
16322         (cancel_fma_deferring): New function.
16323         (result_of_phi): Likewise.
16324         (last_fma_candidate_feeds_initial_phi): Likewise.
16325         (convert_mult_to_fma): Added deferring logic, split actual
16326         transformation to convert_mult_to_fma_1.
16327         (math_opts_dom_walker): New type.
16328         (math_opts_dom_walker::after_dom_children): New method, body moved
16329         here from pass_optimize_widening_mul::execute, added deferring logic
16330         bits.
16331         (pass_optimize_widening_mul::execute): Moved most of code to
16332         math_opts_dom_walker::after_dom_children.
16333         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16334         * config/i386/i386.c (ix86_option_override_internal): Added
16335         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16337 2018-01-12  Richard Biener  <rguenther@suse.de>
16339         PR debug/83157
16340         * dwarf2out.c (gen_variable_die): Do not reset old_die for
16341         inline instance vars.
16343 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
16345         PR target/81819
16346         * config/rx/rx.c (rx_is_restricted_memory_address):
16347         Handle SUBREG case.
16349 2018-01-12  Richard Biener  <rguenther@suse.de>
16351         PR tree-optimization/80846
16352         * target.def (split_reduction): New target hook.
16353         * targhooks.c (default_split_reduction): New function.
16354         * targhooks.h (default_split_reduction): Declare.
16355         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16356         target requests first reduce vectors by combining low and high
16357         parts.
16358         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16359         (get_vectype_for_scalar_type_and_size): Export.
16360         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16361         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16362         * doc/tm.texi: Regenerate.
16363         * config/i386/i386.c (ix86_split_reduction): Implement
16364         TARGET_VECTORIZE_SPLIT_REDUCTION.
16366 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
16368         PR target/83368
16369         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16370         in PIC mode except for TARGET_VXWORKS_RTP.
16371         * config/sparc/sparc.c: Include cfgrtl.h.
16372         (TARGET_INIT_PIC_REG): Define.
16373         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16374         (sparc_pic_register_p): New predicate.
16375         (sparc_legitimate_address_p): Use it.
16376         (sparc_legitimize_pic_address): Likewise.
16377         (sparc_delegitimize_address): Likewise.
16378         (sparc_mode_dependent_address_p): Likewise.
16379         (gen_load_pcrel_sym): Remove 4th parameter.
16380         (load_got_register): Adjust call to above.  Remove obsolete stuff.
16381         (sparc_expand_prologue): Do not call load_got_register here.
16382         (sparc_flat_expand_prologue): Likewise.
16383         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16384         (sparc_use_pseudo_pic_reg): New function.
16385         (sparc_init_pic_reg): Likewise.
16386         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16387         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16389 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
16391         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16392         Add item for branch_cost.
16394 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
16396         PR rtl-optimization/83565
16397         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16398         not extend the result to a larger mode for rotate operations.
16399         (num_sign_bit_copies1): Likewise.
16401 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16403         PR target/40411
16404         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16405         -symbolic.
16406         Use values-Xc.o for -pedantic.
16407         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16409 2018-01-12  Martin Liska  <mliska@suse.cz>
16411         PR ipa/83054
16412         * ipa-devirt.c (final_warning_record::grow_type_warnings):
16413         New function.
16414         (possible_polymorphic_call_targets): Use it.
16415         (ipa_devirt): Likewise.
16417 2018-01-12  Martin Liska  <mliska@suse.cz>
16419         * profile-count.h (enum profile_quality): Use 0 as invalid
16420         enum value of profile_quality.
16422 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
16424         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16425         -mext-string options.
16427 2018-01-12  Richard Biener  <rguenther@suse.de>
16429         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16430         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16431         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16432         Likewise.
16433         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16435 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
16437         * configure.ac (--with-long-double-format): Add support for the
16438         configuration option to change the default long double format on
16439         PowerPC systems.
16440         * config.gcc (powerpc*-linux*-*): Likewise.
16441         * configure: Regenerate.
16442         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16443         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16444         used without modification.
16446 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
16448         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16449         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16450         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16451         MISC_BUILTIN_SPEC_BARRIER.
16452         (rs6000_init_builtins): Likewise.
16453         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16454         enum value.
16455         (speculation_barrier): New define_insn.
16456         * doc/extend.texi: Document __builtin_speculation_barrier.
16458 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
16460         PR target/83203
16461         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16462         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16463         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16464         iterators.
16465         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
16466         integral modes instead of "ss" and "sd".
16467         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16468         vectors with 32-bit and 64-bit elements.
16469         (vecdupssescalarmodesuffix): New mode attribute.
16470         (vec_dup<mode>): Use it.
16472 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
16474         PR target/83330
16475         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16476         frame if argument is passed on stack.
16478 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
16480         PR target/82682
16481         * ree.c (combine_reaching_defs): Optimize also
16482         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16483         reg2=any_extend(exp); reg1=reg2;, formatting fix.
16485 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
16487         PR middle-end/83189
16488         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16490 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
16492         PR middle-end/83718
16493         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16494         after they are computed.
16496 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
16498         PR tree-optimization/83695
16499         * gimple-loop-linterchange.cc
16500         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16501         reset cached scev information after interchange.
16502         (pass_linterchange::execute): Remove call to scev_reset_htab.
16504 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16506         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16507         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16508         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16509         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16510         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16511         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16512         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16513         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16514         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16515         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16516         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16517         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16518         (V_lane_reg): Likewise.
16519         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16520         New define_expand.
16521         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16522         (vfmal_lane_low<mode>_intrinsic,
16523         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16524         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16525         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16526         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16527         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16528         vfmsl_lane_high<mode>_intrinsic): New define_insns.
16530 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16532         * config/arm/arm-cpus.in (fp16fml): New feature.
16533         (ALL_SIMD): Add fp16fml.
16534         (armv8.2-a): Add fp16fml as an option.
16535         (armv8.3-a): Likewise.
16536         (armv8.4-a): Add fp16fml as part of fp16.
16537         * config/arm/arm.h (TARGET_FP16FML): Define.
16538         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16539         when appropriate.
16540         * config/arm/arm-modes.def (V2HF): Define.
16541         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16542         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16543         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16544         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16545         vfmsl_low, vfmsl_high): New set of builtins.
16546         * config/arm/iterators.md (PLUSMINUS): New code iterator.
16547         (vfml_op): New code attribute.
16548         (VFMLHALVES): New int iterator.
16549         (VFML, VFMLSEL): New mode attributes.
16550         (V_reg): Define mapping for V2HF.
16551         (V_hi, V_lo): New mode attributes.
16552         (VF_constraint): Likewise.
16553         (vfml_half, vfml_half_selector): New int attributes.
16554         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16555         define_expand.
16556         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16557         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16558         New define_insn.
16559         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16560         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16561         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16562         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
16563         documentation.
16564         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16565         Document new effective target and option set.
16567 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16569         * config/arm/arm-cpus.in (armv8_4): New feature.
16570         (ARMv8_4a): New fgroup.
16571         (armv8.4-a): New arch.
16572         * config/arm/arm-tables.opt: Regenerate.
16573         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16574         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16575         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16576         Add matching rules for -march=armv8.4-a and extensions.
16577         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16579 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
16581         PR target/81821
16582         * config/rx/rx.md (BW): New mode attribute.
16583         (sync_lock_test_and_setsi): Add mode suffix to insn output.
16585 2018-01-11  Richard Biener  <rguenther@suse.de>
16587         PR tree-optimization/83435
16588         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16589         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16590         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16592 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16593             Alan Hayward  <alan.hayward@arm.com>
16594             David Sherwood  <david.sherwood@arm.com>
16596         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16597         field.
16598         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
16599         (aarch64_print_address_internal): Use it to check for a zero offset.
16601 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16602             Alan Hayward  <alan.hayward@arm.com>
16603             David Sherwood  <david.sherwood@arm.com>
16605         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16606         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16607         Return a poly_int64 rather than a HOST_WIDE_INT.
16608         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16609         rather than a HOST_WIDE_INT.
16610         * config/aarch64/aarch64.h (aarch64_frame): Protect with
16611         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
16612         hard_fp_offset, frame_size, initial_adjust, callee_offset and
16613         final_offset from HOST_WIDE_INT to poly_int64.
16614         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16615         to_constant when getting the number of units in an Advanced SIMD
16616         mode.
16617         (aarch64_builtin_vectorized_function): Check for a constant number
16618         of units.
16619         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16620         GET_MODE_SIZE.
16621         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16622         attribute instead of GET_MODE_NUNITS.
16623         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16624         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16625         GET_MODE_SIZE for fixed-size registers.
16626         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16627         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16628         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16629         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16630         (aarch64_print_operand, aarch64_print_address_internal)
16631         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16632         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16633         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16634         Handle polynomial GET_MODE_SIZE.
16635         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
16636         wider than SImode without modification.
16637         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16638         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16639         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16640         passing and returning SVE modes.
16641         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16642         rather than GEN_INT.
16643         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16644         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16645         (aarch64_allocate_and_probe_stack_space): Likewise.
16646         (aarch64_layout_frame): Cope with polynomial offsets.
16647         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16648         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
16649         polynomial offsets.
16650         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16651         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16652         poly_int64 rather than a HOST_WIDE_INT.
16653         (aarch64_get_separate_components, aarch64_process_components)
16654         (aarch64_expand_prologue, aarch64_expand_epilogue)
16655         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16656         (aarch64_anchor_offset): New function, split out from...
16657         (aarch64_legitimize_address): ...here.
16658         (aarch64_builtin_vectorization_cost): Handle polynomial
16659         TYPE_VECTOR_SUBPARTS.
16660         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16661         GET_MODE_NUNITS.
16662         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16663         number of elements from the PARALLEL rather than the mode.
16664         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16665         rather than GET_MODE_BITSIZE.
16666         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16667         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16668         (aarch64_expand_vec_perm_const_1): Handle polynomial
16669         d->perm.length () and d->perm elements.
16670         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
16671         Apply to_constant to d->perm elements.
16672         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16673         polynomial CONST_VECTOR_NUNITS.
16674         (aarch64_move_pointer): Take amount as a poly_int64 rather
16675         than an int.
16676         (aarch64_progress_pointer): Avoid temporary variable.
16677         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16678         the mode attribute instead of GET_MODE.
16680 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16681             Alan Hayward  <alan.hayward@arm.com>
16682             David Sherwood  <david.sherwood@arm.com>
16684         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16685         x exists before using it.
16686         (aarch64_add_constant_internal): Rename to...
16687         (aarch64_add_offset_1): ...this.  Replace regnum with separate
16688         src and dest rtxes.  Handle the case in which they're different,
16689         including when the offset is zero.  Replace scratchreg with an rtx.
16690         Use 2 additions if there is no spare register into which we can
16691         move a 16-bit constant.
16692         (aarch64_add_constant): Delete.
16693         (aarch64_add_offset): Replace reg with separate src and dest
16694         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
16695         Use aarch64_add_offset_1.
16696         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16697         an rtx rather than an int.  Take the delta as a poly_int64
16698         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
16699         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16700         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16701         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16702         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16703         and aarch64_add_sp.
16704         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16705         aarch64_add_constant.
16707 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16709         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16710         Use scalar_float_mode.
16712 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16714         * config/aarch64/aarch64-simd.md
16715         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16716         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16717         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16718         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16719         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16720         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16721         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16722         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16723         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16724         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16726 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16728         PR target/83514
16729         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16730         targ_options->x_arm_arch_string is non NULL.
16732 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
16734         * config/aarch64/aarch64.h
16735         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
16737 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
16739         PR target/82096
16740         * expmed.c (emit_store_flag_force): Swap if const op0
16741         and change VOIDmode to mode of op0.
16743 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
16745         PR rtl-optimization/83761
16746         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16747         than bytes to mode_for_size.
16749 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
16751         PR middle-end/83189
16752         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16753         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16754         profile.
16756 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
16758         PR middle-end/83575
16759         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16760         when in layout mode.
16761         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16762         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16763         partition fixup.
16765 2018-01-10  Michael Collison  <michael.collison@arm.com>
16767         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16768         * config/aarch64/aarch64-option-extension.def: Add
16769         AARCH64_OPT_EXTENSION of 'fp16fml'.
16770         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16771         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16772         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16773         * config/aarch64/constraints.md (Ui7): New constraint.
16774         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16775         (VFMLA_SEL_W): Ditto.
16776         (f16quad): Ditto.
16777         (f16mac1): Ditto.
16778         (VFMLA16_LOW): New int iterator.
16779         (VFMLA16_HIGH): Ditto.
16780         (UNSPEC_FMLAL): New unspec.
16781         (UNSPEC_FMLSL): Ditto.
16782         (UNSPEC_FMLAL2): Ditto.
16783         (UNSPEC_FMLSL2): Ditto.
16784         (f16mac): New code attribute.
16785         * config/aarch64/aarch64-simd-builtins.def
16786         (aarch64_fmlal_lowv2sf): Ditto.
16787         (aarch64_fmlsl_lowv2sf): Ditto.
16788         (aarch64_fmlalq_lowv4sf): Ditto.
16789         (aarch64_fmlslq_lowv4sf): Ditto.
16790         (aarch64_fmlal_highv2sf): Ditto.
16791         (aarch64_fmlsl_highv2sf): Ditto.
16792         (aarch64_fmlalq_highv4sf): Ditto.
16793         (aarch64_fmlslq_highv4sf): Ditto.
16794         (aarch64_fmlal_lane_lowv2sf): Ditto.
16795         (aarch64_fmlsl_lane_lowv2sf): Ditto.
16796         (aarch64_fmlal_laneq_lowv2sf): Ditto.
16797         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16798         (aarch64_fmlalq_lane_lowv4sf): Ditto.
16799         (aarch64_fmlsl_lane_lowv4sf): Ditto.
16800         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16801         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16802         (aarch64_fmlal_lane_highv2sf): Ditto.
16803         (aarch64_fmlsl_lane_highv2sf): Ditto.
16804         (aarch64_fmlal_laneq_highv2sf): Ditto.
16805         (aarch64_fmlsl_laneq_highv2sf): Ditto.
16806         (aarch64_fmlalq_lane_highv4sf): Ditto.
16807         (aarch64_fmlsl_lane_highv4sf): Ditto.
16808         (aarch64_fmlalq_laneq_highv4sf): Ditto.
16809         (aarch64_fmlsl_laneq_highv4sf): Ditto.
16810         * config/aarch64/aarch64-simd.md:
16811         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16812         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16813         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16814         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16815         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16816         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16817         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16818         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16819         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16820         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16821         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16822         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16823         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16824         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16825         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16826         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16827         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16828         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16829         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16830         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16831         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16832         (vfmlsl_low_u32): Ditto.
16833         (vfmlalq_low_u32): Ditto.
16834         (vfmlslq_low_u32): Ditto.
16835         (vfmlal_high_u32): Ditto.
16836         (vfmlsl_high_u32): Ditto.
16837         (vfmlalq_high_u32): Ditto.
16838         (vfmlslq_high_u32): Ditto.
16839         (vfmlal_lane_low_u32): Ditto.
16840         (vfmlsl_lane_low_u32): Ditto.
16841         (vfmlal_laneq_low_u32): Ditto.
16842         (vfmlsl_laneq_low_u32): Ditto.
16843         (vfmlalq_lane_low_u32): Ditto.
16844         (vfmlslq_lane_low_u32): Ditto.
16845         (vfmlalq_laneq_low_u32): Ditto.
16846         (vfmlslq_laneq_low_u32): Ditto.
16847         (vfmlal_lane_high_u32): Ditto.
16848         (vfmlsl_lane_high_u32): Ditto.
16849         (vfmlal_laneq_high_u32): Ditto.
16850         (vfmlsl_laneq_high_u32): Ditto.
16851         (vfmlalq_lane_high_u32): Ditto.
16852         (vfmlslq_lane_high_u32): Ditto.
16853         (vfmlalq_laneq_high_u32): Ditto.
16854         (vfmlslq_laneq_high_u32): Ditto.
16855         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16856         (AARCH64_FL_FOR_ARCH8_4): New.
16857         (AARCH64_ISA_F16FML): New ISA flag.
16858         (TARGET_F16FML): New feature flag for fp16fml.
16859         (doc/invoke.texi): Document new fp16fml option.
16861 2018-01-10  Michael Collison  <michael.collison@arm.com>
16863         * config/aarch64/aarch64-builtins.c:
16864         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16865         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16866         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16867         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16868         (AARCH64_ISA_SHA3): New ISA flag.
16869         (TARGET_SHA3): New feature flag for sha3.
16870         * config/aarch64/iterators.md (sha512_op): New int attribute.
16871         (CRYPTO_SHA512): New int iterator.
16872         (UNSPEC_SHA512H): New unspec.
16873         (UNSPEC_SHA512H2): Ditto.
16874         (UNSPEC_SHA512SU0): Ditto.
16875         (UNSPEC_SHA512SU1): Ditto.
16876         * config/aarch64/aarch64-simd-builtins.def
16877         (aarch64_crypto_sha512hqv2di): New builtin.
16878         (aarch64_crypto_sha512h2qv2di): Ditto.
16879         (aarch64_crypto_sha512su0qv2di): Ditto.
16880         (aarch64_crypto_sha512su1qv2di): Ditto.
16881         (aarch64_eor3qv8hi): Ditto.
16882         (aarch64_rax1qv2di): Ditto.
16883         (aarch64_xarqv2di): Ditto.
16884         (aarch64_bcaxqv8hi): Ditto.
16885         * config/aarch64/aarch64-simd.md:
16886         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16887         (aarch64_crypto_sha512su0qv2di): Ditto.
16888         (aarch64_crypto_sha512su1qv2di): Ditto.
16889         (aarch64_eor3qv8hi): Ditto.
16890         (aarch64_rax1qv2di): Ditto.
16891         (aarch64_xarqv2di): Ditto.
16892         (aarch64_bcaxqv8hi): Ditto.
16893         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16894         (vsha512h2q_u64): Ditto.
16895         (vsha512su0q_u64): Ditto.
16896         (vsha512su1q_u64): Ditto.
16897         (veor3q_u16): Ditto.
16898         (vrax1q_u64): Ditto.
16899         (vxarq_u64): Ditto.
16900         (vbcaxq_u16): Ditto.
16901         * config/arm/types.md (crypto_sha512): New type attribute.
16902         (crypto_sha3): Ditto.
16903         (doc/invoke.texi): Document new sha3 option.
16905 2018-01-10  Michael Collison  <michael.collison@arm.com>
16907         * config/aarch64/aarch64-builtins.c:
16908         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16909         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16910         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16911         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16912         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16913         (AARCH64_ISA_SM4): New ISA flag.
16914         (TARGET_SM4): New feature flag for sm4.
16915         * config/aarch64/aarch64-simd-builtins.def
16916         (aarch64_sm3ss1qv4si): Ditto.
16917         (aarch64_sm3tt1aq4si): Ditto.
16918         (aarch64_sm3tt1bq4si): Ditto.
16919         (aarch64_sm3tt2aq4si): Ditto.
16920         (aarch64_sm3tt2bq4si): Ditto.
16921         (aarch64_sm3partw1qv4si): Ditto.
16922         (aarch64_sm3partw2qv4si): Ditto.
16923         (aarch64_sm4eqv4si): Ditto.
16924         (aarch64_sm4ekeyqv4si): Ditto.
16925         * config/aarch64/aarch64-simd.md:
16926         (aarch64_sm3ss1qv4si): Ditto.
16927         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16928         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16929         (aarch64_sm4eqv4si): Ditto.
16930         (aarch64_sm4ekeyqv4si): Ditto.
16931         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16932         (sm3part_op): Ditto.
16933         (CRYPTO_SM3TT): Ditto.
16934         (CRYPTO_SM3PART): Ditto.
16935         (UNSPEC_SM3SS1): New unspec.
16936         (UNSPEC_SM3TT1A): Ditto.
16937         (UNSPEC_SM3TT1B): Ditto.
16938         (UNSPEC_SM3TT2A): Ditto.
16939         (UNSPEC_SM3TT2B): Ditto.
16940         (UNSPEC_SM3PARTW1): Ditto.
16941         (UNSPEC_SM3PARTW2): Ditto.
16942         (UNSPEC_SM4E): Ditto.
16943         (UNSPEC_SM4EKEY): Ditto.
16944         * config/aarch64/constraints.md (Ui2): New constraint.
16945         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16946         * config/arm/types.md (crypto_sm3): New type attribute.
16947         (crypto_sm4): Ditto.
16948         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16949         (vsm3tt1aq_u32): Ditto.
16950         (vsm3tt1bq_u32): Ditto.
16951         (vsm3tt2aq_u32): Ditto.
16952         (vsm3tt2bq_u32): Ditto.
16953         (vsm3partw1q_u32): Ditto.
16954         (vsm3partw2q_u32): Ditto.
16955         (vsm4eq_u32): Ditto.
16956         (vsm4ekeyq_u32): Ditto.
16957         (doc/invoke.texi): Document new sm4 option.
16959 2018-01-10  Michael Collison  <michael.collison@arm.com>
16961         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16962         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16963         (AARCH64_FL_FOR_ARCH8_4): New.
16964         (AARCH64_FL_V8_4): New flag.
16965         (doc/invoke.texi): Document new armv8.4-a option.
16967 2018-01-10  Michael Collison  <michael.collison@arm.com>
16969         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16970         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16971         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16972         * config/aarch64/aarch64-option-extension.def: Add
16973         AARCH64_OPT_EXTENSION of 'sha2'.
16974         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16975         (crypto): Disable sha2 and aes if crypto disabled.
16976         (crypto): Enable aes and sha2 if enabled.
16977         (simd): Disable sha2 and aes if simd disabled.
16978         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16979         New flags.
16980         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16981         (TARGET_SHA2): New feature flag for sha2.
16982         (TARGET_AES): New feature flag for aes.
16983         * config/aarch64/aarch64-simd.md:
16984         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16985         conditional on TARGET_AES.
16986         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16987         (aarch64_crypto_sha1hsi): Make pattern conditional
16988         on TARGET_SHA2.
16989         (aarch64_crypto_sha1hv4si): Ditto.
16990         (aarch64_be_crypto_sha1hv4si): Ditto.
16991         (aarch64_crypto_sha1su1v4si): Ditto.
16992         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16993         (aarch64_crypto_sha1su0v4si): Ditto.
16994         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16995         (aarch64_crypto_sha256su0v4si): Ditto.
16996         (aarch64_crypto_sha256su1v4si): Ditto.
16997         (doc/invoke.texi): Document new aes and sha2 options.
16999 2018-01-10  Martin Sebor  <msebor@redhat.com>
17001         PR tree-optimization/83781
17002         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17003         as string arrays.
17005 2018-01-11  Martin Sebor  <msebor@gmail.com>
17006             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17008         PR tree-optimization/83501
17009         PR tree-optimization/81703
17011         * tree-ssa-strlen.c (get_string_cst): Rename...
17012         (get_string_len): ...to this.  Handle global constants.
17013         (handle_char_store): Adjust.
17015 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
17016             Jim Wilson  <jimw@sifive.com>
17018         * config/riscv/riscv-protos.h (riscv_output_return): New.
17019         * config/riscv/riscv.c (struct machine_function): New naked_p field.
17020         (riscv_attribute_table, riscv_output_return),
17021         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17022         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17023         (riscv_compute_frame_info): Only compute frame->mask if not a naked
17024         function.
17025         (riscv_expand_prologue): Add early return for naked function.
17026         (riscv_expand_epilogue): Likewise.
17027         (riscv_function_ok_for_sibcall): Return false for naked function.
17028         (riscv_set_current_function): New.
17029         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17030         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17031         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17032         * doc/extend.texi (RISC-V Function Attributes): New.
17034 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17036         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17037         check for 128-bit long double before checking TCmode.
17038         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17039         128-bit long doubles before checking TFmode or TCmode.
17040         (FLOAT128_IBM_P): Likewise.
17042 2018-01-10  Martin Sebor  <msebor@redhat.com>
17044         PR tree-optimization/83671
17045         * builtins.c (c_strlen): Unconditionally return zero for the empty
17046         string.
17047         Use -Warray-bounds for warnings.
17048         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17049         for non-constant array indices with COMPONENT_REF, arrays of
17050         arrays, and pointers to arrays.
17051         (gimple_fold_builtin_strlen): Determine and set length range for
17052         non-constant character arrays.
17054 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
17056         PR middle-end/81897
17057         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17058         empty blocks.
17060 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
17062         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17064 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
17066         PR target/83399
17067         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17068         VECTOR_MEM_ALTIVEC_OR_VSX_P.
17069         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17070         indexed_or_indirect_operand predicate.
17071         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17072         (*vsx_le_perm_load_v8hi): Likewise.
17073         (*vsx_le_perm_load_v16qi): Likewise.
17074         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17075         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17076         (*vsx_le_perm_store_v8hi): Likewise.
17077         (*vsx_le_perm_store_v16qi): Likewise.
17078         (eight unnamed splitters): Likewise.
17080 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
17082         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17083         * config/rs6000/emmintrin.h: Likewise.
17084         * config/rs6000/mmintrin.h: Likewise.
17085         * config/rs6000/xmmintrin.h: Likewise.
17087 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
17089         PR c++/43486
17090         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17091         "public_flag".
17092         * tree.c (tree_nop_conversion): Return true for location wrapper
17093         nodes.
17094         (maybe_wrap_with_location): New function.
17095         (selftest::check_strip_nops): New function.
17096         (selftest::test_location_wrappers): New function.
17097         (selftest::tree_c_tests): Call it.
17098         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17099         (maybe_wrap_with_location): New decl.
17100         (EXPR_LOCATION_WRAPPER_P): New macro.
17101         (location_wrapper_p): New inline function.
17102         (tree_strip_any_location_wrapper): New inline function.
17104 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
17106         PR target/83735
17107         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17108         stack_realign_offset for the largest alignment of stack slot
17109         actually used.
17110         (ix86_find_max_used_stack_alignment): New function.
17111         (ix86_finalize_stack_frame_flags): Use it.  Set
17112         max_used_stack_alignment if we don't realign stack.
17113         * config/i386/i386.h (machine_function): Add
17114         max_used_stack_alignment.
17116 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
17118         * config/arm/arm.opt (-mbranch-cost): New option.
17119         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17120         account.
17122 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
17124         PR target/83629
17125         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17126         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17128 2018-01-10  Richard Biener  <rguenther@suse.de>
17130         PR debug/83765
17131         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17132         early out so it also covers the case where we have a non-NULL
17133         origin.
17135 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
17137         PR tree-optimization/83753
17138         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17139         for non-strided grouped accesses if the number of elements is 1.
17141 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
17143         PR target/81616
17144         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17145         * i386.h (TARGET_USE_GATHER): Define.
17146         * x86-tune.def (X86_TUNE_USE_GATHER): New.
17148 2018-01-10  Martin Liska  <mliska@suse.cz>
17150         PR bootstrap/82831
17151         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17152         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17153         partitioning.
17154         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17155         CLEANUP_NO_PARTITIONING is not set.
17157 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
17159         * doc/rtl.texi: Remove documentation of (const ...) wrappers
17160         for vectors, as a partial revert of r254296.
17161         * rtl.h (const_vec_p): Delete.
17162         (const_vec_duplicate_p): Don't test for vector CONSTs.
17163         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17164         * expmed.c (make_tree): Likewise.
17166         Revert:
17167         * common.md (E, F): Use CONSTANT_P instead of checking for
17168         CONST_VECTOR.
17169         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17170         checking for CONST_VECTOR.
17172 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
17174         PR middle-end/83575
17175         * predict.c (force_edge_cold): Handle in more sane way edges
17176         with no prediction.
17178 2018-01-09  Carl Love  <cel@us.ibm.com>
17180         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17181         V4SI, V4SF types.
17182         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17183         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17184         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17185         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
17186         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17187         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
17188         * config/rs6000/rs6000-protos.h: Add extern defition for
17189         rs6000_generate_float2_double_code.
17190         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17191         function.
17192         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17193         (float2_v2df): Add define_expand.
17195 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
17197         PR target/83628
17198         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17199         op_mode in the force_to_mode call.
17201 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
17203         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17204         instead of checking each element individually.
17205         (aarch64_evpc_uzp): Likewise.
17206         (aarch64_evpc_zip): Likewise.
17207         (aarch64_evpc_ext): Likewise.
17208         (aarch64_evpc_rev): Likewise.
17209         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17210         instead of checking each element individually.  Return true without
17211         generating rtl if
17212         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17213         whether all selected elements come from the same input, instead of
17214         checking each element individually.  Remove calls to gen_rtx_REG,
17215         start_sequence and end_sequence and instead assert that no rtl is
17216         generated.
17218 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
17220         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17221         order of HIGH and CONST checks.
17223 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
17225         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17226         if the destination isn't an SSA_NAME.
17228 2018-01-09  Richard Biener  <rguenther@suse.de>
17230         PR tree-optimization/83668
17231         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17232         move prologue...
17233         (canonicalize_loop_form): ... here, renamed from ...
17234         (canonicalize_loop_closed_ssa_form): ... this and amended to
17235         swap successor edges for loop exit blocks to make us use
17236         the RPO order we need for initial schedule generation.
17238 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
17240         PR tree-optimization/64811
17241         * match.pd: When optimizing comparisons with Inf, avoid
17242         introducing or losing exceptions from comparisons with NaN.
17244 2018-01-09  Martin Liska  <mliska@suse.cz>
17246         PR sanitizer/82517
17247         * asan.c (shadow_mem_size): Add gcc_assert.
17249 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
17251         Don't save registers in main().
17253         PR target/83738
17254         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17255         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17256         * config/avr/avr.c (avr_set_current_function): Don't error if
17257         naked, OS_task or OS_main are specified at the same time.
17258         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17259         OS_main.
17260         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17261         attribute.
17262         * common/config/avr/avr-common.c (avr_option_optimization_table):
17263         Switch on -mmain-is-OS_task for optimizing compilations.
17265 2018-01-09  Richard Biener  <rguenther@suse.de>
17267         PR tree-optimization/83572
17268         * graphite.c: Include cfganal.h.
17269         (graphite_transform_loops): Connect infinite loops to exit
17270         and remove fake edges at the end.
17272 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
17274         * ipa-inline.c (edge_badness): Revert accidental checkin.
17276 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
17278         PR ipa/80763
17279         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17280         symbols; not inline clones.
17282 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
17284         PR target/83507
17285         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17286         hard registers.  Formatting fixes.
17288         PR preprocessor/83722
17289         * gcc.c (try_generate_repro): Pass
17290         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17291         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17292         do_report_bug.
17294 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
17295             Kito Cheng  <kito.cheng@gmail.com>
17297         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17298         (riscv_leaf_function_p): Delete.
17299         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17301 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17303         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17304         function.
17305         (do_ifelse): New function.
17306         (do_isel): New function.
17307         (do_sub3): New function.
17308         (do_add3): New function.
17309         (do_load_mask_compare): New function.
17310         (do_overlap_load_compare): New function.
17311         (expand_compare_loop): New function.
17312         (expand_block_compare): Call expand_compare_loop() when appropriate.
17313         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17314         option description.
17315         (-mblock-compare-inline-loop-limit): New option.
17317 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17319         PR target/83677
17320         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17321         Reverse order of second and third operands in first alternative.
17322         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17323         of first and second elements in UNSPEC_VPERMR vector.
17324         (altivec_expand_vec_perm_le): Likewise.
17326 2018-01-08  Jeff Law  <law@redhat.com>
17328         PR rtl-optimizatin/81308
17329         * tree-switch-conversion.c (cfg_altered): New file scoped static.
17330         (process_switch): If group_case_labels makes a change, then set
17331         cfg_altered.
17332         (pass_convert_switch::execute): If a switch is converted, then
17333         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
17335         PR rtl-optimization/81308
17336         * recog.c (split_all_insns): Conditionally cleanup the CFG after
17337         splitting insns.
17339 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
17341         PR target/83663 - Revert r255946
17342         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17343         generation for cases where splatting a value is not useful.
17344         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17345         across a vec_duplicate and a paradoxical subreg forming a vector
17346         mode to a vec_concat.
17348 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17350         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17351         -march=armv8.3-a variants.
17352         * config/arm/t-multilib: Likewise.
17353         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
17355 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17357         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17358         to generate rtl.
17359         (cceq_ior_compare_complement): Give it a name so I can use it, and
17360         change boolean_or_operator predicate to boolean_operator so it can
17361         be used to generate a crand.
17362         (eqne): New code iterator.
17363         (bd/bd_neg): New code_attrs.
17364         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17365         a single define_insn.
17366         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17367         decrement (bdnzt/bdnzf/bdzt/bdzf).
17368         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17369         with the new names of the branch decrement patterns, and added the
17370         names of the branch decrement conditional patterns.
17372 2018-01-08  Richard Biener  <rguenther@suse.de>
17374         PR tree-optimization/83563
17375         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17376         cache.
17378 2018-01-08  Richard Biener  <rguenther@suse.de>
17380         PR middle-end/83713
17381         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17383 2018-01-08  Richard Biener  <rguenther@suse.de>
17385         PR tree-optimization/83685
17386         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17387         references to abnormals.
17389 2018-01-08  Richard Biener  <rguenther@suse.de>
17391         PR lto/83719
17392         * dwarf2out.c (output_indirect_strings): Handle empty
17393         skeleton_debug_str_hash.
17394         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17396 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
17398         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17399         (emit_store_direct): Likewise.
17400         (arc_trampoline_adjust_address): Likewise.
17401         (arc_asm_trampoline_template): New function.
17402         (arc_initialize_trampoline): Use asm_trampoline_template.
17403         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17404         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17405         * config/arc/arc.md (flush_icache): Delete pattern.
17407 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
17409         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17410         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17411         munaligned-access.
17413 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
17415         PR target/83681
17416         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17417         by not USED_FOR_TARGET.
17418         (make_pass_resolve_sw_modes): Likewise.
17420 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
17422         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17423         USED_FOR_TARGET.
17425 2018-01-08  Richard Biener  <rguenther@suse.de>
17427         PR middle-end/83580
17428         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17430 2018-01-08  Richard Biener  <rguenther@suse.de>
17432         PR middle-end/83517
17433         * match.pd ((t * 2) / 2) -> t): Add missing :c.
17435 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
17437         PR middle-end/81897
17438         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17439         basic blocks with a small number of successors.
17440         (convert_control_dep_chain_into_preds): Improve handling of
17441         forwarder blocks.
17442         (dump_predicates): Split apart into...
17443         (dump_pred_chain): ...here...
17444         (dump_pred_info): ...and here.
17445         (can_one_predicate_be_invalidated_p): Add debugging printfs.
17446         (can_chain_union_be_invalidated_p): Improve check for invalidation
17447         of paths.
17448         (uninit_uses_cannot_happen): Avoid unnecessary if
17449         convert_control_dep_chain_into_preds yielded nothing.
17451 2018-01-06  Martin Sebor  <msebor@redhat.com>
17453         PR tree-optimization/83640
17454         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17455         subtracting negative offset from size.
17456         (builtin_access::overlap): Adjust offset bounds of the access to fall
17457         within the size of the object if possible.
17459 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
17461         PR rtl-optimization/83699
17462         * expmed.c (extract_bit_field_1): Restrict the vector usage of
17463         extract_bit_field_as_subreg to cases in which the extracted
17464         value is also a vector.
17466         * lra-constraints.c (process_alt_operands): Test for the equivalence
17467         substitutions when detecting a possible reload cycle.
17469 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
17471         PR debug/83480
17472         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17473         by default if flag_selective_schedling{,2}.  Formatting fixes.
17475         PR rtl-optimization/83682
17476         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17477         if it has non-VECTOR_MODE element mode.
17478         (vec_duplicate_p): Likewise.
17480         PR middle-end/83694
17481         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17482         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17484 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
17486         PR target/83604
17487         * config/i386/i386-builtin.def
17488         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17489         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17490         Require also OPTION_MASK_ISA_AVX512F in addition to
17491         OPTION_MASK_ISA_GFNI.
17492         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17493         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17494         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17495         to OPTION_MASK_ISA_GFNI.
17496         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17497         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17498         OPTION_MASK_ISA_AVX512BW.
17499         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17500         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17501         addition to OPTION_MASK_ISA_GFNI.
17502         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17503         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17504         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17505         to OPTION_MASK_ISA_GFNI.
17506         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17507         a requirement for all ISAs rather than any of them with a few
17508         exceptions.
17509         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17510         processing.
17511         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17512         bitmasks to be enabled with 3 exceptions, instead of requiring any
17513         enabled ISA with lots of exceptions.
17514         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17515         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17516         Change avx512bw in isa attribute to avx512f.
17517         * config/i386/sgxintrin.h: Add license boilerplate.
17518         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
17519         to __AVX512F__ and __AVX512VL to __AVX512VL__.
17520         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17521         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17522         defined.
17523         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17524         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17525         temporarily sse2 rather than sse if not enabled already.
17527         PR target/83604
17528         * config/i386/sse.md (VI248_VLBW): Rename to ...
17529         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
17530         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17531         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17532         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17533         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17534         mode iterator instead of VI248_VLBW.
17536 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
17538         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17539         (record_modified): Skip clobbers; add debug output.
17540         (param_change_prob): Use sreal frequencies.
17542 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
17544         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17545         punt for user-aligned variables.
17547 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
17549         * tree-chrec.c (chrec_contains_symbols): Return true for
17550         POLY_INT_CST.
17552 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
17554         PR target/82439
17555         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17556         of (x|y) == x for BICS pattern.
17558 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
17560         PR tree-optimization/83605
17561         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17562         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17563         can throw.
17565 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
17567         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17568         * config/epiphany/rtems.h: New file.
17570 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
17571             Uros Bizjak  <ubizjak@gmail.com>
17573         PR target/83554
17574         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17575         QIreg_operand instead of register_operand predicate.
17576         * config/i386/i386.c (ix86_rop_should_change_byte_p,
17577         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17578         comments instead of -fmitigate[-_]rop.
17580 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17582         PR bootstrap/81926
17583         * cgraphunit.c (symbol_table::compile): Switch to text_section
17584         before calling assembly_start debug hook.
17585         * run-rtl-passes.c (run_rtl_passes): Likewise.
17586         Include output.h.
17588 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17590         * tree-vrp.c (extract_range_from_binary_expr_1): Check
17591         range_int_cst_p rather than !symbolic_range_p before calling
17592         extract_range_from_multiplicative_op_1.
17594 2018-01-04  Jeff Law  <law@redhat.com>
17596         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17597         redundant test in assertion.
17599 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17601         * doc/rtl.texi: Document machine_mode wrapper classes.
17603 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17605         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17606         using tree_to_uhwi.
17608 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17610         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17611         the VEC_PERM_EXPR fold to fail.
17613 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
17615         PR debug/83585
17616         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17617         to switched_sections.
17619 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17621         PR target/83680
17622         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17623         test for d.testing.
17625 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
17627         PR target/83387
17628         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17629         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17631 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
17633         PR debug/83666
17634         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17635         is BLKmode and bitpos not zero or mode change is needed.
17637 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
17639         PR target/83675
17640         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17641         TARGET_VIS2.
17643 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
17645         PR target/83628
17646         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17647         instead of MULT rtx.  Update all corresponding splitters.
17648         (*saddl_se): Ditto.
17649         (*ssub<modesuffix>): Ditto.
17650         (*ssubl_se): Ditto.
17651         (*cmp_sadd_di): Update split patterns.
17652         (*cmp_sadd_si): Ditto.
17653         (*cmp_sadd_sidi): Ditto.
17654         (*cmp_ssub_di): Ditto.
17655         (*cmp_ssub_si): Ditto.
17656         (*cmp_ssub_sidi): Ditto.
17657         * config/alpha/predicates.md (const23_operand): New predicate.
17658         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17659         Look for ASHIFT, not MULT inner operand.
17660         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17662 2018-01-04  Martin Liska  <mliska@suse.cz>
17664         PR gcov-profile/83669
17665         * gcov.c (output_intermediate_file): Add version to intermediate
17666         gcov file.
17667         * doc/gcov.texi: Document new field 'version' in intermediate
17668         file format. Fix location of '-k' option of gcov command.
17670 2018-01-04  Martin Liska  <mliska@suse.cz>
17672         PR ipa/82352
17673         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17675 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
17677         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17679 2018-01-03  Martin Sebor  <msebor@redhat.com>
17681         PR tree-optimization/83655
17682         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17683         checking calls with invalid arguments.
17685 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17687         * tree-vect-stmts.c (vect_get_store_rhs): New function.
17688         (vectorizable_mask_load_store): Delete.
17689         (vectorizable_call): Return false for masked loads and stores.
17690         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
17691         instead of gimple_assign_rhs1.
17692         (vectorizable_load): Handle IFN_MASK_LOAD.
17693         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17695 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17697         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17698         split out from..,
17699         (vectorizable_mask_load_store): ...here.
17700         (vectorizable_load): ...and here.
17702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17704         * tree-vect-stmts.c (vect_build_all_ones_mask)
17705         (vect_build_zero_merge_argument): New functions, split out from...
17706         (vectorizable_load): ...here.
17708 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17710         * tree-vect-stmts.c (vect_check_store_rhs): New function,
17711         split out from...
17712         (vectorizable_mask_load_store): ...here.
17713         (vectorizable_store): ...and here.
17715 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17717         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17718         split out from...
17719         (vectorizable_mask_load_store): ...here.
17721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17723         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17724         (vect_model_store_cost): Take a vec_load_store_type instead of a
17725         vect_def_type.
17726         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17727         (vect_model_store_cost): Take a vec_load_store_type instead of a
17728         vect_def_type.
17729         (vectorizable_mask_load_store): Update accordingly.
17730         (vectorizable_store): Likewise.
17731         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17735         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17736         IFN_MASK_LOAD calls here rather than...
17737         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17739 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17740             Alan Hayward  <alan.hayward@arm.com>
17741             David Sherwood  <david.sherwood@arm.com>
17743         * expmed.c (extract_bit_field_1): For vector extracts,
17744         fall back to extract_bit_field_as_subreg if vec_extract
17745         isn't available.
17747 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17748             Alan Hayward  <alan.hayward@arm.com>
17749             David Sherwood  <david.sherwood@arm.com>
17751         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17752         they are variable or constant sized.
17753         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17754         slots for constant-sized data.
17756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17757             Alan Hayward  <alan.hayward@arm.com>
17758             David Sherwood  <david.sherwood@arm.com>
17760         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17761         handling COND_EXPRs with boolean comparisons, try to find a better
17762         basis for the mask type than the boolean itself.
17764 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17766         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17767         is calculated and how it can be overridden.
17768         * genmodes.c (max_bitsize_mode_any_mode): New variable.
17769         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17770         if defined.
17771         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17772         if nonzero.
17774 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17775             Alan Hayward  <alan.hayward@arm.com>
17776             David Sherwood  <david.sherwood@arm.com>
17778         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17779         Remove the mode argument.
17780         (aarch64_simd_valid_immediate): Remove the mode and inverse
17781         arguments.
17782         * config/aarch64/iterators.md (bitsize): New iterator.
17783         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17784         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17785         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17786         aarch64_simd_valid_immediate.
17787         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17788         (aarch64_reg_or_bic_imm): Likewise.
17789         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17790         with an insn_type enum and msl with a modifier_type enum.
17791         Replace element_width with a scalar_mode.  Change the shift
17792         to unsigned int.  Add constructors for scalar_float_mode and
17793         scalar_int_mode elements.
17794         (aarch64_vect_float_const_representable_p): Delete.
17795         (aarch64_can_const_movi_rtx_p)
17796         (aarch64_simd_scalar_immediate_valid_for_move)
17797         (aarch64_simd_make_constant): Update call to
17798         aarch64_simd_valid_immediate.
17799         (aarch64_advsimd_valid_immediate_hs): New function.
17800         (aarch64_advsimd_valid_immediate): Likewise.
17801         (aarch64_simd_valid_immediate): Remove mode and inverse
17802         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
17803         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17804         and aarch64_float_const_representable_p on the result.
17805         (aarch64_output_simd_mov_immediate): Remove mode argument.
17806         Update call to aarch64_simd_valid_immediate and use of
17807         simd_immediate_info.
17808         (aarch64_output_scalar_simd_mov_immediate): Update call
17809         accordingly.
17811 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17812             Alan Hayward  <alan.hayward@arm.com>
17813             David Sherwood  <david.sherwood@arm.com>
17815         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17816         (mode_nunits): Likewise CONST_MODE_NUNITS.
17817         * machmode.def (ADJUST_NUNITS): Document.
17818         * genmodes.c (mode_data::need_nunits_adj): New field.
17819         (blank_mode): Update accordingly.
17820         (adj_nunits): New variable.
17821         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17822         parameter.
17823         (emit_mode_size_inline): Set need_bytesize_adj for all modes
17824         listed in adj_nunits.
17825         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17826         listed in adj_nunits.  Don't emit case statements for such modes.
17827         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17828         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
17829         nothing if adj_nunits is nonnull.
17830         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17831         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17832         (emit_mode_fbit): Update use of print_maybe_const_decl.
17833         (emit_move_size): Likewise.  Treat the array as non-const
17834         if adj_nunits.
17835         (emit_mode_adjustments): Handle adj_nunits.
17837 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17839         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17840         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17841         (VECTOR_MODES): Use it.
17842         (make_vector_modes): Take the prefix as an argument.
17844 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17845             Alan Hayward  <alan.hayward@arm.com>
17846             David Sherwood  <david.sherwood@arm.com>
17848         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17849         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17850         for MODE_VECTOR_BOOL.
17851         * machmode.def (VECTOR_BOOL_MODE): Document.
17852         * genmodes.c (VECTOR_BOOL_MODE): New macro.
17853         (make_vector_bool_mode): New function.
17854         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17855         MODE_VECTOR_BOOL.
17856         * lto-streamer-in.c (lto_input_mode_table): Likewise.
17857         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17858         Likewise.
17859         * stor-layout.c (int_mode_for_mode): Likewise.
17860         * tree.c (build_vector_type_for_mode): Likewise.
17861         * varasm.c (output_constant_pool_2): Likewise.
17862         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17863         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
17864         for MODE_VECTOR_BOOL.
17865         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17866         of mode class checks.
17867         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17868         instead of a list of mode class checks.
17869         (expand_vector_scalar_condition): Likewise.
17870         (type_for_widest_vector_mode): Handle BImode as an inner mode.
17872 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17873             Alan Hayward  <alan.hayward@arm.com>
17874             David Sherwood  <david.sherwood@arm.com>
17876         * machmode.h (mode_size): Change from unsigned short to
17877         poly_uint16_pod.
17878         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17879         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17880         or if measurement_type is not polynomial.
17881         (fixed_size_mode::includes_p): Check for constant-sized modes.
17882         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17883         return a poly_uint16 rather than an unsigned short.
17884         (emit_mode_size): Change the type of mode_size from unsigned short
17885         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
17886         (emit_mode_adjustments): Cope with polynomial vector sizes.
17887         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17888         for GET_MODE_SIZE.
17889         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17890         for GET_MODE_SIZE.
17891         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17892         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17893         * caller-save.c (setup_save_areas): Likewise.
17894         (replace_reg_with_saved_mem): Likewise.
17895         * calls.c (emit_library_call_value_1): Likewise.
17896         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17897         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17898         (gen_lowpart_for_combine): Likewise.
17899         * convert.c (convert_to_integer_1): Likewise.
17900         * cse.c (equiv_constant, cse_insn): Likewise.
17901         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17902         (cselib_subst_to_values): Likewise.
17903         * dce.c (word_dce_process_block): Likewise.
17904         * df-problems.c (df_word_lr_mark_ref): Likewise.
17905         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17906         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17907         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17908         (rtl_for_decl_location): Likewise.
17909         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17910         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17911         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17912         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17913         (expand_expr_real_1): Likewise.
17914         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17915         (pad_below): Likewise.
17916         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17917         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17918         * ira.c (get_subreg_tracking_sizes): Likewise.
17919         * ira-build.c (ira_create_allocno_objects): Likewise.
17920         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17921         (ira_sort_regnos_for_alter_reg): Likewise.
17922         * ira-costs.c (record_operand_costs): Likewise.
17923         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17924         (resolve_simple_move): Likewise.
17925         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17926         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17927         (lra_constraints): Likewise.
17928         (CONST_POOL_OK_P): Reject variable-sized modes.
17929         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17930         (add_pseudo_to_slot, lra_spill): Likewise.
17931         * omp-low.c (omp_clause_aligned_alignment): Likewise.
17932         * optabs-query.c (get_best_extraction_insn): Likewise.
17933         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17934         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17935         (expand_mult_highpart, valid_multiword_target_p): Likewise.
17936         * recog.c (offsettable_address_addr_space_p): Likewise.
17937         * regcprop.c (maybe_mode_change): Likewise.
17938         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17939         * regrename.c (build_def_use): Likewise.
17940         * regstat.c (dump_reg_info): Likewise.
17941         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17942         (find_reloads, find_reloads_subreg_address): Likewise.
17943         * reload1.c (eliminate_regs_1): Likewise.
17944         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17945         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17946         (simplify_binary_operation_1, simplify_subreg): Likewise.
17947         * targhooks.c (default_function_arg_padding): Likewise.
17948         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17949         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17950         (verify_gimple_assign_ternary): Likewise.
17951         * tree-inline.c (estimate_move_cost): Likewise.
17952         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17953         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17954         (get_address_cost_ainc): Likewise.
17955         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17956         (vect_supportable_dr_alignment): Likewise.
17957         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17958         (vectorizable_reduction): Likewise.
17959         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17960         (vectorizable_operation, vectorizable_load): Likewise.
17961         * tree.c (build_same_sized_truth_vector_type): Likewise.
17962         * valtrack.c (cleanup_auto_inc_dec): Likewise.
17963         * var-tracking.c (emit_note_insn_var_location): Likewise.
17964         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17965         (ADDR_VEC_ALIGN): Likewise.
17967 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17968             Alan Hayward  <alan.hayward@arm.com>
17969             David Sherwood  <david.sherwood@arm.com>
17971         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17972         unsigned short.
17973         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17974         or if measurement_type is polynomial.
17975         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17976         * combine.c (make_extraction): Likewise.
17977         * dse.c (find_shift_sequence): Likewise.
17978         * dwarf2out.c (mem_loc_descriptor): Likewise.
17979         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17980         (extract_bit_field, extract_low_bits): Likewise.
17981         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17982         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17983         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17984         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17985         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17986         * reload.c (find_reloads): Likewise.
17987         * reload1.c (alter_reg): Likewise.
17988         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17989         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17990         * tree-if-conv.c (predicate_mem_writes): Likewise.
17991         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17992         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17993         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17994         * valtrack.c (dead_debug_insert_temp): Likewise.
17995         * varasm.c (mergeable_constant_section): Likewise.
17996         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17998 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17999             Alan Hayward  <alan.hayward@arm.com>
18000             David Sherwood  <david.sherwood@arm.com>
18002         * expr.c (expand_assignment): Cope with polynomial mode sizes
18003         when assigning to a CONCAT.
18005 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18006             Alan Hayward  <alan.hayward@arm.com>
18007             David Sherwood  <david.sherwood@arm.com>
18009         * machmode.h (mode_precision): Change from unsigned short to
18010         poly_uint16_pod.
18011         (mode_to_precision): Return a poly_uint16 rather than an unsigned
18012         short.
18013         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18014         or if measurement_type is not polynomial.
18015         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
18016         in which the mode is already known to be a scalar_int_mode.
18017         * genmodes.c (emit_mode_precision): Change the type of mode_precision
18018         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
18019         initializer.
18020         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18021         for GET_MODE_PRECISION.
18022         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18023         for GET_MODE_PRECISION.
18024         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18025         as polynomial.
18026         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18027         (expand_field_assignment, make_extraction): Likewise.
18028         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18029         (get_last_value): Likewise.
18030         * convert.c (convert_to_integer_1): Likewise.
18031         * cse.c (cse_insn): Likewise.
18032         * expr.c (expand_expr_real_1): Likewise.
18033         * lra-constraints.c (simplify_operand_subreg): Likewise.
18034         * optabs-query.c (can_atomic_load_p): Likewise.
18035         * optabs.c (expand_atomic_load): Likewise.
18036         (expand_atomic_store): Likewise.
18037         * ree.c (combine_reaching_defs): Likewise.
18038         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18039         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18040         * tree.h (type_has_mode_precision_p): Likewise.
18041         * ubsan.c (instrument_si_overflow): Likewise.
18043 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18044             Alan Hayward  <alan.hayward@arm.com>
18045             David Sherwood  <david.sherwood@arm.com>
18047         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18048         polynomial numbers of units.
18049         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18050         (valid_vector_subparts_p): New function.
18051         (build_vector_type): Remove temporary shim and take the number
18052         of units as a poly_uint64 rather than an int.
18053         (build_opaque_vector_type): Take the number of units as a
18054         poly_uint64 rather than an int.
18055         * tree.c (build_vector_from_ctor): Handle polynomial
18056         TYPE_VECTOR_SUBPARTS.
18057         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18058         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18059         (build_vector_from_val): If the number of units is variable,
18060         use build_vec_duplicate_cst for constant operands and
18061         VEC_DUPLICATE_EXPR otherwise.
18062         (make_vector_type): Remove temporary is_constant ().
18063         (build_vector_type, build_opaque_vector_type): Take the number of
18064         units as a poly_uint64 rather than an int.
18065         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18066         VECTOR_CST_NELTS.
18067         * cfgexpand.c (expand_debug_expr): Likewise.
18068         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18069         (store_constructor, expand_expr_real_1): Likewise.
18070         (const_scalar_mask_from_tree): Likewise.
18071         * fold-const-call.c (fold_const_reduction): Likewise.
18072         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18073         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18074         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18075         (fold_relational_const): Likewise.
18076         (native_interpret_vector): Likewise.  Change the size from an
18077         int to an unsigned int.
18078         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18079         TYPE_VECTOR_SUBPARTS.
18080         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18081         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18082         duplicating a non-constant operand into a variable-length vector.
18083         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18084         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18085         * ipa-icf.c (sem_variable::equals): Likewise.
18086         * match.pd: Likewise.
18087         * omp-simd-clone.c (simd_clone_subparts): Likewise.
18088         * print-tree.c (print_node): Likewise.
18089         * stor-layout.c (layout_type): Likewise.
18090         * targhooks.c (default_builtin_vectorization_cost): Likewise.
18091         * tree-cfg.c (verify_gimple_comparison): Likewise.
18092         (verify_gimple_assign_binary): Likewise.
18093         (verify_gimple_assign_ternary): Likewise.
18094         (verify_gimple_assign_single): Likewise.
18095         * tree-pretty-print.c (dump_generic_node): Likewise.
18096         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18097         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18098         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18099         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18100         (vect_shift_permute_load_chain): Likewise.
18101         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18102         (expand_vector_condition, optimize_vector_constructor): Likewise.
18103         (lower_vec_perm, get_compute_type): Likewise.
18104         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18105         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18106         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18107         (vect_recog_mask_conversion_pattern): Likewise.
18108         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18109         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18110         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18111         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18112         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18113         (vectorizable_shift, vectorizable_operation, vectorizable_store)
18114         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18115         (supportable_widening_operation): Likewise.
18116         (supportable_narrowing_operation): Likewise.
18117         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18118         Likewise.
18119         * varasm.c (output_constant): Likewise.
18121 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18122             Alan Hayward  <alan.hayward@arm.com>
18123             David Sherwood  <david.sherwood@arm.com>
18125         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18126         so that both the length == 3 and length != 3 cases set up their
18127         own permute vectors.  Add comments explaining why we know the
18128         number of elements is constant.
18129         (vect_permute_load_chain): Likewise.
18131 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18132             Alan Hayward  <alan.hayward@arm.com>
18133             David Sherwood  <david.sherwood@arm.com>
18135         * machmode.h (mode_nunits): Change from unsigned char to
18136         poly_uint16_pod.
18137         (ONLY_FIXED_SIZE_MODES): New macro.
18138         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18139         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18140         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18141         New typedefs.
18142         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18143         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18144         or if measurement_type is not polynomial.
18145         * genmodes.c (ZERO_COEFFS): New macro.
18146         (emit_mode_nunits_inline): Make mode_nunits_inline return a
18147         poly_uint16.
18148         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18149         Use ZERO_COEFFS when emitting initializers.
18150         * data-streamer.h (bp_pack_poly_value): New function.
18151         (bp_unpack_poly_value): Likewise.
18152         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18153         for GET_MODE_NUNITS.
18154         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18155         for GET_MODE_NUNITS.
18156         * tree.c (make_vector_type): Remove temporary shim and make
18157         the real function take the number of units as a poly_uint64
18158         rather than an int.
18159         (build_vector_type_for_mode): Handle polynomial nunits.
18160         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18161         * emit-rtl.c (const_vec_series_p_1): Likewise.
18162         (gen_rtx_CONST_VECTOR): Likewise.
18163         * fold-const.c (test_vec_duplicate_folding): Likewise.
18164         * genrecog.c (validate_pattern): Likewise.
18165         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18166         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18167         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18168         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18169         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18170         * rtlanal.c (subreg_get_info): Likewise.
18171         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18172         (vect_grouped_load_supported): Likewise.
18173         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18174         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18175         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18176         (simplify_const_unary_operation, simplify_binary_operation_1)
18177         (simplify_const_binary_operation, simplify_ternary_operation)
18178         (test_vector_ops_duplicate, test_vector_ops): Likewise.
18179         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18180         instead of CONST_VECTOR_NUNITS.
18181         * varasm.c (output_constant_pool_2): Likewise.
18182         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18183         explicit-encoded elements in the XVEC for variable-length vectors.
18185 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18187         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18189 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18190             Alan Hayward  <alan.hayward@arm.com>
18191             David Sherwood  <david.sherwood@arm.com>
18193         * coretypes.h (fixed_size_mode): Declare.
18194         (fixed_size_mode_pod): New typedef.
18195         * builtins.h (target_builtins::x_apply_args_mode)
18196         (target_builtins::x_apply_result_mode): Change type to
18197         fixed_size_mode_pod.
18198         * builtins.c (apply_args_size, apply_result_size, result_vector)
18199         (expand_builtin_apply_args_1, expand_builtin_apply)
18200         (expand_builtin_return): Update accordingly.
18202 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18204         * cse.c (hash_rtx_cb): Hash only the encoded elements.
18205         * cselib.c (cselib_hash_rtx): Likewise.
18206         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18207         CONST_VECTOR encoding.
18209 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
18210             Jeff Law  <law@redhat.com>
18212         PR target/83641
18213         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18214         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18215         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18216         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18218         PR target/83641
18219         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18220         explicitly probe *sp in a noreturn function if there were any callee
18221         register saves or frame pointer is needed.
18223 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
18225         PR debug/83621
18226         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18227         BLKmode for ternary, binary or unary expressions.
18229         PR debug/83645
18230         * var-tracking.c (delete_vta_debug_insn): New inline function.
18231         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18232         insns from get_insns () to NULL instead of each bb separately.
18233         Use delete_vta_debug_insn.  No longer static.
18234         (vt_debug_insns_local, variable_tracking_main_1): Adjust
18235         delete_vta_debug_insns callers.
18236         * rtl.h (delete_vta_debug_insns): Declare.
18237         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18238         instead of variable_tracking_main.
18240 2018-01-03  Martin Sebor  <msebor@redhat.com>
18242         PR tree-optimization/83603
18243         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18244         arguments past the endof the argument list in functions declared
18245         without a prototype.
18246         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18247         Avoid checking when arguments are null.
18249 2018-01-03  Martin Sebor  <msebor@redhat.com>
18251         PR c/83559
18252         * doc/extend.texi (attribute const): Fix a typo.
18253         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18254         issuing -Wsuggest-attribute for void functions.
18256 2018-01-03  Martin Sebor  <msebor@redhat.com>
18258         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18259         offset_int::from instead of wide_int::to_shwi.
18260         (maybe_diag_overlap): Remove assertion.
18261         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18262         * gimple-ssa-sprintf.c (format_directive): Same.
18263         (parse_directive): Same.
18264         (sprintf_dom_walker::compute_format_length): Same.
18265         (try_substitute_return_value): Same.
18267 2018-01-03  Jeff Law  <law@redhat.com>
18269         PR middle-end/83654
18270         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18271         non-constant residual for zero at runtime and avoid probing in
18272         that case.  Reorganize code for trailing problem to mirror handling
18273         of the residual.
18275 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18277         PR tree-optimization/83501
18278         * tree-ssa-strlen.c (get_string_cst): New.
18279         (handle_char_store): Call get_string_cst.
18281 2018-01-03  Martin Liska  <mliska@suse.cz>
18283         PR tree-optimization/83593
18284         * tree-ssa-strlen.c: Include tree-cfg.h.
18285         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18286         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18287         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18288         to false.
18289         (strlen_dom_walker::before_dom_children): Call
18290         gimple_purge_dead_eh_edges. Dump tranformation with details
18291         dump flags.
18292         (strlen_dom_walker::before_dom_children): Update call by adding
18293         new argument cleanup_eh.
18294         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18296 2018-01-03  Martin Liska  <mliska@suse.cz>
18298         PR ipa/83549
18299         * cif-code.def (VARIADIC_THUNK): New enum value.
18300         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18301         thunks.
18303 2018-01-03  Jan Beulich  <jbeulich@suse.com>
18305         * sse.md (mov<mode>_internal): Tighten condition for when to use
18306         vmovdqu<ssescalarsize> for TI and OI modes.
18308 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
18310         Update copyright years.
18312 2018-01-03  Martin Liska  <mliska@suse.cz>
18314         PR ipa/83594
18315         * ipa-visibility.c (function_and_variable_visibility): Skip
18316         functions with noipa attribure.
18318 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
18320         * gcc.c (process_command): Update copyright notice dates.
18321         * gcov-dump.c (print_version): Ditto.
18322         * gcov.c (print_version): Ditto.
18323         * gcov-tool.c (print_version): Ditto.
18324         * gengtype.c (create_file): Ditto.
18325         * doc/cpp.texi: Bump @copying's copyright year.
18326         * doc/cppinternals.texi: Ditto.
18327         * doc/gcc.texi: Ditto.
18328         * doc/gccint.texi: Ditto.
18329         * doc/gcov.texi: Ditto.
18330         * doc/install.texi: Ditto.
18331         * doc/invoke.texi: Ditto.
18333 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18335         * vector-builder.h (vector_builder::m_full_nelts): Change from
18336         unsigned int to poly_uint64.
18337         (vector_builder::full_nelts): Update prototype accordingly.
18338         (vector_builder::new_vector): Likewise.
18339         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18340         (vector_builder::operator ==): Likewise.
18341         (vector_builder::finalize): Likewise.
18342         * int-vector-builder.h (int_vector_builder::int_vector_builder):
18343         Take the number of elements as a poly_uint64 rather than an
18344         unsigned int.
18345         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18346         from unsigned int to poly_uint64.
18347         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18348         (vec_perm_indices::new_vector): Likewise.
18349         (vec_perm_indices::length): Likewise.
18350         (vec_perm_indices::nelts_per_input): Likewise.
18351         (vec_perm_indices::input_nelts): Likewise.
18352         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18353         number of elements per input as a poly_uint64 rather than an
18354         unsigned int.  Use the original encoding for variable-length
18355         vectors, rather than clamping each individual element.
18356         For the second and subsequent elements in each pattern,
18357         clamp the step and base before clamping their sum.
18358         (vec_perm_indices::series_p): Handle polynomial element counts.
18359         (vec_perm_indices::all_in_range_p): Likewise.
18360         (vec_perm_indices_to_tree): Likewise.
18361         (vec_perm_indices_to_rtx): Likewise.
18362         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18363         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18364         (tree_vector_builder::new_binary_operation): Handle polynomial
18365         element counts.  Return false if we need to know the number
18366         of elements at compile time.
18367         * fold-const.c (fold_vec_perm): Punt if the number of elements
18368         isn't known at compile time.
18370 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18372         * vec-perm-indices.h (vec_perm_builder): Change element type
18373         from HOST_WIDE_INT to poly_int64.
18374         (vec_perm_indices::element_type): Update accordingly.
18375         (vec_perm_indices::clamp): Handle polynomial element_types.
18376         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18377         (vec_perm_indices::all_in_range_p): Likewise.
18378         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18379         than shwi trees.
18380         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18381         polynomial vec_perm_indices element types.
18382         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18383         * fold-const.c (fold_vec_perm): Likewise.
18384         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18385         * tree-vect-generic.c (lower_vec_perm): Likewise.
18386         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18387         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18388         element type to HOST_WIDE_INT.
18390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18391             Alan Hayward  <alan.hayward@arm.com>
18392             David Sherwood  <david.sherwood@arm.com>
18394         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18395         rather than an int.  Use plus_constant.
18396         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18397         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18400             Alan Hayward  <alan.hayward@arm.com>
18401             David Sherwood  <david.sherwood@arm.com>
18403         * calls.c (emit_call_1, expand_call): Change struct_value_size from
18404         a HOST_WIDE_INT to a poly_int64.
18406 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18407             Alan Hayward  <alan.hayward@arm.com>
18408             David Sherwood  <david.sherwood@arm.com>
18410         * calls.c (load_register_parameters): Cope with polynomial
18411         mode sizes.  Require a constant size for BLKmode parameters
18412         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
18413         forces a parameter to be padded at the lsb end in order to
18414         fill a complete number of words, require the parameter size
18415         to be ordered wrt UNITS_PER_WORD.
18417 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18418             Alan Hayward  <alan.hayward@arm.com>
18419             David Sherwood  <david.sherwood@arm.com>
18421         * reload1.c (spill_stack_slot_width): Change element type
18422         from unsigned int to poly_uint64_pod.
18423         (alter_reg): Treat mode sizes as polynomial.
18425 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18426             Alan Hayward  <alan.hayward@arm.com>
18427             David Sherwood  <david.sherwood@arm.com>
18429         * reload.c (complex_word_subreg_p): New function.
18430         (reload_inner_reg_of_subreg, push_reload): Use it.
18432 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18433             Alan Hayward  <alan.hayward@arm.com>
18434             David Sherwood  <david.sherwood@arm.com>
18436         * lra-constraints.c (process_alt_operands): Reject matched
18437         operands whose sizes aren't ordered.
18438         (match_reload): Refer to this check here.
18440 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18441             Alan Hayward  <alan.hayward@arm.com>
18442             David Sherwood  <david.sherwood@arm.com>
18444         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18445         that the mode size is in the set {1, 2, 4, 8, 16}.
18447 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18448             Alan Hayward  <alan.hayward@arm.com>
18449             David Sherwood  <david.sherwood@arm.com>
18451         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18452         Use plus_constant instead of gen_rtx_PLUS.
18454 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18455             Alan Hayward  <alan.hayward@arm.com>
18456             David Sherwood  <david.sherwood@arm.com>
18458         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18459         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18460         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18461         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18462         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18463         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18464         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18465         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18466         * config/i386/i386.c (ix86_push_rounding): ...this new function.
18467         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18468         a poly_int64.
18469         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18470         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18471         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18472         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18473         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18474         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18475         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18476         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18477         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18478         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18479         function.
18480         * expr.c (emit_move_resolve_push): Treat the input and result
18481         of PUSH_ROUNDING as a poly_int64.
18482         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18483         (emit_push_insn): Likewise.
18484         * lra-eliminations.c (mark_not_eliminable): Likewise.
18485         * recog.c (push_operand): Likewise.
18486         * reload1.c (elimination_effects): Likewise.
18487         * rtlanal.c (nonzero_bits1): Likewise.
18488         * calls.c (store_one_arg): Likewise.  Require the padding to be
18489         known at compile time.
18491 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18492             Alan Hayward  <alan.hayward@arm.com>
18493             David Sherwood  <david.sherwood@arm.com>
18495         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18496         Use plus_constant instead of gen_rtx_PLUS.
18498 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18499             Alan Hayward  <alan.hayward@arm.com>
18500             David Sherwood  <david.sherwood@arm.com>
18502         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18503         rather than an int.
18505 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18506             Alan Hayward  <alan.hayward@arm.com>
18507             David Sherwood  <david.sherwood@arm.com>
18509         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18510         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18511         via stack temporaries.  Treat the mode size as polynomial too.
18513 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18514             Alan Hayward  <alan.hayward@arm.com>
18515             David Sherwood  <david.sherwood@arm.com>
18517         * expr.c (expand_expr_real_2): When handling conversions involving
18518         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18519         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
18520         as a poly_uint64 too.
18522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18523             Alan Hayward  <alan.hayward@arm.com>
18524             David Sherwood  <david.sherwood@arm.com>
18526         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18528 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18529             Alan Hayward  <alan.hayward@arm.com>
18530             David Sherwood  <david.sherwood@arm.com>
18532         * combine.c (can_change_dest_mode): Handle polynomial
18533         REGMODE_NATURAL_SIZE.
18534         * expmed.c (store_bit_field_1): Likewise.
18535         * expr.c (store_constructor): Likewise.
18536         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18537         and polynomial REGMODE_NATURAL_SIZE.
18538         (gen_lowpart_common): Likewise.
18539         * reginfo.c (record_subregs_of_mode): Likewise.
18540         * rtlanal.c (read_modify_subreg_p): Likewise.
18542 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18543             Alan Hayward  <alan.hayward@arm.com>
18544             David Sherwood  <david.sherwood@arm.com>
18546         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18547         numbers of elements.
18549 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18550             Alan Hayward  <alan.hayward@arm.com>
18551             David Sherwood  <david.sherwood@arm.com>
18553         * match.pd: Cope with polynomial numbers of vector elements.
18555 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18556             Alan Hayward  <alan.hayward@arm.com>
18557             David Sherwood  <david.sherwood@arm.com>
18559         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18560         in a POINTER_PLUS_EXPR.
18562 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18563             Alan Hayward  <alan.hayward@arm.com>
18564             David Sherwood  <david.sherwood@arm.com>
18566         * omp-simd-clone.c (simd_clone_subparts): New function.
18567         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18568         (ipa_simd_modify_function_body): Likewise.
18570 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18571             Alan Hayward  <alan.hayward@arm.com>
18572             David Sherwood  <david.sherwood@arm.com>
18574         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18575         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18576         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18577         (expand_vector_condition, vector_element): Likewise.
18578         (subparts_gt): New function.
18579         (get_compute_type): Use subparts_gt.
18580         (count_type_subparts): Delete.
18581         (expand_vector_operations_1): Use subparts_gt instead of
18582         count_type_subparts.
18584 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18585             Alan Hayward  <alan.hayward@arm.com>
18586             David Sherwood  <david.sherwood@arm.com>
18588         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18589         (vect_compile_time_alias): ...this new function.  Do the calculation
18590         on poly_ints rather than trees.
18591         (vect_prune_runtime_alias_test_list): Update call accordingly.
18593 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18594             Alan Hayward  <alan.hayward@arm.com>
18595             David Sherwood  <david.sherwood@arm.com>
18597         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18598         numbers of units.
18599         (vect_schedule_slp_instance): Likewise.
18601 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18602             Alan Hayward  <alan.hayward@arm.com>
18603             David Sherwood  <david.sherwood@arm.com>
18605         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18606         constant and extern definitions for variable-length vectors.
18607         (vect_get_constant_vectors): Note that the number of units
18608         is known to be constant.
18610 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18611             Alan Hayward  <alan.hayward@arm.com>
18612             David Sherwood  <david.sherwood@arm.com>
18614         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18615         of units as polynomial.  Choose between WIDE and NARROW based
18616         on multiple_p.
18618 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18619             Alan Hayward  <alan.hayward@arm.com>
18620             David Sherwood  <david.sherwood@arm.com>
18622         * tree-vect-stmts.c (simd_clone_subparts): New function.
18623         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18625 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18626             Alan Hayward  <alan.hayward@arm.com>
18627             David Sherwood  <david.sherwood@arm.com>
18629         * tree-vect-stmts.c (vectorizable_call): Treat the number of
18630         vectors as polynomial.  Use build_index_vector for
18631         IFN_GOMP_SIMD_LANE.
18633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18634             Alan Hayward  <alan.hayward@arm.com>
18635             David Sherwood  <david.sherwood@arm.com>
18637         * tree-vect-stmts.c (get_load_store_type): Treat the number of
18638         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18639         for variable-length vectors.
18640         (vectorizable_mask_load_store): Treat the number of units as
18641         polynomial, asserting that it is constant if the condition has
18642         already been enforced.
18643         (vectorizable_store, vectorizable_load): Likewise.
18645 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18646             Alan Hayward  <alan.hayward@arm.com>
18647             David Sherwood  <david.sherwood@arm.com>
18649         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18650         of units as polynomial.  Punt if we can't tell at compile time
18651         which vector contains the final result.
18653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18654             Alan Hayward  <alan.hayward@arm.com>
18655             David Sherwood  <david.sherwood@arm.com>
18657         * tree-vect-loop.c (vectorizable_induction): Treat the number
18658         of units as polynomial.  Punt on SLP inductions.  Use an integer
18659         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
18660         cast of such a series for variable-length floating-point
18661         reductions.
18663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18664             Alan Hayward  <alan.hayward@arm.com>
18665             David Sherwood  <david.sherwood@arm.com>
18667         * tree.h (build_index_vector): Declare.
18668         * tree.c (build_index_vector): New function.
18669         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18670         of units as polynomial, forcibly converting it to a constant if
18671         vectorizable_reduction has already enforced the condition.
18672         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
18673         to create a {1,2,3,...} vector.
18674         (vectorizable_reduction): Treat the number of units as polynomial.
18675         Choose vectype_in based on the largest scalar element size rather
18676         than the smallest number of units.  Enforce the restrictions
18677         relied on above.
18679 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18680             Alan Hayward  <alan.hayward@arm.com>
18681             David Sherwood  <david.sherwood@arm.com>
18683         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18684         number of units as polynomial.
18686 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18687             Alan Hayward  <alan.hayward@arm.com>
18688             David Sherwood  <david.sherwood@arm.com>
18690         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18691         * target.def (autovectorize_vector_sizes): Return the vector sizes
18692         by pointer, using vector_sizes rather than a bitmask.
18693         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18694         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18695         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18696         Likewise.
18697         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18698         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18699         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18700         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18701         * omp-general.c (omp_max_vf): Likewise.
18702         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18703         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18704         * tree-vect-loop.c (vect_analyze_loop): Likewise.
18705         * tree-vect-slp.c (vect_slp_bb): Likewise.
18706         * doc/tm.texi: Regenerate.
18707         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18708         to a poly_uint64.
18709         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18710         the vector size as a poly_uint64 rather than an unsigned int.
18711         (current_vector_size): Change from an unsigned int to a poly_uint64.
18712         (get_vectype_for_scalar_type): Update accordingly.
18713         * tree.h (build_truth_vector_type): Take the size and number of
18714         units as a poly_uint64 rather than an unsigned int.
18715         (build_vector_type): Add a temporary overload that takes
18716         the number of units as a poly_uint64 rather than an unsigned int.
18717         * tree.c (make_vector_type): Likewise.
18718         (build_truth_vector_type): Take the number of units as a poly_uint64
18719         rather than an unsigned int.
18721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18722             Alan Hayward  <alan.hayward@arm.com>
18723             David Sherwood  <david.sherwood@arm.com>
18725         * target.def (get_mask_mode): Take the number of units and length
18726         as poly_uint64s rather than unsigned ints.
18727         * targhooks.h (default_get_mask_mode): Update accordingly.
18728         * targhooks.c (default_get_mask_mode): Likewise.
18729         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18730         * doc/tm.texi: Regenerate.
18732 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18733             Alan Hayward  <alan.hayward@arm.com>
18734             David Sherwood  <david.sherwood@arm.com>
18736         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18737         * omp-general.c (omp_max_vf): Likewise.
18738         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18739         (expand_omp_simd): Handle polynomial safelen.
18740         * omp-low.c (omplow_simd_context): Add a default constructor.
18741         (omplow_simd_context::max_vf): Change from int to poly_uint64.
18742         (lower_rec_simd_input_clauses): Update accordingly.
18743         (lower_rec_input_clauses): Likewise.
18745 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18746             Alan Hayward  <alan.hayward@arm.com>
18747             David Sherwood  <david.sherwood@arm.com>
18749         * tree-vectorizer.h (vect_nunits_for_cost): New function.
18750         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18751         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18752         (vect_analyze_slp_cost): Likewise.
18753         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18754         (vect_model_load_cost): Likewise.
18756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18757             Alan Hayward  <alan.hayward@arm.com>
18758             David Sherwood  <david.sherwood@arm.com>
18760         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18761         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18762         from an unsigned int * to a poly_uint64_pod *.
18763         (calculate_unrolling_factor): New function.
18764         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
18766 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18767             Alan Hayward  <alan.hayward@arm.com>
18768             David Sherwood  <david.sherwood@arm.com>
18770         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18771         from an unsigned int to a poly_uint64.
18772         (_loop_vec_info::slp_unrolling_factor): Likewise.
18773         (_loop_vec_info::vectorization_factor): Change from an int
18774         to a poly_uint64.
18775         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18776         (vect_get_num_vectors): New function.
18777         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18778         (vect_get_num_copies): Use vect_get_num_vectors.
18779         (vect_analyze_data_ref_dependences): Change max_vf from an int *
18780         to an unsigned int *.
18781         (vect_analyze_data_refs): Change min_vf from an int * to a
18782         poly_uint64 *.
18783         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18784         than an unsigned HOST_WIDE_INT.
18785         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18786         (vect_analyze_data_ref_dependence): Change max_vf from an int *
18787         to an unsigned int *.
18788         (vect_analyze_data_ref_dependences): Likewise.
18789         (vect_compute_data_ref_alignment): Handle polynomial vf.
18790         (vect_enhance_data_refs_alignment): Likewise.
18791         (vect_prune_runtime_alias_test_list): Likewise.
18792         (vect_shift_permute_load_chain): Likewise.
18793         (vect_supportable_dr_alignment): Likewise.
18794         (dependence_distance_ge_vf): Take the vectorization factor as a
18795         poly_uint64 rather than an unsigned HOST_WIDE_INT.
18796         (vect_analyze_data_refs): Change min_vf from an int * to a
18797         poly_uint64 *.
18798         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18799         vfm1 as a poly_uint64 rather than an int.  Make the same change
18800         for the returned bound_scalar.
18801         (vect_gen_vector_loop_niters): Handle polynomial vf.
18802         (vect_do_peeling): Likewise.  Update call to
18803         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18804         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18805         be constant.
18806         * tree-vect-loop.c (vect_determine_vectorization_factor)
18807         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18808         (vect_get_known_peeling_cost): Likewise.
18809         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18810         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18811         (vect_transform_loop): Likewise.  Use the lowest possible VF when
18812         updating the upper bounds of the loop.
18813         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
18814         rather than an int.
18815         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18816         polynomial unroll factors.
18817         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18818         (vect_make_slp_decision): Likewise.
18819         (vect_supported_load_permutation_p): Likewise, and polynomial
18820         vf too.
18821         (vect_analyze_slp_cost): Handle polynomial vf.
18822         (vect_slp_analyze_node_operations): Likewise.
18823         (vect_slp_analyze_bb_1): Likewise.
18824         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18825         than an unsigned HOST_WIDE_INT.
18826         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18827         (vectorizable_load): Handle polynomial vf.
18828         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18829         a poly_uint64.
18830         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18832 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18833             Alan Hayward  <alan.hayward@arm.com>
18834             David Sherwood  <david.sherwood@arm.com>
18836         * match.pd: Handle bit operations involving three constants
18837         and try to fold one pair.
18839 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18841         * tree-vect-loop-manip.c: Include gimple-fold.h.
18842         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18843         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
18844         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18845         Add a path that uses a step of VF instead of 1, but disable it
18846         for now.
18847         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18848         and niters_no_overflow parameters.  Update calls to
18849         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18850         Create a new SSA name if the latter choses to use a ste other
18851         than zero, and return it via niters_vector_mult_vf_var.
18852         * tree-vect-loop.c (vect_transform_loop): Update calls to
18853         vect_do_peeling, vect_gen_vector_loop_niters and
18854         slpeel_make_loop_iterate_ntimes.
18855         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18856         (vect_gen_vector_loop_niters): Update declarations after above changes.
18858 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
18860         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18861         128-bit round to integer instructions.
18862         (ceil<mode>2): Likewise.
18863         (btrunc<mode>2): Likewise.
18864         (round<mode>2): Likewise.
18866 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18868         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18869         unaligned VSX load/store on P8/P9.
18870         (expand_block_clear): Allow the use of unaligned VSX
18871         load/store on P8/P9.
18873 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18875         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18876         New function.
18877         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18878         swap associated with both a load and a store.
18880 2018-01-02  Andrew Waterman  <andrew@sifive.com>
18882         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18883         * config/riscv/riscv.md (clear_cache): Use it.
18885 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
18887         * web.c: Remove out-of-date comment.
18889 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18891         * expr.c (fixup_args_size_notes): Check that any existing
18892         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18893         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18894         (emit_single_push_insn): ...here.
18896 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18898         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18899         (const_vector_encoded_nelts): New function.
18900         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18901         (const_vector_int_elt, const_vector_elt): Declare.
18902         * emit-rtl.c (const_vector_int_elt_1): New function.
18903         (const_vector_elt): Likewise.
18904         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18905         of CONST_VECTOR_ELT.
18907 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18909         * expr.c: Include rtx-vector-builder.h.
18910         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18911         directly on the tree encoding.
18912         (const_vector_from_tree): Likewise.
18913         * optabs.c: Include rtx-vector-builder.h.
18914         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18915         sequence of "u" values.
18916         * vec-perm-indices.c: Include rtx-vector-builder.h.
18917         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18918         directly on the vec_perm_indices encoding.
18920 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18922         * doc/rtl.texi (const_vector): Describe new encoding scheme.
18923         * Makefile.in (OBJS): Add rtx-vector-builder.o.
18924         * rtx-vector-builder.h: New file.
18925         * rtx-vector-builder.c: Likewise.
18926         * rtl.h (rtx_def::u2): Add a const_vector field.
18927         (CONST_VECTOR_NPATTERNS): New macro.
18928         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18929         (CONST_VECTOR_DUPLICATE_P): Likewise.
18930         (CONST_VECTOR_STEPPED_P): Likewise.
18931         (CONST_VECTOR_ENCODED_ELT): Likewise.
18932         (const_vec_duplicate_p): Check for a duplicated vector encoding.
18933         (unwrap_const_vec_duplicate): Likewise.
18934         (const_vec_series_p): Check for a non-duplicated vector encoding.
18935         Say that the function only returns true for integer vectors.
18936         * emit-rtl.c: Include rtx-vector-builder.h.
18937         (gen_const_vec_duplicate_1): Delete.
18938         (gen_const_vector): Call gen_const_vec_duplicate instead of
18939         gen_const_vec_duplicate_1.
18940         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18941         (gen_const_vec_duplicate): Use rtx_vector_builder.
18942         (gen_const_vec_series): Likewise.
18943         (gen_rtx_CONST_VECTOR): Likewise.
18944         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18945         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18946         Build a new vector rather than modifying a CONST_VECTOR in-place.
18947         (handle_special_swappables): Update call accordingly.
18948         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18949         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18950         Build a new vector rather than modifying a CONST_VECTOR in-place.
18951         (handle_special_swappables): Update call accordingly.
18953 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18955         * simplify-rtx.c (simplify_const_binary_operation): Use
18956         CONST_VECTOR_ELT instead of XVECEXP.
18958 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18960         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18961         the selector elements to be different from the data elements
18962         if the selector is a VECTOR_CST.
18963         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18964         ssizetype for the selector.
18966 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18968         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18969         before testing each element individually.
18970         * tree-vect-generic.c (lower_vec_perm): Likewise.
18972 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18974         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18975         * selftest-run-tests.c (selftest::run_tests): Call it.
18976         * vector-builder.h (vector_builder::operator ==): New function.
18977         (vector_builder::operator !=): Likewise.
18978         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18979         (vec_perm_indices::all_from_input_p): New function.
18980         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18981         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18982         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18983         instead of reading the VECTOR_CST directly.  Detect whether both
18984         vector inputs are the same before constructing the vec_perm_indices,
18985         and update the number of inputs argument accordingly.  Use the
18986         utility functions added above.  Only construct sel2 if we need to.
18988 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
18990         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18991         the broadcast of the low byte.
18992         (expand_mult_highpart): Use an explicit encoding for the permutes.
18993         * optabs-query.c (can_mult_highpart_p): Likewise.
18994         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18995         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18996         (vectorizable_bswap): Likewise.
18997         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18998         explicit encoding for the power-of-2 permutes.
18999         (vect_permute_store_chain): Likewise.
19000         (vect_grouped_load_supported): Likewise.
19001         (vect_permute_load_chain): Likewise.
19003 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19005         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19006         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19007         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19008         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19009         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19010         (vect_gen_perm_mask_any): Likewise.
19012 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19014         * int-vector-builder.h: New file.
19015         * vec-perm-indices.h: Include int-vector-builder.h.
19016         (vec_perm_indices): Redefine as an int_vector_builder.
19017         (auto_vec_perm_indices): Delete.
19018         (vec_perm_builder): Redefine as a stand-alone class.
19019         (vec_perm_indices::vec_perm_indices): New function.
19020         (vec_perm_indices::clamp): Likewise.
19021         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19022         (vec_perm_indices::new_vector): New function.
19023         (vec_perm_indices::new_expanded_vector): Update for new
19024         vec_perm_indices class.
19025         (vec_perm_indices::rotate_inputs): New function.
19026         (vec_perm_indices::all_in_range_p): Operate directly on the
19027         encoded form, without computing elided elements.
19028         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19029         encoding.  Update for new vec_perm_indices class.
19030         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19031         the given vec_perm_builder.
19032         (expand_vec_perm_var): Update vec_perm_builder constructor.
19033         (expand_mult_highpart): Use vec_perm_builder instead of
19034         auto_vec_perm_indices.
19035         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19036         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
19037         or double series encoding as appropriate.
19038         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19039         vec_perm_indices instead of auto_vec_perm_indices.
19040         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19041         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19042         (vect_permute_store_chain): Likewise.
19043         (vect_grouped_load_supported): Likewise.
19044         (vect_permute_load_chain): Likewise.
19045         (vect_shift_permute_load_chain): Likewise.
19046         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19047         (vect_transform_slp_perm_load): Likewise.
19048         (vect_schedule_slp_instance): Likewise.
19049         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19050         (vectorizable_mask_load_store): Likewise.
19051         (vectorizable_bswap): Likewise.
19052         (vectorizable_store): Likewise.
19053         (vectorizable_load): Likewise.
19054         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19055         vec_perm_indices instead of auto_vec_perm_indices.  Use
19056         tree_to_vec_perm_builder to read the vector from a tree.
19057         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19058         vec_perm_builder instead of a vec_perm_indices.
19059         (have_whole_vector_shift): Use vec_perm_builder and
19060         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
19061         truncation to calc_vec_perm_mask_for_shift.
19062         (vect_create_epilog_for_reduction): Likewise.
19063         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19064         from auto_vec_perm_indices to vec_perm_indices.
19065         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19066         instead of changing individual elements.
19067         (aarch64_vectorize_vec_perm_const): Use new_vector to install
19068         the vector in d.perm.
19069         * config/arm/arm.c (expand_vec_perm_d::perm): Change
19070         from auto_vec_perm_indices to vec_perm_indices.
19071         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19072         instead of changing individual elements.
19073         (arm_vectorize_vec_perm_const): Use new_vector to install
19074         the vector in d.perm.
19075         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19076         Update vec_perm_builder constructor.
19077         (rs6000_expand_interleave): Likewise.
19078         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19079         (rs6000_expand_interleave): Likewise.
19081 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19083         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19084         to qimode could truncate the indices.
19085         * optabs.c (expand_vec_perm_var): Likewise.
19087 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19089         * Makefile.in (OBJS): Add vec-perm-indices.o.
19090         * vec-perm-indices.h: New file.
19091         * vec-perm-indices.c: Likewise.
19092         * target.h (vec_perm_indices): Replace with a forward class
19093         declaration.
19094         (auto_vec_perm_indices): Move to vec-perm-indices.h.
19095         * optabs.h: Include vec-perm-indices.h.
19096         (expand_vec_perm): Delete.
19097         (selector_fits_mode_p, expand_vec_perm_var): Declare.
19098         (expand_vec_perm_const): Declare.
19099         * target.def (vec_perm_const_ok): Replace with...
19100         (vec_perm_const): ...this new hook.
19101         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19102         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19103         * doc/tm.texi: Regenerate.
19104         * optabs.def (vec_perm_const): Delete.
19105         * doc/md.texi (vec_perm_const): Likewise.
19106         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19107         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19108         expand_vec_perm for constant permutation vectors.  Assert that
19109         the mode of variable permutation vectors is the integer equivalent
19110         of the mode that is being permuted.
19111         * optabs-query.h (selector_fits_mode_p): Declare.
19112         * optabs-query.c: Include vec-perm-indices.h.
19113         (selector_fits_mode_p): New function.
19114         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19115         is defined, instead of checking whether the vec_perm_const_optab
19116         exists.  Use targetm.vectorize.vec_perm_const instead of
19117         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
19118         fit in the vector mode before using a variable permute.
19119         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19120         vec_perm_indices instead of an rtx.
19121         (expand_vec_perm): Replace with...
19122         (expand_vec_perm_const): ...this new function.  Take the selector
19123         as a vec_perm_indices rather than an rtx.  Also take the mode of
19124         the selector.  Update call to shift_amt_for_vec_perm_mask.
19125         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19126         Use vec_perm_indices::new_expanded_vector to expand the original
19127         selector into bytes.  Check whether the indices fit in the vector
19128         mode before using a variable permute.
19129         (expand_vec_perm_var): Make global.
19130         (expand_mult_highpart): Use expand_vec_perm_const.
19131         * fold-const.c: Includes vec-perm-indices.h.
19132         * tree-ssa-forwprop.c: Likewise.
19133         * tree-vect-data-refs.c: Likewise.
19134         * tree-vect-generic.c: Likewise.
19135         * tree-vect-loop.c: Likewise.
19136         * tree-vect-slp.c: Likewise.
19137         * tree-vect-stmts.c: Likewise.
19138         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19139         Delete.
19140         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19141         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19142         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19143         (aarch64_vectorize_vec_perm_const): ...this new function.
19144         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19145         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19146         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19147         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19148         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19149         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19150         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19151         into...
19152         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
19153         check for NEON modes.
19154         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19155         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19156         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19157         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19158         into...
19159         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
19160         the old VEC_PERM_CONST conditions.
19161         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19162         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19163         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19164         (ia64_vectorize_vec_perm_const_ok): Merge into...
19165         (ia64_vectorize_vec_perm_const): ...this new function.
19166         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19167         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19168         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19169         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19170         * config/mips/mips.c (mips_expand_vec_perm_const)
19171         (mips_vectorize_vec_perm_const_ok): Merge into...
19172         (mips_vectorize_vec_perm_const): ...this new function.
19173         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19174         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19175         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19176         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19177         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19178         (rs6000_expand_vec_perm_const): Delete.
19179         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19180         Delete.
19181         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19182         (altivec_expand_vec_perm_const_le): Take each operand individually.
19183         Operate on constant selectors rather than rtxes.
19184         (altivec_expand_vec_perm_const): Likewise.  Update call to
19185         altivec_expand_vec_perm_const_le.
19186         (rs6000_expand_vec_perm_const): Delete.
19187         (rs6000_vectorize_vec_perm_const_ok): Delete.
19188         (rs6000_vectorize_vec_perm_const): New function.
19189         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19190         an element count and rtx array.
19191         (rs6000_expand_extract_even): Update call accordingly.
19192         (rs6000_expand_interleave): Likewise.
19193         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19194         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19195         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19196         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19197         (rs6000_expand_vec_perm_const): Delete.
19198         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19199         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19200         (altivec_expand_vec_perm_const_le): Take each operand individually.
19201         Operate on constant selectors rather than rtxes.
19202         (altivec_expand_vec_perm_const): Likewise.  Update call to
19203         altivec_expand_vec_perm_const_le.
19204         (rs6000_expand_vec_perm_const): Delete.
19205         (rs6000_vectorize_vec_perm_const_ok): Delete.
19206         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
19207         reference to the SPE evmerge intructions.
19208         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19209         an element count and rtx array.
19210         (rs6000_expand_extract_even): Update call accordingly.
19211         (rs6000_expand_interleave): Likewise.
19212         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19213         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19214         new function.
19215         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19217 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19219         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19220         vector mode and that that mode matches the mode of the data
19221         being permuted.
19222         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19223         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
19224         directly using expand_vec_perm_1 when forcing selectors into
19225         registers.
19226         (expand_vec_perm_var): New function, split out from expand_vec_perm.
19228 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19230         * optabs-query.h (can_vec_perm_p): Delete.
19231         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19232         * optabs-query.c (can_vec_perm_p): Split into...
19233         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19234         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19235         particular selector is valid.
19236         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19237         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19238         (vect_grouped_load_supported): Likewise.
19239         (vect_shift_permute_load_chain): Likewise.
19240         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19241         (vect_transform_slp_perm_load): Likewise.
19242         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19243         (vectorizable_bswap): Likewise.
19244         (vect_gen_perm_mask_checked): Likewise.
19245         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
19246         implementations of variable permutation vectors into account
19247         when deciding which selector to use.
19248         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19249         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19250         with a false third argument.
19251         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19252         to test whether the constant selector is valid and can_vec_perm_var_p
19253         to test whether a variable selector is valid.
19255 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19257         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19258         * optabs-query.c (can_vec_perm_p): Likewise.
19259         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19260         instead of vec_perm_indices.
19261         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19262         (vect_gen_perm_mask_checked): Likewise,
19263         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19264         (vect_gen_perm_mask_checked): Likewise,
19266 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
19268         * optabs-query.h (qimode_for_vec_perm): Declare.
19269         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19270         (qimode_for_vec_perm): ...this new function.
19271         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19273 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19275         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19276         does not have a conditional at the top.
19278 2018-01-02  Richard Biener  <rguenther@suse.de>
19280         * ipa-inline.c (big_speedup_p): Fix expression.
19282 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
19284         PR target/81616
19285         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19286         for generic 4->6.
19288 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
19290         PR target/81616
19291         Generic tuning.
19292         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19293         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19294         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19295         cond_taken_branch_cost 3->4.
19297 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
19299         PR tree-optimization/83581
19300         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19301         TODO_cleanup_cfg if any changes have been made.
19303         PR middle-end/83608
19304         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19305         convert_modes if target mode has the right side, but different mode
19306         class.
19308         PR middle-end/83609
19309         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19310         last argument when extracting from CONCAT.  If either from_real or
19311         from_imag is NULL, use expansion through memory.  If result is not
19312         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19313         the parts directly to inner mode, if even that fails, use expansion
19314         through memory.
19316         PR middle-end/83623
19317         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19318         check for bswap in mode rather than HImode and use that in expand_unop
19319         too.
19321 Copyright (C) 2018 Free Software Foundation, Inc.
19323 Copying and distribution of this file, with or without modification,
19324 are permitted in any medium without royalty provided the copyright
19325 notice and this notice are preserved.