1 2018-07-05 Martin Sebor <msebor@redhat.com>
4 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5 than its domain to compute its the upper bound of a char array.
7 2018-07-03 Carl Love <cel@us.ibm.com>
10 2018-07-03 Carl Love <cel@us.ibm.com>
12 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
14 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
15 VSX_BUILTIN_DOUBLEL_V4SF.
17 2018-07-04 Luis Machado <luis.machado@linaro.org>
19 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
21 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
23 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
26 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
29 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
30 ieee128_float_type_node to long_double_type_node unless
31 TARGET_LONG_DOUBLE_128 is set.
33 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
35 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
36 -fno-omit-frame-pointer when not optimizing.
38 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
41 * dwarf2out.c (loc_list_has_views): Adjust comments.
42 (dw_loc_list): Split single cross-partition range with
45 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
47 Backport from mainline
48 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
50 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
51 behavior of vec_pack (vector double, vector double) to match
52 behavior of vec_float2 (vector double, vector double).
54 2018-06-29 Jakub Jelinek <jakub@redhat.com>
56 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
58 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
60 Backport from mainline
61 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
63 * doc/sourcebuild.texi (Endianness): New subsubsection.
65 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
67 Backport from mainline
68 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
70 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
71 behavior of vec_packsu (vector unsigned long long, vector unsigned
72 long long) to match behavior of vec_packs with same signature.
74 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
77 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
79 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
80 unaligned vsx for 16B memset.
82 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
85 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
88 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
91 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
94 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
96 PR rtl-optimization/85645
97 * regrename.c (build_def_use): Also kill the chains that include the
98 destination of a REG_CFA_REGISTER note.
100 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
103 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
105 PR rtl-optimization/85645
106 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
107 insn that has a REG_CFA_REGISTER note.
109 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
111 * config/s390/s390.h (enum processor_flags): Do not use
112 default tune parameter when -march was specified.
114 2018-06-26 Jakub Jelinek <jakub@redhat.com>
117 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
118 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
121 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
124 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
126 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
127 added IEEE/IBM long double multilib support on PowerPC little
128 endian Linux systems.
129 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
130 (MULTILIB_DEFAULTS): Likewise.
131 * config/rs6000/rs6000.c (rs6000_option_override_internal):
133 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
134 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
135 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
137 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
140 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
142 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
144 2018-05-02 Jan Hubicka <jh@suse.cz>
146 Backport from mainline
147 2018-04-30 Jan Hubicka <jh@suse.cz>
149 * lto-wrapper.c (ltrans_priorities): New static var.
151 (run_gcc): Read priorities and if doing parallel build order
152 the Makefile by them.
154 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
157 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
160 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
161 ieee128 argument takes up only one (vector) register, not two (floating
164 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
166 PR tree-optimization/85989
167 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
169 (backprop::intersect_uses): Check it when deciding whether this
170 is a backedge reference.
171 (backprop::process_block): Add each phi to m_visited_phis
172 after visiting it, then clear it at the end.
174 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
176 Backport from mainline
178 2018-06-19 Tony Reix <tony.reix@atos.com>
179 Damien Bergamini <damien.bergamini@atos.com>
180 David Edelsohn <dje.gcc@gmail.com>
182 * collect2.c (static_obj): New variable.
183 (static_libs): New variable.
184 (is_in_list): Uncomment declaration.
185 (main): Track AIX libraries linked statically.
186 (is_in_list): Uncomment definition.
187 (scan_prog_file): Don't add AIX shared libraries initializer
188 to constructor list if linking statically.
190 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
193 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
195 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
196 conversion insn that shows up when pr85657-3.c is compiled using
197 IEEE 128-bit long double.
200 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
202 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
203 multiply and divide external functions from being created more
207 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
209 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
210 mode to check whether the mode is IBM extended.
212 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
215 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
216 floating point modes, so that IFmode is numerically greater than
217 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
218 to declare the ordering. This prevents IFmode from being
219 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
220 machine. Include rs6000-modes.h to share the fractional values
221 between genmodes* and the rest of the compiler.
225 * config/rs6000/rs6000-modes.h: New file.
226 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
227 meaning of rs6000_long_double_size so that 126..128 selects an
228 appropriate 128-bit floating point type.
229 (rs6000_option_override_internal): Likewise.
230 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
231 (TARGET_LONG_DOUBLE_128): Change the meaning of
232 rs6000_long_double_size so that 126..128 selects an appropriate
233 128-bit floating point type.
234 (LONG_DOUBLE_TYPE_SIZE): Update comment.
235 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
236 source and destination to match the standard usage.
237 (truncifkf2): Likewise.
238 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
239 ISA 2.07 to use an explicit clobber, instead of passing in a
241 (copysign<mode>3_soft): Likewise.
244 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
246 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
250 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
252 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
253 track if we pass or return IEEE 128-bit floating point.
254 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
255 C++ mangling that is compatible with GCC 8.1.
256 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
257 (init_cumulative_args): Note if we pass or return IEEE 128-bit
258 floating point types.
259 (rs6000_function_arg_advance_1): Likewise.
260 (rs6000_mangle_type): Optionally generate mangled names that match
261 what GCC 8.1 generated for IEEE 128-bit floating point types.
262 (rs6000_globalize_decl_name): If we have an external function that
263 passes or returns IEEE 128-bit types, generate a weak reference
264 from the mangled name used in GCC 8.1 to the current mangled
266 (rs6000_init_builtins): Make __ibm128 use the long double type if
267 long double is IBM extended double. Make __float128 use the long
268 double type if long double is IEEE 128-bit.
271 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
272 macro for __ibm128 built-in functions.
273 (PACK_IF): Add __ibm128 pack/unpack functions.
274 (UNPACK_IF): Likewise.
275 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
276 enable long double built-in functions if long double is IEEE
277 128-bit floating point.
278 (rs6000_invalid_builtin): Update long double built-in function
280 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
281 functions, adjust the built-in function to use the long double
282 built-in function if __ibm128 and long double are the same type.
283 * doc/extend.texi (PowerPC builtins): Update documention for
284 __builtin_{,un}pack_longdouble. Add documentation for
285 __builtin_{,un}pack_ibm128.
288 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
290 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
291 the 128-bit floating point types. Fix function comment.
293 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
296 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
299 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
300 define __ibm128 as long double.
301 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
302 as a distinct type with IEEE 128-bit floating point is supported.
303 (init_float128_ieee): Fix up conversions between IFmode and IEEE
304 128-bit types to use the correct functions.
305 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
306 convert between 128-bit floating point types that have different
307 modes but the same representation, instead of using gen_lowpart to
309 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
311 (IFKF_reg): New attributes to give the register constraints for
313 (extend<mode>tf2_internal): New insns to mark an explicit
314 conversion between 128-bit floating point types that have a
315 different mode but share the same representation.
317 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
320 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
321 -x assembler-with-cpp.
323 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
325 Backported from mainline
326 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
328 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
329 multilibs for *-*-rtems*.
330 * config/riscv/t-rtems: New file.
332 2018-06-20 Jakub Jelinek <jakub@redhat.com>
334 Backported from mainline
335 2018-06-16 Jakub Jelinek <jakub@redhat.com>
337 PR rtl-optimization/86108
338 * bb-reorder.c (create_forwarder_block): Renamed to ...
339 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
340 jump from new landing pad to the second part.
341 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
344 2018-06-20 Jakub Jelinek <jakub@redhat.com>
347 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
350 PR tree-optimization/86231
351 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
352 anti-range don't overwrite *vr0min before using it to compute *vr0max.
354 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
356 Backport from mainline
357 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
359 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
361 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
363 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
365 * gimplify.c (gimplify_init_constructor): Really never clear for an
366 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
368 2018-06-18 Martin Sebor <msebor@redhat.com>
371 * calls.c (alloc_max_size): Correct a logic error/typo.
372 Treat excessive arguments as infinite. Warn for invalid arguments.
373 * doc/invoke.texi (-Walloc-size-larger-than): Update.
375 2018-06-16 Jakub Jelinek <jakub@redhat.com>
378 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
379 documented as preserved for backward compatibility only.
380 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
382 2018-06-15 Jakub Jelinek <jakub@redhat.com>
385 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
386 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
387 Only call store_expr for halves if the mode is the same.
390 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
391 Fix up comment formatting.
393 2018-06-14 Jakub Jelinek <jakub@redhat.com>
396 * match.pd ((A +- CST1) +- CST2): Punt if last resort
397 unsigned_type_for returns NULL.
400 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
401 subregs of multi-word pseudos unless the float mode has word size.
403 2018-06-14 Richard Biener <rguenther@suse.de>
406 * tree-vect-generic.c (build_word_mode_vector_type): Remove
407 duplicate and harmful type_hash_canon.
408 * tree.c (type_hash_canon): Assert we didn't find ourselves.
410 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
412 Backport from mainline
413 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
415 * config/rtems.h (STDINT_LONG32): Define.
417 2018-05-05 Roland McGrath <mcgrathr@google.com>
420 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
421 any section for which we don't know a specific type it should have,
422 regardless of name. Previously this was done only for the exact
423 names ".init_array", ".fini_array", and ".preinit_array".
424 (default_elf_asm_named_section): Add comment about
425 relationship with default_section_type_flags and SECTION_NOTYPE.
426 (get_section): Don't consider it a type conflict if one side has
427 SECTION_NOTYPE and the other doesn't, as long as neither has the
428 SECTION_BSS et al used in the default_section_type_flags logic.
430 2018-06-13 Martin Sebor <msebor@redhat.com>
432 PR tree-optimization/86114
433 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
435 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
437 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
440 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
441 offsets for register save directives. Emit a second batch of save
442 directives, if need be, when the function accesses prior frames.
444 2018-06-11 Jason Merrill <jason@redhat.com>
446 PR c++/80485 - inline function non-zero address.
447 * symtab.c (nonzero_address): Check DECL_COMDAT.
449 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
451 Backport from mainline
452 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
455 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
458 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
461 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
464 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
465 on the correct operand.
466 (*movdi_internal64): Ditto.
468 2018-06-08 Carl Love <cel@us.ibm.com>
470 Backport from mainline
471 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
472 vextract_fp_from_shortl): Add BE support.
474 2018-06-08 Carl Love <cel@us.ibm.com>
476 Backport from mainline
477 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
478 Calculate index using natureal element order.
479 (first_match_or_eos_index_<mode>):
480 Calculate index using natural element order.
481 (first_match_index_<mode>):
482 Calculate index using natural element order.
483 (first_match_or_eos_index_<mode>):
484 Calculate index using natural order.
485 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
487 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
488 P9V_BUILTIN_VCLZLSBB_V16QI.
489 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
492 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
494 Backport from mainline
495 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
498 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
499 qualifier_void_pointer and qualifier_const_void_pointer.
500 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
501 (arm_init_builtins): Handle the above.
502 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
503 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
504 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
507 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
510 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
511 Generate SImode target register for null target.
512 <case IX86_BUILTIN_XGETBV>: Ditto.
513 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
514 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
516 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
518 Backport from mainline
519 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
522 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
523 Don't handle -mcpu=power8 if -mpower9-vector is also used.
525 2018-06-06 Richard Biener <rguenther@suse.de>
527 Backport from mainline
528 2018-06-04 Richard Biener <rguenther@suse.de>
530 PR tree-optimization/86038
531 * tracer.c (find_best_successor): Check probability for
532 being initialized, bail out if not.
534 2018-05-30 Richard Biener <rguenther@suse.de>
536 PR tree-optimization/85964
537 * tracer.c (better_p): Drop initialized count check, we only
538 call the function with initialized counts now.
539 (find_best_successor): Do find a best edge if one
540 has uninitialized count.
541 (find_best_predecessor): Likewise. Do BB frequency check only
542 if count is initialized.
544 2018-05-28 Richard Biener <rguenther@suse.de>
546 PR tree-optimization/85934
547 * tree-vect-generic.c (expand_vector_operations_1): Hoist
548 vector boolean check before scalar optimization.
550 2018-05-22 Richard Biener <rguenther@suse.de>
552 PR tree-optimization/85863
553 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
554 comparisons when vectype is specified.
555 (vectorizable_condition): Do not specify vectype for
556 vect_is_simple_cond when SLP vectorizing.
558 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
560 Backport from mainline
561 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
563 * config/s390/s390-builtin-types.def: Add void function type.
564 * config/s390/s390-builtins.def: Use the function type for the
567 2018-06-04 Martin Sebor <msebor@redhat.com>
570 * calls.c (maybe_warn_nonstring_arg): Use string length to set
571 or ajust the presumed bound on an operation to avoid unnecessary
574 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
577 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
578 of bits to ignore when comparing architectures.
580 2018-06-04 Jakub Jelinek <jakub@redhat.com>
583 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
585 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
587 PR tree-optimization/85712
588 Backport from mainline:
589 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
591 PR tree-optimization/85712
592 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
594 (alloc_cand_and_find_basis): Initialize first_interp field.
595 (slsr_process_mul): Modify first_interp field.
596 (slsr_process_add): Likewise.
597 (slsr_process_cast): Modify first_interp field for each new
599 (slsr_process_copy): Likewise.
600 (dump_candidate): Dump first_interp field.
601 (replace_mult_candidate): Process all interpretations, not just
603 (replace_rhs_if_not_dup): Likewise.
604 (replace_one_candidate): Likewise.
606 Backport from mainline:
607 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
609 PR tree-optimization/85712
610 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
611 this candidate has already been replaced in-situ by a copy.
613 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
616 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
617 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
619 (sse4_1_round<mode>2): Use nonimmediate_operand
620 for operand 1 predicate.
622 2018-05-31 Jakub Jelinek <jakub@redhat.com>
625 * bb-reorder.c (pass_partition_blocks::gate): Return false for
626 functions with naked attribute.
628 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
631 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
632 and movx for Haswell.
634 2018-05-30 Jakub Jelinek <jakub@redhat.com>
636 Backported from mainline
637 2018-05-14 Jakub Jelinek <jakub@redhat.com>
640 * config/i386/i386.md: Disallow non-commutative arithmetics in
641 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
642 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
643 in the peephole2 before it.
645 2018-05-08 Jakub Jelinek <jakub@redhat.com>
648 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
649 after cmpelim optimization.
651 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
653 Backport from mainline
654 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
658 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
660 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
664 * varasm.c (assemble_alias): Check ifunc_resolver only on
667 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
670 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
672 (cgraph_node::create_alias): Likewise.
673 (cgraph_node::get_availability): Check ifunc_resolver instead
674 of looking up ifunc attribute.
675 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
676 * varasm.c (do_assemble_alias): Likewise.
677 (assemble_alias): Likewise.
678 (default_binds_local_p_3): Likewise.
679 * cgraph.h (cgraph_node): Add ifunc_resolver.
680 (cgraph_node::only_called_directly_or_aliased_p): Return false
682 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
684 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
685 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
686 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
687 instead of looking up ifunc attribute.
689 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
691 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
693 2018-05-25 Richard Biener <rguenther@suse.de>
696 * tree-dump.c (dequeue_and_dump): Remove access to removed
697 operand 2 of a SWITCH_EXPR.
699 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
701 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
702 Add {q} suffix to insn mnemonic.
704 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
707 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
708 when memory input operand is handled.
710 2018-05-23 Bin Cheng <bin.cheng@arm.com>
712 Backport from mainline
713 2018-05-17 Bin Cheng <bin.cheng@arm.com>
714 Richard Biener <rguenther@suse.de>
716 PR tree-optimization/85793
717 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
718 for VMAT_ELEMENTWISE.
720 2018-05-23 Richard Biener <rguenther@suse.de>
723 * tree-data-ref.c (create_runtime_alias_checks): Defer
724 and ignore overflow warnings.
726 2018-05-23 Martin Liska <mliska@suse.cz>
728 Backport from mainline
729 2018-05-23 Yury Gribov <tetra2005@gmail.com>
731 PR tree-optimization/85822
732 * tree-vrp.c (is_masked_range_test): Fix handling of negative
735 2018-05-23 Martin Liska <mliska@suse.cz>
737 Backport from mainline
738 2018-05-11 Martin Liska <mliska@suse.cz>
741 * doc/extend.texi: Document LLVM style format for no_sanitize
744 2018-05-22 Martin Sebor <msebor@redhat.com>
747 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
749 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
751 PR tree-optimization/85814
752 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
753 a null return from get_strinfo when unsharing the next
754 strinfo in the chain.
756 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
758 Backport from mainline
759 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
760 Segher Boessenkool <segher@kernel.crashing.org>
763 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
765 2018-05-18 Toon Moene <toon@moene.org>
767 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
768 directly after that of -floop-interchange. Indicate that both
769 options are enabled by default when specifying -O3.
771 2018-05-11 Martin Jambor <mjambor@suse.cz>
774 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
777 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
780 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
782 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
785 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
787 (cortex-m0): Use armv6s-m isa.
788 (cortex-m0plus): Likewise.
789 (cortex-m1): Likewise.
790 (cortex-m0.small-multiply): Likewise.
791 (cortex-m0plus.small-multiply): Likewise.
792 (cortex-m1.small-multiply): Likewise.
794 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
796 PR rtl-optimization/85638
797 * bb-reorder.c: Include common/common-target.h.
798 (create_forwarder_block): New function extracted from...
799 (fix_up_crossing_landing_pad): ...here. Rename into...
800 (dw2_fix_up_crossing_landing_pad): ...this.
801 (sjlj_fix_up_crossing_landing_pad): New function.
802 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
803 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
804 from both partitions and exit the loop after one iteration.
806 2018-05-08 Richard Biener <rguenther@suse.de>
808 Backport from mainline
809 2018-05-04 Richard Biener <rguenther@suse.de>
812 * fold-const.c (negate_expr_p): Restrict negation of operand
813 zero of a division to when we know that can happen without
815 (fold_negate_expr_1): Likewise.
817 2018-05-03 Richard Biener <rguenther@suse.de>
819 PR tree-optimization/85615
820 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
821 to loops not nested in BBs loop father to avoid creating multi-entry
824 2018-05-02 Richard Biener <rguenther@suse.de>
827 * gimplify.c (gimplify_save_expr): When in SSA form allow
828 SAVE_EXPRs to compute to SSA vars.
830 2018-05-02 Richard Biener <rguenther@suse.de>
832 PR tree-optimization/85597
833 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
834 do not use split vect_get_vec_defs call but call vect_get_slp_defs
837 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
840 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
841 (check_arch): Likewise.
842 (check_fpu): Return the result rather than printing it.
843 (end arch): Fix operator precedence.
845 (END): Print the result from check_fpu.
847 2018-05-08 Richard Biener <rguenther@suse.de>
849 Backport from mainline
850 2018-05-04 Richard Biener <rguenther@suse.de>
852 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
853 * gimple-ssa-store-merging.c
854 (imm_store_chain_info::output_merged_store): Remove redundant create,
855 release split_store vector contents on failure.
856 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
857 scalar stmt vector on cache hit.
859 2018-05-04 Richard Biener <rguenther@suse.de>
862 * tree-complex.c (tree_lower_complex): Handle removed blocks.
864 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
866 Backport from mainline.
867 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
869 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
870 GET_MODE_MASK before any checking.
871 (nds32_can_use_bset_p): Likewise.
872 (nds32_can_use_btgl_p): Likewise.
874 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
876 Backport from mainline.
877 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
879 * config/nds32/nds32-doubleword.md: New define_split pattern for
880 illegal register number.
882 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
884 Backport from mainline.
885 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
887 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
889 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
891 Backport from mainline.
892 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
894 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
896 2018-05-06 Jakub Jelinek <jakub@redhat.com>
899 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
900 the type is addressable. Don't force op into register if it has
903 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
906 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
909 * config/aarch64/constraints.md (Usg): Limit to 31.
912 2018-05-02 Jakub Jelinek <jakub@redhat.com>
914 * BASE-VER: Set to 8.1.1.
916 2018-05-02 Release Manager
918 * GCC 8.1.0 released.
920 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
922 PR tree-optimization/85586
923 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
924 exit early for statements in the same group if the accesses are
927 2018-05-01 Jakub Jelinek <jakub@redhat.com>
930 * doc/install.texi2html: Replace _002d with - and _002a with * in
931 generated html files using sed.
933 2018-04-27 Martin Jambor <mjambor@suse.cz>
936 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
937 the jump function allows for passing through aggregate values.
939 2018-04-27 Jeff Law <law@redhat.com>
941 * config.gcc: Mark tile* targets as deprecated/obsolete.
943 2018-04-27 Jakub Jelinek <jakub@redhat.com>
945 PR tree-optimization/85529
946 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
947 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
948 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
949 zero extension or masking of the MSB bit.
950 (optimize_range_tests): Add FIRST_BB argument, pass it through
951 to optimize_range_tests_var_bound.
952 (maybe_optimize_range_tests, reassociate_bb): Adjust
953 optimize_range_tests callers.
955 2018-04-26 Richard Biener <rguenther@suse.de>
956 Jakub Jelinek <jakub@redhat.com>
958 * cgraph.h (symbol_table): Just declare debug method here.
959 * symtab.c (symbol_table::debug): Define.
961 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
963 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
965 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
967 Backport from mainline
968 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
971 * config/i386/i386.c (ix86_expand_builtin): Change memory
972 operand to XI, extend p0 to Pmode.
973 * config/i386/i386.md: Change unspec volatile and operand
974 1 mode to XI, change operand 0 mode to P.
976 2018-04-25 Jakub Jelinek <jakub@redhat.com>
978 * DEV-PHASE: Set to prerelease.
980 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
982 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
983 in __abskf2 and __powikf2.
985 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
988 * config/aarch64/constraints.md (Usg, Usj): New constraints.
989 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
990 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
991 Use the above on operand 2. Reindent.
992 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
994 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
997 * common/config/i386/i386-common.c (ix86_handle_option): Don't
999 * config/i386/i386.opt (mcet): Removed.
1000 * doc/install.texi: Remove -mcet documentation.
1001 * doc/invoke.texi: Likewise.
1003 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1006 * doc/install.texi: Remove -mcet from bootstrap-cet.
1008 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1011 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1012 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1016 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1017 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1018 containing a CONST_VECTOR.
1020 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1022 * doc/install.texi: Update newlib dependency for nvptx.
1024 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1027 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1028 instead of INTVAL when shifting x left.
1030 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1032 PR tree-optimization/85478
1033 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1034 vect_grouped_store_supported for single element vectors.
1036 2018-04-24 Richard Biener <rguenther@suse.de>
1039 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1040 load cost increase to the case of non-constant step.
1042 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1045 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1046 destination if any_malformed_asm.
1048 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1051 * expr.c (store_field): In the bitfield case, if the value comes from
1052 a function call and is returned in registers by means of a PARALLEL,
1053 do not change the mode of the temporary unless BLKmode and VOIDmode.
1055 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1057 PR rtl-optimization/85423
1058 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1059 dependencies to debug insns when the previous insn is non-debug.
1061 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1063 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1064 enums into a single definition.
1065 (fls): Fix predicates and printing.
1068 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1070 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1071 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1072 and short u6 immediate.
1073 (check_if_valid_sleep_operand): Remove.
1074 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1076 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1078 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1079 flag_always_save_lp condition.
1080 * config/nds32/nds32.opt (malways-save-lp): New option.
1082 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1084 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1085 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1086 * config/nds32/nds32.h
1087 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1088 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1090 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1092 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1093 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1095 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1096 Chung-Ju Wu <jasonwucj@gmail.com>
1098 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1100 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1101 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1103 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1105 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1107 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1109 * config/nds32/nds32-protos.h (nds32_data_alignment,
1110 nds32_local_alignment): Declare.
1111 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1112 nds32_local_alignment): New functions.
1113 (TARGET_CONSTANT_ALIGNMENT): Define.
1114 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1116 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1118 * config/nds32/nds32.c
1119 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1120 (TARGET_MODES_TIEABLE_P): Likewise.
1122 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1127 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1128 Chung-Ju Wu <jasonwucj@gmail.com>
1130 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1131 for unaligned access.
1132 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1134 * config/nds32/nds32-intrinsic.md: Likewise.
1135 * config/nds32/nds32_intrinsic.h: Likewise.
1136 * config/nds32/nds32.h (nds32_builtins): Likewise.
1137 * config/nds32/nds32.opt (munaligned-access): New option.
1138 * config/nds32/nds32.c (nds32_asm_file_start): Display
1139 flag_unaligned_access status.
1141 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1143 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1144 -mno-relax is present.
1145 * config/riscv/linux.h (LINK_SPEC): Ditto.
1147 2018-04-20 Martin Sebor <msebor@redhat.com>
1150 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1152 (gimple_fold_builtin_stxcpy_chk): Same.
1153 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1155 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1158 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1159 __powikf2 when long double is IEEE 128-bit.
1161 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1163 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1164 step to make sure stack always aligned.
1166 2018-04-20 Carl Love <cel@us.ibm.com>
1169 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1170 size check for arg0.
1172 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1173 Tom de Vries <tom@codesourcery.com>
1176 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1177 Emit insns for calls too.
1178 (nvptx_find_par): Always look for worker-level predecessor insn.
1179 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1181 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1182 (nvptx_process_pars): Propagate frames for calls.
1184 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1187 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1189 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1190 (ix86_handle_option): Don't handle OPT_mibt.
1191 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1193 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1195 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1196 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1197 (ix86_target_macros): Define __CET__ with flag_cf_protection
1198 for -fcf-protection.
1199 * config/i386/i386.c (isa2_opts): Remove -mibt.
1200 * config/i386/i386.h (TARGET_IBT): Removed.
1201 (TARGET_IBT_P): Likewise.
1202 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1203 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1204 * config/i386/i386.opt (mcet): Update help message.
1207 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1208 -mcet as an alias for -mshstk.
1210 2018-04-20 Richard Biener <rguenther@suse.de>
1213 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1214 complexity by forcing a single use of the multiply operand.
1216 2018-04-20 Martin Jambor <mjambor@suse.cz>
1219 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1220 recursion dependency to only apply to non-clones.
1222 2018-04-20 Martin Jambor <mjambor@suse.cz>
1225 * ipa-cp.c (create_specialized_node): Check that clones of
1226 self-recursive edges exist during IPA-CP.
1228 2018-04-19 Toon Moene <toon@moene.org>
1230 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1233 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1235 PR tree-optimization/85467
1236 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1237 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1238 VECTOR_CST element to type.
1240 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1243 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1244 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1245 (builtin_longjmp): Likewise.
1246 (save_stack_nonlocal): New pattern.
1247 (restore_stack_nonlocal): Likewise.
1249 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1252 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1253 Replace ASM_OUTPUT_LABEL with fprintf.
1255 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1258 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1259 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1260 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1261 define __IBT__ and __SHSTK__ for -fcf-protection.
1262 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1264 (ix86_trampoline_init): Likewise.
1265 (x86_output_mi_thunk): Likewise.
1266 (ix86_notrack_prefixed_insn_p): Likewise.
1267 (ix86_option_override_internal): Don't disallow -fcf-protection.
1268 * config/i386/i386.md (rdssp<mode>): Also enable for
1270 (incssp<mode>): Likewise.
1271 (nop_endbr): Likewise.
1272 * config/i386/i386.opt (mcet): Change help message to built-in
1276 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1277 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1278 enable CET built-in functions.
1280 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1282 * common/config/i386/i386-common.c
1283 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1284 OPTION_MASK_ISA_MOVDIRI_UNSET,
1285 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1286 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1287 * config.gcc (movdirintrin.h): New header.
1288 * config/i386/cpuid.h (bit_MOVDIRI,
1289 bit_MOVDIR64B): New bits.
1290 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1292 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1293 (VOID, PVOID, PCVOID)): New function types.
1294 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1295 __builtin_ia32_directstoreu_u64,
1296 __builtin_ia32_movdir64b): New builtins.
1297 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1298 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1300 (ix86_valid_target_attribute_inner_p): Ditto.
1301 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1302 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1303 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1304 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1305 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1306 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1307 (movdiri<mode>, movdir64b_<mode>): New.
1308 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1309 * config/i386/immintrin.h: Include movdirintrin.h.
1310 * config/i386/movdirintrin.h: New file.
1311 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1313 2018-04-19 Richard Biener <rguenther@suse.de>
1316 * cfg.c (clear_bb_flags): When loop state says we have
1317 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1319 2018-04-19 Richard Biener <rguenther@suse.de>
1321 PR tree-optimization/84737
1322 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1323 copying restrict info.
1324 (vect_setup_realignment): Use it.
1325 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1326 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1327 the first DR to all generated stores.
1328 (vectorizable_load): Likewise for loads.
1330 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1332 PR tree-optimization/85446
1333 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1334 the integral and pointer types to have the same precision.
1336 * doc/install.texi: Document --disable-cet being the default and
1339 2018-04-18 Martin Liska <mliska@suse.cz>
1341 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1344 2018-04-18 Martin Liska <mliska@suse.cz>
1347 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1350 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1351 arguments if they are comparable.
1353 2018-04-18 Martin Liska <mliska@suse.cz>
1356 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1359 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1362 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1365 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1366 ENDBR after calling __morestack.
1368 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1371 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1372 by using gcc_base_ver to generate a gcc_driver_version, and use
1373 it when generating GCC_DRIVER_NAME.
1374 * configure: Regenerate.
1376 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1379 * config.gcc: Obsolete powerpc*-*-*spe*.
1381 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1384 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1385 (stabstr_D): Change type of unum from unsigned int to
1386 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1389 2018-04-17 Jim Wilson <jimw@sifive.com>
1392 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1393 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1394 Set arg_pointer_offset after using pretend_args_size.
1396 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1398 PR rtl-optimization/85431
1399 * dse.c (record_store): Ignore zero width stores.
1402 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1403 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1404 __builtin_stack_restore rather than after it.
1405 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1406 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1407 argument instead of virtual_dynamic_stack_rtx.
1409 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1411 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1413 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1414 Add note to error message to explain internal mapping of overloaded
1415 built-in function name to non-overloaded built-in function name.
1416 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1419 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1422 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1423 where the inputs overlap with the output.
1425 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1427 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1428 (=v, v) alternative and explicit "memory" attribute.
1429 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1430 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1432 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1433 "sselog1" type instead of "sselog".
1434 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1435 "sselog". Remove explicit "memory" attribute.
1436 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1437 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1439 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1440 "sselog1" type instead of "sselog". Remove explicit "memory"
1442 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1443 use "sselog1" type instead of "sselog". Remove explicit "memory"
1445 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1446 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1448 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1449 "sselog1" type instead of "sselog". Remove explicit "memory"
1451 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1452 use "sselog1" type instead of "sselog". Remove explicit "memory"
1456 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1459 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1462 2018-04-17 Martin Jambor <mjambor@suse.cz>
1465 * ipa-cp.c (create_specialized_node): Call
1466 expand_all_artificial_thunks if necessary.
1468 2018-04-17 Martin Liska <mliska@suse.cz>
1471 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1472 in message, remote space in between '_G' and '('.
1474 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1477 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1478 avx512f_vmcmp<mode>3<round_saeonly_name>,
1479 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1480 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1481 avx512f_rndscale<mode><round_saeonly_name>,
1482 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1483 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1484 Use %<iptr>2 instead of %2 for -masm=intel.
1485 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1486 avx512f_vcvttss2usi<round_saeonly_name>,
1487 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1489 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1490 avx512f_vcvttsd2usi<round_saeonly_name>,
1491 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1492 Use %q1 instead of %1 for -masm=intel.
1493 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1494 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1495 of %3 for -masm=intel.
1496 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1498 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1500 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1502 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1503 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1505 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1507 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1508 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1509 %g1 and one with %0 and %1.
1510 (avx512er_vmrcp28<mode><round_saeonly_name>,
1511 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1513 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1514 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1515 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1516 of %0 and %{%4%} for -masm=intel.
1517 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1518 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1519 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1520 order of %0 and %{%5%}%{z%} for -masm=intel.
1522 2018-04-17 Jan Hubicka <jh@suse.cz>
1525 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1527 2018-04-17 Martin Liska <mliska@suse.cz>
1530 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1531 for target_clone error message. Make default implementation
1532 clone to be a local declaration.
1533 (separate_attrs): Add new argument and check for an empty
1535 (expand_target_clones): Handle it.
1536 (ipa_target_clone): Make redirection just for target_clones
1539 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1540 Tom de Vries <tom@codesourcery.com>
1543 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1544 tiled basic blocks without omp continue statements.
1546 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1549 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1550 vec_extract expression as having side effects to make sure it gets
1553 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1556 * config/i386/i386.c (get_builtin_code_for_version): Check
1559 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1562 * gcc/config.gcc: Support "skylake".
1563 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1565 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1566 (processor_target_table): Add "skylake".
1567 (ix86_option_override_internal): Add "skylake".
1568 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1569 PROCESSOR_CANNONLAKE.
1570 (get_builtin_code_for_version): Fix priority for
1571 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1572 PROCESSOR_SKYLAKE-AVX512.
1573 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1574 (processor_type): Add PROCESSOR_SKYLAKE.
1576 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1577 Jason Merrill <jason@redhat.com>
1580 * convert.c (convert_to_integer_1): Use direct recursion for
1581 enumeral types and types with a precision less than the number
1582 of bits in their mode.
1584 2018-04-16 Julia Koval <julia.koval@intel.com>
1587 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1588 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1590 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1593 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1594 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1595 and -mno-direct-move.
1597 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1600 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1601 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1602 Ensure negative shifts result in {0}.
1604 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1606 PR rtl-optimization/79916
1607 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1608 regs (if any) to define how to gnerate SD moves when LRA is in
1611 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1613 PR rtl-optimization/85393
1614 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1615 * except.c (expand_dw2_landing_pad_for_region): Make static.
1616 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1617 a label and unconditional jump to old_bb, rather than
1618 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1621 PR rtl-optimization/85376
1622 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1623 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1624 instead of a specific value.
1626 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1627 Bin Cheng <bin.cheng@arm.com>
1629 PR tree-optimization/82965
1630 PR tree-optimization/83991
1631 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1632 by_profile_only parameter.
1633 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1634 information if the loop was predicted to iterate too many times.
1635 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1637 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1640 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1643 2018-04-13 Martin Liska <mliska@suse.cz>
1644 Jakub Jelinek <jakub@redhat.com>
1647 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1648 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1649 * builtins.c (expand_builtin_memory_copy_args): Use
1650 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1651 handle dest_addr == pc_rtx.
1653 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1656 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1657 asked to not generate direct moves.
1658 (fix_trunc<mode>si2_stfiwx): Similar.
1659 (fix_trunc<mode>si2_internal): Similar.
1661 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1664 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1665 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1666 lookup if dest in some wider mode is known to be const0_rtx and
1667 if so, record permanent equivalence for it to be ZERO_EXTEND of
1668 the narrower mode destination.
1670 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1672 * lto-streamer-out.c (output_function): Revert 259346.
1673 * omp-expand.c (expand_oacc_for): Likewise.
1675 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1677 PR rtl-optimization/85354
1678 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1679 * sel-sched.c (sel_global_init): ... here.
1681 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1684 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1685 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1686 mode for PE-COFF targets.
1687 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1688 (i386_pe_asm_lto_end): Likewise.
1689 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1690 (TARGET_ASM_LTO_END): Likewise.
1691 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1692 (i386_pe_asm_lto_start): New function.
1693 (i386_pe_asm_lto_end): Likewise.
1695 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1696 Richard Biener <rguenther@suse.de>
1699 * lto-streamer-out.c (output_function): Fix CFG loop state before
1701 * omp-expand.c (expand_oacc_for): Handle calls to internal
1702 functions like regular functions.
1704 2018-04-12 Richard Biener <rguenther@suse.de>
1707 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1708 for the early LTO debug to properly generate references to it
1709 during DIE emission. Do not re-use that for the skeleton for
1711 (dwarf2out_early_finish): Likewise.
1713 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1716 * config/i386/sse.md
1717 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1718 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1719 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1720 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1721 and output is a reg, avoid creating invalid lowpart subreg, but
1722 instead split into a 512-bit move. Don't split if not AVX512VL,
1723 input is xmm16+ reg and output is a mem.
1724 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1725 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1726 xmm16+ reg and output is a mem.
1728 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1730 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1731 also for flag_dwarf2_cfi_asm.
1733 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1735 PR rtl-optimization/85342
1736 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1737 a bool scalar var inside of the loop instead. Don't try to update
1738 recog_data.operand after failed apply_change_group.
1740 2018-04-12 Tom de Vries <tom@codesourcery.com>
1743 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1744 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1745 array with flexible array member as array without given dimension.
1746 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1747 argument for undefined param to true.
1749 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1752 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1753 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1754 from PowerPC section.
1755 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1756 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1757 help text that is too long.
1758 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1759 help text that is too long.
1760 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1761 help text that is too long.
1763 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1765 * config/alpha/alpha.md (stack_probe_internal): Rename
1766 from "probe_stack". Update all callers.
1768 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1770 PR rtl-optimization/84566
1771 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1772 sched_macro_fuse_insns.
1774 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1777 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1778 (compute_block_dependences): ... from here.
1780 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1782 PR tree-optimization/85331
1783 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1784 from int to HOST_WIDE_INT.
1786 2018-04-11 Martin Jambor <mjambor@suse.cz>
1789 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1790 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1791 not the same as the source val.
1792 (cgraph_edge_brings_value_p): New parameter.
1793 (gather_edges_for_value): Pass destination value to
1794 cgraph_edge_brings_value_p.
1795 (perhaps_add_new_callers): Likewise.
1796 (get_info_about_necessary_edges): Likewise and exclude values brought
1797 only by self-recursive edges.
1798 (create_specialized_node): Redirect only clones of self-calling edges.
1799 (+self_recursive_pass_through_p): New function.
1800 (find_more_scalar_values_for_callers_subset): Use it.
1801 (find_aggregate_values_for_callers_subset): Likewise.
1802 (known_aggs_to_agg_replacement_list): Removed.
1803 (decide_whether_version_node): Re-calculate known constants for all
1804 remaining context clones.
1806 2018-04-11 Richard Biener <rguenther@suse.de>
1809 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1810 from early DWARF output.
1811 (dwarf2out_early_finish): Output line info unconditionally into
1812 early DWARF and add reference to it.
1814 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1817 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1818 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1819 other than V2DFmode using iptr mode attribute.
1820 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1822 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1824 PR rtl-optimization/84659
1825 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1827 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1830 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1832 (output_loc_list): Pass address of a dummy size variable even in the
1833 locview handling loop.
1834 (index_location_lists): Add comment on why skip_loc_list_entry can't
1837 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1840 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1843 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1846 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1847 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1848 and -mstring-compare-inline-limit.
1850 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1853 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1854 for stack clash protection in a register whenever we need it to be in
1857 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1859 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1860 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1862 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1865 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1867 (mlong-double-): Ditto.
1868 * config/rs6000/sysv4.opt (msdata=): Ditto.
1869 (mtls-size=): Ditto.
1871 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1873 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1874 erroneous entries for
1875 "vector int vec_ldl (int, long int *)", and
1876 "vector unsigned int vec_ldl (int, unsigned long int *)".
1877 Add comments and entries for
1878 "vector bool char vec_ldl (int, bool char *)",
1879 "vector bool short vec_ldl (int, bool short *)",
1880 "vector bool int vec_ldl (int, bool int *)",
1881 "vector bool long long vec_ldl (int, bool long long *)",
1882 "vector pixel vec_ldl (int, pixel *)",
1883 "vector long long vec_ldl (int, long long *)",
1884 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1885 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1886 type tree bool_long_long_type_node and correct definition of
1887 bool_V2DI_type_node to make reference to this new type tree.
1888 (rs6000_mangle_type): Replace erroneous reference to
1889 bool_long_type_node with bool_long_long_type_node.
1890 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1891 comments to emphasize sign distinctions for char and int types and
1892 replace RS6000_BTI_bool_long constant with
1893 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1894 use of RS6000_BTI_pixel.
1895 (bool_long_type_node): Remove this macro definition.
1896 (bool_long_long_type_node): New macro definition
1898 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1900 PR rtl-optimization/85300
1901 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1902 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1903 simplify_unary_operation fails.
1905 2018-04-10 Martin Liska <mliska@suse.cz>
1907 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1908 cgraph_edge and ipa_ref.
1910 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1914 * config/i386/sse.md
1915 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1916 computation of the VEC_MERGE selector from mask.
1917 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1918 Fix decoding of the VEC_MERGE selector into mask.
1920 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1922 PR tree-optimization/85286
1923 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1925 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1927 * final.c (final_1): Set insn_last_address as well as
1928 insn_current_address.
1930 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1933 * explow.c (emit_stack_probe): Call validize_mem on memory location
1934 before passing it to gen_probe_stack. Create address operand and
1935 legitimize it for the probe_stack_address case.
1937 2018-04-09 Jan Hubicka <jh@suse.cz>
1940 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1941 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1942 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1944 (free_lang_data): Rebuild type inheritance graph.
1946 2018-04-09 Martin Sebor <msebor@redhat.com>
1948 * invoke.texi (-finline-small-functions): Mention other optimization
1950 (-findirect-inlining, -fpartial-inlining): Same.
1951 (-finline-functions-called-once): Same.
1952 (-freorder-blocks-and-partition): Same.
1954 2018-04-09 Jan Hubicka <jh@suse.cz>
1957 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1958 jumps; choose last target that matches the criteria (i.e.
1959 no partition changes for non-crossing jumps).
1960 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1961 support for redirecting crossing jumps to non-crossing.
1963 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1965 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1966 also for naked functions.
1968 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1970 * config/arc/arc.md (add_shift): New pattern.
1971 (add_shift2): Likewise.
1972 (sub_shift): Likewise.
1973 (sub_shift_cmp0_noout): Likewise.
1974 (compare_si_ashiftsi): Likewise.
1975 (xbfu_cmp0_noout): New combine pattern.
1976 (xbfu_cmp0"): Likewise.
1977 (movsi_set_cc_insn): Place the predicable variant first.
1978 (commutative_binary_cmp0_noout): Remove clobber.
1979 (commutative_binary_cmp0): New pattern.
1980 (noncommutative_binary_cmp0): Likewise.
1981 (noncommutative_binary_cmp0_noout): Likewise.
1982 (noncommutative_binary_comparison_result_used): Removed.
1983 (rsub_cmp0): New pattern.
1984 (rsub_cmp0_noout): Likewise.
1985 (extzvsi): Changed, keep only meaningful variants.
1986 (SQH, SEZ): New iterators.
1987 (SQH_postfix): New mode attribute.
1988 (SEZ_prefix): New code attribute.
1989 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1990 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1991 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1993 (noncommutative_operator): Check the availability of barrel
1996 2018-04-09 Richard Biener <rguenther@suse.de>
1998 PR tree-optimization/85284
1999 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2000 Only use the niter constraining form of simple_iv when the exit
2003 2018-04-09 Tom de Vries <tom@codesourcery.com>
2006 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2007 (define_expand "*memory_barrier"): New define_expand.
2008 (define_insn "memory_barrier"): New insn.
2010 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2012 PR rtl-optimization/80463
2013 PR rtl-optimization/83972
2014 PR rtl-optimization/83480
2016 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2017 correct producer for the insn.
2018 (tidy_control_flow): Fixup seqnos in case of debug insns.
2020 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2022 PR rtl-optimization/83913
2024 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2025 different sched-times when merging exprs.
2027 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2029 PR rtl-optimization/83962
2031 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2032 tidy_fallthru_edge and tidy_control_flow.
2034 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2036 PR rtl-optimization/83530
2038 * sel-sched.c (force_next_insn): New global variable.
2039 (remove_insn_for_debug): When force_next_insn is true, also leave only
2040 next insn in the ready list.
2041 (sel_sched_region): When the region wasn't scheduled, make another pass
2042 over it with force_next_insn set to 1.
2044 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2046 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2048 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2049 for interrupt control.
2050 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2051 functions for interrupt control.
2052 * config/nds32/nds32-intrinsic.md: Likewise.
2053 * config/nds32/nds32_intrinsic.h: Likewise.
2054 * config/nds32/nds32.h (nds32_builtins): Likewise.
2056 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2058 * config/nds32/nds32.c (nds32_init_machine_status,
2059 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2060 strict_aligned_p field.
2061 (nds32_expand_to_rtl_hook): New function.
2062 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2063 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2065 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2066 Chung-Ju Wu <jasonwucj@gmail.com>
2068 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2069 * config/nds32/nds32-n7.md: New file.
2070 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2071 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2073 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2074 * config/nds32/nds32.md (pipeline_model): Add n7.
2075 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2076 * config/nds32/pipelines.md: Include n7 settings.
2078 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2079 Chung-Ju Wu <jasonwucj@gmail.com>
2081 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2082 * config/nds32/nds32-e8.md: New file.
2083 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2084 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2086 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2087 * config/nds32/nds32.md (pipeline_model): Add e8.
2088 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2089 * config/nds32/pipelines.md: Include e8 settings.
2091 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2092 Chung-Ju Wu <jasonwucj@gmail.com>
2094 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2095 * config/nds32/nds32-n8.md: New file.
2096 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2097 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2099 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2100 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2101 * config/nds32/nds32.md (pipeline_model): Add n8.
2102 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2103 * config/nds32/pipelines.md: Include n8 settings.
2105 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2106 Chung-Ju Wu <jasonwucj@gmail.com>
2108 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2109 * config/nds32/nds32-n9-2r1w.md: New file.
2110 * config/nds32/nds32-n9-3r2w.md: New file.
2111 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2112 nds32_register_ports): New or modify for cpu n9.
2113 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2115 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2116 * config/nds32/nds32-utils.c: New file.
2117 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2118 TARGET_MUL_SLOW): Define.
2119 * config/nds32/nds32.md (pipeline_model): New attribute.
2120 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2121 New options that support cpu n9.
2122 * config/nds32/pipelines.md: Include n9 settings.
2123 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2125 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2127 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2128 information if necessary.
2129 (output_cond_branch_compare_zero): Likewise.
2130 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2131 (nds32_target_alignment): Refine for alignment.
2132 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2133 (FUNCTION_BOUNDARY): Modify.
2134 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2136 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2138 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2140 * config/nds32/constants.md (unspec_volatile_element): Add values for
2141 TLB operation and data prefetch.
2142 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2143 functions for TLB operation and data prefetch.
2144 * config/nds32/nds32-intrinsic.md: Likewise.
2145 * config/nds32/nds32_intrinsic.h: Likewise.
2146 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2147 (nds32_print_operand): Likewise.
2148 * config/nds32/nds32.h (nds32_builtins): Likewise.
2150 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2151 Andrew Pinski <pinsika@gcc.gnu.org>
2154 * match.pd: Use constant_boolean_node of correct type instead of
2155 boolean_true_node or boolean_false_node for simplifying
2156 pointer comparisons to zero.
2158 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2160 PR tree-optimization/80021
2161 * tree.c (verify_type_variant): Make error call in verify_variant_match
2162 translatable and remove final full stop.
2164 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2166 * config/nds32/constants.md (unspec_volatile_element): Add
2167 UNSPEC_VOLATILE_EH_RETURN.
2168 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2169 nds32_output_stack_pop): Support dwarf exception handling process.
2170 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2171 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2172 exception handling process.
2173 (nds32_compute_stack_frame): Likewise.
2174 (nds32_return_addr_rtx): Likewise.
2175 (nds32_initial_elimination_offset): Likewise.
2176 (nds32_expand_prologue): Likewise.
2177 (nds32_expand_epilogue): Likewise.
2178 (nds32_dynamic_chain_address): New function.
2179 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2181 (DYNAMIC_CHAIN_ADDRESS): Define.
2182 (EH_RETURN_DATA_REGNO): Define.
2183 (EH_RETURN_STACKADJ_RTX): Define.
2184 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2185 patterns for dwarf exception handling.
2187 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2189 * config/nds32/nds32.h: Clean up obsolete macros.
2191 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2193 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2194 Add enum values for particular instructions.
2195 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2196 particular intrinsic functions.
2197 * config/nds32/nds32-intrinsic.md: Likewise.
2198 * config/nds32/nds32_intrinsic.h: Likewise.
2199 * config/nds32/nds32.h (nds32_builtins): Likewise.
2200 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2201 (btst, ave): New patterns for particular instructions.
2203 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2205 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2206 Add enum values for atomic load/store and memory sync.
2207 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2209 * config/nds32/nds32-intrinsic.md: Likewise.
2210 * config/nds32/nds32_intrinsic.h: Likewise.
2211 * config/nds32/nds32.h (nds32_builtins): Likewise.
2213 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2215 PR tree-optimization/85257
2216 * fold-const.c (native_encode_vector): If not all elts could fit
2217 and off is -1, return 0 rather than offset.
2218 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2219 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2220 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2221 adjust buffer in native_interpret_expr call.
2223 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2225 * config/nds32/constants.md (unspec_volatile_element): Add cache
2226 control enum values.
2227 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2228 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2229 * config/nds32/nds32.c (nds32_cctl_names): New.
2230 (nds32_print_operand): Handle cache control register names.
2231 * config/nds32/nds32.h (nds32_builtins): New enum values.
2232 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2234 * config/nds32/nds32.md (type): Add mmu.
2235 * config/nds32/pipelines.md (simple_insn): Add mmu.
2237 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2239 * config/nds32/nds32.md (type): Remove call.
2240 * config/nds32/pipelines.md (simple_insn): Likewise.
2242 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2244 * config/nds32/constants.md (unspec_volatile_element): Add
2245 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2246 UNSPEC_VOLATILE_FMFCFG.
2247 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2248 description for fmfcfg and fmfcsr.
2249 (bdesc_1arg): Add fmtcsr.
2250 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2251 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2252 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2253 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2254 unspec_fmfcfg): New patterns.
2255 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2256 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2257 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2258 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2259 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2260 __nds32__fmfcfg): Define.
2262 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2264 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2265 intrinsic register names.
2266 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2267 intrinsic register enum values and macros.
2269 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2271 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2272 for load/store addressing form.
2273 (nds32_print_operand_address): Likewise.
2275 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2278 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2279 based on LABEL_REF. Remove useless assertion.
2280 (pic_address_needs_scratch): Fix formatting.
2281 (sparc_legitimize_pic_address): Minor tweaks.
2282 (sparc_delegitimize_address): Adjust assertion accordingly.
2283 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2284 into symbolic_operand.
2285 (movsi_high_pic_label_ref): Likewise.
2286 (movsi_lo_sum_pic_label_ref): Likewise.
2287 (movdi_pic_label_ref): Likewise.
2288 (movdi_high_pic_label_ref): Likewise.
2289 (movdi_lo_sum_pic_label_ref): Likewise.
2291 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2293 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2294 custom LIB_SPEC setup.
2296 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2297 Kito Cheng <kito.cheng@gmail.com>
2299 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2300 * config/riscv/freebsd.h: New.
2302 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2304 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2305 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2308 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2309 Kito Cheng <kito.cheng@gmail.com>
2311 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2312 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2313 * config/nds32/nds32-protos.h (nds32_output_call,
2314 nds32_output_return): Declare.
2315 * config/nds32/nds32.md: Refine all the call and return patterns.
2317 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2320 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2321 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2323 PR rtl-optimization/84872
2324 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2325 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2328 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2330 * expr.c (copy_blkmode_to_reg): Revert 254862.
2331 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2333 2018-04-06 Richard Biener <rguenther@suse.de>
2336 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2337 after seeing a component reference with an adjacent field. Treat
2338 refs to arrays at struct end of external decls similar to
2339 refs to unconstrained commons.
2341 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2344 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2345 look through SAVE_EXPRs with non-side-effects argument. Adjust
2347 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2350 2018-04-06 Richard Biener <rguenther@suse.de>
2353 * alias.c (find_base_term): New wrapper around find_base_term
2354 unwinding CSELIB_VAL_PTR changes.
2355 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2358 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2360 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2362 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2363 constant definitions.
2364 ("nop"): lr 0,0 -> nopr r0
2365 ("nop_lr0", "nop_lr1"): New insn definitions.
2367 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2369 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2370 NDS32_V3PUSH_AVAILABLE_P macro.
2372 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2373 Chung-Ju Wu <jasonwucj@gmail.com>
2375 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2376 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2377 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2378 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2379 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2380 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2381 * config/nds32/constraints.md: New constraints and checking for hard
2382 float configuration.
2383 * config/nds32/iterators.md: New mode iterator and attribute for hard
2384 float configuration.
2385 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2387 * config/nds32/nds32-fpu.md: New file.
2388 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2389 deal with hard float code generation.
2390 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2392 (abi_type, float_reg_number): New enum type.
2393 * config/nds32/nds32-predicates.c: New predicates for hard float.
2394 * config/nds32/nds32-protos.h: Declare functions for hard float.
2395 * config/nds32/nds32.c: Implementation for hard float configuration.
2396 * config/nds32/nds32.h: Definitions for hard float configuration.
2397 * config/nds32/nds32.md: Include hard float machine description and
2398 modify patterns for hard float configuration.
2399 * config/nds32/nds32.opt: New options for hard float configuration.
2400 * config/nds32/predicates.md: New predicates for hard float
2403 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2405 * common/config/nds32/nds32-common.c
2406 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2408 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2411 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2412 CONSTRUCTOR_ELT (ctor, ...)->value.
2414 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2417 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2419 2018-04-05 Tom de Vries <tom@codesourcery.com>
2422 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2425 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2426 Kito Cheng <kito.cheng@gmail.com>
2428 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2429 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2430 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2431 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2433 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2434 Kito Cheng <kito.cheng@gmail.com>
2436 * config/nds32/constraints.md (Ufe): New memory constraint.
2437 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2438 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2439 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2441 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2442 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2444 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2446 * config/nds32/nds32.md: Use optimize_size in the condition for
2447 alu-shift instructions.
2449 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2451 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2453 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2455 * config/nds32/nds32.md (negsi2): Refine pattern.
2457 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2458 Chung-Ju Wu <jasonwucj@gmail.com>
2460 * config/nds32/iterators.md (shift_rotate): New code iterator.
2461 (shift): New code attribute.
2462 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2463 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2464 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2465 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2466 bit-wise operations.
2467 (andsi3, *andsi3): Ditto.
2468 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2469 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2470 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2471 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2472 nds32_ior_operand, nds32_xor_operand): New predicates.
2474 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2476 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2477 (addsi3, subsi3): ... this.
2479 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2481 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2483 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2485 * config/nds32/nds32.md: Adjust indention.
2487 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2489 * config/nds32/nds32.md (feature): New attribute.
2491 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2493 * config/nds32/nds32.md (subtype): New attribute.
2495 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2498 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2499 expansion to perform a bitwise AND of the argument followed by a
2500 boolean negation of the result.
2502 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2504 PR rtl-optimization/84878
2505 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2506 the basic block. Assert the use reference is not artificial and that
2507 it has an associated insn.
2509 2018-04-04 Michael Matz <matz@suse.de>
2511 * builtins.c (compute_objsize): Pass correct operand
2512 to array_at_struct_end_p.
2514 2018-04-04 Richard Biener <rguenther@suse.de>
2517 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2518 from contexts for DINFO_LEVEL_TERSE and below.
2520 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2522 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2523 resiter_operand condition.
2524 * config/nds32/nds32.md (*move<mode>): Ditto.
2526 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2527 Monk Chiang <sh.chiang04@gmail.com>
2529 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2531 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2533 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2535 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2536 Kito Cheng <kito.cheng@gmail.com>
2538 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2539 nds32_cond_code_str, output_cond_branch,
2540 output_cond_branch_compare_zero, nds32_expand_cbranch,
2541 nds32_expand_cstore, nds32_expand_movcc,
2542 nds32_output_cbranchsi4_equality_zero,
2543 nds32_output_cbranchsi4_equality_reg,
2544 nds32_output_cbranchsi4_equality_reg_or_const_int,
2545 nds32_output_cbranchsi4_greater_less_zero: New functions.
2546 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2547 nds32_expand_cstore, nds32_expand_movcc,
2548 nds32_output_cbranchsi4_equality_zero,
2549 nds32_output_cbranchsi4_equality_reg,
2550 nds32_output_cbranchsi4_equality_reg_or_const_int,
2551 nds32_output_cbranchsi4_greater_less_zero): Declare.
2552 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2553 nds32_rimm11s_operand): New predicates.
2554 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2555 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2558 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2560 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2561 * config/nds32/nds32.md: Ditto.
2562 * config/nds32/pipelines.md: Ditto.
2564 2018-04-04 Richard Biener <rguenther@suse.de>
2566 PR tree-optimization/85168
2567 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2568 propagating abnormals.
2570 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2572 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2574 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2575 Kito Cheng <kito.cheng@gmail.com>
2577 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2578 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2579 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2580 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2581 * config/nds32/nds32.md (sibcall_internal): New.
2582 (sibcall_register): Remove.
2583 (sibcall_immediate): Remove.
2584 (sibcall_value_internal): New.
2585 (sibcall_value_register): Remove.
2586 (sibcall_value_immediate): Remove.
2587 * config/nds32/predicates.md (nds32_general_register_operand): New.
2588 (nds32_call_address_operand): New.
2590 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2592 PR rtl-optimization/85167
2593 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2594 bb_defs if *split_p, instead preinitialize it to NULL.
2596 PR tree-optimization/85156
2597 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2598 evaluating the argument multiple times.
2600 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2602 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2604 (_mm_cvtpd_ps): Likewise.
2605 (_mm_cvttpd_epi32): Likewise.
2606 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2607 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2608 vector, pixel, and bool following altivec.h include.
2610 2018-04-03 Martin Sebor <msebor@redhat.com>
2612 * doc/extend.texi (Common Function Attributes): Clarify.
2613 (const attribute): Likewise.
2614 (pure attribute): Likewise.
2616 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2619 * config/i386/i386.c (ix86_expand_vector_set): Use
2620 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2622 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2624 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2625 instructions when changing rounding bits to preserve precision bits
2626 in the x87 control word.
2628 2018-04-03 Martin Liska <mliska@suse.cz>
2630 PR tree-optimization/82491
2631 * rtl.h (strip_offset_and_add): Replace += suboffset with
2632 poly_uint64 () + suboffset.
2634 2018-03-29 Martin Liska <mliska@suse.cz>
2635 Martin Jambor <mjambor@suse.cz>
2638 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2639 param_type is not an integral or pointer type.
2641 2018-04-03 Richard Biener <rguenther@suse.de>
2643 * sese.h (recompute_all_dominators): Remove.
2645 2018-04-02 Martin Sebor <msebor@redhat.com>
2647 * doc/invoke.texi (-Wrestrict): Fix typos.
2649 2018-04-02 Jim Wilson <jimw@sifive.com>
2651 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2652 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2653 (<optab>di3, <optab>si3_extend): Likewise.
2654 (<optab>si3_mask, <optab>si3_mask_1): New.
2655 (<optab>di3_mask, <optab>di3_mask_1): New.
2656 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2657 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2658 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2660 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2662 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2665 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2667 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2668 (nds32_canonicalize_comparison): New function.
2670 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2671 Kito Cheng <kito.cheng@gmail.com>
2672 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2674 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2675 * config/nds32/constants.md (unspec_volatile_element): Add
2676 UNSPEC_VOLATILE_RELAX_GROUP.
2677 * config/nds32/nds32-relax-opt.c: New file.
2678 * config/nds32/nds32-predicates.c
2679 (nds32_symbol_load_store_p): New function.
2680 * config/nds32/nds32-protos.h
2681 (nds32_symbol_load_store_p): Declare function.
2682 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2683 * config/nds32/nds32.c
2684 (nds32_register_pass): New function to register pass.
2685 (nds32_register_passes): New function to register passes.
2686 * config/nds32/nds32.md (relax_group): New pattern.
2687 * config/nds32/nds32.opt (mrelax-hint): New option.
2688 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2690 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2692 * config/nds32/t-nds32: Modify files dependency.
2694 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2696 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2697 (PROFILE_HOOK): Define its implementation.
2699 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2701 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2702 type and 32-bit size.
2704 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2707 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2708 (V_128_256): New mode iterator.
2709 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2710 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2711 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2713 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2714 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2716 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2719 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2720 NaN inputs correctly.
2722 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2725 * config/rs6000/vsx.md (??r): New mode attribute.
2726 (*vsx_mov<mode>_64bit): Use it.
2727 (*vsx_mov<mode>_32bit): Likewise.
2729 2018-03-30 Martin Sebor <msebor@redhat.com>
2731 PR tree-optimization/84818
2732 * builtins.c (check_access): Use warning_n.
2734 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2737 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2739 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2742 2018-03-30 Julia Koval <julia.koval@intel.com>
2745 * x86-tune.def (movx, partial_reg_dependency): Enable for
2748 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2751 * lra-constraints.c (process_alt_operands): Move setting
2752 this_alternative_matches below.
2754 2018-03-29 Martin Liska <mliska@suse.cz>
2757 * doc/invoke.texi: Document how LTO works with debug info.
2758 Describe auto-load support of binutils. Mention 'x86-64'
2759 as valid option value of -march option.
2761 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2763 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2766 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2767 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2768 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2771 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2774 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2775 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2776 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2777 for RS6000_BTM_POWERPC64.
2778 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2779 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2780 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2785 2018-03-28 Carl Love <cel@us.ibm.com>
2788 2017-09-27 Carl Love <cel@us.ibm.com>
2790 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2791 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2792 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2795 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2797 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2798 instead of __vector bool.
2799 (_mm_max_pu8): Likewise.
2800 (_mm_min_pi16): Likewise.
2802 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2805 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2806 (DIVWEUO): Likewise.
2808 (DIVDEUO): Likewise.
2809 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2810 DIVWEUO and DIVDEUO.
2811 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2812 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2813 (div_extend): Likewise.
2814 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2816 (__builtin_divweuo): Likewise.
2817 (__builtin_divdeo): Likewise.
2818 (__builtin_divdeuo): Likewise.
2820 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2823 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2824 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2826 PR tree-optimization/82004
2827 * gimple-match-head.c (optimize_pow_to_exp): New function.
2828 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2829 Don't fold to exp if optimize_pow_to_exp is false.
2831 2018-03-28 Martin Liska <mliska@suse.cz>
2834 * calls.c (initialize_argument_information): Fix trailing space.
2835 * common.opt: Fix typo and provide better explanation for
2836 -fsanitize-coverage option.
2837 * config/i386/i386.opt: Fix typo.
2839 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2840 Martin Liska <mliska@suse.cz>
2843 * gimplify.c (asan_poison_variable): Don't do the check for
2844 gimplify_omp_ctxp here.
2845 (gimplify_decl_expr): Do it here.
2846 (gimplify_target_expr): Likewise.
2848 2018-03-28 Martin Liska <mliska@suse.cz>
2851 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2852 chkp_type_bounds_count if MPX is not enabled.
2854 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2856 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2858 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2861 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2862 function to create the function decl for complex long double
2863 multiply and divide for -mabi=ieeelongdouble.
2864 (init_float128_ieee): Call it.
2866 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2869 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2870 -fcf-protection=branch -mibt.
2871 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2873 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2876 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2878 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2881 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2882 extern array declarations.
2884 2018-03-27 Richard Biener <rguenther@suse.de>
2887 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2888 explicit single_use checks.
2890 2018-03-27 Richard Biener <rguenther@suse.de>
2892 PR tree-optimization/85082
2893 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2896 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2898 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2899 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2900 Turn on fasynchronous-unwind-tables and funwind-tables.
2902 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2905 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2906 (*bmi_blsr_<mode>_ccz): Ditto.
2908 2018-03-26 Tom de Vries <tom@codesourcery.com>
2910 PR tree-optimization/85063
2911 * omp-general.c (offloading_function_p): New function. Factor out
2913 * omp-offload.c (pass_omp_target_link::gate): ... here.
2914 * omp-general.h (offloading_function_p): Declare.
2915 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2916 with attribute omp declare target for offloading functions.
2918 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2920 PR tree-optimization/84005
2921 * tree-data-ref.h (get_base_for_alignment): Declare.
2922 * tree-data-ref.c (get_base_for_alignment_1): New function.
2923 (get_base_for_alignment): Likewise.
2924 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2925 get_base_for_alignment to find a suitable base object, instead
2926 of always using drb->base_address.
2928 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2931 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2932 known size by default.
2934 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2937 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2938 and non BLKmode operands.
2940 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2943 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2944 Clean up attributes.
2946 2018-03-23 Richard Biener <rguenther@suse.de>
2949 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2950 we are going to emit early debug for LTO.
2952 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2955 * function.c (match_asm_constraints_1): Don't optimize if input
2956 doesn't satisfy general_operand predicate for output's mode.
2959 * alias.c (write_dependence_p): Don't require for x_canonicalized
2960 non-VOIDmode if x has VOIDmode.
2963 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2964 just don't try to optimize it rather than assert it never happens.
2966 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2968 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2969 macro expansions for definition of ST_INTERNAL_<mode> and
2970 LD_INTERNAL_<mode> builtins.
2971 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2973 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2975 (altivec_expand_st_builtin): Likewise.
2976 (altivec_expand_builtin): Remove calls to deleted functions.
2977 (rs6000_address_for_altivec): Delete this function.
2978 * config/rs6000/vector.md: Remove expands for
2979 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2981 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2984 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2985 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2986 re-computing once computed.
2987 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2988 (arm_init_machine_status): Initialize
2989 machine->static_chain_stack_bytes.
2991 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2994 * doc/extend.texi: Add four new prototypes for vec_ld.
2995 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2996 definitions for more logical presentation.
2997 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2998 entries for V1TI variants of __builtin_altivec_ld builtin.
2999 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3000 handling of V1TI variant of LVX icode pattern.
3001 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3002 (rs6000_gimple_fold_builtin): Likewise.
3003 (altivec_init_builtins): Add code to define
3004 __builtin_altivec_lvx_v1ti function.
3006 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3009 * function.c (match_asm_constraints_1): Don't do the optimization
3010 if input isn't a REG, SUBREG, MEM or constant.
3012 2018-03-22 Tom de Vries <tom@codesourcery.com>
3014 PR tree-optimization/84956
3015 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3016 bb_has_abnormal_pred.
3018 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3021 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3022 DECL_INITIAL (decl) to decl at the end.
3023 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3026 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3028 * doc/extend.texi (__builtin_tgmath): Document when complex
3029 integer types are treated as _Complex _Float64.
3031 2018-03-21 Tom de Vries <tom@codesourcery.com>
3033 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3035 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3037 PR tree-optimization/84960
3038 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3039 if it is ENTRY block, move them into single succ of ENTRY in that case.
3041 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3043 PR tree-optimization/84811
3044 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3045 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3046 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3047 (known_subrange_p): Update accordingly. Cast each value involved
3048 in the size comparison, rather than casting the result of the
3051 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3053 PR tree-optimization/84982
3054 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3055 by flipping the least significant bit rather than all bits from
3056 bitpos to bitpos + bitsize - 1.
3058 2018-03-21 Nathan Sidwell <nathan@acm.org>
3060 * doc/extend.texi (Deprecated Features): Remove mention of
3061 long-deleted deprecations.
3063 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3066 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3067 * configure: Regenerate.
3069 2018-03-21 Tom de Vries <tom@codesourcery.com>
3071 PR tree-optimization/83126
3072 * tree-parloops.c (num_phis): New function.
3073 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3075 2018-03-21 Nathan Sidwell <nathan@acm.org>
3077 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3078 mention anon-struct/union members and trailing attributes.
3080 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3082 PR tree-optimization/84969
3083 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3084 builtin memset partitions if they set different rhs values.
3086 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3088 PR rtl-optimization/84989
3089 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3090 VEC_DUPLICATE with scalar result mode.
3092 2018-03-21 Martin Liska <mliska@suse.cz>
3095 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3096 not intended return statement.
3098 2018-03-21 Martin Liska <mliska@suse.cz>
3101 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3102 (chkp_find_bound_slots_1): Limit number of iterations.
3104 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3107 * Minor grammar fixes for x86 options.
3109 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3112 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3113 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3115 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3118 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3119 (altivec_lvx_<mode>_1op): Likewise.
3120 (altivec_stvx_<mode>_2op): Likewise.
3121 (altivec_stvx_<mode>_1op): Likewise.
3122 (altivec_lvx_<VM2:mode>): New define_expand.
3123 (altivec_stvx_<VM2:mode>): Likewise.
3124 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3125 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3126 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3127 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3128 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3129 (rs6000_gen_lvx): Likewise.
3130 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3131 (altivec_expand_stv_builtin): Likewise.
3132 (altivec_expand_builtin): Likewise.
3133 * config/rs6000/vector.md: Likewise.
3135 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3138 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3141 2018-03-20 Richard Biener <rguenther@suse.de>
3144 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3145 sign-conversions as zero, fall back to standard scalar_stmt
3148 2018-03-20 Martin Liska <mliska@suse.cz>
3151 * predict.c (rebuild_frequencies): Handle case when we have
3152 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3154 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3157 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3158 flag_section_anchors.
3159 * varasm.c (use_blocks_for_decl_p): Remove hack for
3160 dw2_force_const_mem.
3163 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3165 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3166 be created, use lowpart_subreg of operands[0] rather than operands[0]
3168 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3169 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3170 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3171 and n constraint instead of aarch64_shift_imm_di and Usd.
3172 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3173 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3175 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3178 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3179 to favor GPR over NEON registers.
3180 (<shift>di3_neon): Likewise.
3182 2018-03-20 Tom de Vries <tom@codesourcery.com>
3185 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3186 (nvptx_process_pars): Emit bar.sync asap and alap.
3188 2018-03-20 Tom de Vries <tom@codesourcery.com>
3191 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3192 seen_label if seen_label is already set.
3194 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3197 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3198 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3199 Use 1U instead of 1. Formatting fixes.
3202 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3203 instead of TREE_TYPE (s1) for the return value.
3205 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3207 PR tree-optimization/84946
3208 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3209 bitsize + bitsize in poly_uint64 rather than poly_int64.
3212 * dwarf2asm.c: Include fold-const.c.
3213 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3214 of decl rather than decl itself.
3216 PR rtl-optimization/84643
3217 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3219 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3222 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3223 calling assemble_variable.
3225 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3228 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3229 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3231 2018-03-19 Jim Wilson <jimw@sifive.com>
3234 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3235 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3236 (riscv_first_stack_step): Likewise.
3237 (riscv_option_override): Use STACK_BOUNDARY instead of
3239 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3241 (BIGGEST_ALIGNMENT): Set to 128.
3242 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3243 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3246 2018-03-19 Richard Biener <rguenther@suse.de>
3248 PR tree-optimization/84933
3249 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3250 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3252 2018-03-19 Richard Biener <rguenther@suse.de>
3254 PR tree-optimization/84859
3255 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3256 (cond_if_else_store_replacement): Perform sinking operation on
3257 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3258 Generalize what a BB with a single eligible store is.
3260 2018-03-19 Richard Biener <rguenther@suse.de>
3262 PR tree-optimization/84929
3263 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3264 chrec_is_positive against non-chrec arg.
3266 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3269 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3271 2018-03-18 Martin Liska <mliska@suse.cz>
3273 PR rtl-optimization/84635
3274 * regrename.c (build_def_use): Use matches_mode only when
3277 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3279 PR tree-optimization/84913
3280 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3281 vectorize chains of COND_EXPRs.
3283 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3287 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3289 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3291 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3293 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3295 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3296 Kito Cheng <kito.cheng@gmail.com>
3298 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3299 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3300 (nds32_adjust_reg_alloc_order): New function.
3301 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3303 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3305 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3306 nds32_print_operand, nds32_print_operand_address): Use
3307 HOST_WIDE_INT_PRINT_DEC instead.
3309 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3311 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3313 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3316 * config/i386/i386.c (initial_ix86_tune_features,
3317 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3319 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3320 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3321 rather than 1u << ix86_tune. Formatting fix.
3322 (ix86_option_override_internal): Change ix86_arch_mask from
3323 unsigned int to unsigned HOST_WIDE_INT, initialize to
3324 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3325 (ix86_function_specific_restore): Likewise.
3327 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3330 * postreload.c (reload_combine_recognize_pattern): Perform
3331 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3332 truncate_int_for_mode the result for the destination's mode.
3335 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3336 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3338 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3341 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3344 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3345 regno_allocno_class_array and sorted_pseudos.
3346 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3347 insns where regno is used.
3349 2018-03-16 Martin Liska <mliska@suse.cz>
3352 * multiple_target.c (create_dispatcher_calls): Redirect
3353 reference in the symbol table.
3355 2018-03-16 Martin Liska <mliska@suse.cz>
3358 * multiple_target.c (create_dispatcher_calls): Redirect also
3361 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3365 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3366 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3367 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3369 2018-03-16 Julia Koval <julia.koval@intel.com>
3371 * doc/invoke.texi (Skylake Server): Add CLWB.
3372 Cannonlake): Remove CLWB.
3374 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3376 PR tree-optimization/84841
3377 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3379 (FLOAT_ONE_CONST_TYPE): Define.
3380 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3381 (sort_by_operand_rank): Put entries with higher constant_type last
3382 rather than first to match comments.
3384 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3386 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3389 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3392 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3393 check and use address of target always.
3395 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3398 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3399 (indirect_thunk_bnd_needed): Likewise.
3400 (indirect_thunks_used): Likewise.
3401 (indirect_thunks_bnd_used): Likewise.
3402 (indirect_return_needed): New.
3403 (indirect_return_bnd_needed): Likewise.
3404 (output_indirect_thunk_function): Add a bool argument for
3406 (output_indirect_thunk_function): Don't generate alias for
3407 function return thunk.
3408 (ix86_code_end): Call output_indirect_thunk_function to generate
3409 function return thunks.
3410 (ix86_output_function_return): Set indirect_return_bnd_needed
3411 and indirect_return_needed instead of indirect_thunk_bnd_needed
3412 and indirect_thunk_needed.
3414 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3416 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3417 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3418 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3420 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3421 Paul Hua <paul.hua.gm@gmail.com>
3424 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3426 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3428 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3429 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3432 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3435 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3436 instead of GET_MODE_SIZE when comparing Units.
3438 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3441 * varasm.c (hash_section): Return an unchangeble hash value
3442 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3443 Return !aarch64_can_use_per_function_literal_pools_p ().
3445 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3448 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3449 rather than address of cmode as last argument to prepare_cmp_insn.
3451 2018-03-15 Julia Koval <julia.koval@intel.com>
3453 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3454 F_AVX512VNNI, F_AVX512BITALG): New.
3456 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3459 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3460 insn for floating-point loads and stores.
3462 2018-03-14 Carl Love <cel@us.ibm.com>
3464 * config/rs6000/rs6000-c.c: Add macro definitions for
3465 ALTIVEC_BUILTIN_VEC_PERMXOR.
3466 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3467 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3468 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3469 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3471 * config/doc/extend.texi: Add prototypes for vec_permxor.
3473 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3476 * diagnostic-show-locus.c (class layout_point): Convert m_line
3477 from int to linenum_type.
3478 (line_span::comparator): Use linenum "compare" function when
3479 comparing line numbers.
3480 (test_line_span): New function.
3481 (layout_range::contains_point): Convert param "row" from int to
3483 (layout_range::intersects_line_p): Likewise.
3484 (layout::will_show_line_p): Likewise.
3485 (layout::print_source_line): Likewise.
3486 (layout::should_print_annotation_line_p): Likewise.
3487 (layout::print_annotation_line): Likewise.
3488 (layout::print_leading_fixits): Likewise.
3489 (layout::annotation_line_showed_range_p): Likewise.
3490 (struct line_corrections): Likewise for field m_row.
3491 (line_corrections::line_corrections): Likewise for param "row".
3492 (layout::print_trailing_fixits): Likewise.
3493 (layout::get_state_at_point): Likewise.
3494 (layout::get_x_bound_for_row): Likewise.
3495 (layout::print_line): Likewise.
3496 (diagnostic_show_locus): Likewise for locals "last_line" and
3498 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3499 * input.c (selftest::test_linenum_comparisons): New function.
3500 (selftest::input_c_tests): Call it.
3501 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3502 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3503 * selftest.h (ASSERT_GT): New macro.
3504 (ASSERT_GT_AT): New macro.
3505 (ASSERT_LT): New macro.
3506 (ASSERT_LT_AT): New macro.
3508 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3510 PR rtl-optimization/84780
3511 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3513 2018-03-14 Martin Liska <mliska@suse.cz>
3515 * tree.c (record_node_allocation_statistics): Use
3516 get_stats_node_kind.
3517 (get_stats_node_kind): New function extracted from
3518 record_node_allocation_statistics.
3519 (free_node): Use get_stats_node_kind.
3521 2018-03-14 Richard Biener <rguenther@suse.de>
3523 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3524 that the value-set of ANTIC_IN doesn't grow.
3527 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3529 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3530 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3532 2018-03-14 Julia Koval <julia.koval@intel.com>
3534 * config.gcc (icelake-client, icelake-server): New.
3536 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3537 (initial_ix86_arch_features): Ditto.
3538 (PTA_SKYLAKE): Add SGX.
3539 (PTA_ICELAKE): Remove.
3540 (PTA_ICELAKE_CLIENT): New.
3541 (PTA_ICELAKE_SERVER): New.
3542 (ix86_option_override_internal): Split up icelake on icelake client and
3544 (get_builtin_code_for_version): Ditto.
3545 (fold_builtin_cpu): Ditto.
3546 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3547 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3548 * config/i386/i386.h (processor_type): Ditto.
3549 * doc/invoke.texi: Ditto.
3551 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3554 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3555 INTEGER_CST offset, add it together with bitpos / 8 and
3556 sign extend based on POINTER_SIZE.
3560 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3563 * config/i386/constraints.md (Yc): New register constraint.
3564 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3565 Use Yc constraint for alternative 2 of operand 0. Remove
3566 preferred_for_speed attribute.
3568 2018-03-14 Richard Biener <rguenther@suse.de>
3570 PR tree-optimization/84830
3571 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3572 with the old one to avoid oscillations.
3574 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3577 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3579 (assign_by_spills): Return a flag of reload assignment failure.
3580 Do not process the reload assignment failures. Do not spill other
3581 reload pseudos if they has the same reg class. Update n if
3583 (lra_assign): Add a return arg. Set up from the result of
3584 assign_by_spills call.
3585 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3586 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3587 usage_insns if it is not NULL.
3588 (spill_hard_reg_in_range): New function.
3589 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3590 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3591 function prototypes.
3592 (lra_assign): Change prototype.
3593 * lra.c (lra): Add code to deal with fails by splitting hard reg
3596 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3598 * config/riscv/riscv.opt (mrelax): New option.
3599 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3600 "norelax" when riscv_mrelax is disabled.
3601 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3603 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3606 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3607 reassociation for int modes.
3609 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3611 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3612 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3614 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3615 * config/aarch64/aarch64-sve.md
3616 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3617 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3618 (*extend<mode><Vwide>2): Rename to...
3619 (aarch64_sve_extend<mode><Vwide>2): ...this.
3620 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3621 renaming the old pattern to...
3622 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3624 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3625 define_expand, renaming the old pattern to...
3626 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3627 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3628 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3629 account when deciding which SVE instruction the optab should use.
3630 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3632 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3634 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3635 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3636 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3637 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3638 (tlsdesc_small_<mode>): Turn a define_expand and use
3639 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3640 (tlsdesc_small_advsimd_<mode>): ...this.
3641 (tlsdesc_small_sve_<mode>): New pattern.
3643 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3645 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3646 (UNSPEC_UMUL_HIGHPART): New constants.
3647 (MUL_HIGHPART): New int iteraor.
3648 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3649 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3651 (*<su>mul<mode>3_highpart): New define_insn.
3653 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3656 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3659 2018-03-13 Martin Liska <mliska@suse.cz>
3662 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3664 (sem_item_optimizer::~sem_item_optimizer): Release it.
3665 (sem_item_optimizer::merge_classes): Register variable aliases.
3666 (sem_item_optimizer::fixup_pt_set): New function.
3667 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3668 * ipa-icf.h: Declare new variables and functions.
3670 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3673 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3674 integer_pow2p@2 and test integer_pow2p in condition.
3675 (A < 0 ? C : 0): Similarly for @1.
3678 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3679 characters starting at p contain '\0' character, don't look beyond
3683 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3684 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3687 * reg-stack.c (change_stack): Change update_end var from int to
3688 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3689 also call set_block_for_insn on the newly added insns and rescan.
3692 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3693 on the last operand.
3696 * tree.c (stabilize_reference_1): Return save_expr (e) for
3697 STATEMENT_LIST even if it doesn't have side-effects.
3699 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3701 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3703 2018-03-12 Renlin Li <renlin.li@arm.com>
3705 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3706 aarch64_output_scalar_simd_mov_immediate.
3708 2018-03-12 Martin Sebor <msebor@redhat.com>
3710 PR tree-optimization/83456
3711 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3712 for perfectly overlapping calls to memcpy.
3713 (gimple_fold_builtin_memory_chk): Same.
3714 (gimple_fold_builtin_strcpy): Handle no-warning.
3715 (gimple_fold_builtin_stxcpy_chk): Same.
3716 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3718 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3720 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3721 parameter. Use it for SFmode.
3722 (rs6000_function_arg_advance_1): Adjust.
3723 (rs6000_function_arg): Adjust.
3724 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3726 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3728 PR rtl-optimization/84169
3729 PR rtl-optimization/84780
3730 * combine.c (can_combine_p): Check for a 2-insn combination whether
3731 the destination register is used between the two insns, too.
3733 2018-03-12 Richard Biener <rguenther@suse.de>
3735 PR tree-optimization/84803
3736 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3737 for refs DR analysis didn't process.
3739 2018-03-12 Richard Biener <rguenther@suse.de>
3741 PR tree-optimization/84777
3742 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3743 force-vectorize loops ignore whether we are optimizing for size.
3745 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3747 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3748 (TARGET_MD_ASM_ADJUST): Define.
3750 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3751 Kito Cheng <kito.cheng@gmail.com>
3752 Chung-Ju Wu <jasonwucj@gmail.com>
3754 * config/nds32/nds32.c (nds32_compute_stack_frame,
3755 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3756 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3757 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3758 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3759 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3760 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3761 * config/nds32/nds32.md (prologue, epilogue): Use macro
3762 NDS32_V3PUSH_AVAILABLE_P to do checking.
3764 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3767 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3768 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3769 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3770 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3771 addition of most attributes on !orig_type_die or the attribute not
3772 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3774 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3775 Chung-Ju Wu <jasonwucj@gmail.com>
3777 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3779 * config/nds32/nds32.opt (mvh): New option.
3781 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3782 Chung-Ju Wu <jasonwucj@gmail.com>
3784 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3786 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3787 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3790 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3791 Chung-Ju Wu <jasonwucj@gmail.com>
3793 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3795 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3796 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3798 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3799 Kito Cheng <kito.cheng@gmail.com>
3800 Chung-Ju Wu <jasonwucj@gmail.com>
3802 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3803 UNSPEC_FFMISM and UNSPEC_FLMISM.
3804 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3805 for ffb, ffmism and flmism.
3806 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3807 (unspec_ffmism): Ditto.
3808 (unspec_flmism): Ditto.
3809 (nds32_expand_builtin_impl): Check if string extension is available.
3810 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3811 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3813 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3816 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3819 * lra-assigns.c (assign_by_spills): Return a flag of reload
3820 assignment failure. Do not process the reload assignment
3821 failures. Do not spill other reload pseudos if they has the same
3823 (lra_assign): Add a return arg. Set up from the result of
3824 assign_by_spills call.
3825 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3826 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3827 usage_insns if it is not NULL.
3828 (spill_hard_reg_in_range): New function.
3829 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3830 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3831 function prototypes.
3832 (lra_assign): Change prototype.
3833 * lra.c (lra): Add code to deal with fails by splitting hard reg
3836 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3839 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3841 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3844 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3845 (dw_val_node): Add val_symbolic_view.
3846 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3847 (symview_upper_bound): New.
3848 (new_line_info_table): Initialize symviews_since_reset.
3849 (dwarf2out_source_line): Count symviews_since_reset and set
3850 symview_upper_bound.
3851 (dw_val_equal_p): Handle symview.
3852 (add_AT_symview): New.
3853 (print_dw_val): Handle symview.
3854 (attr_checksum, attr_checksum_ordered): Likewise.
3855 (same_dw_val_p, size_of_die): Likewise.
3856 (value_format, output_die): Likewise.
3857 (add_high_low_attributes): Use add_AT_symview for entry_view.
3858 (dwarf2out_finish): Reset symview_upper_bound, clear
3861 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3864 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3865 Add strict argument and use it.
3866 (rs6000_split_multireg_move): Update for new strict argument.
3867 (mem_operand_gpr): Disallow all non-offsettable addresses.
3868 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3870 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3873 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3874 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3875 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3878 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3879 decl, use remap_type if we want to use the type.
3881 2018-03-09 Martin Sebor <msebor@redhat.com>
3883 PR tree-optimization/84526
3884 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3886 (builtin_access::generic_overlap): Be prepared to handle non-array
3889 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3891 PR rtl-optimization/84682
3892 * lra-constraints.c (process_address_1): Check is_address flag
3893 for address constraints.
3894 (process_alt_operands): Likewise.
3895 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3896 preprocess_constraints.
3897 * recog.h (preprocess_constraints): Add oploc parameter.
3899 * recog.c (preprocess_constraints): Test address_operand for
3900 CT_ADDRESS constraints.
3902 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3905 * lra-assigns.c (assign_by_spills): Return a flag of reload
3906 assignment failure. Do not process the reload assignment
3907 failures. Do not spill other reload pseudos if they has the same
3909 (lra_assign): Add a return arg. Set up from the result of
3910 assign_by_spills call.
3911 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3912 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3913 usage_insns if it is not NULL.
3914 (spill_hard_reg_in_range): New function.
3915 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3916 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3917 function prototypes.
3918 (lra_assign): Change prototype.
3919 * lra.c (lra): Add code to deal with fails by splitting hard reg
3922 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3925 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3926 Accept complain bool parameter. Only emit errors if it is true.
3927 (arm_parse_cpu_option_name): Likewise.
3928 (arm_target_thumb_only): Adjust callers of the above.
3929 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3930 prototype to take a default true bool parameter.
3931 (arm_parse_arch_option_name): Likewise.
3933 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3934 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3938 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3939 * configure: Regenerate.
3940 * configure.ac ("linker --version-script option"): New.
3941 ("linker soname option"): New.
3943 2018-03-09 Richard Biener <rguenther@suse.de>
3945 PR tree-optimization/84775
3946 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3947 immediate uses of predicate stmts and mark them modified.
3950 PR tree-optimization/84178
3951 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3953 (version_loop_for_if_conversion): Delay update_ssa call.
3954 (tree_if_conversion): Delay update_ssa until after predicate
3957 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3960 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3961 when the function accesses prior frames.
3963 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3966 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3967 gen_llsym, otherwise call maybe_gen_llsym.
3970 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3971 has ',' character inside of it.
3973 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3976 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3977 as clobbering CC_REGNUM.
3979 2018-03-08 Richard Biener <rguenther@suse.de>
3982 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3983 (follow_copies_to_constant): Do not follow SSA names registered
3986 2018-03-08 Richard Biener <rguenther@suse.de>
3988 PR tree-optimization/84178
3989 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3991 (version_loop_for_if_conversion): Delay update_ssa call.
3992 (tree_if_conversion): Delay update_ssa until after predicate
3995 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3997 PR tree-optimization/84178
3998 * tree-if-conv.c (release_bb_predicate): Remove the
3999 the assertion that the stmts have NULL use_ops.
4000 Discard the statements, asserting that they haven't
4001 yet been added to a BB.
4003 2018-03-08 Richard Biener <rguenther@suse.de>
4005 PR tree-optimization/84746
4006 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4007 (phi_translate): Pass in destination ANTIC_OUT set.
4008 (phi_translate_1): Likewise. For a simplified result lookup
4009 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4010 (phi_translate_set): Adjust.
4011 (do_pre_regular_insertion): Likewise.
4012 (do_pre_partial_partial_insertion): Likewise.
4014 2018-03-08 Martin Liska <mliska@suse.cz>
4016 PR gcov-profile/84735
4017 * doc/gcov.texi: Document usage of profile files.
4018 * gcov-io.h: Document changes in the format.
4020 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4024 * dwarf2out.c (struct dw_line_info_table): Update comments for
4026 (FORCE_RESET_NEXT_VIEW): New.
4027 (FORCE_RESETTING_VIEW_P): New.
4028 (RESETTING_VIEW_P): Check for -1 too.
4029 (ZERO_VIEW_P): Likewise.
4030 (new_line_info_table): Force-reset next view.
4031 (dwarf2out_begin_function): Likewise.
4032 (dwarf2out_source_line): Simplify zero_view_p initialization.
4033 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4034 view directly. Omit view when omitting .loc at line 0.
4036 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4038 PR tree-optimization/84740
4039 * tree-switch-conversion.c (process_switch): Call build_constructors
4040 only if info.phi_count is non-zero.
4042 PR tree-optimization/84739
4043 * tree-tailcall.c (find_tail_calls): Check call arguments against
4044 DECL_ARGUMENTS (current_function_decl) rather than
4045 DECL_ARGUMENTS (func) when checking for tail recursion.
4047 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4049 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4050 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4051 Volker Reichelt's entry and add entries for people that perform
4052 GCC fuzzy testing and report numerous bugs.
4054 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4057 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4058 readonly data in sdata, if that is disabled.
4059 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4060 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4061 -mreadonly-in-sdata option.
4063 2018-03-07 Martin Sebor <msebor@redhat.com>
4065 PR tree-optimization/84468
4066 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4067 basic block when looking for nul assignment.
4069 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4072 * except.h (output_function_exception_table): Adjust prototype.
4073 * except.c (output_function_exception_table): Remove FNNAME parameter
4074 and add SECTION parameter. Ouput one part of the table at a time.
4075 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4076 the first part of the exception table and emit unwind directives.
4077 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4078 (i386_pe_seh_cold_init): Likewise.
4079 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4080 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4081 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4082 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4083 * config/i386/winnt.c: Include except.h.
4084 (struct seh_frame_state): Add reg_offset, after_prologue and
4085 in_cold_section fields.
4086 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4087 (i386_pe_seh_cold_init): New function.
4088 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4089 to seh->in_cold_section.
4090 (seh_emit_push): Record the offset of the push.
4091 (seh_emit_save): Record the offet of the save.
4092 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4093 Test seh->after_prologue to disregard the epilogue.
4094 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4095 (i386_pe_end_cold_function): New function.
4097 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4100 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4101 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4104 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4106 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4107 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4110 * multiple_target.c: Include tree-inline.h and intl.h.
4111 (expand_target_clones): Diagnose and fail if node->definition and
4112 !tree_versionable_function_p (node->decl).
4114 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4116 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4118 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4119 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4120 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4122 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4125 * combine.c (try_combine): Use reg_or_subregno instead of handling
4126 just paradoxical SUBREGs and REGs.
4128 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4130 * config/arc/arc.c (arc_finalize_pic): Remove function.
4131 (arc_must_save_register): We use single base PIC register, remove
4132 checks to save/restore the PIC register.
4133 (arc_expand_prologue): Likewise.
4134 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4136 (arc_verify_short): Likewise.
4137 (arc_attr_type): Likewise.
4138 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4139 (walk_stores): Likewise.
4140 (arc_address_cost): Make it static.
4141 (arc_verify_short): Likewise.
4142 (branch_dest): Likewise.
4143 (arc_attr_type): Likewise.
4144 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4145 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4146 (arc_final_prescan_insn): Remove inserting the nops due to
4147 hardware hazards. It is done in reorg step.
4148 (insn_length_variant_t): Remove.
4149 (insn_length_parameters_t): Likewise.
4150 (arc_insn_length_parameters): Likewise.
4151 (arc_get_insn_variants): Likewise.
4152 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4154 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4157 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4160 PR tree-optimization/84687
4161 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4163 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4165 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4167 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4168 Rename to ppc_speculation_barrier.
4169 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4170 __builtin_ppc_speculation_barrier.
4172 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4175 * combine.c (combine_simplify_rtx): Don't try to simplify if
4176 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4179 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4181 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4182 to 32 bytes when compiling for POWER9.
4184 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4187 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4188 regparm >= 3 with no arg reg available also for calls with
4189 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4192 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4194 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4196 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4199 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4201 2018-03-05 Richard Biener <rguenther@suse.de>
4203 PR tree-optimization/84486
4204 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4205 When inserting a __builtin_assume_aligned call set the LHS
4206 SSA name alignment info accordingly.
4208 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4210 PR tree-optimization/84114
4211 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4212 Avoid reassociation of FLOAT_MODE addition.
4214 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4216 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4217 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4218 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4219 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4220 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4221 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4222 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4224 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4225 __builtin_ia32_wbinvd): New builtins.
4226 (SPECIAL_ARGS2): New.
4227 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4228 (SPECIAL_ARGS2): New.
4229 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4230 (ix86_valid_target_attribute_inner_p): Ditto.
4231 (ix86_init_mmx_sse_builtins): Add special_args2.
4232 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4233 TARGET_WBNOINVD_P): New.
4234 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4235 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4236 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4237 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4238 * config/i386/pconfigintrin.h: New file.
4239 * config/i386/wbnoinvdintrin.h: Ditto.
4240 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4241 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4243 2018-03-05 Richard Biener <rguenther@suse.de>
4245 PR tree-optimization/84670
4246 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4248 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4249 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4250 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4251 doesn't grow if all successors (recursively) were visited at least
4254 2018-03-05 Richard Biener <rguenther@suse.de>
4256 PR tree-optimization/84650
4257 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4258 if executed in the loop pipeline.
4260 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4262 * doc/configfiles.texi (Configuration Files): Move info about
4263 conditionalizing $target-protos.h to...
4264 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4265 differs from $target-protos.h.
4267 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4268 Chung-Ju Wu <jasonwucj@gmail.com>
4270 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4271 * config/nds32/nds32-multiple.md (setmemsi): Define.
4272 * config/nds32/nds32-memory-manipulation.c
4273 (nds32_gen_dup_4_byte_to_word_value): New.
4274 (emit_setmem_word_loop): New.
4275 (emit_setmem_byte_loop): New.
4276 (nds32_expand_setmem_loop): New.
4277 (nds32_expand_setmem_loop_v3m): New.
4278 (nds32_expand_setmem_unroll): New.
4279 (nds32_expand_setmem): New.
4281 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4282 Chung-Ju Wu <jasonwucj@gmail.com>
4284 * config/nds32/nds32-memory-manipulation.c
4285 (nds32_emit_load_store): New.
4286 (nds32_emit_post_inc_load_store): New.
4287 (nds32_emit_mem_move): New.
4288 (nds32_emit_mem_move_block): New.
4289 (nds32_expand_movmemsi_loop_unknown_size): New.
4290 (nds32_expand_movmemsi_loop_known_size): New.
4291 (nds32_expand_movmemsi_loop): New.
4292 (nds32_expand_movmemsi_unroll): New.
4293 (nds32_expand_movmemqi): Rename ...
4294 (nds32_expand_movmemsi): ... to this.
4295 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4296 (movmemsi): ... to this.
4297 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4298 (nds32_expand_movmemsi): ... to this.
4300 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4301 Monk Chiang <sh.chiang04@gmail.com>
4302 Chung-Ju Wu <jasonwucj@gmail.com>
4304 * config/nds32/nds32-protos.h
4305 (nds32_expand_load_multiple): New arguments.
4306 (nds32_expand_store_multiple): Ditto.
4307 (nds32_valid_multiple_load_store): Rename ...
4308 (nds32_valid_multiple_load_store_p): ... to this.
4309 * config/nds32/nds32-memory-manipulation.c
4310 (nds32_expand_load_multiple): Refine implementation.
4311 (nds32_expand_store_multiple): Ditto.
4312 * config/nds32/nds32-multiple.md
4313 (load_multiple): Update nds32_expand_load_multiple interface.
4314 (store_multiple): Update nds32_expand_store_multiple interface.
4315 * config/nds32/nds32-predicates.c
4316 (nds32_valid_multiple_load_store): Rename ...
4317 (nds32_valid_multiple_load_store_p): ... to this and refine
4319 * config/nds32/predicates.md
4320 (nds32_load_multiple_and_update_address_operation): New predicate.
4321 (nds32_store_multiple_and_update_address_operation): New predicate.
4323 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4324 Chung-Ju Wu <jasonwucj@gmail.com>
4326 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4327 (combo): New attribute.
4328 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4330 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4332 * config/nds32/nds32.opt: Change -mcmodel= default value.
4334 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4335 Monk Chiang <sh.chiang04@gmail.com>
4336 Chung-Ju Wu <jasonwucj@gmail.com>
4338 * config/nds32/constants.md (unspec_element): New enum.
4339 * config/nds32/constraints.md (Umw): New constraint.
4340 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4341 * config/nds32/nds32-intrinsic.md: Likewise.
4342 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4343 (nds32_valid_smw_lwm_base_p): New.
4344 (nds32_output_smw_single_word): New.
4345 (nds32_output_lmw_single_word): New.
4346 (nds32_expand_unaligned_load): New.
4347 (nds32_expand_unaligned_store): New.
4348 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4349 (nds32_output_smw_single_word): Declare.
4350 (nds32_output_lmw_single_word): Declare.
4351 (nds32_expand_unaligned_load): Declare.
4352 (nds32_expand_unaligned_store): Declare.
4353 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4354 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4355 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4356 NDS32_BUILTIN_UASTORE_DW.
4357 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4360 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4361 Kito Cheng <kito.cheng@gmail.com>
4362 Chung-Ju Wu <jasonwucj@gmail.com>
4364 * config/nds32/nds32-intrinsic.c
4365 (nds32_expand_builtin_null_ftype_reg): Delete.
4366 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4367 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4368 (nds32_read_argument): New.
4369 (nds32_legitimize_target): Ditto.
4370 (nds32_legitimize_argument): Ditto.
4371 (nds32_check_constant_argument): Ditto.
4372 (nds32_expand_unop_builtin): Ditto.
4373 (nds32_expand_unopimm_builtin): Ditto.
4374 (nds32_expand_binop_builtin): Ditto.
4375 (nds32_builtin_decl_impl): Ditto.
4376 (builtin_description): Ditto.
4377 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4378 (nds32_init_builtins_impl): Ditto.
4379 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4380 (nds32_builtin_decl): New.
4381 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4382 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4384 2018-03-02 Jeff Law <law@redhat.com>
4386 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4387 (redundant_insn, fill_simple_delay_slots): Likewise.
4388 (fill_slots_from_thread): Likewise.
4389 * resource.c (mark_referenced_resources): Likewise.
4390 (mark_set_resources, find_dead_or_set_registers): Likewise.
4392 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4394 * substring-locations.h (format_warning_va): Formatting fix for
4396 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4398 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4399 * substring-locations.c: Include intl.h.
4400 (format_warning_va): Turned into small wrapper around
4401 format_warning_n_va, renamed to ...
4402 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4403 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4405 (format_warning_at_substring_n): New function.
4406 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4407 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4408 format_warning_at_substring with just a shorter name instead of
4409 const function pointer.
4410 (fmtwarn_n): New function.
4411 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4412 appropriate, get rid of all the fmtstr temporaries, move conditionals
4413 with G_() wrapped string literals directly into fmtwarn arguments,
4414 cast dir.len to (int), formatting fixes.
4416 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4418 * doc/invoke.texi: Remove "Cilk Plus" references.
4420 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4421 Richard Biener <rguenther@suse.de>
4424 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4425 for error or warning attributes if CALL_FROM_THUNK_P is set.
4428 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4431 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4432 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4435 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4436 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4438 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4439 instead of -1U in last predictors element's probability member.
4441 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4444 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4445 arguments if they are comparable.
4447 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4449 PR tree-optimization/84634
4450 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4451 masks and masked_loop_p with a single loop_masks, making sure it's
4452 null for bb vectorization.
4454 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4456 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4457 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4458 loop->force_vectorize to check whether there is no alias.
4460 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4463 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4465 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4467 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4468 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4469 instead of a loop around prev_real_insn.
4470 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4474 * config/i386/i386.c (ix86_print_operand): Use conditional
4475 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4478 2018-03-02 Richard Biener <rguenther@suse.de>
4480 PR tree-optimization/84427
4481 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4482 (bitmap_set_subtract_values): Rewrite to handle multiple
4485 (prune_clobbered_mems): Likewise.
4486 (phi_translate): Take edge instead of pred/phiblock.
4487 (phi_translate_1): Likewise.
4488 (phi_translate_set): Likewise. Insert all translated
4489 exprs for a value into the set, keeping possibly multiple
4490 expressions per value.
4491 (compute_antic_aux): Adjust for phi_translate changes.
4492 When intersecting union the expressions and prune those
4493 not in the final value set, keeping possibly multiple
4494 expressions per value. Do not use value-insertion
4495 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4496 all expressions. Add verification that the value-sets
4497 only shrink during iteration.
4498 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4499 (do_pre_regular_insertion): Likewise.
4500 (do_pre_partial_partial_insertion): Likewise.
4502 2018-03-02 Richard Biener <rguenther@suse.de>
4505 * config/darwin.c (saved_debug_info_level): New static global.
4506 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4507 (darwin_asm_lto_end): Restore debug info generation settings.
4509 2018-03-01 Martin Liska <mliska@suse.cz>
4512 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4515 2018-03-01 Richard Biener <rguenther@suse.de>
4518 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4519 type annotation with LTO when debug was disabled at compile-time.
4521 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4523 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4525 (mips_final_postscan_insn): Likewise.
4527 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4529 PR rtl-optimization/84528
4530 * alias.c (init_alias_target): Add commentary.
4531 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4532 a unique base value if the frame pointer is not eliminated
4533 to the stack pointer.
4535 2018-03-01 Tom de Vries <tom@codesourcery.com>
4537 PR rtl-optimization/83327
4538 * lra-int.h (hard_regs_spilled_into): Declare.
4539 * lra.c (hard_regs_spilled_into): Define.
4540 (init_reg_info): Init hard_regs_spilled_into.
4541 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4542 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4543 (process_bb_lives): Handle hard_regs_spilled_into.
4544 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4545 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4547 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4549 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4550 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4551 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4552 * config/rs6000/aix72.h: New file.
4554 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4556 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4557 instead of warning_at with conditional singular and plural messages
4561 * stor-layout.c (update_alignment_for_field): For
4562 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4563 && !DECL_PACKED (field), do the alignment update, just use
4564 only desired_align instead of MAX (type_align, desired_align)
4566 (place_field): Don't do known_align < desired_align handling
4567 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4568 is non-NULL, instead do it after rli->prev_field handling and
4569 only if not within a bitfield word. For DECL_PACKED (field)
4570 use type_align of BITS_PER_UNIT.
4572 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4574 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4575 superfluous parentheses and trailing spaces.
4577 2018-02-28 Richard Biener <rguenther@suse.de>
4579 PR tree-optimization/84584
4580 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4581 SCoPs with fake exit edge.
4583 2018-02-28 Martin Liska <mliska@suse.cz>
4586 * timevar.c (timer::print): Fix format to properly print 100%
4589 2018-02-28 Richard Biener <rguenther@suse.de>
4592 * genmatch.c (capture_info::walk_match): Do not mark
4593 captured expressions without operands as expr_p given
4594 they act more like predicates and should be subject to
4595 "lost tail" side-effect preserving.
4597 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4599 PR rtl-optimization/81611
4600 * auto-inc-dec.c (attempt_change): Move dead note from
4601 mem_insn if it's the next use of regno
4602 (find_address): Take address use of reg holding
4603 non-incremented value. Add parm to limit search to the named
4605 (merge_in_block): Attempt to use a mem insn that is the next
4606 use of the original regno.
4608 2018-02-27 Martin Sebor <msebor@redhat.com>
4611 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4612 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4614 2018-02-27 Martin Sebor <msebor@redhat.com>
4616 PR translation/84207
4617 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4618 n argument to unsigned HOST_WIDE_INT.
4619 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4620 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4621 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4622 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4624 2018-02-27 Richard Biener <rguenther@suse.de>
4626 PR tree-optimization/84512
4627 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4628 Do not use the estimate returned from record_stmt_cost for
4629 the scalar iteration cost but sum properly using add_stmt_cost.
4631 2018-02-27 Richard Biener <rguenther@suse.de>
4633 PR tree-optimization/84466
4634 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4635 Adjust last change to less strictly validate use operands.
4637 2018-02-27 Martin Liska <mliska@suse.cz>
4639 PR gcov-profile/84548
4640 * gcov.c (process_file): Allow partial overlap and consider it
4641 also as group functions.
4642 (output_lines): Properly calculate range of lines for a group.
4644 2018-02-27 Martin Liska <mliska@suse.cz>
4646 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4647 'ggc' suffixes. Change first column width.
4648 (timer::print): Fix formatting of the column.
4650 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4652 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4653 preserve inline entry blocks for the sake of debug inline
4654 entry point markers alone.
4655 (remove_unused_locals): Suggest in comments a better place to
4656 force the preservation of inline entry blocks that are
4657 otherwise unused, but do not preserve them.
4659 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4661 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4663 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4666 * config/i386/constraints.md (Bs): Replace
4667 ix86_indirect_branch_register with
4668 TARGET_INDIRECT_BRANCH_REGISTER.
4670 * config/i386/i386.md (indirect_jump): Likewise.
4671 (tablejump): Likewise.
4672 (*sibcall_memory): Likewise.
4673 (*sibcall_value_memory): Likewise.
4674 Peepholes of indirect call and jump via memory: Likewise.
4675 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4676 (*sibcall_value_GOT_32): Likewise.
4677 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4678 (GOT_memory_operand): Likewise.
4679 (call_insn_operand): Likewise.
4680 (sibcall_insn_operand): Likewise.
4681 (GOT32_symbol_operand): Likewise.
4682 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4684 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4686 PR rtl-optimization/83496
4687 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4688 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4689 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4690 redundant insn, if any.
4691 (relax_delay_slots): Likewise.
4692 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4694 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4696 PR tree-optimization/83965
4697 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4698 that grouped statements are part of a reduction chain. Return
4699 true if the statement is not marked as a reduction itself but
4701 (vect_recog_dot_prod_pattern): Don't check whether the statement
4702 is part of a group here.
4703 (vect_recog_sad_pattern): Likewise.
4704 (vect_recog_widen_sum_pattern): Likewise.
4706 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4709 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4711 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4714 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4716 (ix86_output_indirect_function_return): New prototype.
4717 (ix86_split_simple_return_pop_internal): Likewise.
4718 * config/i386/i386.c (indirect_return_via_cx): New.
4719 (indirect_return_via_cx_bnd): Likewise.
4720 (indirect_thunk_name): Handle return va CX_REG.
4721 (output_indirect_thunk_function): Create alias for
4722 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4723 (ix86_output_indirect_jmp): Remove the bool argument.
4724 (ix86_output_indirect_function_return): New function.
4725 (ix86_split_simple_return_pop_internal): Likewise.
4726 * config/i386/i386.md (*indirect_jump): Don't pass false
4727 to ix86_output_indirect_jmp.
4728 (*tablejump_1): Likewise.
4729 (simple_return_pop_internal): Change it to define_insn_and_split.
4730 Call ix86_split_simple_return_pop_internal to split it for
4732 (simple_return_indirect_internal): Call
4733 ix86_output_indirect_function_return instead of
4734 ix86_output_indirect_jmp.
4736 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4739 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4740 memset and value initialization afterwards.
4742 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4744 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4746 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4749 * common/config/aarch64/aarch64-common.c
4750 (aarch_option_optimization_table[]): Switch
4751 off fomit-frame-pointer
4753 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4754 Chung-Ju Wu <jasonwucj@gmail.com>
4756 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4758 (store_multiple): Ditto.
4760 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4762 * config.gcc: Add --with-cpu support for nds32 target.
4763 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4764 * config/nds32/nds32.opt: Add -mcpu= option.
4766 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4768 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4769 isel=yes): Warn for these deprecated options.
4771 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4773 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4774 ISA_2_5_MASKS_EMBEDDED.
4776 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4778 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4779 p->max as pointers rather than using iterative_hash_expr.
4781 2018-02-23 Carl Love <cel@us.ibm.com>
4783 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4784 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4786 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4787 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4788 P8V_BUILTIN_VEC_VUNSIGNED2.
4790 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4793 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4794 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4796 * lra-constraints.c (curr_insn_transform): Set up
4797 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4799 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4800 * lra-eliminations.c (spill_pseudos): Ditto.
4801 (process_insn_for_elimination): Ditto.
4802 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4803 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4804 LRA_NON_CLOBBERED_ALT.
4806 2018-02-22 Martin Sebor <msebor@redhat.com>
4808 PR tree-optimization/84480
4809 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4810 to maybe_diag_stxncpy_trunc. Call it.
4811 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4812 from gimple_fold_builtin_strcpy. Print inlining stack.
4813 (handle_builtin_stxncpy): Print inlining stack.
4814 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4816 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4819 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4820 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4821 and -fcheck-pointer-bounds are used together.
4822 (indirect_thunk_prefix): New enum.
4823 (indirect_thunk_need_prefix): New function.
4824 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4825 "_nt" instead of "_bnd" for NOTRACK prefix.
4826 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4827 (output_indirect_thunk_function): Likewise.
4829 (ix86_code_end): Update output_indirect_thunk_function calls.
4830 (ix86_output_indirect_branch_via_reg): Replace
4831 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4832 (ix86_output_indirect_branch_via_push): Likewise.
4833 (ix86_output_function_return): Likewise.
4834 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4835 incompatible with -fcf-protection=branch and
4836 -fcheck-pointer-bounds.
4838 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4841 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4842 Change gcc_assert call to output_operand_lossage.
4844 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4846 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4848 2018-02-22 DJ Delorie <dj@redhat.com>
4849 Sebastian Perta <sebastian.perta@renesas.com>
4850 Oleg Endo <olegendo@gcc.gnu.org>
4852 * config/rx/rx.c (rx_rtx_costs): New function.
4853 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4855 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4857 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4859 2018-02-22 Martin Liska <mliska@suse.cz>
4862 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4863 Add "native" as a possible value.
4865 2018-02-22 Martin Liska <mliska@suse.cz>
4868 * config/i386/i386.c (ix86_option_override_internal):
4869 Add "native" as a possible value for -march and -mtune.
4871 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4874 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4875 to all type variants.
4877 PR tree-optimization/84503
4878 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4879 width as info->bitpos + info->bitsize - start.
4880 (merged_store_group::merge_overlapping): Simplify width computation.
4881 (check_no_overlap): New function.
4882 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4883 start + width and last_order of the group, fail if check_no_overlap
4885 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4886 to group if check_no_overlap fails.
4888 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4890 * config/rs6000/altivec.md: Delete contraint arguments to
4891 define_expand, define_split, and define_peephole2, and in
4892 define_insn_and_split if always unused.
4893 * config/rs6000/darwin.md: Ditto.
4894 * config/rs6000/dfp.md: Ditto.
4895 * config/rs6000/rs6000.md: Ditto.
4896 * config/rs6000/sync.md: Ditto.
4897 * config/rs6000/vector.md: Ditto.
4898 * config/rs6000/vsx.md: Ditto.
4900 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4902 * config/rs6000/altivec.md: Write output control strings as braced
4903 blocks instead of double-quoted strings.
4904 * config/rs6000/darwin.md: Ditto.
4905 * config/rs6000/rs6000.md: Ditto.
4906 * config/rs6000/vector.md: Ditto.
4907 * config/rs6000/vsx.md: Ditto.
4909 2018-02-21 Jason Merrill <jason@redhat.com>
4911 PR c++/84314 - ICE with templates and fastcall attribute.
4912 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4914 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4916 * ipa-cp.c (determine_versionability): Fix comment typos.
4918 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4921 * ipa-cp.c (determine_versionability): Do not version functions caling
4924 2018-02-21 Martin Liska <mliska@suse.cz>
4927 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4928 Add "native" as a possible value.
4929 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4930 the macro when native cpu detection is available.
4932 2018-02-21 Martin Liska <mliska@suse.cz>
4935 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4936 Add "native" as a possible value.
4937 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4938 when native cpu detection is available.
4940 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4941 Martin Sebor <msebor@redhat.com>
4943 PR tree-optimization/84478
4944 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4946 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4947 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4948 support which is conservatively correct, for 2 only stay conservative
4949 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4950 argument to the 2 argument get_range_strlen, adjust 6 arg
4951 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4953 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4954 (gimple_fold_builtin_strlen): Pass true as last argument to
4957 2018-02-20 Martin Sebor <msebor@redhat.com>
4960 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4961 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4962 (builtin_memref::builtin_memref): Factor out parts into
4963 set_base_and_offset and call it.
4965 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4968 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4969 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4970 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4971 search at the associated MODE_INT.
4973 2018-02-20 Jeff Law <law@redhat.com>
4976 PR tree-optimization/81592
4978 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4979 for range data rather than using global data.
4980 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4981 range data rather than using global data.
4982 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4983 pass it to children as needed.
4984 (struct directive::fmtresult): Similarly.
4985 (struct directive::set_width): Similarly.
4986 (struct directive::set_precision): Similarly.
4987 (format_integer, format_directive, parse_directive): Similarly.
4988 (format_none): Accept unnamed vr_values parameter.
4989 (format_percent, format_floating, format_character): Similarly.
4990 (format_string, format_plain): Similarly.
4991 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4992 the EVRP range analyzer for range data rather than using global data.
4993 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4994 gimple-ssa-evrp-analyze.h
4995 (class sprintf_dom_walker): Add after_dom_children member function.
4996 Add evrp_range_analyzer member.
4997 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4998 range analyzer as needed.
4999 (sprintf_dom_walker::after_dom_children): New member function.
5000 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5002 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5003 (evrp_range_analyzer::pop_to_marker): Likewise.
5005 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5007 PR tree-optimization/84419
5008 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5009 with the required type if its current type is compatible but
5012 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5015 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5016 after vectorization.
5018 2018-02-20 Martin Liska <mliska@suse.cz>
5021 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5022 possible values if we don't have a hint.
5024 2018-02-20 Martin Liska <mliska@suse.cz>
5028 * final.c (shorten_branches): Build align_tab array with one
5030 * opts.c (finish_options): Add alignment option limit check.
5031 (MAX_CODE_ALIGN): Likewise.
5032 (MAX_CODE_ALIGN_VALUE): Likewise.
5033 * doc/invoke.texi: Document maximum allowed option value for
5034 all -falign-* options.
5036 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5039 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5040 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5041 * var-tracking.c (emit_note_insn_var_location): Remove all references
5042 to NOTE_INSN_CALL_ARG_LOCATION.
5043 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5044 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5045 Use copy_rtx_if_shared.
5046 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5047 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5048 (dwarf2out_var_location): Remove handling of
5049 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5051 * final.c (final_scan_insn): Remove all references to
5052 NOTE_INSN_CALL_ARG_LOCATION.
5053 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5054 before dumping final insns.
5055 * except.c (emit_note_eh_region_end): Remove all references to
5056 NOTE_INSN_CALL_ARG_LOCATION.
5057 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5058 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5059 * config/arc/arc.c (hwloop_optimize): Likewise.
5060 * config/arm/arm.c (create_fix_barrier): Likewise.
5061 * config/s390/s390.c (s390_chunkify_start): Likewise.
5062 * config/sh/sh.c (find_barrier): Likewise.
5063 * config/i386/i386.c (rest_of_insert_endbranch,
5064 ix86_seh_fixup_eh_fallthru): Likewise.
5065 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5066 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5067 * config/frv/frv.c (frv_function_prologue): Likewise.
5068 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5070 (note_outside_basic_block_p): Remove all references to
5071 NOTE_INSN_CALL_ARG_LOCATION.
5072 * gengtype.c (adjust_field_rtx_def): Likewise.
5073 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5075 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5076 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5079 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5082 PR tree-optimization/84452
5083 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5084 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5087 2018-02-19 Martin Liska <mliska@suse.cz>
5090 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5092 2018-02-19 Martin Liska <mliska@suse.cz>
5093 Richard Sandiford <richard.sandiford@linaro.org>
5095 PR tree-optimization/82491
5096 * gimple-fold.c (get_base_constructor): Make earlier bail out
5099 2018-02-19 Carl Love <cel@us.ibm.com>
5101 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5102 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5104 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5105 P8V_BUILTIN_VEC_NEG.
5107 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5109 * config/rl78/rl78.md (movdf): New define expand.
5111 2018-02-19 Martin Liska <mliska@suse.cz>
5114 * doc/invoke.texi: Fix typo.
5115 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5117 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5119 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5120 handle rs6000_single_float and rs6000_double_float specially for
5123 2018-02-16 Jeff Law <law@redhat.com>
5125 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5128 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5131 * ipa-inline.c (inline_small_functions): Fix a typo.
5133 2018-02-16 Nathan Sidwell <nathan@acm.org>
5135 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5137 2018-02-16 Carl Love <cel@us.ibm.com>
5139 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5140 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5141 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5142 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5143 expansion to P8V_BUILTIN_VEC_FLOAT2.
5145 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5147 PR rtl-optimization/70023
5148 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5149 src_regno into account.
5151 2018-02-16 Carl Love <cel@us.ibm.com>
5153 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5154 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5155 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5156 * config/rs6000/rs6000.c: Remove case statements for
5157 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5158 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5159 and P9V_BUILTIN_VEC_VINSERT4B.
5160 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5161 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5162 * config/rs6000/vsx.md:
5163 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5166 2018-02-16 Carl Love <cel@us.ibm.com>
5168 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5170 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5172 * config/rs6000/rs6000-c.c: Add the definitions for
5173 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5174 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5175 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5176 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5177 definition for insert4b and define insn *insert3b_internal.
5178 * doc/extend.texi: Add documentation for vec_extract4b.
5180 2018-02-16 Nathan Sidwell <nathan@acm.org>
5182 * doc/extend.texi (Backwards Compatibility): Mention friend
5183 injection. Note for-scope is deprecated.
5184 * doc/invoke.texi (-ffriend-injection): Deprecate.
5186 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5188 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5189 that moved to I2, also allow destinations that are a paradoxical
5190 subreg (instead of a normal reg).
5192 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5195 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5198 2018-02-16 Richard Biener <rguenther@suse.de>
5200 PR tree-optimization/84037
5201 PR tree-optimization/84016
5203 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5204 Adjust vec_construct for the fact we need additional higher latency
5205 128bit inserts for AVX256 and AVX512 vector builds.
5206 (ix86_add_stmt_cost): Scale vector construction cost for
5209 2018-02-16 Richard Biener <rguenther@suse.de>
5211 PR tree-optimization/84417
5212 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5213 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5214 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5216 2018-02-16 Martin Liska <mliska@suse.cz>
5219 * internal-fn.def (ASAN_CHECK): Set proper flags.
5220 (ASAN_MARK): Likewise.
5222 2018-02-16 Julia Koval <julia.koval@intel.com>
5224 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5225 from PTA_CANNONLAKE.
5227 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5230 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5231 Use ++iter rather than iter++ for std::list iterators.
5232 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5233 defer deleting them until all nodes in the forest are processed. Do
5234 free even leaf nodes. Change to_process into auto_vec.
5237 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5238 * vec.h (vec_default_construct): Use memset instead of placement new
5239 if BROKEN_VALUE_INITIALIZATION is defined.
5240 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5241 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5244 PR rtl-optimization/83723
5245 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5246 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5247 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5249 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5251 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5253 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5255 PR rtl-optimization/81443
5256 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5257 from inner REGs to paradoxical SUBREGs.
5259 2018-02-16 Richard Biener <rguenther@suse.de>
5261 PR tree-optimization/84399
5262 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5263 For operands we can analyze at their definition make sure we can
5264 analyze them at each use as well.
5266 2018-02-16 Richard Biener <rguenther@suse.de>
5268 PR tree-optimization/84190
5269 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5270 volatile accesses if the decl isn't volatile.
5272 2018-02-15 Jason Merrill <jason@redhat.com>
5274 PR c++/84314 - ICE with templates and fastcall attribute.
5275 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5276 TYPE_CANONICAL on an existing type.
5278 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5280 PR tree-optimization/84383
5281 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5282 dstoff nor call operand_equal_p if dstbase is NULL.
5284 PR tree-optimization/84334
5285 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5286 also a CONSTANT_CLASS_P, punt.
5288 2018-02-14 Jim Wilson <jimw@sifive.com>
5290 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5291 first SMALL_OPERAND check. New local min_second_step. Move assert
5292 to where locals are set. Add TARGET_RVC support.
5293 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5295 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5297 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5299 2018-02-14 Martin Sebor <msebor@redhat.com>
5301 PR tree-optimization/83698
5302 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5303 arrays constrain the offset range to their bounds.
5304 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5305 (builtin_access::overlap): Avoid setting the size of overlap if it's
5307 (maybe_diag_overlap): Also consider arrays when deciding what values
5308 of offsets to include in diagnostics.
5310 2018-02-14 Martin Sebor <msebor@redhat.com>
5313 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5314 that correspond to the kind of a declaration.
5316 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5319 * config/pa/pa.md: Load address of PIC label using the linkage table
5320 if the label is nonlocal.
5322 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5324 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5325 warning message if user requests -maltivec=be.
5326 * doc/invoke.texi: Document deprecation of -maltivec=be.
5328 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5331 * config/rs6000/rs6000-c.c: Update definitions for
5332 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5333 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5335 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5338 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5339 add _get_ssp intrinsics. Remove argument from
5340 __builtin_ia32_rdssp[d|q].
5341 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5342 * config/i386/i386-builtin.def: Remove argument from
5343 __builtin_ia32_rdssp[d|q].
5344 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5345 ix86_expand_special_args_builtin for _rdssp[d|q].
5346 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5347 Clear register before usage.
5348 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5349 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5351 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5353 PR tree-optimization/84357
5354 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5355 operand 1 of an ARRAY_REF too.
5357 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5360 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5361 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5363 (set_of_reg): New struct.
5364 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5365 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5366 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5368 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5369 Split into bitclr, bitset, bitinvert patterns if appropriate.
5370 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5371 use rx_fuse_in_memory_bitop.
5372 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5373 to named insn, correct maximum insn length.
5375 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5378 * machmode.def: Define a complex mode for PARTIAL_INT.
5379 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5381 * doc/rtl.texi: Document CSPImode.
5382 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5384 (msp430_hard_regno_nregs_with_padding): Likewise.
5386 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5389 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5391 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5393 PR rtl-optimization/84169
5394 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5395 we generated a parallel as new i3 and we split that to new i2 and i3
5396 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5397 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5398 those to i2, not i1. Partially rewrite this scan code.
5400 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5403 * stor-layout.c (place_field): For variable length fields, adjust
5404 offset_align afterwards not just based on the field's alignment,
5405 but also on the size.
5408 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5409 of exps and logs in the use_exp2 case.
5411 2018-02-13 Jeff Law <law@redhat.com>
5413 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5416 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5419 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5423 * common.opt (gas-loc-support, gas-locview-support): New.
5424 (ginline-points, ginternal-reset-location-views): New.
5425 * doc/invoke.texi: Document them. Use @itemx where intended.
5426 (gvariable-location-views): Adjust.
5427 * target.def (reset_location_view): New.
5428 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5429 (TARGET_RESET_LOCATION_VIEW): New.
5430 * doc/tm.texi: Rebuilt.
5431 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5432 (dwarf2out_default_as_locview_support): New.
5433 (output_asm_line_debug_info): Use option variables.
5434 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5435 (output_loc_list): Likewise.
5436 (add_high_low_attributes): Check option variables.
5437 Don't output entry view attribute in strict mode.
5438 (gen_inlined_subroutine_die): Check option variables.
5439 (dwarf2out_inline_entry): Likewise.
5440 (init_sections_and_labels): Likewise.
5441 (dwarf2out_early_finish): Likewise.
5442 (maybe_reset_location_view): New, from...
5443 (dwarf2out_var_location): ... here. Call it.
5444 * debug.h (dwarf2out_default_as_loc_support): Declare.
5445 (dwarf2out_default_as_locview_support): Declare.
5446 * hooks.c (hook_int_rtx_insn_0): New.
5447 * hooks.h (hook_int_rtx_insn_0): Declare.
5448 * toplev.c (process_options): Take -gas-loc-support and
5449 -gas-locview-support from dwarf2out. Enable
5450 -gvariable-location-views by default only with locview
5451 assembler support. Enable -ginternal-reset-location-views by
5452 default only if the target defines the corresponding hook.
5453 Enable -ginline-points by default if location views are
5454 enabled; force it disabled if statement frontiers are
5456 * tree-inline.c (expand_call_inline): Check option variables.
5457 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5459 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5461 PR tree-optimization/84321
5462 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5463 handling. Also check whether the anti-range contains any values
5464 that satisfy the mask; switch to a VR_RANGE if not.
5466 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5469 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5471 2018-02-13 Martin Jambor <mjambor@suse.cz>
5474 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5475 of call statements, also set location of a load to a temporary.
5477 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5479 * config/rl78/rl78.c (add_vector_labels): New function.
5480 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5481 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5482 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5483 which checks that no arguments are passed.
5484 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5485 * doc/extend.texi: Documentation for the new attribute.
5487 2018-02-13 Andreas Schwab <schwab@suse.de>
5489 * config/riscv/linux.h (CPP_SPEC): Define.
5491 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5494 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5495 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5496 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5497 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5498 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5499 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5500 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5501 temporarily for AES and PCLMUL builtins.
5503 PR tree-optimization/84339
5504 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5505 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5509 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5510 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5511 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5513 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5515 * omp-simd-clone.h: New file.
5516 * omp-simd-clone.c: Include omp-simd-clone.h.
5517 (expand_simd_clones): No longer static.
5518 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5519 cgraph.h and omp-simd-clone.h.
5520 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5521 (vect_recog_widen_shift_pattern): Formatting fix.
5522 (vect_pattern_recog_1): Don't check optab for calls.
5525 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5526 operands[2] into a REG before using gen_lowpart on it.
5528 2018-02-12 Jeff Law <law@redhat.com>
5531 * config/sh/sh.c (find_barrier): Consider a sibling call
5534 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5535 successfully back substituting a reg.
5537 2018-02-12 Richard Biener <rguenther@suse.de>
5539 PR tree-optimization/84037
5540 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5541 parameter, move visited init to caller.
5542 (vect_slp_analyze_operations): Separate cost from validity
5543 check, initialize visited once for all instances.
5544 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5546 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5548 (vect_model_promotion_demotion_cost): Likewise.
5549 (vectorizable_bswap): Guard cost modeling with !slp_node
5550 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5552 (vectorizable_call): Likewise.
5553 (vectorizable_conversion): Likewise.
5554 (vectorizable_assignment): Likewise.
5555 (vectorizable_shift): Likewise.
5556 (vectorizable_operation): Likewise.
5557 (vectorizable_store): Likewise.
5558 (vectorizable_load): Likewise.
5559 (vectorizable_condition): Likewise.
5560 (vectorizable_comparison): Likewise.
5562 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5565 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5566 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5567 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5569 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5572 * params.def (inline-min-speedup): Increase from 8 to 15.
5573 (max-inline-insns-auto): Decrease from 40 to 30.
5574 * ipa-split.c (consider_split): Add some buffer for function to
5575 be considered inlining candidate.
5576 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5579 2018-02-12 Richard Biener <rguenther@suse.de>
5581 PR tree-optimization/84037
5582 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5583 matched stmts if we cannot swap the non-matched ones.
5585 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5587 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5588 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5589 _mm_maskz_scalef_round_ss): New intrinsics.
5590 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5591 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5592 __builtin_ia32_scalefss_round): Remove.
5593 (__builtin_ia32_scalefsd_mask_round,
5594 __builtin_ia32_scalefss_mask_round): New intrinsics.
5595 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5596 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5597 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5598 "<round_constraint>")): Changed to ...
5599 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5600 "<round_scalar_constraint>")): ... this.
5601 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5602 %0, %1, %2<round_op3>}"): Changed to ...
5603 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5604 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5605 %2<round_scalar_mask_op3>}"): ... this.
5606 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5608 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5610 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5611 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5612 (_mm_maskz_sqrt_round_ss): New intrinsics.
5613 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5614 (__builtin_ia32_sqrtsd_mask_round)
5615 (__builtin_ia32_sqrtss_mask_round): New builtins.
5616 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5617 (__builtin_ia32_sqrtss_round): Remove.
5618 (__builtin_ia32_sqrtsd_mask_round)
5619 (__builtin_ia32_sqrtss_mask_round): New builtins.
5620 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5621 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5622 ((match_operand:VF_128 1 "vector_operand"
5623 "xBm,<round_constraint>")): Changed to ...
5624 ((match_operand:VF_128 1 "vector_operand"
5625 "xBm,<round_scalar_constraint>")): ... this.
5626 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5627 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5628 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5629 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5630 %<iptr>1<round_scalar_mask_op3>}): ... this.
5631 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5632 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5634 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5637 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5638 Cast vec_cmpeq result to correct type.
5639 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5640 Cast vec_cmpgt result to correct type.
5642 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5644 * final.c (final_scan_insn_1): Renamed from...
5645 (final_scan_insn): ... this. New wrapper, to recover
5646 seen from the outermost call in recursive ones.
5647 * config/sparc/sparc.c (output_return): Drop seen from call.
5648 (output_sibcall): Likewise.
5649 * config/visium/visium.c (output_branch): Likewise.
5651 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5653 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5656 2018-02-10 Alan Modra <amodra@gmail.com>
5659 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5660 Specify LR as an input.
5662 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5665 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5666 remove_member_access_dummy_vars): New functions.
5667 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5668 lower_omp_1, execute_lower_omp): Use them.
5670 PR rtl-optimization/84308
5671 * shrink-wrap.c (spread_components): Release todo vector.
5673 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5675 PR rtl-optimization/57193
5676 * ira-color.c (struct allocno_color_data): Add member
5677 conflict_allocno_hard_prefs.
5678 (update_conflict_allocno_hard_prefs): New.
5679 (bucket_allocno_compare_func): Add a preference based on
5680 conflict_allocno_hard_prefs.
5681 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5682 (color_allocnos): Remove a dead code. Initiate
5683 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5685 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5688 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5689 constraint from =wa to wa. Avoid a subreg on the output operand,
5690 instead use a pseudo and subreg it in a move.
5691 (p9_xxbrd_<mode>): Changed to ...
5692 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5693 (p9_xxbrd_v2df): New expander.
5694 (p9_xxbrw_<mode>): Changed to ...
5695 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5696 (p9_xxbrw_v4sf): New expander.
5698 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5700 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5702 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5705 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5706 multiply in 32-bit mode.
5707 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5708 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5711 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5713 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5714 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5715 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5716 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5718 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5721 * dwarf2out.c (is_trivial_indirect_ref): New function.
5722 (dwarf2out_late_global_decl): Do not generate a location
5723 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5724 and that are not defined in the current unit.
5726 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5728 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5729 instead of a libcall for UNORDERED.
5731 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5734 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5735 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5737 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5740 * config/s390/s390.c (s390_set_current_function): Invoke
5741 s390_indirect_branch_settings also if fndecl didn't change.
5743 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5745 * config/rs6000/rs6000.md (blockage): Set length to zero.
5747 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5749 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5751 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5754 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5755 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5759 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5760 PARALLEL incoming that failed vt_get_decl_and_offset check.
5763 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5764 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5766 (get_variable_section): For decls in named .bss* sections pass true as
5767 second argument to bss_initializer_p.
5769 2018-02-09 Marek Polacek <polacek@redhat.com>
5770 Jakub Jelinek <jakub@redhat.com>
5773 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5774 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5775 Sync some changes from cxx_fold_indirect_ref.
5777 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5779 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5781 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5782 (BLOCK_INLINE_ENTRY_LABEL): New.
5783 (dwarf2out_var_location): Disregard inline entry markers.
5784 (inline_entry_data): New struct.
5785 (inline_entry_data_hasher): New hashtable type.
5786 (inline_entry_data_hasher::hash): New.
5787 (inline_entry_data_hasher::equal): New.
5788 (inline_entry_data_table): New variable.
5789 (add_high_low_attributes): Add DW_AT_entry_pc and
5790 DW_AT_GNU_entry_view attributes if a pending entry is found
5791 in inline_entry_data_table. Add old entry_pc attribute only
5792 if debug nonbinding markers are disabled.
5793 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5794 markers are enabled.
5795 (block_within_block_p, dwarf2out_inline_entry): New.
5796 (dwarf2out_finish): Check that no entries remained in
5797 inline_entry_data_table.
5798 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5799 (final_scan_insn, notice_source_line): Likewise.
5800 (rest_of_clean_state): Skip inline entry markers.
5801 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5803 * gimple.c (gimple_build_debug_inline_entry): New.
5804 * gimple.h (enum gimple_debug_subcode): Add
5805 GIMPLE_DEBUG_INLINE_ENTRY.
5806 (gimple_build_debug_inline_entry): Declare.
5807 (gimple_debug_inline_entry_p): New.
5808 (gimple_debug_nonbind_marker_p): Adjust.
5809 * insn-notes.def (INLINE_ENTRY): New.
5810 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5811 inline entry marker notes.
5812 (print_insn): Likewise.
5813 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5814 (INSN_DEBUG_MARKER_KIND): Likewise.
5815 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5816 * tree-inline.c (expand_call_inline): Build and insert
5817 debug_inline_entry stmt.
5818 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5819 inline entry blocks early, if nonbind markers are enabled.
5820 (dump_scope_block): Dump fragment info.
5821 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5822 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5823 (gimple_build_debug_inline_entry): New.
5824 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5825 Enable/disable inline entry points too.
5826 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5827 (DEBUG_INSN): Describe inline entry markers.
5829 * common.opt (gvariable-location-views): New.
5830 (gvariable-location-views=incompat5): New.
5831 * config.in: Rebuilt.
5832 * configure: Rebuilt.
5833 * configure.ac: Test assembler for view support.
5834 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5835 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5836 * dwarf2out.c (var_loc_view): New typedef.
5837 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5838 (dwarf2out_locviews_in_attribute): New.
5839 (dwarf2out_locviews_in_loclist): New.
5840 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5841 (enum dw_line_info_opcode): Add LI_adv_address.
5842 (struct dw_line_info_table): Add view.
5843 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5844 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5845 (zero_view_p): New variable.
5846 (ZERO_VIEW_P): New macro.
5847 (output_asm_line_debug_info): New.
5848 (struct var_loc_node): Add view.
5849 (add_AT_view_list, AT_loc_list): New.
5850 (add_var_loc_to_decl): Add view param. Test it against last.
5851 (new_loc_list): Add view params. Record them.
5852 (AT_loc_list_ptr): Handle loc and view lists.
5853 (view_list_to_loc_list_val_node): New.
5854 (print_dw_val): Handle dw_val_class_view_list.
5855 (size_of_die): Likewise.
5856 (value_format): Likewise.
5857 (loc_list_has_views): New.
5858 (gen_llsym): Set vl_symbol too.
5859 (maybe_gen_llsym, skip_loc_list_entry): New.
5860 (dwarf2out_maybe_output_loclist_view_pair): New.
5861 (output_loc_list): Output view list or entries too.
5862 (output_view_list_offset): New.
5863 (output_die): Handle dw_val_class_view_list.
5864 (output_dwarf_version): New.
5865 (output_compilation_unit_header): Use it.
5866 (output_skeleton_debug_sections): Likewise.
5867 (output_rnglists, output_line_info): Likewise.
5868 (output_pubnames, output_aranges): Update version comments.
5869 (output_one_line_info_table): Output view numbers in asm comments.
5870 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5871 Call maybe_gen_llsym.
5872 (loc_list_from_tree_1): Adjust.
5873 (add_AT_location_description): Create view list attribute if
5874 needed, check it's absent otherwise.
5875 (convert_cfa_to_fb_loc_list): Adjust.
5876 (maybe_emit_file): Call output_asm_line_debug_info for test.
5877 (dwarf2out_var_location): Reset views as needed. Precompute
5878 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5879 attribute. Set view.
5880 (new_line_info_table): Reset next view.
5881 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5882 (dwarf2out_source_line): Likewise. Output view resets and labels to
5883 the assembler, or select appropriate line info opcodes.
5884 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5885 (optimize_string_length): Catch it. Adjust.
5886 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5887 dw_val_class_view_list, and remove it if no longer needed.
5888 (hash_loc_list): Hash view numbers.
5889 (loc_list_hasher::equal): Compare them.
5890 (optimize_location_lists): Check whether a view list symbol is
5891 needed, and whether the locview attribute is present, and
5892 whether they match. Remove the locview attribute if no longer
5894 (index_location_lists): Call skip_loc_list_entry for test.
5895 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5896 Use output_dwarf_version.
5897 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5898 (struct dw_val_node): Add val_view_list.
5899 * final.c (SEEN_NEXT_VIEW): New.
5900 (set_next_view_needed): New.
5901 (clear_next_view_needed): New.
5902 (maybe_output_next_view): New.
5903 (final_start_function): Rename to...
5904 (final_start_function_1): ... this. Take pointer to FIRST,
5905 add SEEN parameter. Emit param bindings in the initial view.
5906 (final_start_function): Reintroduce SEEN-less interface.
5907 (final): Rename to...
5908 (final_1): ... this. Take SEEN parameter. Output final pending
5909 next view at the end.
5910 (final): Reintroduce seen-less interface.
5911 (final_scan_insn): Output pending next view before switching
5912 sections or ending a block. Mark the next view as needed when
5913 outputting variable locations. Notify debug backend of section
5914 changes, and of location view changes.
5915 (rest_of_handle_final): Adjust.
5916 * toplev.c (process_options): Autodetect value for debug variable
5917 location views option. Warn on incompat5 without -gdwarf-5.
5918 * doc/invoke.texi (gvariable-location-views): New.
5919 (gvariable-location-views=incompat5): New.
5920 (gno-variable-location-views): New.
5922 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5924 PR tree-optimization/84136
5925 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5926 that the result of find_edge is non-NULL.
5928 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5931 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5932 storing integer register in SImode. Fix cost of 256 and 512
5933 byte aligned SSE register store.
5935 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5937 * config/i386/i386.c (ix86_multiplication_cost): Fix
5938 multiplication cost for TARGET_AVX512DQ.
5940 2018-02-08 Marek Polacek <polacek@redhat.com>
5942 PR tree-optimization/84238
5943 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5946 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5948 PR tree-optimization/84265
5949 * tree-vect-stmts.c (vectorizable_store): Don't treat
5950 VMAT_CONTIGUOUS accesses as grouped.
5951 (vectorizable_load): Likewise.
5953 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5955 PR tree-optimization/81635
5956 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5957 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5958 (test_round_for_mask): New functions.
5959 (wide_int_cc_tests): Call test_round_for_mask.
5960 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5961 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5962 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5963 range returned by get_range_info.
5965 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5968 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5969 * symtab.c: Include builtins.h
5970 (symtab_node::output_to_lto_symbol_table_p): Move here
5971 from lto-streamer-out.c:output_symbol_p.
5972 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5973 (output_symbol_p): Move all logic to symtab.c
5974 (produce_symtab): Update.
5976 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5978 * config/s390/s390-opts.h (enum indirect_branch): Define.
5979 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5980 (s390_indirect_branch_via_thunk)
5981 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5982 (enum s390_indirect_branch_type): Define.
5983 * config/s390/s390.c (struct s390_frame_layout, struct
5984 machine_function): Remove.
5985 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5986 (indirect_branch_table_label_no, indirect_branch_table_name):
5988 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5989 (enum s390_indirect_branch_option): Define.
5990 (s390_return_addr_from_memory): New function.
5991 (s390_handle_string_attribute): New function.
5992 (s390_attribute_table): Add new attribute handler.
5993 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5994 (s390_indirect_branch_via_thunk): New function.
5995 (s390_indirect_branch_via_inline_thunk): New function.
5996 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5997 sibling call optimization for non z10 compiles.
5998 (s390_emit_call): Force indirect branch target to be a single
5999 register. Add r1 clobber for non-z10 compiles.
6000 (s390_emit_epilogue): Emit return jump via return_use expander.
6001 (s390_reorg): Handle JUMP_INSNs as execute targets.
6002 (s390_option_override_internal): Perform validity checks for the
6003 new command line options.
6004 (s390_indirect_branch_attrvalue): New function.
6005 (s390_indirect_branch_settings): New function.
6006 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6007 (s390_output_indirect_thunk_function): New function.
6008 (s390_code_end): Implement target hook.
6009 (s390_case_values_threshold): Implement target hook.
6010 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6012 * config/s390/s390.h (struct s390_frame_layout)
6013 (struct machine_function): Move here from s390.c.
6014 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6015 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6016 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6017 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6018 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6019 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6020 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6021 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6022 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6023 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6024 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6025 (mnemonic attribute): Add values which aren't recognized
6027 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6028 pattern for branch conversion. Fix mnemonic attribute.
6029 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6030 indirect branch via thunk if requested.
6031 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6032 ("*indirect_jump"): Disable for branch conversion using out of
6034 ("indirect_jump_via_thunk<mode>_z10")
6035 ("indirect_jump_via_thunk<mode>")
6036 ("indirect_jump_via_inlinethunk<mode>_z10")
6037 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6038 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6039 ("casesi_jump_via_inlinethunk<mode>_z10")
6040 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6041 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6042 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6043 ("*indirect2_jump"): Disable for branch conversion.
6044 ("casesi_jump"): Turn into expander and expand patterns for branch
6046 ("return_use"): New expander.
6047 ("*return"): Emit return via thunk and rename it to ...
6048 ("*return<mode>"): ... this one.
6049 * config/s390/s390.opt: Add new options and and enum for the
6052 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6054 * lra-constraints.c (match_reload): Unconditionally use
6055 gen_lowpart_SUBREG, rather than selecting between that
6056 and equivalent gen_rtx_SUBREG code.
6058 2018-02-08 Richard Biener <rguenther@suse.de>
6060 PR tree-optimization/84233
6061 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6062 changed flag instead of boguously re-using phi_inserted.
6064 2018-02-08 Martin Jambor <mjambor@suse.cz>
6066 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6067 static local variables.
6069 2018-02-08 Richard Biener <rguenther@suse.de>
6071 PR tree-optimization/84278
6072 * tree-vect-stmts.c (vectorizable_store): When looking for
6073 smaller vector types to perform grouped strided loads/stores
6074 make sure the mode is supported by the target.
6075 (vectorizable_load): Likewise.
6077 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6079 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6080 Increase LDP/STP opportunities by adding adjacent callee-saves.
6082 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6084 PR rtl-optimization/84068
6085 PR rtl-optimization/83459
6086 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6088 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6090 PR tree-optimization/84224
6091 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6092 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6095 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6098 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6099 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6100 count, remove one USE.
6102 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6104 * doc/install.texi (Configuration): Document the
6105 --with-long-double-format={ibm,ieee} PowerPC configuration
6109 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6110 Convert from define_expand to be define_insn_and_split. Rework
6111 float/double/_Float128 conversions to QI/HI/SImode to work with
6112 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6113 conversions to QI/HImode types did a store and then a load to
6114 truncate the value. For conversions to VSX registers, don't split
6115 the insn, instead emit the code directly. Use the code iterator
6116 any_fix to combine signed and unsigned conversions.
6117 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6118 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6119 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6120 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6121 (fix_<mode>di2_hw): Likewise.
6122 (fixuns_<mode>di2_hw): Likewise.
6123 (fix_<mode>si2_hw): Likewise.
6124 (fixuns_<mode>si2_hw): Likewise.
6125 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6126 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6127 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6128 fix<uns>_trunc<SFDF:mode>si2_p8.
6129 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6131 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6132 (fix<uns>_<mode>_mem): Likewise.
6133 (fctiw<u>z_<mode>_mem): Likewise.
6134 (fix<uns>_<mode>_mem): Likewise.
6135 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6136 the register allocator from doing a direct move to the GPRs to do
6137 a store, and instead use the ISA 3.0 store byte/half-word from
6138 vector register instruction. For IEEE 128-bit floating point,
6139 also optimize stores of 32-bit ints.
6140 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6142 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6144 * genextract.c (push_pathstr_operand): New function to support
6146 (walk_rtx): Call push_pathstr_operand.
6147 (print_path): Support [a-zA-Z].
6149 2018-02-07 Richard Biener <rguenther@suse.de>
6151 PR tree-optimization/84037
6152 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6153 (cse_and_gimplify_to_preheader): Declare.
6154 (vect_get_place_in_interleaving_chain): Likewise.
6155 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6157 (_loop_vec_info::~_loop_vec_info): Delete it.
6158 (cse_and_gimplify_to_preheader): New function.
6159 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6160 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6161 (vectorizable_load): Likewise. For grouped stores always base
6162 the IV on the first element.
6163 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6164 condition before gimplifying.
6166 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6168 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6169 *DIV_EXPR and *MOD_EXPR.
6171 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6174 * config/i386/i386.c (ix86_option_override_internal): Mask out
6175 the CF_SET bit when checking -fcf-protection.
6177 2018-02-07 Tom de Vries <tom@codesourcery.com>
6180 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6183 2018-02-07 Richard Biener <rguenther@suse.de>
6185 PR tree-optimization/84204
6186 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6189 PR tree-optimization/84205
6190 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6191 special-case isl_ast_op_zdiv_r.
6193 PR tree-optimization/84223
6194 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6195 Only add conditions from within the region.
6196 (gather_bbs::after_dom_children): Adjust.
6198 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6201 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6202 * config/avr/avr.md: Only post-reload split REG-REG moves if
6203 either register is GENERAL_REG_P.
6205 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6207 PR tree-optimization/84235
6208 * tree-ssa-scopedtables.c
6209 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6210 if the subtraction is performed in floating point type where NaNs are
6211 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6212 build 1. Formatting fix.
6214 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6217 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6218 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6219 and skip it regardless of bb boundaries. Use CALL_P macro,
6220 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6221 unnecessarily, formatting fix.
6223 2018-02-06 Michael Collison <michael.collison@arm.com>
6225 * config/arm/thumb2.md:
6226 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6227 (*thumb_mov_notscc): Ditto.
6229 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6232 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6233 unsigned_fix, not "s".
6235 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6237 * configure.ac (gcc_fn_eh_frame_ro): New function.
6238 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6239 correct .eh_frame permissions.
6240 * configure: Regenerate.
6242 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6244 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6247 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6249 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6250 Display warning message for -mno-speculate-indirect-jumps.
6252 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6254 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6256 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6258 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6260 PR tree-optimization/84225
6261 * tree-eh.c (find_trapping_overflow): Only call
6262 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6264 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6267 * config/i386/i386.c: Reimplement the check of possible options
6268 -mibt/-mshstk conbination. Change error messages.
6269 * doc/invoke.texi: Fix a typo: remove extra '='.
6271 2018-02-06 Marek Polacek <polacek@redhat.com>
6273 PR tree-optimization/84228
6274 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6276 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6279 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6280 emitted arch directives.
6281 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6282 __ARM_FEATURE_COPROC before changing architectures.
6284 2018-02-06 Richard Biener <rguenther@suse.de>
6286 * config/i386/i386.c (print_reg): Fix typo.
6287 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6289 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6291 * configure: Regenerate.
6293 2018-02-05 Martin Sebor <msebor@redhat.com>
6295 PR tree-optimization/83369
6296 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6299 2018-02-05 Martin Liska <mliska@suse.cz>
6301 * doc/invoke.texi: Cherry-pick upstream r323995.
6303 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6305 * ira.c (ira_init_register_move_cost): Adjust comment.
6307 2018-02-05 Martin Liska <mliska@suse.cz>
6309 PR gcov-profile/84137
6310 * doc/gcov.texi: Fix typo in documentation.
6312 2018-02-05 Martin Liska <mliska@suse.cz>
6314 PR gcov-profile/83879
6315 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6316 using dlopen functionality.
6318 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6320 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6321 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6322 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6323 _mm_maskz_range_round_ss): New intrinsics.
6324 (__builtin_ia32_rangesd128_round)
6325 (__builtin_ia32_rangess128_round): Remove.
6326 (__builtin_ia32_rangesd128_mask_round,
6327 __builtin_ia32_rangess128_mask_round): New builtins.
6328 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6329 __builtin_ia32_rangess128_round): Remove.
6330 (__builtin_ia32_rangesd128_mask_round,
6331 __builtin_ia32_rangess128_mask_round): New builtins.
6332 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6333 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6334 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6335 "<round_saeonly_constraint>")): Changed to ...
6336 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6337 "<round_saeonly_scalar_constraint>")): ... this.
6338 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6339 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6340 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6341 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6342 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6344 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6346 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6348 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6349 Remove all values except native, 8540 and 8548.
6351 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6353 * config/i386/i386.c (ix86_output_function_return): Pass
6354 INVALID_REGNUM, instead of -1, as invalid register number to
6355 indirect_thunk_name and output_indirect_thunk.
6357 2018-02-02 Julia Koval <julia.koval@intel.com>
6359 * config.gcc: Add -march=icelake.
6360 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6361 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6362 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6363 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6364 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6365 (processor_target_table): Add icelake.
6366 (ix86_option_override_internal): Handle new PTAs.
6367 (get_builtin_code_for_version): Handle icelake.
6368 (M_INTEL_COREI7_ICELAKE): New.
6369 (fold_builtin_cpu): Handle icelake.
6370 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6371 * doc/invoke.texi: Add -march=icelake.
6373 2018-02-02 Julia Koval <julia.koval@intel.com>
6375 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6376 to wide_int_bitmask.
6377 * wide-int-bitmask.h: New.
6379 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6382 * config/i386/i386.md: Replace Pmode with word_mode in
6383 builtin_setjmp_setup and builtin_longjmp to support x32.
6385 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6389 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6391 (rs6000_supported_cpu_names): New static variable.
6392 (linux_cpu_translation_table): Likewise.
6393 (elf_platform) <cpu>: Define new static variable and use it.
6394 Translate kernel AT_PLATFORM name to canonical name if needed.
6395 Error if platform name is unknown.
6397 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6400 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6402 2018-02-01 Jeff Law <law@redhat.com>
6405 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6406 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6407 the scratch if RELEASE_VIA_POP is false.
6408 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6409 If we have to save a temporary register, decrement SIZE appropriately.
6410 Pass new arguments to release_scratch_register_on_entry.
6411 (ix86_adjust_stack_and_probe): Likewise.
6412 (ix86_emit_probe_stack_range): Pass new arguments to
6413 release_scratch_register_on_entry.
6415 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6417 PR rtl-optimization/84157
6418 * combine.c (change_zero_ext): Use REG_P predicate in
6419 front of HARD_REGISTER_P predicate.
6421 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6423 * config/avr/avr.c (avr_option_override): Move disabling of
6424 -fdelete-null-pointer-checks to...
6425 * common/config/avr/avr-common.c (avr_option_optimization_table):
6428 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6430 PR tree-optimization/81635
6431 * tree-data-ref.c (split_constant_offset_1): For types that
6432 wrap on overflow, try to use range info to prove that wrapping
6435 2018-02-01 Renlin Li <renlin.li@arm.com>
6438 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6440 (aarch64_register_move_cost): Likewise.
6441 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6443 (REG_CLASS_NAMES): Likewise.
6444 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6445 TAILCALL_ADDR_REGS. Remove IP registers.
6446 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6448 2018-02-01 Richard Biener <rguenther@suse.de>
6450 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6451 for specifying RPO order and allow NULL for that.
6452 * domwalk.c (dom_walker::dom_walker): Likewise.
6453 (dom_walker::walk): Handle NULL RPO order.
6454 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6456 (rewrite_update_dom_walker): Likewise.
6457 (mark_def_dom_walker): Likewise.
6459 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6461 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6462 (aarch64_maybe_expand_sve_subreg_move): Declare.
6463 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6464 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6466 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6467 that are semantically a reverse operation.
6468 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6469 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6470 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6472 (aarch64_can_change_mode_class): For big-endian, forbid changes
6473 between two SVE modes if they have different element sizes.
6475 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6477 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6478 the TImode handling for big-endian targets.
6480 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6482 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6483 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6485 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6486 Remove BSWAP handing for big-endian targets and use the form of
6487 LD1RQ appropariate for the mode.
6489 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6491 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6492 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6495 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6498 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6499 check for operands that need to go through aarch64_sve_reload_be.
6501 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6503 PR tree-optimization/81661
6504 PR tree-optimization/84117
6505 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6506 * tree-eh.c: Include gimplify.h.
6507 (find_trapping_overflow, replace_trapping_overflow,
6508 rewrite_to_non_trapping_overflow): New functions.
6509 * tree-vect-loop.c: Include tree-eh.h.
6510 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6511 * tree-data-ref.c: Include tree-eh.h.
6512 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6514 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6516 PR rtl-optimization/84123
6517 * combine.c (change_zero_ext): Check if hard register satisfies
6518 can_change_dest_mode before calling gen_lowpart_SUBREG.
6520 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6523 * ira.c (ira_init_register_move_cost): Remove assert.
6525 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6527 PR rtl-optimization/84071
6528 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6529 * doc/tm.texi: Regenerate.
6531 2018-01-31 Richard Biener <rguenther@suse.de>
6533 PR tree-optimization/84132
6534 * tree-data-ref.c (analyze_miv_subscript): Properly
6535 check whether evolution_function_is_affine_multivariate_p
6536 before calling gcd_of_steps_may_divide_p.
6538 2018-01-31 Julia Koval <julia.koval@intel.com>
6541 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6542 * config/i386/i386.md (rdpid_rex64) New.
6543 (rdpid): Make 32bit only.
6545 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6548 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6549 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6551 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6554 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6556 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6558 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6560 PR rtl-optimization/84071
6561 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6562 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6564 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6566 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6567 (arc_attribute_table): Add 'aux' attribute.
6568 (arc_in_small_data_p): Consider aux like variables.
6569 (arc_is_aux_reg_p): New function.
6570 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6571 (arc_get_aux_arg): New function.
6572 (prepare_move_operands): Handle aux-register access.
6573 (arc_handle_aux_attribute): New function.
6574 * doc/extend.texi (ARC Variable attributes): Add subsection.
6576 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6578 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6579 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6580 (arc_attribute_table): Add 'uncached' attribute.
6581 (arc_print_operand): Print '.di' flag for uncached memory
6583 (arc_in_small_data_p): Do not consider for small data the uncached
6585 (arc_is_uncached_mem_p): New function.
6586 * config/arc/predicates.md (compact_store_memory_operand): Check
6587 for uncached memory accesses.
6588 (nonvol_nonimm_operand): Likewise.
6589 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6591 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6594 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6595 falign-loops=): Add Optimization flag.
6597 2018-01-30 Jeff Law <law@redhat.com>
6600 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6601 INT_REGISTERS_SAVED. Check it prior to calling
6602 get_scratch_register_on_entry.
6603 (ix86_adjust_stack_and_probe): Similarly.
6604 (ix86_emit_probe_stack_range): Similarly.
6605 (ix86_expand_prologue): Corresponding changes.
6607 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6610 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6611 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6613 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6616 * lra-constraints.c (curr_insn_transform): Process AND in the
6619 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6621 PR rtl-optimization/83986
6622 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6623 dependence against last_pending_memory_flush in addition to
6626 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6628 PR tree-optimization/81611
6629 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6632 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6635 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6638 2018-01-30 Richard Biener <rguenther@suse.de>
6639 Jakub Jelinek <jakub@redhat.com>
6641 PR tree-optimization/84111
6642 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6643 inner loops added during recursion, as they don't have up-to-date
6646 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6649 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6650 (can_inline_edge_by_limits_p): ... here.
6651 (can_early_inline_edge_p, check_callers,
6652 update_caller_keys, update_callee_keys, recursive_inlining,
6653 add_new_edges_to_heap, speculation_useful_p,
6654 inline_small_functions,
6655 inline_small_functions, flatten_function,
6656 inline_to_all_callers_1): Update.
6658 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6660 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6663 2018-01-30 Richard Biener <rguenther@suse.de>
6665 PR tree-optimization/83008
6666 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6667 invariant and constant vector uses in stmts when they need
6670 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6673 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6674 * configure: Regenerate.
6676 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6678 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6680 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6681 Use gen_rtx_REG rather than gen_lowpart.
6683 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6685 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6686 rather than 0 when creating partial subregs.
6688 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6690 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6693 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6696 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6697 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6698 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6699 flags. This restores the settings used before the 2017-07-24.
6700 Turning off pre increment/decrement/modify allows IVOPTS to
6701 optimize DF/SF loops where the index is an int.
6703 2018-01-29 Richard Biener <rguenther@suse.de>
6704 Kelvin Nilsen <kelvin@gcc.gnu.org>
6707 * tree-vect-stmts.c (vectorizable_call): Don't call
6708 targetm.vectorize_builtin_md_vectorized_function if callee is
6711 2018-01-22 Carl Love <cel@us.ibm.com>
6713 * doc/extend.tex: Fix typo in second arg in
6714 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6716 2018-01-29 Richard Biener <rguenther@suse.de>
6718 PR tree-optimization/84086
6719 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6720 (flush_ssaname_freelist): When SSA names were released reset
6721 the SCEV hash table.
6723 2018-01-29 Richard Biener <rguenther@suse.de>
6725 PR tree-optimization/84057
6726 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6727 removed paths when removing edges.
6729 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6731 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6732 -mfunction-return=@var{choice}.
6734 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6737 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6739 (layout::print_source_line): Likewise.
6740 (test_get_line_width_without_trailing_whitespace): Add test cases.
6742 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6745 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6748 2018-01-26 Jim Wilson <jimw@sifive.com>
6750 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6752 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6755 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6757 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6758 and CMP + SUB-immediate -> SUBS.
6760 2018-01-26 Martin Sebor <msebor@redhat.com>
6762 PR tree-optimization/83896
6763 * tree-ssa-strlen.c (get_string_len): Rename...
6764 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6765 Avoid assuming length is constant.
6766 (handle_char_store): Use HOST_WIDE_INT for string length.
6768 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6771 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6772 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6774 2018-01-26 Richard Biener <rguenther@suse.de>
6776 PR rtl-optimization/84003
6777 * dse.c (record_store): Only record redundant stores when
6778 the earlier store aliases at least all accesses the later one does.
6780 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6782 PR rtl-optimization/83985
6783 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6784 REG_CFA_RESTORE insns.
6785 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6786 REG_CFA_RESTORE insns here.
6789 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6790 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6792 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6794 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6795 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6796 (arc_init): Likewise.
6797 (arc_override_options): Likewise.
6798 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6800 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6802 * config/arc/arc.h (TARGET_DBNZ): Define.
6803 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6804 properly set the tune attribute.
6805 (dbnz): Use TARGET_DBNZ guard.
6806 * config/arc/arc.opt (mtune): Add core3 option.
6808 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6810 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6811 recognize new pic like addresses.
6812 (arc_delegitimize_address): Clean up.
6814 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6816 * config/arc/arc-arches.def: Option mrf16 valid for all
6818 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6819 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6820 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6821 * config/arc/arc-tables.opt: Regenerate.
6822 * config/arc/arc.c (arc_conditional_register_usage): Handle
6823 reduced register file case.
6824 (arc_file_start): Set must have build attributes.
6825 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6827 * config/arc/arc.opt (mrf16): Add new option.
6828 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6829 * config/arc/genmultilib.awk: Handle new mrf16 option.
6830 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6831 * config/arc/t-multilib: Regenerate.
6832 * doc/invoke.texi (ARC Options): Document mrf16 option.
6834 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6836 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6837 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6838 (arc_attribute_table): Add 'secure_call' attribute.
6839 (arc_print_operand): Print secure call operand.
6840 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6842 (arc_is_secure_call_p): New function. * config/arc/arc.md
6843 (call_i): Add support for sjli instruction.
6844 (call_value_i): Likewise.
6845 * config/arc/constraints.md (Csc): New constraint.
6847 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6848 John Eric Martin <John.Martin@emmicro-us.com>
6850 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6851 * config/arc/arc.c (_arc_jli_section): New struct.
6852 (arc_jli_section): New type.
6853 (rc_jli_sections): New static variable.
6854 (arc_handle_jli_attribute): New function.
6855 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6856 (arc_file_end): New function.
6857 (TARGET_ASM_FILE_END): Define.
6858 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6859 (arc_add_jli_section): New function.
6860 (jli_call_scan): Likewise.
6861 (arc_reorg): Call jli_call_scan.
6862 (arc_output_addsi): Remove 'S' from printing asm operand.
6863 (arc_is_jli_call_p): New function.
6864 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6866 (movhi_insn): Likewise.
6867 (movsi_insn): Likewise.
6868 (movsi_set_cc_insn): Likewise.
6869 (loadqi_update): Likewise.
6870 (load_zeroextendqisi_update): Likewise.
6871 (load_signextendqisi_update): Likewise.
6872 (loadhi_update): Likewise.
6873 (load_zeroextendhisi_update): Likewise.
6874 (load_signextendhisi_update): Likewise.
6875 (loadsi_update): Likewise.
6876 (loadsf_update): Likewise.
6877 (movsicc_insn): Likewise.
6878 (bset_insn): Likewise.
6879 (bxor_insn): Likewise.
6880 (bclr_insn): Likewise.
6881 (bmsk_insn): Likewise.
6882 (bicsi3_insn): Likewise.
6883 (cmpsi_cc_c_insn): Likewise.
6884 (movsi_ne): Likewise.
6885 (movsi_cond_exec): Likewise.
6886 (clrsbsi2): Likewise.
6897 (call_i): Remove 'S' asm letter, add jli instruction.
6898 (call_value_i): Likewise.
6899 * config/arc/arc.op (mjli-always): New option.
6900 * config/arc/constraints.md (Cji): New constraint.
6901 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6903 (subsf3_fpx): Likewise.
6904 (mulsf3_fpx): Likewise.
6905 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6907 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6909 * doc/invoke.texi (ARC): Document mjli-always option.
6911 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6913 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6914 avoid addition with 0 and use incw and decw where possible.
6916 2018-01-26 Richard Biener <rguenther@suse.de>
6918 PR tree-optimization/81082
6919 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6920 association if it requires casting to unsigned.
6921 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6922 from fold_plusminus_mult_expr to catch important cases late when
6923 range info is available.
6925 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6927 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6928 * configure.ac (hidden_linkonce): New test.
6929 * configure: Regenerate.
6930 * config.in: Regenerate.
6932 2018-01-26 Julia Koval <julia.koval@intel.com>
6934 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6935 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6936 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6937 _mm_mask_bitshuffle_epi64_mask): Fix type.
6938 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6939 USI_FTYPE_V4DI_V4DI_USI): Remove.
6940 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6941 __builtin_ia32_vpshufbitqmb256_mask,
6942 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6943 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6944 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6946 2018-01-26 Alan Modra <amodra@gmail.com>
6949 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6950 UNSPEC_VBPERMQ. Sort other unspecs.
6952 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6954 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6956 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6959 * predict.c (drop_profile): Do not push/pop cfun; update also
6961 (handle_missing_profiles): Fix logic looking for zero profiles.
6963 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6966 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6967 on functions with #pragma omp declare simd or functions with simd
6969 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6970 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6971 Remove trailing \n from warning_at calls.
6973 2018-01-25 Tom de Vries <tom@codesourcery.com>
6976 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6977 for neutered workers.
6979 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6982 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6983 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6985 2018-01-24 Jeff Law <law@redhat.com>
6988 * i386.c (get_probe_interval): Move to earlier point.
6989 (ix86_compute_frame_layout): If -fstack-clash-protection and
6990 the frame is larger than the probe interval, then use pushes
6991 to save registers rather than reg->mem moves.
6992 (ix86_expand_prologue): Remove conditional for int_registers_saved
6995 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6998 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6999 min/max for never referenced object.
7001 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7004 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7006 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7007 attributes from DECL_ATTRIBUTES (decl) without affecting
7008 DECL_ATTRIBUTES (current_function_decl).
7009 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7010 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7012 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7014 PR tree-optimization/83979
7015 * fold-const.c (fold_comparison): Use constant_boolean_node
7016 instead of boolean_{true,false}_node.
7018 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7020 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7023 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7025 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7026 Simplify the clause that sets the length attribute.
7027 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7028 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7029 clause that sets the length attribute.
7030 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7032 2018-01-24 Tom de Vries <tom@codesourcery.com>
7035 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7036 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7037 Add strict parameter.
7038 (prevent_branch_around_nothing): Insert dummy insn between branch to
7039 label and label with no ptx insn inbetween.
7040 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7042 2018-01-24 Tom de Vries <tom@codesourcery.com>
7045 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7046 for neutered threads in warp.
7047 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7049 2018-01-24 Richard Biener <rguenther@suse.de>
7051 PR tree-optimization/83176
7052 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7055 2018-01-24 Richard Biener <rguenther@suse.de>
7057 PR tree-optimization/82819
7058 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7059 code generating pluses that are no-ops in the target precision.
7061 2018-01-24 Richard Biener <rguenther@suse.de>
7064 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7066 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7068 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7069 to merge probabilities.
7070 * predict.c (probably_never_executed): Also mark as cold functions
7071 with global 0 profile and guessed local profile.
7072 * profile-count.c (profile_probability::combine_with_count): New
7074 * profile-count.h (profile_probability::operator*,
7075 profile_probability::operator*=, profile_probability::operator/,
7076 profile_probability::operator/=): Reduce precision to adjusted
7077 and set value to guessed on contradictory divisions.
7078 (profile_probability::combine_with_freq): Remove.
7079 (profile_probability::combine_wiht_count): Declare.
7080 (profile_count::force_nonzero):: Set to adjusted.
7081 (profile_count::probability_in):: Set quality to adjusted.
7082 * tree-ssa-tail-merge.c (replace_block_by): Use
7085 2018-01-23 Andrew Waterman <andrew@sifive.com>
7086 Jim Wilson <jimw@sifive.com>
7088 * config/riscv/riscv.c (riscv_stack_boundary): New.
7089 (riscv_option_override): Set riscv_stack_boundary. Handle
7090 riscv_preferred_stack_boundary_arg.
7091 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7092 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7093 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7094 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7095 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7096 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7098 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7101 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7102 of struct ix86_frame.
7103 (ix86_expand_epilogue): Likewise. Add a local variable for
7104 the reg_save_offset field in struct ix86_frame.
7106 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7108 PR tree-optimization/82604
7109 * tree-loop-distribution.c (enum partition_kind): New enum item
7110 PKIND_PARTIAL_MEMSET.
7111 (partition_builtin_p): Support above new enum item.
7112 (generate_code_for_partition): Ditto.
7113 (compute_access_range): Differentiate cases that equality can be
7114 proven at all loops, the innermost loops or no loops.
7115 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7116 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7117 (finalize_partitions, distribute_loop): Don't fuse partition of
7118 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7119 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7122 2018-01-23 Martin Liska <mliska@suse.cz>
7124 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7125 order to ignore the predictor.
7126 (PRED_POLYMORPHIC_CALL): Likewise.
7127 (PRED_RECURSIVE_CALL): Likewise.
7129 2018-01-23 Martin Liska <mliska@suse.cz>
7131 * tree-profile.c (tree_profiling): Print function header to
7132 aware reader which function we are working on.
7133 * value-prof.c (gimple_find_values_to_profile): Do not print
7134 not interesting value histograms.
7136 2018-01-23 Martin Liska <mliska@suse.cz>
7138 * profile-count.h (enum profile_quality): Add
7139 profile_uninitialized as the first value. Do not number values
7140 as they are zero based.
7141 (profile_count::verify): Update sanity check.
7142 (profile_probability::verify): Likewise.
7144 2018-01-23 Nathan Sidwell <nathan@acm.org>
7146 * doc/invoke.texi (ffor-scope): Deprecate.
7148 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7150 PR tree-optimization/83510
7151 * domwalk.c (set_all_edges_as_executable): New function.
7152 (dom_walker::dom_walker): Convert bool param
7153 "skip_unreachable_blocks" to enum reachability. Move setup of
7154 edge flags to set_all_edges_as_executable and only do it when
7155 reachability is REACHABLE_BLOCKS.
7156 * domwalk.h (enum dom_walker::reachability): New enum.
7157 (dom_walker::dom_walker): Convert bool param
7158 "skip_unreachable_blocks" to enum reachability.
7159 (set_all_edges_as_executable): New decl.
7160 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7161 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7163 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7164 but converting true to REACHABLE_BLOCKS.
7165 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7167 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7168 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7169 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7171 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7172 if check_all_array_refs will be called.
7174 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7176 * tree.c (selftest::test_location_wrappers): Add more test
7179 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7181 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7182 (selftest::test_bit_in_range): Likewise.
7184 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7187 * doc/sourcebuild.texi (vect_float): Say that the selector
7188 only describes the situation when -funsafe-math-optimizations is on.
7189 (vect_float_strict): Document.
7191 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7193 PR tree-optimization/83965
7194 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7195 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7196 instead of checking only for a reduction.
7197 (vect_recog_widen_sum_pattern): Likewise.
7199 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7201 * predict.c (probably_never_executed): Only use precise profile info.
7202 (compute_function_frequency): Skip after inlining hack since we now
7203 have quality checking.
7205 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7207 * profile-count.h (profile_probability::very_unlikely,
7208 profile_probability::unlikely, profile_probability::even): Set
7209 precision to guessed.
7211 2018-01-23 Richard Biener <rguenther@suse.de>
7213 PR tree-optimization/83963
7214 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7215 Properly terminate dominator walk when crossing the exit edge not
7216 when visiting its source block.
7218 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7221 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7222 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7224 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7226 PR tree-optimization/83957
7227 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7228 semicolon after for body surrounded by braces.
7230 PR tree-optimization/83081
7231 * profile-count.h (profile_probability::split): New method.
7232 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7233 Use profile_probability::split.
7234 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7235 when splitting a single conditional jump into 2.
7237 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7239 PR tree-optimization/69452
7240 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7243 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7245 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7246 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7247 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7249 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7251 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7252 * config/rl78/rl78.md (movdi): New define_expand.
7253 * config/rl78/rl78.c (rl78_split_movdi): New function.
7255 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7258 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7260 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7261 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7262 128-bit to produce an UNSPEC move to get the double word with the
7263 signbit and then a shift directly to do signbit.
7264 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7265 implementation with a new version that just does either a direct
7266 move or a regular move. Move memory interface to separate insns.
7267 Move insns so they are next to the expander.
7268 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7269 with signbit move. Split big and little endian case.
7270 (signbit<mode>2_dm_mem_le): Likewise.
7271 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7272 (signbit<mode>2_dm2): Likewise.
7274 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7276 * config/rl78/rl78.md (anddi3): New define_expand.
7278 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7280 * config/rl78/rl78.md (umindi3): New define_expand.
7282 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7284 * config/rl78/rl78.md (smindi3): New define_expand.
7286 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7288 * config/rl78/rl78.md (smaxdi3): New define_expand.
7290 2018-01-22 Carl Love <cel@us.ibm.com>
7292 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7293 LVX_V1TI): Add macro expansion.
7294 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7295 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7296 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7297 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7298 Change check to determine if the instruction is a byte reversing
7299 entry. Fix typo in comment.
7300 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7301 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7302 Add def_builtin calls for new builtins.
7303 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7304 Add define_insn expansion.
7306 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7308 * config/rl78/rl78.md (umaxdi3): New define_expand.
7310 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7312 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7313 for non-QImode registers.
7315 2018-01-22 Richard Biener <rguenther@suse.de>
7317 PR tree-optimization/83963
7318 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7319 including the loop exit block.
7320 (scop_detection::merge_sese): Likewise.
7321 (scop_detection::add_scop): Do it here instead.
7323 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7325 * doc/sourcebuild.texi (arm_softfloat): Document.
7327 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7330 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7331 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7332 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7334 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7335 David Edelsohn <dje.gcc@gmail.com>
7338 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7339 Change "crset eq" to "crset 2".
7340 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7341 (*call_indirect_aix<mode>_nospec): Likewise.
7342 (*call_value_indirect_aix<mode>_nospec): Likewise.
7343 (*call_indirect_elfv2<mode>_nospec): Likewise.
7344 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7345 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7346 change assembly output from . to $.
7347 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7348 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7349 (*tablejump<mode>_internal1_nospec): Likewise.
7351 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7354 * config/sh/sh_optimize_sett_clrt.cc:
7355 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7357 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7359 PR tree-optimization/83940
7360 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7361 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7362 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7363 use it to pass back the definition type.
7364 (vect_check_store_rhs): Likewise rhs_dt_out.
7365 (vect_build_gather_load_calls): Add a mask_dt argument and use
7366 it instead of a call to vect_is_simple_use.
7367 (vectorizable_store): Update calls to vect_check_load_store_mask
7368 and vect_check_store_rhs. Use the dt returned by the latter instead
7369 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7370 instead of calls to vect_is_simple_use. Pass the scalar rather
7371 than the vector operand to vect_is_simple_use when handling
7372 second and subsequent copies of an rhs value.
7373 (vectorizable_load): Update calls to vect_check_load_store_mask
7374 and vect_build_gather_load_calls. Use the cached mask_dt and
7375 gs_info.offset_dt instead of calls to vect_is_simple_use.
7377 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7380 * tree-emutls.c: Include gimplify.h.
7381 (lower_emutls_2): New function.
7382 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7383 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7384 it before further processing.
7387 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7388 UINTVAL (trueop1) instead of INTVAL (op1).
7390 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7394 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7395 INCOMING_FRAME_SP_OFFSET if not defined.
7396 (scan_trace): Add ENTRY argument. If true and
7397 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7398 emit a note to adjust the CFA offset.
7399 (create_cfi_notes): Adjust scan_trace callers.
7400 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7401 INCOMING_FRAME_SP_OFFSET in the CIE.
7402 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7403 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7405 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7406 * doc/tm.texi: Regenerated.
7408 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7410 PR rtl-optimization/83147
7411 * lra-constraints.c (remove_inheritance_pseudos): Use
7412 lra_substitute_pseudo_within_insn.
7414 2018-01-19 Tom de Vries <tom@codesourcery.com>
7415 Cesar Philippidis <cesar@codesourcery.com>
7418 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7420 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7423 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7424 spaces for function labels.
7426 2018-01-19 Martin Liska <mliska@suse.cz>
7428 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7429 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7430 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7431 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7432 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7433 (PRED_CONST_RETURN): Change from 69 to 65.
7434 (PRED_NULL_RETURN): Change from 91 to 71.
7435 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7436 (PRED_LOOP_GUARD): Change from 66 to 73.
7438 2018-01-19 Martin Liska <mliska@suse.cz>
7440 * predict.c (predict_insn_def): Add new assert.
7441 (struct branch_predictor): Change type to signed integer.
7442 (test_prediction_value_range): Amend test to cover
7444 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7445 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7446 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7447 (PRED_LOOP_IV_COMPARE): Likewise.
7448 * predict.h (PROB_UNINITIALIZED): Define new constant.
7450 2018-01-19 Martin Liska <mliska@suse.cz>
7452 * predict.c (dump_prediction): Add new format for
7453 analyze_brprob.py script which is enabled with -details
7455 * profile-count.h (precise_p): New function.
7457 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7459 PR tree-optimization/83922
7460 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7461 there are no statements that need masking.
7462 (vect_active_double_reduction_p): New function.
7463 (vect_analyze_loop_operations): Use it when handling phis that
7464 are not in the loop header.
7466 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7468 PR tree-optimization/83914
7469 * tree-vect-loop.c (vectorizable_induction): Don't convert
7470 init_expr or apply the peeling adjustment for inductions
7471 that are nested within the vectorized loop.
7473 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7475 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7478 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7482 * function.h (gimplify_parameters): Add gimple_seq * argument.
7483 * function.c: Include gimple.h and options.h.
7484 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7485 for the added local temporaries if needed.
7486 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7487 if there are any parameter cleanups, wrap whole body into a
7488 try/finally with the cleanups.
7490 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7493 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7494 Use GET_MODE_CLASS for scalar floating point.
7496 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7500 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7501 Fix call of call_cgraph_insertion_hooks.
7503 2018-01-18 Martin Sebor <msebor@redhat.com>
7505 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7507 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7510 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7513 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7516 * common.opt: (-ffile-prefix-map): New option.
7517 * opts.c (common_handle_option): Defer it.
7518 * opts-global.c (handle_common_deferred_options): Handle it.
7519 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7520 * file-prefix-map.h: New file.
7521 (remap_debug_filename, add_debug_prefix_map): ...here.
7522 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7523 * final.c (debug_prefix_map, add_debug_prefix_map
7524 remap_debug_filename): Move to...
7525 * file-prefix-map.c: New file.
7526 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7527 generalize, get rid of alloca(), use strrchr() instead of strchr().
7528 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7529 Implement in terms of add_prefix_map().
7530 (remap_macro_filename, remap_debug_filename): Implement in term of
7532 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7533 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7534 * dbxout.c: Include file-prefix-map.h.
7535 * varasm.c: Likewise.
7536 * vmsdbgout.c: Likewise.
7537 * xcoffout.c: Likewise.
7538 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7539 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7540 * doc/invoke.texi (-ffile-prefix-map): Document.
7541 (-fdebug-prefix-map): Update description.
7543 2018-01-18 Martin Liska <mliska@suse.cz>
7545 * config/i386/i386.c (indirect_thunk_name): Document that also
7547 (output_indirect_thunk): Document why both instructions
7548 (pause and lfence) are generated.
7550 2018-01-18 Richard Biener <rguenther@suse.de>
7552 PR tree-optimization/83887
7553 * graphite-scop-detection.c
7554 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7555 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7556 (scop_detection::merge_sese): Re-implement with a flood-fill
7557 algorithm that properly finds a SESE region if it exists.
7559 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7562 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7563 pointer_diff optimizations use view_convert instead of convert.
7565 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7567 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7568 Generate different code for -mno-speculate-indirect-jumps.
7569 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7570 (*call_indirect_aix<mode>): Disable for
7571 -mno-speculate-indirect-jumps.
7572 (*call_indirect_aix<mode>_nospec): New define_insn.
7573 (*call_value_indirect_aix<mode>): Disable for
7574 -mno-speculate-indirect-jumps.
7575 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7576 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7577 -mno-speculate-indirect-jumps.
7578 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7580 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7582 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7583 long double type, set the flags for noting the default long double
7584 type, even if we don't pass or return a long double type.
7586 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7589 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7592 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7594 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7595 support for merge[hl].
7596 (fold_mergehl_helper): New helper function.
7597 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7598 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7599 (altivec_vmrglw_direct): Add xxmrglw insn.
7601 2018-01-17 Andrew Waterman <andrew@sifive.com>
7603 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7604 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7606 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7609 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7610 call the lto_location_cache before reading the
7611 DECL_SOURCE_LOCATION of the types.
7613 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7614 Richard Sandiford <richard.sandiford@linaro.org>
7616 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7617 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7618 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7619 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7620 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7622 * config/aarch64/constraints.md (aarch64_movti_operand):
7624 * config/aarch64/predicates.md (Uti): Add new constraint.
7626 2018-01-17 Carl Love <cel@us.ibm.com>
7628 * config/rs6000/vsx.md (define_expand xl_len_r,
7629 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7630 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7632 (define_expand, define_insn): Move the shift left from the
7633 define_insn to the define_expand for lxvl and stxvl instructions.
7634 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7635 and XL_LEN_R definitions to PURE.
7637 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7639 * config/i386/i386.c (indirect_thunk_name): Declare regno
7640 as unsigned int. Compare regno with INVALID_REGNUM.
7641 (output_indirect_thunk): Ditto.
7642 (output_indirect_thunk_function): Ditto.
7643 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7644 in the call to output_indirect_thunk_function.
7646 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7649 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7650 rather than the size of inner_type to determine the stack slot size
7651 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7653 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7656 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7659 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7661 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7662 endian Linux systems to optionally enable multilibs for selecting
7663 the long double type if the user configured an explicit type.
7664 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7665 have no long double multilibs if not defined.
7666 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7667 warn if the user used -mabi={ieee,ibm}longdouble and we built
7668 multilibs for long double.
7669 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7670 appropriate multilib option.
7671 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7673 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7674 for building long double multilibs.
7675 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7677 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7679 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7682 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7684 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7687 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7690 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7693 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7695 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7696 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7698 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7700 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7701 different rtl trees depending on TARGET_64BIT.
7702 (rs6000_gen_lvx): Likewise.
7704 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7706 * config/visium/visium.md (nop): Tweak comment.
7707 (hazard_nop): Likewise.
7709 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7711 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7712 -mspeculate-indirect-jumps.
7713 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7714 for -mno-speculate-indirect-jumps.
7715 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7716 (*call_value_indirect_elfv2<mode>): Disable for
7717 -mno-speculate-indirect-jumps.
7718 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7719 (indirect_jump): Emit different RTL for
7720 -mno-speculate-indirect-jumps.
7721 (*indirect_jump<mode>): Disable for
7722 -mno-speculate-indirect-jumps.
7723 (*indirect_jump<mode>_nospec): New define_insn.
7724 (tablejump): Emit different RTL for
7725 -mno-speculate-indirect-jumps.
7726 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7727 (tablejumpsi_nospec): New define_expand.
7728 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7729 (tablejumpdi_nospec): New define_expand.
7730 (*tablejump<mode>_internal1): Disable for
7731 -mno-speculate-indirect-jumps.
7732 (*tablejump<mode>_internal1_nospec): New define_insn.
7733 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7736 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7738 * caller-save.c (insert_save): Drop unnecessary parameter. All
7741 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7742 Richard Biener <rguenth@suse.de>
7745 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7746 return early, inline manually is_gimple_sizepos. Make sure if we
7747 call gimplify_expr we don't end up with a gimple constant.
7748 * tree.c (variably_modified_type_p): Don't return true for
7749 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7750 * gimplify.h (is_gimple_sizepos): Remove.
7752 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7754 PR tree-optimization/83857
7755 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7756 vectorizable_live_operation for pure SLP statements.
7757 (vectorizable_live_operation): Handle PHIs.
7759 2018-01-16 Richard Biener <rguenther@suse.de>
7761 PR tree-optimization/83867
7762 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7763 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7765 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7768 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7769 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7770 If off is not INTEGER_CST, issue a may not be aligned warning
7771 rather than isn't aligned. Use isn%'t rather than isn't.
7772 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7774 <case MULT_EXPR>: Improve the case when bottom and one of the
7775 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7776 operand, in that case check if the other operand is multiple of
7777 bottom divided by the INTEGER_CST operand.
7779 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7782 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7783 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7784 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7785 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7786 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7787 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7788 (pa_function_arg_size): New function.
7790 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7792 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7793 in a separate statement.
7795 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7797 PR tree-optimization/83847
7798 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7799 group gathers and scatters.
7801 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7803 PR rtl-optimization/86620
7804 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7806 PR rtl-optimization/83213
7807 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7808 to last if both are JUMP_INSNs.
7810 PR tree-optimization/83843
7811 * gimple-ssa-store-merging.c
7812 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7813 store_immediate_info for bswap/nop orig_stores.
7815 2018-01-15 Andrew Waterman <andrew@sifive.com>
7817 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7819 <UDIV>: Increase cost if !TARGET_DIV.
7821 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7823 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7824 (define_attr "cr_logical_3op"): New.
7825 (cceq_ior_compare): Adjust.
7826 (cceq_ior_compare_complement): Adjust.
7827 (*cceq_rev_compare): Adjust.
7828 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7829 (is_cracked_insn): Adjust.
7830 (insn_must_be_first_in_group): Adjust.
7831 * config/rs6000/40x.md: Adjust.
7832 * config/rs6000/440.md: Adjust.
7833 * config/rs6000/476.md: Adjust.
7834 * config/rs6000/601.md: Adjust.
7835 * config/rs6000/603.md: Adjust.
7836 * config/rs6000/6xx.md: Adjust.
7837 * config/rs6000/7450.md: Adjust.
7838 * config/rs6000/7xx.md: Adjust.
7839 * config/rs6000/8540.md: Adjust.
7840 * config/rs6000/cell.md: Adjust.
7841 * config/rs6000/e300c2c3.md: Adjust.
7842 * config/rs6000/e500mc.md: Adjust.
7843 * config/rs6000/e500mc64.md: Adjust.
7844 * config/rs6000/e5500.md: Adjust.
7845 * config/rs6000/e6500.md: Adjust.
7846 * config/rs6000/mpc.md: Adjust.
7847 * config/rs6000/power4.md: Adjust.
7848 * config/rs6000/power5.md: Adjust.
7849 * config/rs6000/power6.md: Adjust.
7850 * config/rs6000/power7.md: Adjust.
7851 * config/rs6000/power8.md: Adjust.
7852 * config/rs6000/power9.md: Adjust.
7853 * config/rs6000/rs64.md: Adjust.
7854 * config/rs6000/titan.md: Adjust.
7856 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7858 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7859 ix86_indirect_branch_register logic.
7861 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7863 * config/i386/constraints.md (Bs): Update
7864 ix86_indirect_branch_register check. Don't check
7865 ix86_indirect_branch_register with GOT_memory_operand.
7867 * config/i386/predicates.md (GOT_memory_operand): Don't check
7868 ix86_indirect_branch_register here.
7869 (GOT32_symbol_operand): Likewise.
7871 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7873 * config/i386/predicates.md (constant_call_address_operand):
7874 Rewrite ix86_indirect_branch_register logic.
7875 (sibcall_insn_operand): Likewise.
7877 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7879 * config/i386/constraints.md (Bs): Replace
7880 ix86_indirect_branch_thunk_register with
7881 ix86_indirect_branch_register.
7883 * config/i386/i386.md (indirect_jump): Likewise.
7884 (tablejump): Likewise.
7885 (*sibcall_memory): Likewise.
7886 (*sibcall_value_memory): Likewise.
7887 Peepholes of indirect call and jump via memory: Likewise.
7888 * config/i386/i386.opt: Likewise.
7889 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7890 (GOT_memory_operand): Likewise.
7891 (call_insn_operand): Likewise.
7892 (sibcall_insn_operand): Likewise.
7893 (GOT32_symbol_operand): Likewise.
7895 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7898 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7899 type rather than type addr's type points to.
7900 (expand_omp_atomic_mutex): Likewise.
7901 (expand_omp_atomic): Likewise.
7903 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7906 * config/i386/i386.c (output_indirect_thunk_function): Use
7907 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7908 for __x86_return_thunk.
7910 2018-01-15 Richard Biener <rguenther@suse.de>
7913 * expmed.c (extract_bit_field_1): Fix typo.
7915 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7918 * config/arm/iterators.md (VF): New mode iterator.
7919 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7920 Remove integer-related logic from pattern.
7921 (neon_vabd<mode>_3): Likewise.
7923 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7926 * common.opt (fstrict-overflow): No longer an alias.
7927 (fwrapv-pointer): New option.
7928 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7929 also for pointer types based on flag_wrapv_pointer.
7930 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7931 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7932 opts->x_flag_wrapv got set.
7933 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7934 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7935 POINTER_TYPE_OVERFLOW_UNDEFINED.
7936 * match.pd: Likewise in address comparison pattern.
7937 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7939 2018-01-15 Richard Biener <rguenther@suse.de>
7942 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7943 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7944 Reset type names to their identifier if their TYPE_DECL doesn't
7945 have linkage (and thus is used for ODR and devirt).
7946 (save_debug_info_for_decl): Remove.
7947 (save_debug_info_for_type): Likewise.
7948 (add_tree_to_fld_list): Adjust.
7949 * tree-pretty-print.c (dump_generic_node): Make dumping of
7950 type names more robust.
7952 2018-01-15 Richard Biener <rguenther@suse.de>
7954 * BASE-VER: Bump to 8.0.1.
7956 2018-01-14 Martin Sebor <msebor@redhat.com>
7959 * builtins.c (check_access): Avoid warning when the no-warning bit
7962 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7964 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7965 * ira-color (allocno_hard_regs_compare): Likewise.
7967 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7970 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7971 Use .pushsection/.popsection.
7973 2018-01-14 Martin Sebor <msebor@redhat.com>
7976 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7978 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7980 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7981 entry from extra_headers.
7982 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7983 extra_headers, make the list bitwise identical to the i?86-*-* one.
7985 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7987 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7988 -mcmodel=large with -mindirect-branch=thunk,
7989 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7990 -mfunction-return=thunk-extern.
7991 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7992 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7993 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7995 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7997 * config/i386/i386.c (print_reg): Print the name of the full
7998 integer register without '%'.
7999 (ix86_print_operand): Handle 'V'.
8000 * doc/extend.texi: Document 'V' modifier.
8002 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8004 * config/i386/constraints.md (Bs): Disallow memory operand for
8005 -mindirect-branch-register.
8007 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8008 (GOT_memory_operand): Likewise.
8009 (call_insn_operand): Likewise.
8010 (sibcall_insn_operand): Likewise.
8011 (GOT32_symbol_operand): Likewise.
8012 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8013 for -mindirect-branch-register.
8014 (tablejump): Likewise.
8015 (*sibcall_memory): Likewise.
8016 (*sibcall_value_memory): Likewise.
8017 Disallow peepholes of indirect call and jump via memory for
8018 -mindirect-branch-register.
8019 (*call_pop): Replace m with Bw.
8020 (*call_value_pop): Likewise.
8021 (*sibcall_pop_memory): Replace m with Bs.
8022 * config/i386/i386.opt (mindirect-branch-register): New option.
8023 * doc/invoke.texi: Document -mindirect-branch-register option.
8025 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8027 * config/i386/i386-protos.h (ix86_output_function_return): New.
8028 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8029 set function_return_type.
8030 (indirect_thunk_name): Add ret_p to indicate thunk for function
8032 (output_indirect_thunk_function): Pass false to
8033 indirect_thunk_name.
8034 (ix86_output_indirect_branch_via_reg): Likewise.
8035 (ix86_output_indirect_branch_via_push): Likewise.
8036 (output_indirect_thunk_function): Create alias for function
8037 return thunk if regno < 0.
8038 (ix86_output_function_return): New function.
8039 (ix86_handle_fndecl_attribute): Handle function_return.
8040 (ix86_attribute_table): Add function_return.
8041 * config/i386/i386.h (machine_function): Add
8042 function_return_type.
8043 * config/i386/i386.md (simple_return_internal): Use
8044 ix86_output_function_return.
8045 (simple_return_internal_long): Likewise.
8046 * config/i386/i386.opt (mfunction-return=): New option.
8047 (indirect_branch): Mention -mfunction-return=.
8048 * doc/extend.texi: Document function_return function attribute.
8049 * doc/invoke.texi: Document -mfunction-return= option.
8051 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8053 * config/i386/i386-opts.h (indirect_branch): New.
8054 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8055 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8056 with local indirect jump when converting indirect call and jump.
8057 (ix86_set_indirect_branch_type): New.
8058 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8059 (indirectlabelno): New.
8060 (indirect_thunk_needed): Likewise.
8061 (indirect_thunk_bnd_needed): Likewise.
8062 (indirect_thunks_used): Likewise.
8063 (indirect_thunks_bnd_used): Likewise.
8064 (INDIRECT_LABEL): Likewise.
8065 (indirect_thunk_name): Likewise.
8066 (output_indirect_thunk): Likewise.
8067 (output_indirect_thunk_function): Likewise.
8068 (ix86_output_indirect_branch_via_reg): Likewise.
8069 (ix86_output_indirect_branch_via_push): Likewise.
8070 (ix86_output_indirect_branch): Likewise.
8071 (ix86_output_indirect_jmp): Likewise.
8072 (ix86_code_end): Call output_indirect_thunk_function if needed.
8073 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8075 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8076 (ix86_attribute_table): Add indirect_branch.
8077 * config/i386/i386.h (machine_function): Add indirect_branch_type
8078 and has_local_indirect_jump.
8079 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8081 (tablejump): Likewise.
8082 (*indirect_jump): Use ix86_output_indirect_jmp.
8083 (*tablejump_1): Likewise.
8084 (simple_return_indirect_internal): Likewise.
8085 * config/i386/i386.opt (mindirect-branch=): New option.
8086 (indirect_branch): New.
8089 (thunk-inline): Likewise.
8090 (thunk-extern): Likewise.
8091 * doc/extend.texi: Document indirect_branch function attribute.
8092 * doc/invoke.texi: Document -mindirect-branch= option.
8094 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8097 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8099 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8101 * ipa-inline.c (want_inline_small_function_p): Return false if
8102 inlining has already failed with CIF_FINAL_ERROR.
8103 (update_caller_keys): Call want_inline_small_function_p before
8105 (update_callee_keys): Likewise.
8107 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8109 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8111 (rs6000_quadword_masked_address_p): Likewise.
8112 (quad_aligned_load_p): Likewise.
8113 (quad_aligned_store_p): Likewise.
8114 (const_load_sequence_p): Add comment to describe the outer-most loop.
8115 (mimic_memory_attributes_and_flags): New function.
8116 (rs6000_gen_stvx): Likewise.
8117 (replace_swapped_aligned_store): Likewise.
8118 (rs6000_gen_lvx): Likewise.
8119 (replace_swapped_aligned_load): Likewise.
8120 (replace_swapped_load_constant): Capitalize argument name in
8121 comment describing this function.
8122 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8123 and stores that access quad-word aligned addresses and replace
8124 with stvx or lvx instructions when appropriate.
8125 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8126 New function prototype.
8127 (rs6000_quadword_masked_address_p): Likewise.
8128 (rs6000_gen_lvx): Likewise.
8129 (rs6000_gen_stvx): Likewise.
8130 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8131 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8132 when memory address is aligned.
8133 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8134 this split to select lvx instruction when memory address is aligned.
8135 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8136 instruction when memory address is aligned.
8137 (*vsx_le_perm_load_v16qi): Likewise.
8138 (four unnamed splitters): Modify to select the stvx instruction
8139 when memory is aligned.
8141 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8143 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8144 which appears in the queue multiple times.
8146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8147 Alan Hayward <alan.hayward@arm.com>
8148 David Sherwood <david.sherwood@arm.com>
8150 * tree-vectorizer.h (vec_lower_bound): New structure.
8151 (_loop_vec_info): Add check_nonzero and lower_bounds.
8152 (LOOP_VINFO_CHECK_NONZERO): New macro.
8153 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8154 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8155 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8156 fields. Make seg_len the distance travelled, not including the
8158 (dr_direction_indicator): Declare.
8159 (dr_zero_step_indicator): Likewise.
8160 (dr_known_forward_stride_p): Likewise.
8161 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8163 (runtime_alias_check_p): Allow runtime alias checks with
8165 (operator ==): Compare access_size and align.
8166 (prune_runtime_alias_test_list): Rework for new distinction between
8167 the access_size and seg_len.
8168 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8170 (get_segment_min_max): New function.
8171 (create_intersect_range_checks): Use it.
8172 (dr_step_indicator): New function.
8173 (dr_direction_indicator): Likewise.
8174 (dr_zero_step_indicator): Likewise.
8175 (dr_known_forward_stride_p): Likewise.
8176 * tree-loop-distribution.c (data_ref_segment_size): Return
8177 DR_STEP * (niters - 1).
8178 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8180 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8181 (vect_preserves_scalar_order_p): New function, split out from...
8182 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8183 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8184 (vect_vfa_access_size): New function.
8185 (vect_vfa_align): Likewise.
8186 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8187 (dump_lower_bound): New function.
8188 (vect_check_lower_bound): Likewise.
8189 (vect_small_gap_p): Likewise.
8190 (vectorizable_with_step_bound_p): Likewise.
8191 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8192 depencies if the vectorization factor is 1. Convert the checks
8193 for nonzero steps into checks on the bounds of DR_STEP. Try using
8194 a bunds check for variable steps if the minimum required step is
8195 relatively small. Update calls to the dr_with_seg_len
8196 constructor and to vect_compile_time_alias.
8197 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8199 (vect_loop_versioning): Call it.
8200 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8202 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8205 Alan Hayward <alan.hayward@arm.com>
8206 David Sherwood <david.sherwood@arm.com>
8208 * doc/sourcebuild.texi (vect_scatter_store): Document.
8209 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8211 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8213 * genopinit.c (main): Add supports_vec_scatter_store and
8214 supports_vec_scatter_store_cached to target_optabs.
8215 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8216 IFN_MASK_SCATTER_STORE.
8217 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8219 * internal-fn.h (internal_store_fn_p): Declare.
8220 (internal_fn_stored_value_index): Likewise.
8221 * internal-fn.c (scatter_store_direct): New macro.
8222 (expand_scatter_store_optab_fn): New function.
8223 (direct_scatter_store_optab_supported_p): New macro.
8224 (internal_store_fn_p): New function.
8225 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8226 IFN_MASK_SCATTER_STORE.
8227 (internal_fn_mask_index): Likewise.
8228 (internal_fn_stored_value_index): New function.
8229 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8231 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8232 * optabs-query.c (supports_vec_scatter_store_p): New function.
8233 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8234 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8235 true for scatter stores.
8236 (vect_gather_scatter_fn_p): Handle scatter stores too.
8237 (vect_check_gather_scatter): Consider using scatter stores if
8238 supports_vec_scatter_store_p.
8239 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8241 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8242 internal_fn_stored_value_index.
8243 (check_load_store_masking): Handle scatter stores too.
8244 (vect_get_store_rhs): Make public.
8245 (vectorizable_call): Use internal_store_fn_p.
8246 (vectorizable_store): Handle scatter store internal functions.
8247 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8248 when deciding whether the end of the group has been reached.
8249 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8250 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8251 (mask_scatter_store<mode>): New insns.
8253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8254 Alan Hayward <alan.hayward@arm.com>
8255 David Sherwood <david.sherwood@arm.com>
8257 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8258 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8259 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8261 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8262 Use vect_truncate_gather_scatter_offset if we can't treat the
8263 operation as a normal gather load or scatter store.
8264 (get_group_load_store_type): Take the gather_scatter_info
8265 as argument. Try using a gather load or scatter store for
8266 single-element groups.
8267 (get_load_store_type): Update calls to get_group_load_store_type
8268 and vect_use_strided_gather_scatters_p.
8270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8271 Alan Hayward <alan.hayward@arm.com>
8272 David Sherwood <david.sherwood@arm.com>
8274 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8275 optional tree argument.
8276 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8278 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8279 but continue to use the current value as a fallback.
8280 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8281 to compare the updates.
8282 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8283 (get_load_store_type): Use it when handling a strided access.
8284 (vect_get_strided_load_store_ops): New function.
8285 (vect_get_data_ptr_increment): Likewise.
8286 (vectorizable_load): Handle strided gather loads. Always pass
8287 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8290 Alan Hayward <alan.hayward@arm.com>
8291 David Sherwood <david.sherwood@arm.com>
8293 * doc/md.texi (gather_load@var{m}): Document.
8294 (mask_gather_load@var{m}): Likewise.
8295 * genopinit.c (main): Add supports_vec_gather_load and
8296 supports_vec_gather_load_cached to target_optabs.
8297 * optabs-tree.c (init_tree_optimization_optabs): Use
8298 ggc_cleared_alloc to allocate target_optabs.
8299 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8300 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8302 * internal-fn.h (internal_load_fn_p): Declare.
8303 (internal_gather_scatter_fn_p): Likewise.
8304 (internal_fn_mask_index): Likewise.
8305 (internal_gather_scatter_fn_supported_p): Likewise.
8306 * internal-fn.c (gather_load_direct): New macro.
8307 (expand_gather_load_optab_fn): New function.
8308 (direct_gather_load_optab_supported_p): New macro.
8309 (direct_internal_fn_optab): New function.
8310 (internal_load_fn_p): Likewise.
8311 (internal_gather_scatter_fn_p): Likewise.
8312 (internal_fn_mask_index): Likewise.
8313 (internal_gather_scatter_fn_supported_p): Likewise.
8314 * optabs-query.c (supports_at_least_one_mode_p): New function.
8315 (supports_vec_gather_load_p): Likewise.
8316 * optabs-query.h (supports_vec_gather_load_p): Declare.
8317 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8318 and memory_type field.
8319 (NUM_PATTERNS): Bump to 15.
8320 * tree-vect-data-refs.c: Include internal-fn.h.
8321 (vect_gather_scatter_fn_p): New function.
8322 (vect_describe_gather_scatter_call): Likewise.
8323 (vect_check_gather_scatter): Try using internal functions for
8324 gather loads. Recognize existing calls to a gather load function.
8325 (vect_analyze_data_refs): Consider using gather loads if
8326 supports_vec_gather_load_p.
8327 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8328 (vect_get_gather_scatter_offset_type): Likewise.
8329 (vect_convert_mask_for_vectype): Likewise.
8330 (vect_add_conversion_to_patterm): Likewise.
8331 (vect_try_gather_scatter_pattern): Likewise.
8332 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8333 (vect_vect_recog_func_ptrs): Add it.
8334 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8335 internal_fn_mask_index and internal_gather_scatter_fn_p.
8336 (check_load_store_masking): Take the gather_scatter_info as an
8337 argument and handle gather loads.
8338 (vect_get_gather_scatter_ops): New function.
8339 (vectorizable_call): Check internal_load_fn_p.
8340 (vectorizable_load): Likewise. Handle gather load internal
8342 (vectorizable_store): Update call to check_load_store_masking.
8343 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8344 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8345 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8346 (aarch64_gather_scale_operand_d): New predicates.
8347 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8348 (mask_gather_load<mode>): New insns.
8350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8351 Alan Hayward <alan.hayward@arm.com>
8352 David Sherwood <david.sherwood@arm.com>
8354 * optabs.def (fold_left_plus_optab): New optab.
8355 * doc/md.texi (fold_left_plus_@var{m}): Document.
8356 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8357 * internal-fn.c (fold_left_direct): Define.
8358 (expand_fold_left_optab_fn): Likewise.
8359 (direct_fold_left_optab_supported_p): Likewise.
8360 * fold-const-call.c (fold_const_fold_left): New function.
8361 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8362 * tree-parloops.c (valid_reduction_p): New function.
8363 (gather_scalar_reductions): Use it.
8364 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8365 (vect_finish_replace_stmt): Declare.
8366 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8367 (needs_fold_left_reduction_p): New function, split out from...
8368 (vect_is_simple_reduction): ...here. Accept reductions that
8369 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8370 (vect_force_simple_reduction): Also store the reduction type in
8371 the assignment's STMT_VINFO_REDUC_TYPE.
8372 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8373 (merge_with_identity): New function.
8374 (vect_expand_fold_left): Likewise.
8375 (vectorize_fold_left_reduction): Likewise.
8376 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8377 scalar phi in place for it. Check for target support and reject
8378 cases that would reassociate the operation. Defer the transform
8379 phase to vectorize_fold_left_reduction.
8380 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8381 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8382 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8386 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8387 call to ifc_temp_var.
8389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8390 Alan Hayward <alan.hayward@arm.com>
8391 David Sherwood <david.sherwood@arm.com>
8393 * target.def (legitimize_address_displacement): Take the original
8394 offset as a poly_int.
8395 * targhooks.h (default_legitimize_address_displacement): Update
8397 * targhooks.c (default_legitimize_address_displacement): Likewise.
8398 * doc/tm.texi: Regenerate.
8399 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8400 as an argument, moving assert of ad->disp == ad->disp_term to...
8401 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8402 Try calling targetm.legitimize_address_displacement before expanding
8403 the address rather than afterwards, and adjust for the new interface.
8404 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8405 Match the new hook interface. Handle SVE addresses.
8406 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8411 * Makefile.in (OBJS): Add early-remat.o.
8412 * target.def (select_early_remat_modes): New hook.
8413 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8414 * doc/tm.texi: Regenerate.
8415 * targhooks.h (default_select_early_remat_modes): Declare.
8416 * targhooks.c (default_select_early_remat_modes): New function.
8417 * timevar.def (TV_EARLY_REMAT): New timevar.
8418 * passes.def (pass_early_remat): New pass.
8419 * tree-pass.h (make_pass_early_remat): Declare.
8420 * early-remat.c: New file.
8421 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8423 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8426 Alan Hayward <alan.hayward@arm.com>
8427 David Sherwood <david.sherwood@arm.com>
8429 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8430 vfm1 with a bound_epilog parameter.
8431 (vect_do_peeling): Update calls accordingly, and move the prologue
8432 call earlier in the function. Treat the base bound_epilog as 0 for
8433 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8434 this base when peeling for gaps.
8435 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8436 with fully-masked loops.
8437 (vect_estimate_min_profitable_iters): Handle the single peeled
8438 iteration in that case.
8440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8441 Alan Hayward <alan.hayward@arm.com>
8442 David Sherwood <david.sherwood@arm.com>
8444 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8445 single-element interleaving even if the size is not a power of 2.
8446 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8447 accesses for single-element interleaving if the group size is
8450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8451 Alan Hayward <alan.hayward@arm.com>
8452 David Sherwood <david.sherwood@arm.com>
8454 * doc/md.texi (fold_extract_last_@var{m}): Document.
8455 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8456 * optabs.def (fold_extract_last_optab): New optab.
8457 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8458 * internal-fn.c (fold_extract_direct): New macro.
8459 (expand_fold_extract_optab_fn): Likewise.
8460 (direct_fold_extract_optab_supported_p): Likewise.
8461 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8462 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8463 EXTRACT_LAST_REDUCTION.
8464 (get_initial_def_for_reduction): Do not create an initial vector
8465 for EXTRACT_LAST_REDUCTION reductions.
8466 (vectorizable_reduction): Leave the scalar phi in place for
8467 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8468 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8469 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8470 transform phase to vectorizable_condition.
8471 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8473 (vect_finish_stmt_generation): ...here.
8474 (vect_finish_replace_stmt): New function.
8475 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8476 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8478 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8481 Alan Hayward <alan.hayward@arm.com>
8482 David Sherwood <david.sherwood@arm.com>
8484 * doc/md.texi (extract_last_@var{m}): Document.
8485 * optabs.def (extract_last_optab): New optab.
8486 * internal-fn.def (EXTRACT_LAST): New internal function.
8487 * internal-fn.c (cond_unary_direct): New macro.
8488 (expand_cond_unary_optab_fn): Likewise.
8489 (direct_cond_unary_optab_supported_p): Likewise.
8490 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8491 loops using EXTRACT_LAST.
8492 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8493 (extract_last_<mode>): ...this optab.
8494 (vec_extract<mode><Vel>): Update accordingly.
8496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8497 Alan Hayward <alan.hayward@arm.com>
8498 David Sherwood <david.sherwood@arm.com>
8500 * target.def (empty_mask_is_expensive): New hook.
8501 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8502 * doc/tm.texi: Regenerate.
8503 * targhooks.h (default_empty_mask_is_expensive): Declare.
8504 * targhooks.c (default_empty_mask_is_expensive): New function.
8505 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8506 if the target says that empty masks are expensive.
8507 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8509 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8512 Alan Hayward <alan.hayward@arm.com>
8513 David Sherwood <david.sherwood@arm.com>
8515 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8516 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8517 (vect_use_loop_mask_for_alignment_p): New function.
8518 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8519 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8520 niters_skip argument. Make sure that the first niters_skip elements
8521 of the first iteration are inactive.
8522 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8523 Update call to vect_set_loop_masks_directly.
8524 (get_misalign_in_elems): New function, split out from...
8525 (vect_gen_prolog_loop_niters): ...here.
8526 (vect_update_init_of_dr): Take a code argument that specifies whether
8527 the adjustment should be added or subtracted.
8528 (vect_update_init_of_drs): Likewise.
8529 (vect_prepare_for_masked_peels): New function.
8530 (vect_do_peeling): Skip prologue peeling if we're using a mask
8531 instead. Update call to vect_update_inits_of_drs.
8532 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8534 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8535 alignment. Do not include the number of peeled iterations in
8536 the minimum threshold in that case.
8537 (vectorizable_induction): Adjust the start value down by
8538 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8539 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8540 Take the number of skipped iterations into account when calculating
8542 * tree-vect-stmts.c (vect_gen_while_not): New function.
8544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8545 Alan Hayward <alan.hayward@arm.com>
8546 David Sherwood <david.sherwood@arm.com>
8548 * doc/sourcebuild.texi (vect_fully_masked): Document.
8549 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8551 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8553 (vect_analyze_loop_2): ...here. Don't check the vectorization
8554 factor against the number of loop iterations if the loop is
8557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8558 Alan Hayward <alan.hayward@arm.com>
8559 David Sherwood <david.sherwood@arm.com>
8561 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8562 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8563 (dump_groups): Update accordingly.
8564 (iv_use::mem_type): New member variable.
8565 (address_p): New function.
8566 (record_use): Add a mem_type argument and initialize the new
8568 (record_group_use): Add a mem_type argument. Use address_p.
8569 Remove obsolete null checks of base_object. Update call to record_use.
8570 (find_interesting_uses_op): Update call to record_group_use.
8571 (find_interesting_uses_cond): Likewise.
8572 (find_interesting_uses_address): Likewise.
8573 (get_mem_type_for_internal_fn): New function.
8574 (find_address_like_use): Likewise.
8575 (find_interesting_uses_stmt): Try find_address_like_use before
8576 calling find_interesting_uses_op.
8577 (addr_offset_valid_p): Use the iv mem_type field as the type
8578 of the addressed memory.
8579 (add_autoinc_candidates): Likewise.
8580 (get_address_cost): Likewise.
8581 (split_small_address_groups_p): Use address_p.
8582 (split_address_groups): Likewise.
8583 (add_iv_candidate_for_use): Likewise.
8584 (autoinc_possible_for_pair): Likewise.
8585 (rewrite_groups): Likewise.
8586 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8587 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8588 (get_alias_ptr_type_for_ptr_address): New function.
8589 (rewrite_use_address): Rewrite address uses in calls that were
8590 identified by find_address_like_use.
8592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8593 Alan Hayward <alan.hayward@arm.com>
8594 David Sherwood <david.sherwood@arm.com>
8596 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8598 * gimple-expr.h (is_gimple_addressable: Likewise.
8599 * gimple-expr.c (is_gimple_address): Likewise.
8600 * internal-fn.c (expand_call_mem_ref): New function.
8601 (expand_mask_load_optab_fn): Use it.
8602 (expand_mask_store_optab_fn): Likewise.
8604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8605 Alan Hayward <alan.hayward@arm.com>
8606 David Sherwood <david.sherwood@arm.com>
8608 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8609 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8610 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8611 (cond_umax@var{mode}): Document.
8612 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8613 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8614 (cond_umin_optab, cond_umax_optab): New optabs.
8615 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8616 (COND_IOR, COND_XOR): New internal functions.
8617 * internal-fn.h (get_conditional_internal_fn): Declare.
8618 * internal-fn.c (cond_binary_direct): New macro.
8619 (expand_cond_binary_optab_fn): Likewise.
8620 (direct_cond_binary_optab_supported_p): Likewise.
8621 (get_conditional_internal_fn): New function.
8622 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8623 Cope with reduction statements that are vectorized as calls rather
8625 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8626 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8627 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8628 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8629 (UNSPEC_COND_EOR): New unspecs.
8630 (optab): Add mappings for them.
8631 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8632 (sve_int_op, sve_fp_op): New int attributes.
8634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8635 Alan Hayward <alan.hayward@arm.com>
8636 David Sherwood <david.sherwood@arm.com>
8638 * optabs.def (while_ult_optab): New optab.
8639 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8640 * internal-fn.def (WHILE_ULT): New internal function.
8641 * internal-fn.h (direct_internal_fn_supported_p): New override
8642 that takes two types as argument.
8643 * internal-fn.c (while_direct): New macro.
8644 (expand_while_optab_fn): New function.
8645 (convert_optab_supported_p): Likewise.
8646 (direct_while_optab_supported_p): New macro.
8647 * wide-int.h (wi::udiv_ceil): New function.
8648 * tree-vectorizer.h (rgroup_masks): New structure.
8649 (vec_loop_masks): New typedef.
8650 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8652 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8653 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8654 (vect_max_vf): New function.
8655 (slpeel_make_loop_iterate_ntimes): Delete.
8656 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8657 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8658 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8659 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8660 internal-fn.h, stor-layout.h and optabs-query.h.
8661 (vect_set_loop_mask): New function.
8662 (add_preheader_seq): Likewise.
8663 (add_header_seq): Likewise.
8664 (interleave_supported_p): Likewise.
8665 (vect_maybe_permute_loop_masks): Likewise.
8666 (vect_set_loop_masks_directly): Likewise.
8667 (vect_set_loop_condition_masked): Likewise.
8668 (vect_set_loop_condition_unmasked): New function, split out from
8669 slpeel_make_loop_iterate_ntimes.
8670 (slpeel_make_loop_iterate_ntimes): Rename to..
8671 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8672 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8673 (vect_do_peeling): Update call accordingly.
8674 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8676 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8677 mask_compare_type, can_fully_mask_p and fully_masked_p.
8678 (release_vec_loop_masks): New function.
8679 (_loop_vec_info): Use it to free the loop masks.
8680 (can_produce_all_loop_masks_p): New function.
8681 (vect_get_max_nscalars_per_iter): Likewise.
8682 (vect_verify_full_masking): Likewise.
8683 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8684 retries, and free the mask rgroups before retrying. Check loop-wide
8685 reasons for disallowing fully-masked loops. Make the final decision
8686 about whether use a fully-masked loop or not.
8687 (vect_estimate_min_profitable_iters): Do not assume that peeling
8688 for the number of iterations will be needed for fully-masked loops.
8689 (vectorizable_reduction): Disable fully-masked loops.
8690 (vectorizable_live_operation): Likewise.
8691 (vect_halve_mask_nunits): New function.
8692 (vect_double_mask_nunits): Likewise.
8693 (vect_record_loop_mask): Likewise.
8694 (vect_get_loop_mask): Likewise.
8695 (vect_transform_loop): Handle the case in which the final loop
8696 iteration might handle a partial vector. Call vect_set_loop_condition
8697 instead of slpeel_make_loop_iterate_ntimes.
8698 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8699 (check_load_store_masking): New function.
8700 (prepare_load_store_mask): Likewise.
8701 (vectorizable_store): Handle fully-masked loops.
8702 (vectorizable_load): Likewise.
8703 (supportable_widening_operation): Use vect_halve_mask_nunits for
8705 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8706 (vect_gen_while): New function.
8707 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8708 (aarch64_uqdec<mode>): New insn.
8710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8711 Alan Hayward <alan.hayward@arm.com>
8712 David Sherwood <david.sherwood@arm.com>
8714 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8715 (reduc_xor_scal_optab): New optabs.
8716 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8717 (reduc_xor_scal_@var{m}): Document.
8718 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8719 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8721 * fold-const-call.c (fold_const_call): Handle them.
8722 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8723 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8724 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8725 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8726 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8727 (UNSPEC_XORV): New unspecs.
8728 (optab): Add entries for them.
8729 (BITWISEV): New int iterator.
8730 (bit_reduc_op): New int attributes.
8732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8733 Alan Hayward <alan.hayward@arm.com>
8734 David Sherwood <david.sherwood@arm.com>
8736 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8737 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8738 * optabs.def (vec_shl_insert_optab): New optab.
8739 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8740 (duplicate_and_interleave): Likewise.
8741 * tree-vect-loop.c: Include internal-fn.h.
8742 (neutral_op_for_slp_reduction): New function, split out from
8743 get_initial_defs_for_reduction.
8744 (get_initial_def_for_reduction): Handle option 2 for variable-length
8745 vectors by loading the neutral value into a vector and then shifting
8746 the initial value into element 0.
8747 (get_initial_defs_for_reduction): Replace the code argument with
8748 the neutral value calculated by neutral_op_for_slp_reduction.
8749 Use gimple_build_vector for constant-length vectors.
8750 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8751 but the first group_size elements have a neutral value.
8752 Use duplicate_and_interleave otherwise.
8753 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8754 Update call to get_initial_defs_for_reduction. Handle SLP
8755 reductions for variable-length vectors by creating one vector
8756 result for each scalar result, with the elements associated
8757 with other scalar results stubbed out with the neutral value.
8758 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8759 Require IFN_VEC_SHL_INSERT for double reductions on
8760 variable-length vectors, or SLP reductions that have
8761 a neutral value. Require can_duplicate_and_interleave_p
8762 support for variable-length unchained SLP reductions if there
8763 is no neutral value, such as for MIN/MAX reductions. Also require
8764 the number of vector elements to be a multiple of the number of
8765 SLP statements when doing variable-length unchained SLP reductions.
8766 Update call to vect_create_epilog_for_reduction.
8767 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8768 and remove initial values.
8769 (duplicate_and_interleave): Make public.
8770 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8771 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8774 Alan Hayward <alan.hayward@arm.com>
8775 David Sherwood <david.sherwood@arm.com>
8777 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8778 (can_duplicate_and_interleave_p): New function.
8779 (vect_get_and_check_slp_defs): Take the vector of statements
8780 rather than just the current one. Remove excess parentheses.
8781 Restriction rejectinon of vect_constant_def and vect_external_def
8782 for variable-length vectors to boolean types, or types for which
8783 can_duplicate_and_interleave_p is false.
8784 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8785 (duplicate_and_interleave): New function.
8786 (vect_get_constant_vectors): Use gimple_build_vector for
8787 constant-length vectors and suitable variable-length constant
8788 vectors. Use duplicate_and_interleave for other variable-length
8789 vectors. Don't defer the update when inserting new statements.
8791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8792 Alan Hayward <alan.hayward@arm.com>
8793 David Sherwood <david.sherwood@arm.com>
8795 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8796 min_profitable_iters doesn't go negative.
8798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8799 Alan Hayward <alan.hayward@arm.com>
8800 David Sherwood <david.sherwood@arm.com>
8802 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8803 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8804 * optabs.def (vec_mask_load_lanes_optab): New optab.
8805 (vec_mask_store_lanes_optab): Likewise.
8806 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8807 (MASK_STORE_LANES): Likewise.
8808 * internal-fn.c (mask_load_lanes_direct): New macro.
8809 (mask_store_lanes_direct): Likewise.
8810 (expand_mask_load_optab_fn): Handle masked operations.
8811 (expand_mask_load_lanes_optab_fn): New macro.
8812 (expand_mask_store_optab_fn): Handle masked operations.
8813 (expand_mask_store_lanes_optab_fn): New macro.
8814 (direct_mask_load_lanes_optab_supported_p): Likewise.
8815 (direct_mask_store_lanes_optab_supported_p): Likewise.
8816 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8818 (vect_load_lanes_supported): Likewise.
8819 * tree-vect-data-refs.c (strip_conversion): New function.
8820 (can_group_stmts_p): Likewise.
8821 (vect_analyze_data_ref_accesses): Use it instead of checking
8822 for a pair of assignments.
8823 (vect_store_lanes_supported): Take a masked_p parameter.
8824 (vect_load_lanes_supported): Likewise.
8825 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8826 vect_store_lanes_supported and vect_load_lanes_supported.
8827 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8828 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8829 parameter. Don't allow gaps for masked accesses.
8830 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8831 and vect_load_lanes_supported.
8832 (get_load_store_type): Take a masked_p parameter and update
8833 call to get_group_load_store_type.
8834 (vectorizable_store): Update call to get_load_store_type.
8835 Handle IFN_MASK_STORE_LANES.
8836 (vectorizable_load): Update call to get_load_store_type.
8837 Handle IFN_MASK_LOAD_LANES.
8839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8840 Alan Hayward <alan.hayward@arm.com>
8841 David Sherwood <david.sherwood@arm.com>
8843 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8845 * config/aarch64/aarch64-protos.h
8846 (aarch64_sve_struct_memory_operand_p): Declare.
8847 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8848 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8849 (VPRED, vpred): Handle SVE structure modes.
8850 * config/aarch64/constraints.md (Utx): New constraint.
8851 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8852 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8853 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8854 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8855 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8856 structure modes. Split into pieces after RA.
8857 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8858 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8860 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8861 SVE structure modes.
8862 (aarch64_classify_address): Likewise.
8863 (sizetochar): Move earlier in file.
8864 (aarch64_print_operand): Handle SVE register lists.
8865 (aarch64_array_mode): New function.
8866 (aarch64_sve_struct_memory_operand_p): Likewise.
8867 (TARGET_ARRAY_MODE): Redefine.
8869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8870 Alan Hayward <alan.hayward@arm.com>
8871 David Sherwood <david.sherwood@arm.com>
8873 * target.def (array_mode): New target hook.
8874 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8875 * doc/tm.texi: Regenerate.
8876 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8877 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8878 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8880 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8884 Alan Hayward <alan.hayward@arm.com>
8885 David Sherwood <david.sherwood@arm.com>
8887 * fold-const.c (fold_binary_loc): Check the argument types
8888 rather than the result type when testing for a vector operation.
8890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8892 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8893 * doc/tm.texi: Regenerate.
8895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8896 Alan Hayward <alan.hayward@arm.com>
8897 David Sherwood <david.sherwood@arm.com>
8899 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8900 (sve): Document new AArch64 extension.
8901 * doc/md.texi (w): Extend the description of the AArch64
8902 constraint to include SVE vectors.
8903 (Upl, Upa): Document new AArch64 predicate constraints.
8904 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8906 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8907 (msve-vector-bits=): New option.
8908 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8909 SVE when these are disabled.
8910 (sve): New extension.
8911 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8912 modes. Adjust their number of units based on aarch64_sve_vg.
8913 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8914 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8915 aarch64_addr_query_type.
8916 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8917 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8918 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8919 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8920 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8921 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8922 (aarch64_simd_imm_zero_p): Delete.
8923 (aarch64_check_zero_based_sve_index_immediate): Declare.
8924 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8925 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8926 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8927 (aarch64_sve_float_mul_immediate_p): Likewise.
8928 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8930 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8931 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8932 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8933 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8934 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8935 (aarch64_regmode_natural_size): Likewise.
8936 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8937 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8939 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8940 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8941 for VG and the SVE predicate registers.
8942 (V_ALIASES): Add a "z"-prefixed alias.
8943 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8944 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8945 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8946 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8947 (REG_CLASS_NAMES): Add entries for them.
8948 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8949 and the predicate registers.
8950 (aarch64_sve_vg): Declare.
8951 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8952 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8953 (REGMODE_NATURAL_SIZE): Define.
8954 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8956 * config/aarch64/aarch64.c: Include cfgrtl.h.
8957 (simd_immediate_info): Add a constructor for series vectors,
8958 and an associated step field.
8959 (aarch64_sve_vg): New variable.
8960 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8961 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8962 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8963 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8964 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8965 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8966 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8967 (aarch64_get_mask_mode): New functions.
8968 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8969 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8970 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8971 predicate modes and predicate registers. Explicitly restrict
8972 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8973 to store a vector mode if it is recognized by
8974 aarch64_classify_vector_mode.
8975 (aarch64_regmode_natural_size): New function.
8976 (aarch64_hard_regno_caller_save_mode): Return the original mode
8978 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8979 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8980 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8981 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8983 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8984 does not overlap dest if the function is frame-related. Handle
8986 (aarch64_split_add_offset): New function.
8987 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8988 them aarch64_add_offset.
8989 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8990 and update call to aarch64_sub_sp.
8991 (aarch64_add_cfa_expression): New function.
8992 (aarch64_expand_prologue): Pass extra temporary registers to the
8993 functions above. Handle the case in which we need to emit new
8994 DW_CFA_expressions for registers that were originally saved
8995 relative to the stack pointer, but now have to be expressed
8996 relative to the frame pointer.
8997 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8999 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9000 IP0 and IP1 values for SVE frames.
9001 (aarch64_expand_vec_series): New function.
9002 (aarch64_expand_sve_widened_duplicate): Likewise.
9003 (aarch64_expand_sve_const_vector): Likewise.
9004 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9005 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9006 into the register, rather than emitting a SET directly.
9007 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9008 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9009 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9010 (offset_9bit_signed_scaled_p): New functions.
9011 (aarch64_replicate_bitmask_imm): New function.
9012 (aarch64_bitmask_imm): Use it.
9013 (aarch64_cannot_force_const_mem): Reject expressions involving
9014 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9015 (aarch64_classify_index): Handle SVE indices, by requiring
9016 a plain register index with a scale that matches the element size.
9017 (aarch64_classify_address): Handle SVE addresses. Assert that
9018 the mode of the address is VOIDmode or an integer mode.
9019 Update call to aarch64_classify_symbol.
9020 (aarch64_classify_symbolic_expression): Update call to
9021 aarch64_classify_symbol.
9022 (aarch64_const_vec_all_in_range_p): New function.
9023 (aarch64_print_vector_float_operand): Likewise.
9024 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9025 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9026 and the FP immediates 1.0 and 0.5.
9027 (aarch64_print_address_internal): Handle SVE addresses.
9028 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9029 (aarch64_regno_regclass): Handle predicate registers.
9030 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9032 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9033 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9034 (aarch64_convert_sve_vector_bits): New function.
9035 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9036 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9038 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9039 Handle SVE vector and predicate modes. Accept VL-based constants
9040 that need only one temporary register, and VL offsets that require
9041 no temporary registers.
9042 (aarch64_conditional_register_usage): Mark the predicate registers
9043 as fixed if SVE isn't available.
9044 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9045 Return true for SVE vector and predicate modes.
9046 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9047 rather than an unsigned int. Handle SVE modes.
9048 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9050 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9052 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9053 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9054 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9055 (aarch64_sve_float_mul_immediate_p): New functions.
9056 (aarch64_sve_valid_immediate): New function.
9057 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9058 Explicitly reject structure modes. Check for INDEX constants.
9059 Handle PTRUE and PFALSE constants.
9060 (aarch64_check_zero_based_sve_index_immediate): New function.
9061 (aarch64_simd_imm_zero_p): Delete.
9062 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9063 vector modes. Accept constants in the range of CNT[BHWD].
9064 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9065 ask for an Advanced SIMD mode.
9066 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9067 (aarch64_simd_vector_alignment): Handle SVE predicates.
9068 (aarch64_vectorize_preferred_vector_alignment): New function.
9069 (aarch64_simd_vector_alignment_reachable): Use it instead of
9071 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9072 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9074 (MAX_VECT_LEN): Delete.
9075 (expand_vec_perm_d): Add a vec_flags field.
9076 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9077 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9078 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9080 (aarch64_evpc_rev): Rename to...
9081 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9082 (aarch64_evpc_rev_global): New function.
9083 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9084 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9086 (aarch64_evpc_sve_tbl): New function.
9087 (aarch64_expand_vec_perm_const_1): Update after rename of
9088 aarch64_evpc_rev. Handle SVE permutes too, trying
9089 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9090 than aarch64_evpc_tbl.
9091 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9092 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9093 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9094 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9095 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9096 (aarch64_expand_sve_vcond): New functions.
9097 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9098 of aarch64_vector_mode_p.
9099 (aarch64_dwarf_poly_indeterminate_value): New function.
9100 (aarch64_compute_pressure_classes): Likewise.
9101 (aarch64_can_change_mode_class): Likewise.
9102 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9103 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9104 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9105 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9106 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9107 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9108 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9109 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9111 (Dn, Dl, Dr): Accept const as well as const_vector.
9112 (Dz): Likewise. Compare against CONST0_RTX.
9113 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9114 of "vector" where appropriate.
9115 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9116 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9117 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9118 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9119 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9120 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9121 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9122 (v_int_equiv): Extend to SVE modes.
9123 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9125 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9126 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9127 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9128 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9129 (SVE_COND_FP_CMP): New int iterators.
9130 (perm_hilo): Handle the new unpack unspecs.
9131 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9133 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9134 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9135 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9136 (aarch64_equality_operator, aarch64_constant_vector_operand)
9137 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9138 (aarch64_sve_nonimmediate_operand): Likewise.
9139 (aarch64_sve_general_operand): Likewise.
9140 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9141 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9142 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9143 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9144 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9145 (aarch64_sve_float_arith_immediate): Likewise.
9146 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9147 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9148 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9149 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9150 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9151 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9152 (aarch64_sve_float_arith_operand): Likewise.
9153 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9154 (aarch64_sve_float_mul_operand): Likewise.
9155 (aarch64_sve_vec_perm_operand): Likewise.
9156 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9157 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9158 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9159 as well as const_vector.
9160 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9161 in file. Use CONST0_RTX and CONSTM1_RTX.
9162 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9163 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9164 Use aarch64_simd_imm_zero.
9165 * config/aarch64/aarch64-sve.md: New file.
9166 * config/aarch64/aarch64.md: Include it.
9167 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9168 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9169 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9170 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9171 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9172 (sve): New attribute.
9173 (enabled): Disable instructions with the sve attribute unless
9175 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9176 aarch64_expand_mov_immediate.
9177 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9178 CNT[BHSD] immediates.
9179 (movti): Split CONST_POLY_INT moves into two halves.
9180 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9181 Split additions that need a temporary here if the destination
9182 is the stack pointer.
9183 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9184 (*add<mode>3_poly_1): New instruction.
9185 (set_clobber_cc): New expander.
9187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9189 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9190 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9191 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9192 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9193 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9194 Change innermode from fixed_mode_size to machine_mode.
9195 (simplify_subreg): Update call accordingly. Handle a constant-sized
9196 subreg of a variable-length CONST_VECTOR.
9198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9199 Alan Hayward <alan.hayward@arm.com>
9200 David Sherwood <david.sherwood@arm.com>
9202 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9203 (add_offset_to_base): New function, split out from...
9204 (create_mem_ref): ...here. When handling a scale other than 1,
9205 check first whether the address is valid without the offset.
9206 Add it into the base if so, leaving the index and scale as-is.
9208 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9211 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9212 fold_for_warn before checking if arg2 is INTEGER_CST.
9214 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9216 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9217 (store_multiple_operation): Delete.
9218 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9219 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9220 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9221 guarded by TARGET_STRING.
9222 (rs6000_output_load_multiple): Delete.
9223 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9224 OPTION_MASK_STRING / TARGET_STRING handling.
9225 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9226 (const rs6000_opt_masks) <"string">: Change mask to 0.
9227 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9228 (MASK_STRING): Delete.
9229 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9231 (load_multiple): Delete.
9238 (store_multiple): Delete.
9245 (movmemsi_8reg): Delete.
9246 (corresponding unnamed define_insn): Delete.
9247 (movmemsi_6reg): Delete.
9248 (corresponding unnamed define_insn): Delete.
9249 (movmemsi_4reg): Delete.
9250 (corresponding unnamed define_insn): Delete.
9251 (movmemsi_2reg): Delete.
9252 (corresponding unnamed define_insn): Delete.
9253 (movmemsi_1reg): Delete.
9254 (corresponding unnamed define_insn): Delete.
9255 * config/rs6000/rs6000.opt (mno-string): New.
9256 (mstring): Replace by deprecation warning stub.
9257 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9259 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9261 * regrename.c (regrename_do_replace): If replacing the same
9262 reg multiple times, try to reuse last created gen_raw_REG.
9265 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9266 main to workaround a bug in GDB.
9268 2018-01-12 Tom de Vries <tom@codesourcery.com>
9271 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9273 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9275 PR rtl-optimization/80481
9276 * ira-color.c (get_cap_member): New function.
9277 (allocnos_conflict_by_live_ranges_p): Use it.
9278 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9279 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9281 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9284 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9285 (*saddl_se_1): Ditto.
9287 (*ssubl_se_1): Ditto.
9289 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9291 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9292 rather than wi::to_widest for DR_INITs.
9293 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9294 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9295 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9297 (vect_analyze_group_access_1): Note that here.
9299 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9301 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9302 polynomial type sizes.
9304 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9306 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9307 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9308 (gimple_add_tmp_var): Likewise.
9310 2018-01-12 Martin Liska <mliska@suse.cz>
9312 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9313 (gimple_alloc_sizes): Likewise.
9314 (dump_gimple_statistics): Use PRIu64 in printf format.
9315 * gimple.h: Change uint64_t to int.
9317 2018-01-12 Martin Liska <mliska@suse.cz>
9319 * tree-core.h: Use uint64_t instead of int.
9320 * tree.c (tree_node_counts): Likewise.
9321 (tree_node_sizes): Likewise.
9322 (dump_tree_statistics): Use PRIu64 in printf format.
9324 2018-01-12 Martin Liska <mliska@suse.cz>
9326 * Makefile.in: As qsort_chk is implemented in vec.c, add
9327 vec.o to linkage of gencfn-macros.
9328 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9329 passing the info to record_node_allocation_statistics.
9330 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9332 * ggc-common.c (struct ggc_usage): Add operator== and use
9333 it in operator< and compare function.
9334 * mem-stats.h (struct mem_usage): Likewise.
9335 * vec.c (struct vec_usage): Remove operator< and compare
9336 function. Can be simply inherited.
9338 2018-01-12 Martin Jambor <mjambor@suse.cz>
9341 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9342 * tree-ssa-math-opts.c: Include domwalk.h.
9343 (convert_mult_to_fma_1): New function.
9344 (fma_transformation_info): New type.
9345 (fma_deferring_state): Likewise.
9346 (cancel_fma_deferring): New function.
9347 (result_of_phi): Likewise.
9348 (last_fma_candidate_feeds_initial_phi): Likewise.
9349 (convert_mult_to_fma): Added deferring logic, split actual
9350 transformation to convert_mult_to_fma_1.
9351 (math_opts_dom_walker): New type.
9352 (math_opts_dom_walker::after_dom_children): New method, body moved
9353 here from pass_optimize_widening_mul::execute, added deferring logic
9355 (pass_optimize_widening_mul::execute): Moved most of code to
9356 math_opts_dom_walker::after_dom_children.
9357 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9358 * config/i386/i386.c (ix86_option_override_internal): Added
9359 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9361 2018-01-12 Richard Biener <rguenther@suse.de>
9364 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9365 inline instance vars.
9367 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9370 * config/rx/rx.c (rx_is_restricted_memory_address):
9373 2018-01-12 Richard Biener <rguenther@suse.de>
9375 PR tree-optimization/80846
9376 * target.def (split_reduction): New target hook.
9377 * targhooks.c (default_split_reduction): New function.
9378 * targhooks.h (default_split_reduction): Declare.
9379 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9380 target requests first reduce vectors by combining low and high
9382 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9383 (get_vectype_for_scalar_type_and_size): Export.
9384 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9385 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9386 * doc/tm.texi: Regenerate.
9387 * config/i386/i386.c (ix86_split_reduction): Implement
9388 TARGET_VECTORIZE_SPLIT_REDUCTION.
9390 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9393 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9394 in PIC mode except for TARGET_VXWORKS_RTP.
9395 * config/sparc/sparc.c: Include cfgrtl.h.
9396 (TARGET_INIT_PIC_REG): Define.
9397 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9398 (sparc_pic_register_p): New predicate.
9399 (sparc_legitimate_address_p): Use it.
9400 (sparc_legitimize_pic_address): Likewise.
9401 (sparc_delegitimize_address): Likewise.
9402 (sparc_mode_dependent_address_p): Likewise.
9403 (gen_load_pcrel_sym): Remove 4th parameter.
9404 (load_got_register): Adjust call to above. Remove obsolete stuff.
9405 (sparc_expand_prologue): Do not call load_got_register here.
9406 (sparc_flat_expand_prologue): Likewise.
9407 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9408 (sparc_use_pseudo_pic_reg): New function.
9409 (sparc_init_pic_reg): Likewise.
9410 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9411 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9413 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9415 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9416 Add item for branch_cost.
9418 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9420 PR rtl-optimization/83565
9421 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9422 not extend the result to a larger mode for rotate operations.
9423 (num_sign_bit_copies1): Likewise.
9425 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9428 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9430 Use values-Xc.o for -pedantic.
9431 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9433 2018-01-12 Martin Liska <mliska@suse.cz>
9436 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9438 (possible_polymorphic_call_targets): Use it.
9439 (ipa_devirt): Likewise.
9441 2018-01-12 Martin Liska <mliska@suse.cz>
9443 * profile-count.h (enum profile_quality): Use 0 as invalid
9444 enum value of profile_quality.
9446 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9448 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9449 -mext-string options.
9451 2018-01-12 Richard Biener <rguenther@suse.de>
9453 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9454 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9455 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9457 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9459 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9461 * configure.ac (--with-long-double-format): Add support for the
9462 configuration option to change the default long double format on
9464 * config.gcc (powerpc*-linux*-*): Likewise.
9465 * configure: Regenerate.
9466 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9467 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9468 used without modification.
9470 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9472 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9473 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9474 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9475 MISC_BUILTIN_SPEC_BARRIER.
9476 (rs6000_init_builtins): Likewise.
9477 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9479 (speculation_barrier): New define_insn.
9480 * doc/extend.texi: Document __builtin_speculation_barrier.
9482 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9485 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9486 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9487 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9489 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9490 integral modes instead of "ss" and "sd".
9491 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9492 vectors with 32-bit and 64-bit elements.
9493 (vecdupssescalarmodesuffix): New mode attribute.
9494 (vec_dup<mode>): Use it.
9496 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9499 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9500 frame if argument is passed on stack.
9502 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9505 * ree.c (combine_reaching_defs): Optimize also
9506 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9507 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9509 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9512 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9514 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9517 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9518 after they are computed.
9520 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9522 PR tree-optimization/83695
9523 * gimple-loop-linterchange.cc
9524 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9525 reset cached scev information after interchange.
9526 (pass_linterchange::execute): Remove call to scev_reset_htab.
9528 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9530 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9531 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9532 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9533 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9534 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9535 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9536 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9537 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9538 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9539 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9540 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9541 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9542 (V_lane_reg): Likewise.
9543 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9545 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9546 (vfmal_lane_low<mode>_intrinsic,
9547 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9548 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9549 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9550 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9551 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9552 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9554 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9556 * config/arm/arm-cpus.in (fp16fml): New feature.
9557 (ALL_SIMD): Add fp16fml.
9558 (armv8.2-a): Add fp16fml as an option.
9559 (armv8.3-a): Likewise.
9560 (armv8.4-a): Add fp16fml as part of fp16.
9561 * config/arm/arm.h (TARGET_FP16FML): Define.
9562 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9564 * config/arm/arm-modes.def (V2HF): Define.
9565 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9566 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9567 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9568 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9569 vfmsl_low, vfmsl_high): New set of builtins.
9570 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9571 (vfml_op): New code attribute.
9572 (VFMLHALVES): New int iterator.
9573 (VFML, VFMLSEL): New mode attributes.
9574 (V_reg): Define mapping for V2HF.
9575 (V_hi, V_lo): New mode attributes.
9576 (VF_constraint): Likewise.
9577 (vfml_half, vfml_half_selector): New int attributes.
9578 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9580 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9581 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9583 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9584 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9585 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9586 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9588 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9589 Document new effective target and option set.
9591 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9593 * config/arm/arm-cpus.in (armv8_4): New feature.
9594 (ARMv8_4a): New fgroup.
9595 (armv8.4-a): New arch.
9596 * config/arm/arm-tables.opt: Regenerate.
9597 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9598 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9599 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9600 Add matching rules for -march=armv8.4-a and extensions.
9601 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9603 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9606 * config/rx/rx.md (BW): New mode attribute.
9607 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9609 2018-01-11 Richard Biener <rguenther@suse.de>
9611 PR tree-optimization/83435
9612 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9613 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9614 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9616 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9617 Alan Hayward <alan.hayward@arm.com>
9618 David Sherwood <david.sherwood@arm.com>
9620 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9622 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9623 (aarch64_print_address_internal): Use it to check for a zero offset.
9625 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9626 Alan Hayward <alan.hayward@arm.com>
9627 David Sherwood <david.sherwood@arm.com>
9629 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9630 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9631 Return a poly_int64 rather than a HOST_WIDE_INT.
9632 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9633 rather than a HOST_WIDE_INT.
9634 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9635 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9636 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9637 final_offset from HOST_WIDE_INT to poly_int64.
9638 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9639 to_constant when getting the number of units in an Advanced SIMD
9641 (aarch64_builtin_vectorized_function): Check for a constant number
9643 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9645 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9646 attribute instead of GET_MODE_NUNITS.
9647 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9648 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9649 GET_MODE_SIZE for fixed-size registers.
9650 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9651 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9652 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9653 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9654 (aarch64_print_operand, aarch64_print_address_internal)
9655 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9656 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9657 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9658 Handle polynomial GET_MODE_SIZE.
9659 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9660 wider than SImode without modification.
9661 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9662 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9663 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9664 passing and returning SVE modes.
9665 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9666 rather than GEN_INT.
9667 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9668 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9669 (aarch64_allocate_and_probe_stack_space): Likewise.
9670 (aarch64_layout_frame): Cope with polynomial offsets.
9671 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9672 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9674 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9675 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9676 poly_int64 rather than a HOST_WIDE_INT.
9677 (aarch64_get_separate_components, aarch64_process_components)
9678 (aarch64_expand_prologue, aarch64_expand_epilogue)
9679 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9680 (aarch64_anchor_offset): New function, split out from...
9681 (aarch64_legitimize_address): ...here.
9682 (aarch64_builtin_vectorization_cost): Handle polynomial
9683 TYPE_VECTOR_SUBPARTS.
9684 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9686 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9687 number of elements from the PARALLEL rather than the mode.
9688 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9689 rather than GET_MODE_BITSIZE.
9690 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9691 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9692 (aarch64_expand_vec_perm_const_1): Handle polynomial
9693 d->perm.length () and d->perm elements.
9694 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9695 Apply to_constant to d->perm elements.
9696 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9697 polynomial CONST_VECTOR_NUNITS.
9698 (aarch64_move_pointer): Take amount as a poly_int64 rather
9700 (aarch64_progress_pointer): Avoid temporary variable.
9701 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9702 the mode attribute instead of GET_MODE.
9704 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9705 Alan Hayward <alan.hayward@arm.com>
9706 David Sherwood <david.sherwood@arm.com>
9708 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9709 x exists before using it.
9710 (aarch64_add_constant_internal): Rename to...
9711 (aarch64_add_offset_1): ...this. Replace regnum with separate
9712 src and dest rtxes. Handle the case in which they're different,
9713 including when the offset is zero. Replace scratchreg with an rtx.
9714 Use 2 additions if there is no spare register into which we can
9715 move a 16-bit constant.
9716 (aarch64_add_constant): Delete.
9717 (aarch64_add_offset): Replace reg with separate src and dest
9718 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9719 Use aarch64_add_offset_1.
9720 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9721 an rtx rather than an int. Take the delta as a poly_int64
9722 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9723 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9724 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9725 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9726 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9728 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9729 aarch64_add_constant.
9731 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9733 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9734 Use scalar_float_mode.
9736 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9738 * config/aarch64/aarch64-simd.md
9739 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9740 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9741 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9742 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9743 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9744 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9745 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9746 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9747 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9748 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9750 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9753 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9754 targ_options->x_arm_arch_string is non NULL.
9756 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9758 * config/aarch64/aarch64.h
9759 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9761 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9764 * expmed.c (emit_store_flag_force): Swap if const op0
9765 and change VOIDmode to mode of op0.
9767 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9769 PR rtl-optimization/83761
9770 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9771 than bytes to mode_for_size.
9773 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9776 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9777 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9780 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9783 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9784 when in layout mode.
9785 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9786 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9789 2018-01-10 Michael Collison <michael.collison@arm.com>
9791 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9792 * config/aarch64/aarch64-option-extension.def: Add
9793 AARCH64_OPT_EXTENSION of 'fp16fml'.
9794 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9795 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9796 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9797 * config/aarch64/constraints.md (Ui7): New constraint.
9798 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9799 (VFMLA_SEL_W): Ditto.
9802 (VFMLA16_LOW): New int iterator.
9803 (VFMLA16_HIGH): Ditto.
9804 (UNSPEC_FMLAL): New unspec.
9805 (UNSPEC_FMLSL): Ditto.
9806 (UNSPEC_FMLAL2): Ditto.
9807 (UNSPEC_FMLSL2): Ditto.
9808 (f16mac): New code attribute.
9809 * config/aarch64/aarch64-simd-builtins.def
9810 (aarch64_fmlal_lowv2sf): Ditto.
9811 (aarch64_fmlsl_lowv2sf): Ditto.
9812 (aarch64_fmlalq_lowv4sf): Ditto.
9813 (aarch64_fmlslq_lowv4sf): Ditto.
9814 (aarch64_fmlal_highv2sf): Ditto.
9815 (aarch64_fmlsl_highv2sf): Ditto.
9816 (aarch64_fmlalq_highv4sf): Ditto.
9817 (aarch64_fmlslq_highv4sf): Ditto.
9818 (aarch64_fmlal_lane_lowv2sf): Ditto.
9819 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9820 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9821 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9822 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9823 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9824 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9825 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9826 (aarch64_fmlal_lane_highv2sf): Ditto.
9827 (aarch64_fmlsl_lane_highv2sf): Ditto.
9828 (aarch64_fmlal_laneq_highv2sf): Ditto.
9829 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9830 (aarch64_fmlalq_lane_highv4sf): Ditto.
9831 (aarch64_fmlsl_lane_highv4sf): Ditto.
9832 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9833 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9834 * config/aarch64/aarch64-simd.md:
9835 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9836 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9837 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9838 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9839 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9840 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9841 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9842 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9843 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9844 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9845 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9846 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9847 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9848 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9849 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9850 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9851 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9852 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9853 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9854 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9855 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9856 (vfmlsl_low_u32): Ditto.
9857 (vfmlalq_low_u32): Ditto.
9858 (vfmlslq_low_u32): Ditto.
9859 (vfmlal_high_u32): Ditto.
9860 (vfmlsl_high_u32): Ditto.
9861 (vfmlalq_high_u32): Ditto.
9862 (vfmlslq_high_u32): Ditto.
9863 (vfmlal_lane_low_u32): Ditto.
9864 (vfmlsl_lane_low_u32): Ditto.
9865 (vfmlal_laneq_low_u32): Ditto.
9866 (vfmlsl_laneq_low_u32): Ditto.
9867 (vfmlalq_lane_low_u32): Ditto.
9868 (vfmlslq_lane_low_u32): Ditto.
9869 (vfmlalq_laneq_low_u32): Ditto.
9870 (vfmlslq_laneq_low_u32): Ditto.
9871 (vfmlal_lane_high_u32): Ditto.
9872 (vfmlsl_lane_high_u32): Ditto.
9873 (vfmlal_laneq_high_u32): Ditto.
9874 (vfmlsl_laneq_high_u32): Ditto.
9875 (vfmlalq_lane_high_u32): Ditto.
9876 (vfmlslq_lane_high_u32): Ditto.
9877 (vfmlalq_laneq_high_u32): Ditto.
9878 (vfmlslq_laneq_high_u32): Ditto.
9879 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9880 (AARCH64_FL_FOR_ARCH8_4): New.
9881 (AARCH64_ISA_F16FML): New ISA flag.
9882 (TARGET_F16FML): New feature flag for fp16fml.
9883 (doc/invoke.texi): Document new fp16fml option.
9885 2018-01-10 Michael Collison <michael.collison@arm.com>
9887 * config/aarch64/aarch64-builtins.c:
9888 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9889 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9890 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9891 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9892 (AARCH64_ISA_SHA3): New ISA flag.
9893 (TARGET_SHA3): New feature flag for sha3.
9894 * config/aarch64/iterators.md (sha512_op): New int attribute.
9895 (CRYPTO_SHA512): New int iterator.
9896 (UNSPEC_SHA512H): New unspec.
9897 (UNSPEC_SHA512H2): Ditto.
9898 (UNSPEC_SHA512SU0): Ditto.
9899 (UNSPEC_SHA512SU1): Ditto.
9900 * config/aarch64/aarch64-simd-builtins.def
9901 (aarch64_crypto_sha512hqv2di): New builtin.
9902 (aarch64_crypto_sha512h2qv2di): Ditto.
9903 (aarch64_crypto_sha512su0qv2di): Ditto.
9904 (aarch64_crypto_sha512su1qv2di): Ditto.
9905 (aarch64_eor3qv8hi): Ditto.
9906 (aarch64_rax1qv2di): Ditto.
9907 (aarch64_xarqv2di): Ditto.
9908 (aarch64_bcaxqv8hi): Ditto.
9909 * config/aarch64/aarch64-simd.md:
9910 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9911 (aarch64_crypto_sha512su0qv2di): Ditto.
9912 (aarch64_crypto_sha512su1qv2di): Ditto.
9913 (aarch64_eor3qv8hi): Ditto.
9914 (aarch64_rax1qv2di): Ditto.
9915 (aarch64_xarqv2di): Ditto.
9916 (aarch64_bcaxqv8hi): Ditto.
9917 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9918 (vsha512h2q_u64): Ditto.
9919 (vsha512su0q_u64): Ditto.
9920 (vsha512su1q_u64): Ditto.
9921 (veor3q_u16): Ditto.
9922 (vrax1q_u64): Ditto.
9924 (vbcaxq_u16): Ditto.
9925 * config/arm/types.md (crypto_sha512): New type attribute.
9926 (crypto_sha3): Ditto.
9927 (doc/invoke.texi): Document new sha3 option.
9929 2018-01-10 Michael Collison <michael.collison@arm.com>
9931 * config/aarch64/aarch64-builtins.c:
9932 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9933 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9934 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9935 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9936 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9937 (AARCH64_ISA_SM4): New ISA flag.
9938 (TARGET_SM4): New feature flag for sm4.
9939 * config/aarch64/aarch64-simd-builtins.def
9940 (aarch64_sm3ss1qv4si): Ditto.
9941 (aarch64_sm3tt1aq4si): Ditto.
9942 (aarch64_sm3tt1bq4si): Ditto.
9943 (aarch64_sm3tt2aq4si): Ditto.
9944 (aarch64_sm3tt2bq4si): Ditto.
9945 (aarch64_sm3partw1qv4si): Ditto.
9946 (aarch64_sm3partw2qv4si): Ditto.
9947 (aarch64_sm4eqv4si): Ditto.
9948 (aarch64_sm4ekeyqv4si): Ditto.
9949 * config/aarch64/aarch64-simd.md:
9950 (aarch64_sm3ss1qv4si): Ditto.
9951 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9952 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9953 (aarch64_sm4eqv4si): Ditto.
9954 (aarch64_sm4ekeyqv4si): Ditto.
9955 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9956 (sm3part_op): Ditto.
9957 (CRYPTO_SM3TT): Ditto.
9958 (CRYPTO_SM3PART): Ditto.
9959 (UNSPEC_SM3SS1): New unspec.
9960 (UNSPEC_SM3TT1A): Ditto.
9961 (UNSPEC_SM3TT1B): Ditto.
9962 (UNSPEC_SM3TT2A): Ditto.
9963 (UNSPEC_SM3TT2B): Ditto.
9964 (UNSPEC_SM3PARTW1): Ditto.
9965 (UNSPEC_SM3PARTW2): Ditto.
9966 (UNSPEC_SM4E): Ditto.
9967 (UNSPEC_SM4EKEY): Ditto.
9968 * config/aarch64/constraints.md (Ui2): New constraint.
9969 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9970 * config/arm/types.md (crypto_sm3): New type attribute.
9971 (crypto_sm4): Ditto.
9972 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9973 (vsm3tt1aq_u32): Ditto.
9974 (vsm3tt1bq_u32): Ditto.
9975 (vsm3tt2aq_u32): Ditto.
9976 (vsm3tt2bq_u32): Ditto.
9977 (vsm3partw1q_u32): Ditto.
9978 (vsm3partw2q_u32): Ditto.
9979 (vsm4eq_u32): Ditto.
9980 (vsm4ekeyq_u32): Ditto.
9981 (doc/invoke.texi): Document new sm4 option.
9983 2018-01-10 Michael Collison <michael.collison@arm.com>
9985 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9986 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9987 (AARCH64_FL_FOR_ARCH8_4): New.
9988 (AARCH64_FL_V8_4): New flag.
9989 (doc/invoke.texi): Document new armv8.4-a option.
9991 2018-01-10 Michael Collison <michael.collison@arm.com>
9993 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9994 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9995 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9996 * config/aarch64/aarch64-option-extension.def: Add
9997 AARCH64_OPT_EXTENSION of 'sha2'.
9998 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9999 (crypto): Disable sha2 and aes if crypto disabled.
10000 (crypto): Enable aes and sha2 if enabled.
10001 (simd): Disable sha2 and aes if simd disabled.
10002 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10004 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10005 (TARGET_SHA2): New feature flag for sha2.
10006 (TARGET_AES): New feature flag for aes.
10007 * config/aarch64/aarch64-simd.md:
10008 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10009 conditional on TARGET_AES.
10010 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10011 (aarch64_crypto_sha1hsi): Make pattern conditional
10013 (aarch64_crypto_sha1hv4si): Ditto.
10014 (aarch64_be_crypto_sha1hv4si): Ditto.
10015 (aarch64_crypto_sha1su1v4si): Ditto.
10016 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10017 (aarch64_crypto_sha1su0v4si): Ditto.
10018 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10019 (aarch64_crypto_sha256su0v4si): Ditto.
10020 (aarch64_crypto_sha256su1v4si): Ditto.
10021 (doc/invoke.texi): Document new aes and sha2 options.
10023 2018-01-10 Martin Sebor <msebor@redhat.com>
10025 PR tree-optimization/83781
10026 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10029 2018-01-11 Martin Sebor <msebor@gmail.com>
10030 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10032 PR tree-optimization/83501
10033 PR tree-optimization/81703
10035 * tree-ssa-strlen.c (get_string_cst): Rename...
10036 (get_string_len): ...to this. Handle global constants.
10037 (handle_char_store): Adjust.
10039 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10040 Jim Wilson <jimw@sifive.com>
10042 * config/riscv/riscv-protos.h (riscv_output_return): New.
10043 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10044 (riscv_attribute_table, riscv_output_return),
10045 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10046 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10047 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10049 (riscv_expand_prologue): Add early return for naked function.
10050 (riscv_expand_epilogue): Likewise.
10051 (riscv_function_ok_for_sibcall): Return false for naked function.
10052 (riscv_set_current_function): New.
10053 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10054 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10055 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10056 * doc/extend.texi (RISC-V Function Attributes): New.
10058 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10060 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10061 check for 128-bit long double before checking TCmode.
10062 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10063 128-bit long doubles before checking TFmode or TCmode.
10064 (FLOAT128_IBM_P): Likewise.
10066 2018-01-10 Martin Sebor <msebor@redhat.com>
10068 PR tree-optimization/83671
10069 * builtins.c (c_strlen): Unconditionally return zero for the empty
10071 Use -Warray-bounds for warnings.
10072 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10073 for non-constant array indices with COMPONENT_REF, arrays of
10074 arrays, and pointers to arrays.
10075 (gimple_fold_builtin_strlen): Determine and set length range for
10076 non-constant character arrays.
10078 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10080 PR middle-end/81897
10081 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10084 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10086 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10088 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10091 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10092 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10093 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10094 indexed_or_indirect_operand predicate.
10095 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10096 (*vsx_le_perm_load_v8hi): Likewise.
10097 (*vsx_le_perm_load_v16qi): Likewise.
10098 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10099 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10100 (*vsx_le_perm_store_v8hi): Likewise.
10101 (*vsx_le_perm_store_v16qi): Likewise.
10102 (eight unnamed splitters): Likewise.
10104 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10106 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10107 * config/rs6000/emmintrin.h: Likewise.
10108 * config/rs6000/mmintrin.h: Likewise.
10109 * config/rs6000/xmmintrin.h: Likewise.
10111 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10114 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10116 * tree.c (tree_nop_conversion): Return true for location wrapper
10118 (maybe_wrap_with_location): New function.
10119 (selftest::check_strip_nops): New function.
10120 (selftest::test_location_wrappers): New function.
10121 (selftest::tree_c_tests): Call it.
10122 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10123 (maybe_wrap_with_location): New decl.
10124 (EXPR_LOCATION_WRAPPER_P): New macro.
10125 (location_wrapper_p): New inline function.
10126 (tree_strip_any_location_wrapper): New inline function.
10128 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10131 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10132 stack_realign_offset for the largest alignment of stack slot
10134 (ix86_find_max_used_stack_alignment): New function.
10135 (ix86_finalize_stack_frame_flags): Use it. Set
10136 max_used_stack_alignment if we don't realign stack.
10137 * config/i386/i386.h (machine_function): Add
10138 max_used_stack_alignment.
10140 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10142 * config/arm/arm.opt (-mbranch-cost): New option.
10143 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10146 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10149 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10150 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10152 2018-01-10 Richard Biener <rguenther@suse.de>
10155 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10156 early out so it also covers the case where we have a non-NULL
10159 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10161 PR tree-optimization/83753
10162 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10163 for non-strided grouped accesses if the number of elements is 1.
10165 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10168 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10169 * i386.h (TARGET_USE_GATHER): Define.
10170 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10172 2018-01-10 Martin Liska <mliska@suse.cz>
10175 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10176 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10178 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10179 CLEANUP_NO_PARTITIONING is not set.
10181 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10183 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10184 for vectors, as a partial revert of r254296.
10185 * rtl.h (const_vec_p): Delete.
10186 (const_vec_duplicate_p): Don't test for vector CONSTs.
10187 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10188 * expmed.c (make_tree): Likewise.
10191 * common.md (E, F): Use CONSTANT_P instead of checking for
10193 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10194 checking for CONST_VECTOR.
10196 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10198 PR middle-end/83575
10199 * predict.c (force_edge_cold): Handle in more sane way edges
10200 with no prediction.
10202 2018-01-09 Carl Love <cel@us.ibm.com>
10204 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10206 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10207 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10208 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10209 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10210 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10211 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10212 * config/rs6000/rs6000-protos.h: Add extern defition for
10213 rs6000_generate_float2_double_code.
10214 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10216 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10217 (float2_v2df): Add define_expand.
10219 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10222 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10223 op_mode in the force_to_mode call.
10225 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10227 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10228 instead of checking each element individually.
10229 (aarch64_evpc_uzp): Likewise.
10230 (aarch64_evpc_zip): Likewise.
10231 (aarch64_evpc_ext): Likewise.
10232 (aarch64_evpc_rev): Likewise.
10233 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10234 instead of checking each element individually. Return true without
10236 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10237 whether all selected elements come from the same input, instead of
10238 checking each element individually. Remove calls to gen_rtx_REG,
10239 start_sequence and end_sequence and instead assert that no rtl is
10242 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10244 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10245 order of HIGH and CONST checks.
10247 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10249 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10250 if the destination isn't an SSA_NAME.
10252 2018-01-09 Richard Biener <rguenther@suse.de>
10254 PR tree-optimization/83668
10255 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10257 (canonicalize_loop_form): ... here, renamed from ...
10258 (canonicalize_loop_closed_ssa_form): ... this and amended to
10259 swap successor edges for loop exit blocks to make us use
10260 the RPO order we need for initial schedule generation.
10262 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10264 PR tree-optimization/64811
10265 * match.pd: When optimizing comparisons with Inf, avoid
10266 introducing or losing exceptions from comparisons with NaN.
10268 2018-01-09 Martin Liska <mliska@suse.cz>
10271 * asan.c (shadow_mem_size): Add gcc_assert.
10273 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10275 Don't save registers in main().
10278 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10279 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10280 * config/avr/avr.c (avr_set_current_function): Don't error if
10281 naked, OS_task or OS_main are specified at the same time.
10282 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10284 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10286 * common/config/avr/avr-common.c (avr_option_optimization_table):
10287 Switch on -mmain-is-OS_task for optimizing compilations.
10289 2018-01-09 Richard Biener <rguenther@suse.de>
10291 PR tree-optimization/83572
10292 * graphite.c: Include cfganal.h.
10293 (graphite_transform_loops): Connect infinite loops to exit
10294 and remove fake edges at the end.
10296 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10298 * ipa-inline.c (edge_badness): Revert accidental checkin.
10300 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10303 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10304 symbols; not inline clones.
10306 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10309 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10310 hard registers. Formatting fixes.
10312 PR preprocessor/83722
10313 * gcc.c (try_generate_repro): Pass
10314 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10315 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10318 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10319 Kito Cheng <kito.cheng@gmail.com>
10321 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10322 (riscv_leaf_function_p): Delete.
10323 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10325 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10327 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10329 (do_ifelse): New function.
10330 (do_isel): New function.
10331 (do_sub3): New function.
10332 (do_add3): New function.
10333 (do_load_mask_compare): New function.
10334 (do_overlap_load_compare): New function.
10335 (expand_compare_loop): New function.
10336 (expand_block_compare): Call expand_compare_loop() when appropriate.
10337 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10338 option description.
10339 (-mblock-compare-inline-loop-limit): New option.
10341 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10344 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10345 Reverse order of second and third operands in first alternative.
10346 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10347 of first and second elements in UNSPEC_VPERMR vector.
10348 (altivec_expand_vec_perm_le): Likewise.
10350 2018-01-08 Jeff Law <law@redhat.com>
10352 PR rtl-optimizatin/81308
10353 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10354 (process_switch): If group_case_labels makes a change, then set
10356 (pass_convert_switch::execute): If a switch is converted, then
10357 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10359 PR rtl-optimization/81308
10360 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10363 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10365 PR target/83663 - Revert r255946
10366 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10367 generation for cases where splatting a value is not useful.
10368 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10369 across a vec_duplicate and a paradoxical subreg forming a vector
10370 mode to a vec_concat.
10372 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10374 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10375 -march=armv8.3-a variants.
10376 * config/arm/t-multilib: Likewise.
10377 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10379 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10381 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10383 (cceq_ior_compare_complement): Give it a name so I can use it, and
10384 change boolean_or_operator predicate to boolean_operator so it can
10385 be used to generate a crand.
10386 (eqne): New code iterator.
10387 (bd/bd_neg): New code_attrs.
10388 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10389 a single define_insn.
10390 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10391 decrement (bdnzt/bdnzf/bdzt/bdzf).
10392 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10393 with the new names of the branch decrement patterns, and added the
10394 names of the branch decrement conditional patterns.
10396 2018-01-08 Richard Biener <rguenther@suse.de>
10398 PR tree-optimization/83563
10399 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10402 2018-01-08 Richard Biener <rguenther@suse.de>
10404 PR middle-end/83713
10405 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10407 2018-01-08 Richard Biener <rguenther@suse.de>
10409 PR tree-optimization/83685
10410 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10411 references to abnormals.
10413 2018-01-08 Richard Biener <rguenther@suse.de>
10416 * dwarf2out.c (output_indirect_strings): Handle empty
10417 skeleton_debug_str_hash.
10418 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10420 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10422 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10423 (emit_store_direct): Likewise.
10424 (arc_trampoline_adjust_address): Likewise.
10425 (arc_asm_trampoline_template): New function.
10426 (arc_initialize_trampoline): Use asm_trampoline_template.
10427 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10428 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10429 * config/arc/arc.md (flush_icache): Delete pattern.
10431 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10433 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10434 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10437 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10440 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10441 by not USED_FOR_TARGET.
10442 (make_pass_resolve_sw_modes): Likewise.
10444 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10446 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10449 2018-01-08 Richard Biener <rguenther@suse.de>
10451 PR middle-end/83580
10452 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10454 2018-01-08 Richard Biener <rguenther@suse.de>
10456 PR middle-end/83517
10457 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10459 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10461 PR middle-end/81897
10462 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10463 basic blocks with a small number of successors.
10464 (convert_control_dep_chain_into_preds): Improve handling of
10466 (dump_predicates): Split apart into...
10467 (dump_pred_chain): ...here...
10468 (dump_pred_info): ...and here.
10469 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10470 (can_chain_union_be_invalidated_p): Improve check for invalidation
10472 (uninit_uses_cannot_happen): Avoid unnecessary if
10473 convert_control_dep_chain_into_preds yielded nothing.
10475 2018-01-06 Martin Sebor <msebor@redhat.com>
10477 PR tree-optimization/83640
10478 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10479 subtracting negative offset from size.
10480 (builtin_access::overlap): Adjust offset bounds of the access to fall
10481 within the size of the object if possible.
10483 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10485 PR rtl-optimization/83699
10486 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10487 extract_bit_field_as_subreg to cases in which the extracted
10488 value is also a vector.
10490 * lra-constraints.c (process_alt_operands): Test for the equivalence
10491 substitutions when detecting a possible reload cycle.
10493 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10496 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10497 by default if flag_selective_schedling{,2}. Formatting fixes.
10499 PR rtl-optimization/83682
10500 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10501 if it has non-VECTOR_MODE element mode.
10502 (vec_duplicate_p): Likewise.
10504 PR middle-end/83694
10505 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10506 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10508 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10511 * config/i386/i386-builtin.def
10512 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10513 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10514 Require also OPTION_MASK_ISA_AVX512F in addition to
10515 OPTION_MASK_ISA_GFNI.
10516 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10517 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10518 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10519 to OPTION_MASK_ISA_GFNI.
10520 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10521 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10522 OPTION_MASK_ISA_AVX512BW.
10523 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10524 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10525 addition to OPTION_MASK_ISA_GFNI.
10526 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10527 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10528 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10529 to OPTION_MASK_ISA_GFNI.
10530 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10531 a requirement for all ISAs rather than any of them with a few
10533 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10535 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10536 bitmasks to be enabled with 3 exceptions, instead of requiring any
10537 enabled ISA with lots of exceptions.
10538 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10539 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10540 Change avx512bw in isa attribute to avx512f.
10541 * config/i386/sgxintrin.h: Add license boilerplate.
10542 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10543 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10544 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10545 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10547 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10548 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10549 temporarily sse2 rather than sse if not enabled already.
10552 * config/i386/sse.md (VI248_VLBW): Rename to ...
10553 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10554 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10555 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10556 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10557 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10558 mode iterator instead of VI248_VLBW.
10560 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10562 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10563 (record_modified): Skip clobbers; add debug output.
10564 (param_change_prob): Use sreal frequencies.
10566 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10568 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10569 punt for user-aligned variables.
10571 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10573 * tree-chrec.c (chrec_contains_symbols): Return true for
10576 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10579 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10580 of (x|y) == x for BICS pattern.
10582 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10584 PR tree-optimization/83605
10585 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10586 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10589 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10591 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10592 * config/epiphany/rtems.h: New file.
10594 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10595 Uros Bizjak <ubizjak@gmail.com>
10598 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10599 QIreg_operand instead of register_operand predicate.
10600 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10601 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10602 comments instead of -fmitigate[-_]rop.
10604 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10607 * cgraphunit.c (symbol_table::compile): Switch to text_section
10608 before calling assembly_start debug hook.
10609 * run-rtl-passes.c (run_rtl_passes): Likewise.
10612 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10614 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10615 range_int_cst_p rather than !symbolic_range_p before calling
10616 extract_range_from_multiplicative_op_1.
10618 2018-01-04 Jeff Law <law@redhat.com>
10620 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10621 redundant test in assertion.
10623 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10625 * doc/rtl.texi: Document machine_mode wrapper classes.
10627 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10629 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10630 using tree_to_uhwi.
10632 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10634 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10635 the VEC_PERM_EXPR fold to fail.
10637 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10640 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10641 to switched_sections.
10643 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10646 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10647 test for d.testing.
10649 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10652 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10653 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10655 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10658 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10659 is BLKmode and bitpos not zero or mode change is needed.
10661 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10664 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10667 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10670 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10671 instead of MULT rtx. Update all corresponding splitters.
10672 (*saddl_se): Ditto.
10673 (*ssub<modesuffix>): Ditto.
10674 (*ssubl_se): Ditto.
10675 (*cmp_sadd_di): Update split patterns.
10676 (*cmp_sadd_si): Ditto.
10677 (*cmp_sadd_sidi): Ditto.
10678 (*cmp_ssub_di): Ditto.
10679 (*cmp_ssub_si): Ditto.
10680 (*cmp_ssub_sidi): Ditto.
10681 * config/alpha/predicates.md (const23_operand): New predicate.
10682 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10683 Look for ASHIFT, not MULT inner operand.
10684 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10686 2018-01-04 Martin Liska <mliska@suse.cz>
10688 PR gcov-profile/83669
10689 * gcov.c (output_intermediate_file): Add version to intermediate
10691 * doc/gcov.texi: Document new field 'version' in intermediate
10692 file format. Fix location of '-k' option of gcov command.
10694 2018-01-04 Martin Liska <mliska@suse.cz>
10697 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10699 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10701 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10703 2018-01-03 Martin Sebor <msebor@redhat.com>
10705 PR tree-optimization/83655
10706 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10707 checking calls with invalid arguments.
10709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10711 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10712 (vectorizable_mask_load_store): Delete.
10713 (vectorizable_call): Return false for masked loads and stores.
10714 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10715 instead of gimple_assign_rhs1.
10716 (vectorizable_load): Handle IFN_MASK_LOAD.
10717 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10721 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10723 (vectorizable_mask_load_store): ...here.
10724 (vectorizable_load): ...and here.
10726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10728 * tree-vect-stmts.c (vect_build_all_ones_mask)
10729 (vect_build_zero_merge_argument): New functions, split out from...
10730 (vectorizable_load): ...here.
10732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10734 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10736 (vectorizable_mask_load_store): ...here.
10737 (vectorizable_store): ...and here.
10739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10741 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10743 (vectorizable_mask_load_store): ...here.
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10747 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10748 (vect_model_store_cost): Take a vec_load_store_type instead of a
10750 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10751 (vect_model_store_cost): Take a vec_load_store_type instead of a
10753 (vectorizable_mask_load_store): Update accordingly.
10754 (vectorizable_store): Likewise.
10755 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10759 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10760 IFN_MASK_LOAD calls here rather than...
10761 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10764 Alan Hayward <alan.hayward@arm.com>
10765 David Sherwood <david.sherwood@arm.com>
10767 * expmed.c (extract_bit_field_1): For vector extracts,
10768 fall back to extract_bit_field_as_subreg if vec_extract
10771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10772 Alan Hayward <alan.hayward@arm.com>
10773 David Sherwood <david.sherwood@arm.com>
10775 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10776 they are variable or constant sized.
10777 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10778 slots for constant-sized data.
10780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10781 Alan Hayward <alan.hayward@arm.com>
10782 David Sherwood <david.sherwood@arm.com>
10784 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10785 handling COND_EXPRs with boolean comparisons, try to find a better
10786 basis for the mask type than the boolean itself.
10788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10790 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10791 is calculated and how it can be overridden.
10792 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10793 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10795 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10799 Alan Hayward <alan.hayward@arm.com>
10800 David Sherwood <david.sherwood@arm.com>
10802 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10803 Remove the mode argument.
10804 (aarch64_simd_valid_immediate): Remove the mode and inverse
10806 * config/aarch64/iterators.md (bitsize): New iterator.
10807 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10808 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10809 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10810 aarch64_simd_valid_immediate.
10811 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10812 (aarch64_reg_or_bic_imm): Likewise.
10813 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10814 with an insn_type enum and msl with a modifier_type enum.
10815 Replace element_width with a scalar_mode. Change the shift
10816 to unsigned int. Add constructors for scalar_float_mode and
10817 scalar_int_mode elements.
10818 (aarch64_vect_float_const_representable_p): Delete.
10819 (aarch64_can_const_movi_rtx_p)
10820 (aarch64_simd_scalar_immediate_valid_for_move)
10821 (aarch64_simd_make_constant): Update call to
10822 aarch64_simd_valid_immediate.
10823 (aarch64_advsimd_valid_immediate_hs): New function.
10824 (aarch64_advsimd_valid_immediate): Likewise.
10825 (aarch64_simd_valid_immediate): Remove mode and inverse
10826 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10827 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10828 and aarch64_float_const_representable_p on the result.
10829 (aarch64_output_simd_mov_immediate): Remove mode argument.
10830 Update call to aarch64_simd_valid_immediate and use of
10831 simd_immediate_info.
10832 (aarch64_output_scalar_simd_mov_immediate): Update call
10835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10836 Alan Hayward <alan.hayward@arm.com>
10837 David Sherwood <david.sherwood@arm.com>
10839 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10840 (mode_nunits): Likewise CONST_MODE_NUNITS.
10841 * machmode.def (ADJUST_NUNITS): Document.
10842 * genmodes.c (mode_data::need_nunits_adj): New field.
10843 (blank_mode): Update accordingly.
10844 (adj_nunits): New variable.
10845 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10847 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10848 listed in adj_nunits.
10849 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10850 listed in adj_nunits. Don't emit case statements for such modes.
10851 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10852 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10853 nothing if adj_nunits is nonnull.
10854 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10855 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10856 (emit_mode_fbit): Update use of print_maybe_const_decl.
10857 (emit_move_size): Likewise. Treat the array as non-const
10859 (emit_mode_adjustments): Handle adj_nunits.
10861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10863 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10864 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10865 (VECTOR_MODES): Use it.
10866 (make_vector_modes): Take the prefix as an argument.
10868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10869 Alan Hayward <alan.hayward@arm.com>
10870 David Sherwood <david.sherwood@arm.com>
10872 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10873 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10874 for MODE_VECTOR_BOOL.
10875 * machmode.def (VECTOR_BOOL_MODE): Document.
10876 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10877 (make_vector_bool_mode): New function.
10878 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10880 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10881 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10883 * stor-layout.c (int_mode_for_mode): Likewise.
10884 * tree.c (build_vector_type_for_mode): Likewise.
10885 * varasm.c (output_constant_pool_2): Likewise.
10886 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10887 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10888 for MODE_VECTOR_BOOL.
10889 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10890 of mode class checks.
10891 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10892 instead of a list of mode class checks.
10893 (expand_vector_scalar_condition): Likewise.
10894 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10897 Alan Hayward <alan.hayward@arm.com>
10898 David Sherwood <david.sherwood@arm.com>
10900 * machmode.h (mode_size): Change from unsigned short to
10902 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10903 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10904 or if measurement_type is not polynomial.
10905 (fixed_size_mode::includes_p): Check for constant-sized modes.
10906 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10907 return a poly_uint16 rather than an unsigned short.
10908 (emit_mode_size): Change the type of mode_size from unsigned short
10909 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10910 (emit_mode_adjustments): Cope with polynomial vector sizes.
10911 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10913 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10915 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10916 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10917 * caller-save.c (setup_save_areas): Likewise.
10918 (replace_reg_with_saved_mem): Likewise.
10919 * calls.c (emit_library_call_value_1): Likewise.
10920 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10921 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10922 (gen_lowpart_for_combine): Likewise.
10923 * convert.c (convert_to_integer_1): Likewise.
10924 * cse.c (equiv_constant, cse_insn): Likewise.
10925 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10926 (cselib_subst_to_values): Likewise.
10927 * dce.c (word_dce_process_block): Likewise.
10928 * df-problems.c (df_word_lr_mark_ref): Likewise.
10929 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10930 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10931 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10932 (rtl_for_decl_location): Likewise.
10933 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10934 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10935 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10936 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10937 (expand_expr_real_1): Likewise.
10938 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10939 (pad_below): Likewise.
10940 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10941 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10942 * ira.c (get_subreg_tracking_sizes): Likewise.
10943 * ira-build.c (ira_create_allocno_objects): Likewise.
10944 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10945 (ira_sort_regnos_for_alter_reg): Likewise.
10946 * ira-costs.c (record_operand_costs): Likewise.
10947 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10948 (resolve_simple_move): Likewise.
10949 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10950 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10951 (lra_constraints): Likewise.
10952 (CONST_POOL_OK_P): Reject variable-sized modes.
10953 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10954 (add_pseudo_to_slot, lra_spill): Likewise.
10955 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10956 * optabs-query.c (get_best_extraction_insn): Likewise.
10957 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10958 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10959 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10960 * recog.c (offsettable_address_addr_space_p): Likewise.
10961 * regcprop.c (maybe_mode_change): Likewise.
10962 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10963 * regrename.c (build_def_use): Likewise.
10964 * regstat.c (dump_reg_info): Likewise.
10965 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10966 (find_reloads, find_reloads_subreg_address): Likewise.
10967 * reload1.c (eliminate_regs_1): Likewise.
10968 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10969 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10970 (simplify_binary_operation_1, simplify_subreg): Likewise.
10971 * targhooks.c (default_function_arg_padding): Likewise.
10972 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10973 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10974 (verify_gimple_assign_ternary): Likewise.
10975 * tree-inline.c (estimate_move_cost): Likewise.
10976 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10977 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10978 (get_address_cost_ainc): Likewise.
10979 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10980 (vect_supportable_dr_alignment): Likewise.
10981 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10982 (vectorizable_reduction): Likewise.
10983 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10984 (vectorizable_operation, vectorizable_load): Likewise.
10985 * tree.c (build_same_sized_truth_vector_type): Likewise.
10986 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10987 * var-tracking.c (emit_note_insn_var_location): Likewise.
10988 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10989 (ADDR_VEC_ALIGN): Likewise.
10991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10992 Alan Hayward <alan.hayward@arm.com>
10993 David Sherwood <david.sherwood@arm.com>
10995 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10997 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10998 or if measurement_type is polynomial.
10999 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11000 * combine.c (make_extraction): Likewise.
11001 * dse.c (find_shift_sequence): Likewise.
11002 * dwarf2out.c (mem_loc_descriptor): Likewise.
11003 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11004 (extract_bit_field, extract_low_bits): Likewise.
11005 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11006 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11007 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11008 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11009 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11010 * reload.c (find_reloads): Likewise.
11011 * reload1.c (alter_reg): Likewise.
11012 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11013 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11014 * tree-if-conv.c (predicate_mem_writes): Likewise.
11015 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11016 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11017 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11018 * valtrack.c (dead_debug_insert_temp): Likewise.
11019 * varasm.c (mergeable_constant_section): Likewise.
11020 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11023 Alan Hayward <alan.hayward@arm.com>
11024 David Sherwood <david.sherwood@arm.com>
11026 * expr.c (expand_assignment): Cope with polynomial mode sizes
11027 when assigning to a CONCAT.
11029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11030 Alan Hayward <alan.hayward@arm.com>
11031 David Sherwood <david.sherwood@arm.com>
11033 * machmode.h (mode_precision): Change from unsigned short to
11035 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11037 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11038 or if measurement_type is not polynomial.
11039 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11040 in which the mode is already known to be a scalar_int_mode.
11041 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11042 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11044 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11045 for GET_MODE_PRECISION.
11046 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11047 for GET_MODE_PRECISION.
11048 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11050 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11051 (expand_field_assignment, make_extraction): Likewise.
11052 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11053 (get_last_value): Likewise.
11054 * convert.c (convert_to_integer_1): Likewise.
11055 * cse.c (cse_insn): Likewise.
11056 * expr.c (expand_expr_real_1): Likewise.
11057 * lra-constraints.c (simplify_operand_subreg): Likewise.
11058 * optabs-query.c (can_atomic_load_p): Likewise.
11059 * optabs.c (expand_atomic_load): Likewise.
11060 (expand_atomic_store): Likewise.
11061 * ree.c (combine_reaching_defs): Likewise.
11062 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11063 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11064 * tree.h (type_has_mode_precision_p): Likewise.
11065 * ubsan.c (instrument_si_overflow): Likewise.
11067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11068 Alan Hayward <alan.hayward@arm.com>
11069 David Sherwood <david.sherwood@arm.com>
11071 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11072 polynomial numbers of units.
11073 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11074 (valid_vector_subparts_p): New function.
11075 (build_vector_type): Remove temporary shim and take the number
11076 of units as a poly_uint64 rather than an int.
11077 (build_opaque_vector_type): Take the number of units as a
11078 poly_uint64 rather than an int.
11079 * tree.c (build_vector_from_ctor): Handle polynomial
11080 TYPE_VECTOR_SUBPARTS.
11081 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11082 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11083 (build_vector_from_val): If the number of units is variable,
11084 use build_vec_duplicate_cst for constant operands and
11085 VEC_DUPLICATE_EXPR otherwise.
11086 (make_vector_type): Remove temporary is_constant ().
11087 (build_vector_type, build_opaque_vector_type): Take the number of
11088 units as a poly_uint64 rather than an int.
11089 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11091 * cfgexpand.c (expand_debug_expr): Likewise.
11092 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11093 (store_constructor, expand_expr_real_1): Likewise.
11094 (const_scalar_mask_from_tree): Likewise.
11095 * fold-const-call.c (fold_const_reduction): Likewise.
11096 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11097 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11098 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11099 (fold_relational_const): Likewise.
11100 (native_interpret_vector): Likewise. Change the size from an
11101 int to an unsigned int.
11102 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11103 TYPE_VECTOR_SUBPARTS.
11104 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11105 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11106 duplicating a non-constant operand into a variable-length vector.
11107 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11108 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11109 * ipa-icf.c (sem_variable::equals): Likewise.
11110 * match.pd: Likewise.
11111 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11112 * print-tree.c (print_node): Likewise.
11113 * stor-layout.c (layout_type): Likewise.
11114 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11115 * tree-cfg.c (verify_gimple_comparison): Likewise.
11116 (verify_gimple_assign_binary): Likewise.
11117 (verify_gimple_assign_ternary): Likewise.
11118 (verify_gimple_assign_single): Likewise.
11119 * tree-pretty-print.c (dump_generic_node): Likewise.
11120 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11121 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11122 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11123 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11124 (vect_shift_permute_load_chain): Likewise.
11125 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11126 (expand_vector_condition, optimize_vector_constructor): Likewise.
11127 (lower_vec_perm, get_compute_type): Likewise.
11128 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11129 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11130 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11131 (vect_recog_mask_conversion_pattern): Likewise.
11132 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11133 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11134 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11135 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11136 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11137 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11138 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11139 (supportable_widening_operation): Likewise.
11140 (supportable_narrowing_operation): Likewise.
11141 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11143 * varasm.c (output_constant): Likewise.
11145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11146 Alan Hayward <alan.hayward@arm.com>
11147 David Sherwood <david.sherwood@arm.com>
11149 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11150 so that both the length == 3 and length != 3 cases set up their
11151 own permute vectors. Add comments explaining why we know the
11152 number of elements is constant.
11153 (vect_permute_load_chain): Likewise.
11155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11156 Alan Hayward <alan.hayward@arm.com>
11157 David Sherwood <david.sherwood@arm.com>
11159 * machmode.h (mode_nunits): Change from unsigned char to
11161 (ONLY_FIXED_SIZE_MODES): New macro.
11162 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11163 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11164 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11166 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11167 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11168 or if measurement_type is not polynomial.
11169 * genmodes.c (ZERO_COEFFS): New macro.
11170 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11172 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11173 Use ZERO_COEFFS when emitting initializers.
11174 * data-streamer.h (bp_pack_poly_value): New function.
11175 (bp_unpack_poly_value): Likewise.
11176 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11177 for GET_MODE_NUNITS.
11178 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11179 for GET_MODE_NUNITS.
11180 * tree.c (make_vector_type): Remove temporary shim and make
11181 the real function take the number of units as a poly_uint64
11182 rather than an int.
11183 (build_vector_type_for_mode): Handle polynomial nunits.
11184 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11185 * emit-rtl.c (const_vec_series_p_1): Likewise.
11186 (gen_rtx_CONST_VECTOR): Likewise.
11187 * fold-const.c (test_vec_duplicate_folding): Likewise.
11188 * genrecog.c (validate_pattern): Likewise.
11189 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11190 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11191 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11192 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11193 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11194 * rtlanal.c (subreg_get_info): Likewise.
11195 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11196 (vect_grouped_load_supported): Likewise.
11197 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11198 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11199 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11200 (simplify_const_unary_operation, simplify_binary_operation_1)
11201 (simplify_const_binary_operation, simplify_ternary_operation)
11202 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11203 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11204 instead of CONST_VECTOR_NUNITS.
11205 * varasm.c (output_constant_pool_2): Likewise.
11206 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11207 explicit-encoded elements in the XVEC for variable-length vectors.
11209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11211 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11214 Alan Hayward <alan.hayward@arm.com>
11215 David Sherwood <david.sherwood@arm.com>
11217 * coretypes.h (fixed_size_mode): Declare.
11218 (fixed_size_mode_pod): New typedef.
11219 * builtins.h (target_builtins::x_apply_args_mode)
11220 (target_builtins::x_apply_result_mode): Change type to
11221 fixed_size_mode_pod.
11222 * builtins.c (apply_args_size, apply_result_size, result_vector)
11223 (expand_builtin_apply_args_1, expand_builtin_apply)
11224 (expand_builtin_return): Update accordingly.
11226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11228 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11229 * cselib.c (cselib_hash_rtx): Likewise.
11230 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11231 CONST_VECTOR encoding.
11233 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11234 Jeff Law <law@redhat.com>
11237 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11238 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11239 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11240 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11243 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11244 explicitly probe *sp in a noreturn function if there were any callee
11245 register saves or frame pointer is needed.
11247 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11250 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11251 BLKmode for ternary, binary or unary expressions.
11254 * var-tracking.c (delete_vta_debug_insn): New inline function.
11255 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11256 insns from get_insns () to NULL instead of each bb separately.
11257 Use delete_vta_debug_insn. No longer static.
11258 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11259 delete_vta_debug_insns callers.
11260 * rtl.h (delete_vta_debug_insns): Declare.
11261 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11262 instead of variable_tracking_main.
11264 2018-01-03 Martin Sebor <msebor@redhat.com>
11266 PR tree-optimization/83603
11267 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11268 arguments past the endof the argument list in functions declared
11269 without a prototype.
11270 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11271 Avoid checking when arguments are null.
11273 2018-01-03 Martin Sebor <msebor@redhat.com>
11276 * doc/extend.texi (attribute const): Fix a typo.
11277 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11278 issuing -Wsuggest-attribute for void functions.
11280 2018-01-03 Martin Sebor <msebor@redhat.com>
11282 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11283 offset_int::from instead of wide_int::to_shwi.
11284 (maybe_diag_overlap): Remove assertion.
11285 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11286 * gimple-ssa-sprintf.c (format_directive): Same.
11287 (parse_directive): Same.
11288 (sprintf_dom_walker::compute_format_length): Same.
11289 (try_substitute_return_value): Same.
11291 2018-01-03 Jeff Law <law@redhat.com>
11293 PR middle-end/83654
11294 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11295 non-constant residual for zero at runtime and avoid probing in
11296 that case. Reorganize code for trailing problem to mirror handling
11299 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11301 PR tree-optimization/83501
11302 * tree-ssa-strlen.c (get_string_cst): New.
11303 (handle_char_store): Call get_string_cst.
11305 2018-01-03 Martin Liska <mliska@suse.cz>
11307 PR tree-optimization/83593
11308 * tree-ssa-strlen.c: Include tree-cfg.h.
11309 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11310 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11311 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11313 (strlen_dom_walker::before_dom_children): Call
11314 gimple_purge_dead_eh_edges. Dump tranformation with details
11316 (strlen_dom_walker::before_dom_children): Update call by adding
11317 new argument cleanup_eh.
11318 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11320 2018-01-03 Martin Liska <mliska@suse.cz>
11323 * cif-code.def (VARIADIC_THUNK): New enum value.
11324 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11327 2018-01-03 Jan Beulich <jbeulich@suse.com>
11329 * sse.md (mov<mode>_internal): Tighten condition for when to use
11330 vmovdqu<ssescalarsize> for TI and OI modes.
11332 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11334 Update copyright years.
11336 2018-01-03 Martin Liska <mliska@suse.cz>
11339 * ipa-visibility.c (function_and_variable_visibility): Skip
11340 functions with noipa attribure.
11342 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11344 * gcc.c (process_command): Update copyright notice dates.
11345 * gcov-dump.c (print_version): Ditto.
11346 * gcov.c (print_version): Ditto.
11347 * gcov-tool.c (print_version): Ditto.
11348 * gengtype.c (create_file): Ditto.
11349 * doc/cpp.texi: Bump @copying's copyright year.
11350 * doc/cppinternals.texi: Ditto.
11351 * doc/gcc.texi: Ditto.
11352 * doc/gccint.texi: Ditto.
11353 * doc/gcov.texi: Ditto.
11354 * doc/install.texi: Ditto.
11355 * doc/invoke.texi: Ditto.
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11359 * vector-builder.h (vector_builder::m_full_nelts): Change from
11360 unsigned int to poly_uint64.
11361 (vector_builder::full_nelts): Update prototype accordingly.
11362 (vector_builder::new_vector): Likewise.
11363 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11364 (vector_builder::operator ==): Likewise.
11365 (vector_builder::finalize): Likewise.
11366 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11367 Take the number of elements as a poly_uint64 rather than an
11369 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11370 from unsigned int to poly_uint64.
11371 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11372 (vec_perm_indices::new_vector): Likewise.
11373 (vec_perm_indices::length): Likewise.
11374 (vec_perm_indices::nelts_per_input): Likewise.
11375 (vec_perm_indices::input_nelts): Likewise.
11376 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11377 number of elements per input as a poly_uint64 rather than an
11378 unsigned int. Use the original encoding for variable-length
11379 vectors, rather than clamping each individual element.
11380 For the second and subsequent elements in each pattern,
11381 clamp the step and base before clamping their sum.
11382 (vec_perm_indices::series_p): Handle polynomial element counts.
11383 (vec_perm_indices::all_in_range_p): Likewise.
11384 (vec_perm_indices_to_tree): Likewise.
11385 (vec_perm_indices_to_rtx): Likewise.
11386 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11387 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11388 (tree_vector_builder::new_binary_operation): Handle polynomial
11389 element counts. Return false if we need to know the number
11390 of elements at compile time.
11391 * fold-const.c (fold_vec_perm): Punt if the number of elements
11392 isn't known at compile time.
11394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11396 * vec-perm-indices.h (vec_perm_builder): Change element type
11397 from HOST_WIDE_INT to poly_int64.
11398 (vec_perm_indices::element_type): Update accordingly.
11399 (vec_perm_indices::clamp): Handle polynomial element_types.
11400 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11401 (vec_perm_indices::all_in_range_p): Likewise.
11402 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11404 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11405 polynomial vec_perm_indices element types.
11406 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11407 * fold-const.c (fold_vec_perm): Likewise.
11408 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11409 * tree-vect-generic.c (lower_vec_perm): Likewise.
11410 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11411 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11412 element type to HOST_WIDE_INT.
11414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 Alan Hayward <alan.hayward@arm.com>
11416 David Sherwood <david.sherwood@arm.com>
11418 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11419 rather than an int. Use plus_constant.
11420 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11421 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11424 Alan Hayward <alan.hayward@arm.com>
11425 David Sherwood <david.sherwood@arm.com>
11427 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11428 a HOST_WIDE_INT to a poly_int64.
11430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11431 Alan Hayward <alan.hayward@arm.com>
11432 David Sherwood <david.sherwood@arm.com>
11434 * calls.c (load_register_parameters): Cope with polynomial
11435 mode sizes. Require a constant size for BLKmode parameters
11436 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11437 forces a parameter to be padded at the lsb end in order to
11438 fill a complete number of words, require the parameter size
11439 to be ordered wrt UNITS_PER_WORD.
11441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11442 Alan Hayward <alan.hayward@arm.com>
11443 David Sherwood <david.sherwood@arm.com>
11445 * reload1.c (spill_stack_slot_width): Change element type
11446 from unsigned int to poly_uint64_pod.
11447 (alter_reg): Treat mode sizes as polynomial.
11449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11450 Alan Hayward <alan.hayward@arm.com>
11451 David Sherwood <david.sherwood@arm.com>
11453 * reload.c (complex_word_subreg_p): New function.
11454 (reload_inner_reg_of_subreg, push_reload): Use it.
11456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11457 Alan Hayward <alan.hayward@arm.com>
11458 David Sherwood <david.sherwood@arm.com>
11460 * lra-constraints.c (process_alt_operands): Reject matched
11461 operands whose sizes aren't ordered.
11462 (match_reload): Refer to this check here.
11464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11465 Alan Hayward <alan.hayward@arm.com>
11466 David Sherwood <david.sherwood@arm.com>
11468 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11469 that the mode size is in the set {1, 2, 4, 8, 16}.
11471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11472 Alan Hayward <alan.hayward@arm.com>
11473 David Sherwood <david.sherwood@arm.com>
11475 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11476 Use plus_constant instead of gen_rtx_PLUS.
11478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11483 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11484 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11485 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11486 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11487 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11488 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11489 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11490 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11491 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11493 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11494 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11495 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11496 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11497 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11498 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11499 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11500 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11501 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11502 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11504 * expr.c (emit_move_resolve_push): Treat the input and result
11505 of PUSH_ROUNDING as a poly_int64.
11506 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11507 (emit_push_insn): Likewise.
11508 * lra-eliminations.c (mark_not_eliminable): Likewise.
11509 * recog.c (push_operand): Likewise.
11510 * reload1.c (elimination_effects): Likewise.
11511 * rtlanal.c (nonzero_bits1): Likewise.
11512 * calls.c (store_one_arg): Likewise. Require the padding to be
11513 known at compile time.
11515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11516 Alan Hayward <alan.hayward@arm.com>
11517 David Sherwood <david.sherwood@arm.com>
11519 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11520 Use plus_constant instead of gen_rtx_PLUS.
11522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11523 Alan Hayward <alan.hayward@arm.com>
11524 David Sherwood <david.sherwood@arm.com>
11526 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11527 rather than an int.
11529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11534 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11535 via stack temporaries. Treat the mode size as polynomial too.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * expr.c (expand_expr_real_2): When handling conversions involving
11542 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11543 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11544 as a poly_uint64 too.
11546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11547 Alan Hayward <alan.hayward@arm.com>
11548 David Sherwood <david.sherwood@arm.com>
11550 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11553 Alan Hayward <alan.hayward@arm.com>
11554 David Sherwood <david.sherwood@arm.com>
11556 * combine.c (can_change_dest_mode): Handle polynomial
11557 REGMODE_NATURAL_SIZE.
11558 * expmed.c (store_bit_field_1): Likewise.
11559 * expr.c (store_constructor): Likewise.
11560 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11561 and polynomial REGMODE_NATURAL_SIZE.
11562 (gen_lowpart_common): Likewise.
11563 * reginfo.c (record_subregs_of_mode): Likewise.
11564 * rtlanal.c (read_modify_subreg_p): Likewise.
11566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11571 numbers of elements.
11573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11574 Alan Hayward <alan.hayward@arm.com>
11575 David Sherwood <david.sherwood@arm.com>
11577 * match.pd: Cope with polynomial numbers of vector elements.
11579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11580 Alan Hayward <alan.hayward@arm.com>
11581 David Sherwood <david.sherwood@arm.com>
11583 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11584 in a POINTER_PLUS_EXPR.
11586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11587 Alan Hayward <alan.hayward@arm.com>
11588 David Sherwood <david.sherwood@arm.com>
11590 * omp-simd-clone.c (simd_clone_subparts): New function.
11591 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11592 (ipa_simd_modify_function_body): Likewise.
11594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11595 Alan Hayward <alan.hayward@arm.com>
11596 David Sherwood <david.sherwood@arm.com>
11598 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11599 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11600 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11601 (expand_vector_condition, vector_element): Likewise.
11602 (subparts_gt): New function.
11603 (get_compute_type): Use subparts_gt.
11604 (count_type_subparts): Delete.
11605 (expand_vector_operations_1): Use subparts_gt instead of
11606 count_type_subparts.
11608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11609 Alan Hayward <alan.hayward@arm.com>
11610 David Sherwood <david.sherwood@arm.com>
11612 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11613 (vect_compile_time_alias): ...this new function. Do the calculation
11614 on poly_ints rather than trees.
11615 (vect_prune_runtime_alias_test_list): Update call accordingly.
11617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11618 Alan Hayward <alan.hayward@arm.com>
11619 David Sherwood <david.sherwood@arm.com>
11621 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11623 (vect_schedule_slp_instance): Likewise.
11625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11626 Alan Hayward <alan.hayward@arm.com>
11627 David Sherwood <david.sherwood@arm.com>
11629 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11630 constant and extern definitions for variable-length vectors.
11631 (vect_get_constant_vectors): Note that the number of units
11632 is known to be constant.
11634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11635 Alan Hayward <alan.hayward@arm.com>
11636 David Sherwood <david.sherwood@arm.com>
11638 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11639 of units as polynomial. Choose between WIDE and NARROW based
11642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11643 Alan Hayward <alan.hayward@arm.com>
11644 David Sherwood <david.sherwood@arm.com>
11646 * tree-vect-stmts.c (simd_clone_subparts): New function.
11647 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11650 Alan Hayward <alan.hayward@arm.com>
11651 David Sherwood <david.sherwood@arm.com>
11653 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11654 vectors as polynomial. Use build_index_vector for
11655 IFN_GOMP_SIMD_LANE.
11657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11658 Alan Hayward <alan.hayward@arm.com>
11659 David Sherwood <david.sherwood@arm.com>
11661 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11662 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11663 for variable-length vectors.
11664 (vectorizable_mask_load_store): Treat the number of units as
11665 polynomial, asserting that it is constant if the condition has
11666 already been enforced.
11667 (vectorizable_store, vectorizable_load): Likewise.
11669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11670 Alan Hayward <alan.hayward@arm.com>
11671 David Sherwood <david.sherwood@arm.com>
11673 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11674 of units as polynomial. Punt if we can't tell at compile time
11675 which vector contains the final result.
11677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11678 Alan Hayward <alan.hayward@arm.com>
11679 David Sherwood <david.sherwood@arm.com>
11681 * tree-vect-loop.c (vectorizable_induction): Treat the number
11682 of units as polynomial. Punt on SLP inductions. Use an integer
11683 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11684 cast of such a series for variable-length floating-point
11687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11688 Alan Hayward <alan.hayward@arm.com>
11689 David Sherwood <david.sherwood@arm.com>
11691 * tree.h (build_index_vector): Declare.
11692 * tree.c (build_index_vector): New function.
11693 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11694 of units as polynomial, forcibly converting it to a constant if
11695 vectorizable_reduction has already enforced the condition.
11696 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11697 to create a {1,2,3,...} vector.
11698 (vectorizable_reduction): Treat the number of units as polynomial.
11699 Choose vectype_in based on the largest scalar element size rather
11700 than the smallest number of units. Enforce the restrictions
11703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11704 Alan Hayward <alan.hayward@arm.com>
11705 David Sherwood <david.sherwood@arm.com>
11707 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11708 number of units as polynomial.
11710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11711 Alan Hayward <alan.hayward@arm.com>
11712 David Sherwood <david.sherwood@arm.com>
11714 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11715 * target.def (autovectorize_vector_sizes): Return the vector sizes
11716 by pointer, using vector_sizes rather than a bitmask.
11717 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11718 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11719 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11721 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11722 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11723 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11724 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11725 * omp-general.c (omp_max_vf): Likewise.
11726 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11727 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11728 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11729 * tree-vect-slp.c (vect_slp_bb): Likewise.
11730 * doc/tm.texi: Regenerate.
11731 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11733 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11734 the vector size as a poly_uint64 rather than an unsigned int.
11735 (current_vector_size): Change from an unsigned int to a poly_uint64.
11736 (get_vectype_for_scalar_type): Update accordingly.
11737 * tree.h (build_truth_vector_type): Take the size and number of
11738 units as a poly_uint64 rather than an unsigned int.
11739 (build_vector_type): Add a temporary overload that takes
11740 the number of units as a poly_uint64 rather than an unsigned int.
11741 * tree.c (make_vector_type): Likewise.
11742 (build_truth_vector_type): Take the number of units as a poly_uint64
11743 rather than an unsigned int.
11745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11746 Alan Hayward <alan.hayward@arm.com>
11747 David Sherwood <david.sherwood@arm.com>
11749 * target.def (get_mask_mode): Take the number of units and length
11750 as poly_uint64s rather than unsigned ints.
11751 * targhooks.h (default_get_mask_mode): Update accordingly.
11752 * targhooks.c (default_get_mask_mode): Likewise.
11753 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11754 * doc/tm.texi: Regenerate.
11756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11761 * omp-general.c (omp_max_vf): Likewise.
11762 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11763 (expand_omp_simd): Handle polynomial safelen.
11764 * omp-low.c (omplow_simd_context): Add a default constructor.
11765 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11766 (lower_rec_simd_input_clauses): Update accordingly.
11767 (lower_rec_input_clauses): Likewise.
11769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11770 Alan Hayward <alan.hayward@arm.com>
11771 David Sherwood <david.sherwood@arm.com>
11773 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11774 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11775 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11776 (vect_analyze_slp_cost): Likewise.
11777 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11778 (vect_model_load_cost): Likewise.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11785 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11786 from an unsigned int * to a poly_uint64_pod *.
11787 (calculate_unrolling_factor): New function.
11788 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11791 Alan Hayward <alan.hayward@arm.com>
11792 David Sherwood <david.sherwood@arm.com>
11794 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11795 from an unsigned int to a poly_uint64.
11796 (_loop_vec_info::slp_unrolling_factor): Likewise.
11797 (_loop_vec_info::vectorization_factor): Change from an int
11799 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11800 (vect_get_num_vectors): New function.
11801 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11802 (vect_get_num_copies): Use vect_get_num_vectors.
11803 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11804 to an unsigned int *.
11805 (vect_analyze_data_refs): Change min_vf from an int * to a
11807 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11808 than an unsigned HOST_WIDE_INT.
11809 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11810 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11811 to an unsigned int *.
11812 (vect_analyze_data_ref_dependences): Likewise.
11813 (vect_compute_data_ref_alignment): Handle polynomial vf.
11814 (vect_enhance_data_refs_alignment): Likewise.
11815 (vect_prune_runtime_alias_test_list): Likewise.
11816 (vect_shift_permute_load_chain): Likewise.
11817 (vect_supportable_dr_alignment): Likewise.
11818 (dependence_distance_ge_vf): Take the vectorization factor as a
11819 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11820 (vect_analyze_data_refs): Change min_vf from an int * to a
11822 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11823 vfm1 as a poly_uint64 rather than an int. Make the same change
11824 for the returned bound_scalar.
11825 (vect_gen_vector_loop_niters): Handle polynomial vf.
11826 (vect_do_peeling): Likewise. Update call to
11827 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11828 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11830 * tree-vect-loop.c (vect_determine_vectorization_factor)
11831 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11832 (vect_get_known_peeling_cost): Likewise.
11833 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11834 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11835 (vect_transform_loop): Likewise. Use the lowest possible VF when
11836 updating the upper bounds of the loop.
11837 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11838 rather than an int.
11839 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11840 polynomial unroll factors.
11841 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11842 (vect_make_slp_decision): Likewise.
11843 (vect_supported_load_permutation_p): Likewise, and polynomial
11845 (vect_analyze_slp_cost): Handle polynomial vf.
11846 (vect_slp_analyze_node_operations): Likewise.
11847 (vect_slp_analyze_bb_1): Likewise.
11848 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11849 than an unsigned HOST_WIDE_INT.
11850 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11851 (vectorizable_load): Handle polynomial vf.
11852 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11854 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11857 Alan Hayward <alan.hayward@arm.com>
11858 David Sherwood <david.sherwood@arm.com>
11860 * match.pd: Handle bit operations involving three constants
11861 and try to fold one pair.
11863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11865 * tree-vect-loop-manip.c: Include gimple-fold.h.
11866 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11867 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11868 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11869 Add a path that uses a step of VF instead of 1, but disable it
11871 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11872 and niters_no_overflow parameters. Update calls to
11873 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11874 Create a new SSA name if the latter choses to use a ste other
11875 than zero, and return it via niters_vector_mult_vf_var.
11876 * tree-vect-loop.c (vect_transform_loop): Update calls to
11877 vect_do_peeling, vect_gen_vector_loop_niters and
11878 slpeel_make_loop_iterate_ntimes.
11879 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11880 (vect_gen_vector_loop_niters): Update declarations after above changes.
11882 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11884 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11885 128-bit round to integer instructions.
11886 (ceil<mode>2): Likewise.
11887 (btrunc<mode>2): Likewise.
11888 (round<mode>2): Likewise.
11890 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11892 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11893 unaligned VSX load/store on P8/P9.
11894 (expand_block_clear): Allow the use of unaligned VSX
11895 load/store on P8/P9.
11897 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11899 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11901 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11902 swap associated with both a load and a store.
11904 2018-01-02 Andrew Waterman <andrew@sifive.com>
11906 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11907 * config/riscv/riscv.md (clear_cache): Use it.
11909 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11911 * web.c: Remove out-of-date comment.
11913 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11915 * expr.c (fixup_args_size_notes): Check that any existing
11916 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11917 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11918 (emit_single_push_insn): ...here.
11920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11922 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11923 (const_vector_encoded_nelts): New function.
11924 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11925 (const_vector_int_elt, const_vector_elt): Declare.
11926 * emit-rtl.c (const_vector_int_elt_1): New function.
11927 (const_vector_elt): Likewise.
11928 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11929 of CONST_VECTOR_ELT.
11931 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11933 * expr.c: Include rtx-vector-builder.h.
11934 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11935 directly on the tree encoding.
11936 (const_vector_from_tree): Likewise.
11937 * optabs.c: Include rtx-vector-builder.h.
11938 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11939 sequence of "u" values.
11940 * vec-perm-indices.c: Include rtx-vector-builder.h.
11941 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11942 directly on the vec_perm_indices encoding.
11944 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11946 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11947 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11948 * rtx-vector-builder.h: New file.
11949 * rtx-vector-builder.c: Likewise.
11950 * rtl.h (rtx_def::u2): Add a const_vector field.
11951 (CONST_VECTOR_NPATTERNS): New macro.
11952 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11953 (CONST_VECTOR_DUPLICATE_P): Likewise.
11954 (CONST_VECTOR_STEPPED_P): Likewise.
11955 (CONST_VECTOR_ENCODED_ELT): Likewise.
11956 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11957 (unwrap_const_vec_duplicate): Likewise.
11958 (const_vec_series_p): Check for a non-duplicated vector encoding.
11959 Say that the function only returns true for integer vectors.
11960 * emit-rtl.c: Include rtx-vector-builder.h.
11961 (gen_const_vec_duplicate_1): Delete.
11962 (gen_const_vector): Call gen_const_vec_duplicate instead of
11963 gen_const_vec_duplicate_1.
11964 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11965 (gen_const_vec_duplicate): Use rtx_vector_builder.
11966 (gen_const_vec_series): Likewise.
11967 (gen_rtx_CONST_VECTOR): Likewise.
11968 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11969 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11970 Build a new vector rather than modifying a CONST_VECTOR in-place.
11971 (handle_special_swappables): Update call accordingly.
11972 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11973 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11974 Build a new vector rather than modifying a CONST_VECTOR in-place.
11975 (handle_special_swappables): Update call accordingly.
11977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11979 * simplify-rtx.c (simplify_const_binary_operation): Use
11980 CONST_VECTOR_ELT instead of XVECEXP.
11982 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11984 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11985 the selector elements to be different from the data elements
11986 if the selector is a VECTOR_CST.
11987 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11988 ssizetype for the selector.
11990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11992 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11993 before testing each element individually.
11994 * tree-vect-generic.c (lower_vec_perm): Likewise.
11996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11998 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11999 * selftest-run-tests.c (selftest::run_tests): Call it.
12000 * vector-builder.h (vector_builder::operator ==): New function.
12001 (vector_builder::operator !=): Likewise.
12002 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12003 (vec_perm_indices::all_from_input_p): New function.
12004 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12005 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12006 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12007 instead of reading the VECTOR_CST directly. Detect whether both
12008 vector inputs are the same before constructing the vec_perm_indices,
12009 and update the number of inputs argument accordingly. Use the
12010 utility functions added above. Only construct sel2 if we need to.
12012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12014 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12015 the broadcast of the low byte.
12016 (expand_mult_highpart): Use an explicit encoding for the permutes.
12017 * optabs-query.c (can_mult_highpart_p): Likewise.
12018 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12019 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12020 (vectorizable_bswap): Likewise.
12021 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12022 explicit encoding for the power-of-2 permutes.
12023 (vect_permute_store_chain): Likewise.
12024 (vect_grouped_load_supported): Likewise.
12025 (vect_permute_load_chain): Likewise.
12027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12029 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12030 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12031 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12032 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12033 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12034 (vect_gen_perm_mask_any): Likewise.
12036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12038 * int-vector-builder.h: New file.
12039 * vec-perm-indices.h: Include int-vector-builder.h.
12040 (vec_perm_indices): Redefine as an int_vector_builder.
12041 (auto_vec_perm_indices): Delete.
12042 (vec_perm_builder): Redefine as a stand-alone class.
12043 (vec_perm_indices::vec_perm_indices): New function.
12044 (vec_perm_indices::clamp): Likewise.
12045 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12046 (vec_perm_indices::new_vector): New function.
12047 (vec_perm_indices::new_expanded_vector): Update for new
12048 vec_perm_indices class.
12049 (vec_perm_indices::rotate_inputs): New function.
12050 (vec_perm_indices::all_in_range_p): Operate directly on the
12051 encoded form, without computing elided elements.
12052 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12053 encoding. Update for new vec_perm_indices class.
12054 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12055 the given vec_perm_builder.
12056 (expand_vec_perm_var): Update vec_perm_builder constructor.
12057 (expand_mult_highpart): Use vec_perm_builder instead of
12058 auto_vec_perm_indices.
12059 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12060 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12061 or double series encoding as appropriate.
12062 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12063 vec_perm_indices instead of auto_vec_perm_indices.
12064 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12065 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12066 (vect_permute_store_chain): Likewise.
12067 (vect_grouped_load_supported): Likewise.
12068 (vect_permute_load_chain): Likewise.
12069 (vect_shift_permute_load_chain): Likewise.
12070 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12071 (vect_transform_slp_perm_load): Likewise.
12072 (vect_schedule_slp_instance): Likewise.
12073 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12074 (vectorizable_mask_load_store): Likewise.
12075 (vectorizable_bswap): Likewise.
12076 (vectorizable_store): Likewise.
12077 (vectorizable_load): Likewise.
12078 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12079 vec_perm_indices instead of auto_vec_perm_indices. Use
12080 tree_to_vec_perm_builder to read the vector from a tree.
12081 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12082 vec_perm_builder instead of a vec_perm_indices.
12083 (have_whole_vector_shift): Use vec_perm_builder and
12084 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12085 truncation to calc_vec_perm_mask_for_shift.
12086 (vect_create_epilog_for_reduction): Likewise.
12087 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12088 from auto_vec_perm_indices to vec_perm_indices.
12089 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12090 instead of changing individual elements.
12091 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12092 the vector in d.perm.
12093 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12094 from auto_vec_perm_indices to vec_perm_indices.
12095 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12096 instead of changing individual elements.
12097 (arm_vectorize_vec_perm_const): Use new_vector to install
12098 the vector in d.perm.
12099 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12100 Update vec_perm_builder constructor.
12101 (rs6000_expand_interleave): Likewise.
12102 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12103 (rs6000_expand_interleave): Likewise.
12105 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12107 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12108 to qimode could truncate the indices.
12109 * optabs.c (expand_vec_perm_var): Likewise.
12111 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12113 * Makefile.in (OBJS): Add vec-perm-indices.o.
12114 * vec-perm-indices.h: New file.
12115 * vec-perm-indices.c: Likewise.
12116 * target.h (vec_perm_indices): Replace with a forward class
12118 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12119 * optabs.h: Include vec-perm-indices.h.
12120 (expand_vec_perm): Delete.
12121 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12122 (expand_vec_perm_const): Declare.
12123 * target.def (vec_perm_const_ok): Replace with...
12124 (vec_perm_const): ...this new hook.
12125 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12126 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12127 * doc/tm.texi: Regenerate.
12128 * optabs.def (vec_perm_const): Delete.
12129 * doc/md.texi (vec_perm_const): Likewise.
12130 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12131 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12132 expand_vec_perm for constant permutation vectors. Assert that
12133 the mode of variable permutation vectors is the integer equivalent
12134 of the mode that is being permuted.
12135 * optabs-query.h (selector_fits_mode_p): Declare.
12136 * optabs-query.c: Include vec-perm-indices.h.
12137 (selector_fits_mode_p): New function.
12138 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12139 is defined, instead of checking whether the vec_perm_const_optab
12140 exists. Use targetm.vectorize.vec_perm_const instead of
12141 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12142 fit in the vector mode before using a variable permute.
12143 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12144 vec_perm_indices instead of an rtx.
12145 (expand_vec_perm): Replace with...
12146 (expand_vec_perm_const): ...this new function. Take the selector
12147 as a vec_perm_indices rather than an rtx. Also take the mode of
12148 the selector. Update call to shift_amt_for_vec_perm_mask.
12149 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12150 Use vec_perm_indices::new_expanded_vector to expand the original
12151 selector into bytes. Check whether the indices fit in the vector
12152 mode before using a variable permute.
12153 (expand_vec_perm_var): Make global.
12154 (expand_mult_highpart): Use expand_vec_perm_const.
12155 * fold-const.c: Includes vec-perm-indices.h.
12156 * tree-ssa-forwprop.c: Likewise.
12157 * tree-vect-data-refs.c: Likewise.
12158 * tree-vect-generic.c: Likewise.
12159 * tree-vect-loop.c: Likewise.
12160 * tree-vect-slp.c: Likewise.
12161 * tree-vect-stmts.c: Likewise.
12162 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12164 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12165 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12166 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12167 (aarch64_vectorize_vec_perm_const): ...this new function.
12168 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12169 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12170 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12171 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12172 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12173 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12174 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12176 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12177 check for NEON modes.
12178 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12179 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12180 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12181 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12183 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12184 the old VEC_PERM_CONST conditions.
12185 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12186 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12187 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12188 (ia64_vectorize_vec_perm_const_ok): Merge into...
12189 (ia64_vectorize_vec_perm_const): ...this new function.
12190 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12191 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12192 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12193 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12194 * config/mips/mips.c (mips_expand_vec_perm_const)
12195 (mips_vectorize_vec_perm_const_ok): Merge into...
12196 (mips_vectorize_vec_perm_const): ...this new function.
12197 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12198 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12199 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12200 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12201 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12202 (rs6000_expand_vec_perm_const): Delete.
12203 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12205 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12206 (altivec_expand_vec_perm_const_le): Take each operand individually.
12207 Operate on constant selectors rather than rtxes.
12208 (altivec_expand_vec_perm_const): Likewise. Update call to
12209 altivec_expand_vec_perm_const_le.
12210 (rs6000_expand_vec_perm_const): Delete.
12211 (rs6000_vectorize_vec_perm_const_ok): Delete.
12212 (rs6000_vectorize_vec_perm_const): New function.
12213 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12214 an element count and rtx array.
12215 (rs6000_expand_extract_even): Update call accordingly.
12216 (rs6000_expand_interleave): Likewise.
12217 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12218 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12219 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12220 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12221 (rs6000_expand_vec_perm_const): Delete.
12222 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12223 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12224 (altivec_expand_vec_perm_const_le): Take each operand individually.
12225 Operate on constant selectors rather than rtxes.
12226 (altivec_expand_vec_perm_const): Likewise. Update call to
12227 altivec_expand_vec_perm_const_le.
12228 (rs6000_expand_vec_perm_const): Delete.
12229 (rs6000_vectorize_vec_perm_const_ok): Delete.
12230 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12231 reference to the SPE evmerge intructions.
12232 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12233 an element count and rtx array.
12234 (rs6000_expand_extract_even): Update call accordingly.
12235 (rs6000_expand_interleave): Likewise.
12236 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12237 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12239 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12241 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12243 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12244 vector mode and that that mode matches the mode of the data
12246 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12247 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12248 directly using expand_vec_perm_1 when forcing selectors into
12250 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12254 * optabs-query.h (can_vec_perm_p): Delete.
12255 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12256 * optabs-query.c (can_vec_perm_p): Split into...
12257 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12258 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12259 particular selector is valid.
12260 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12261 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12262 (vect_grouped_load_supported): Likewise.
12263 (vect_shift_permute_load_chain): Likewise.
12264 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12265 (vect_transform_slp_perm_load): Likewise.
12266 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12267 (vectorizable_bswap): Likewise.
12268 (vect_gen_perm_mask_checked): Likewise.
12269 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12270 implementations of variable permutation vectors into account
12271 when deciding which selector to use.
12272 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12273 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12274 with a false third argument.
12275 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12276 to test whether the constant selector is valid and can_vec_perm_var_p
12277 to test whether a variable selector is valid.
12279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12281 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12282 * optabs-query.c (can_vec_perm_p): Likewise.
12283 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12284 instead of vec_perm_indices.
12285 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12286 (vect_gen_perm_mask_checked): Likewise,
12287 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12288 (vect_gen_perm_mask_checked): Likewise,
12290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12292 * optabs-query.h (qimode_for_vec_perm): Declare.
12293 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12294 (qimode_for_vec_perm): ...this new function.
12295 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12297 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12299 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12300 does not have a conditional at the top.
12302 2018-01-02 Richard Biener <rguenther@suse.de>
12304 * ipa-inline.c (big_speedup_p): Fix expression.
12306 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12309 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12312 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12316 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12317 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12318 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12319 cond_taken_branch_cost 3->4.
12321 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12323 PR tree-optimization/83581
12324 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12325 TODO_cleanup_cfg if any changes have been made.
12327 PR middle-end/83608
12328 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12329 convert_modes if target mode has the right side, but different mode
12332 PR middle-end/83609
12333 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12334 last argument when extracting from CONCAT. If either from_real or
12335 from_imag is NULL, use expansion through memory. If result is not
12336 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12337 the parts directly to inner mode, if even that fails, use expansion
12340 PR middle-end/83623
12341 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12342 check for bswap in mode rather than HImode and use that in expand_unop
12345 Copyright (C) 2018 Free Software Foundation, Inc.
12347 Copying and distribution of this file, with or without modification,
12348 are permitted in any medium without royalty provided the copyright
12349 notice and this notice are preserved.