1 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
4 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
5 Sunil K Pandey <sunil.k.pandey@intel.com>
8 * config/i386/i386.c (m_CORE_AVX512): New.
9 (m_CORE_AVX2): Likewise.
10 (m_CORE_ALL): Add m_CORE_AVX2.
11 * config/i386/x86-tune.def: Replace m_HASWELL with m_CORE_AVX2.
12 Replace m_SKYLAKE_AVX512 with m_CORE_AVX512 on avx256_optimal
13 and remove the rest of m_SKYLAKE_AVX512.
15 2018-07-13 Marek Polacek <polacek@redhat.com>
18 2018-07-03 Marek Polacek <polacek@redhat.com>
21 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
23 2018-07-12 Richard Biener <rguenther@suse.de>
26 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
27 Remove -mieee-fp handling.
29 2018-07-12 Jakub Jelinek <jakub@redhat.com>
31 PR tree-optimization/86492
32 * gimple-ssa-store-merging.c
33 (imm_store_chain_info::coalesce_immediate_stores): Call
34 check_no_overlap even for the merge_overlapping case.
36 2018-07-11 Jakub Jelinek <jakub@redhat.com>
38 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
39 Use __mmask64 type instead of __mmask8 for __M argument.
40 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
41 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
43 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
44 __mmask16 for __M argument.
45 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
46 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
47 to __mmask16 instead of __mmask8.
48 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
49 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
50 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
51 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
52 instead of __mmask16 for __U argument.
53 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
54 __mmask16 instead of __mmask8 for __U argument.
55 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
57 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
59 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
61 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
63 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
65 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
67 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
69 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
71 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
73 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
75 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
76 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
77 return type as well as __M argument type and all casts from __mmask8
79 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
80 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
81 return type as well as __M argument type and all casts from __mmask8
83 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
84 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
85 return type as well as __M argument type and all casts from __mmask8
87 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
88 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
89 return type as well as __M argument type and all casts from __mmask8
91 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
92 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
95 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
97 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
98 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
101 2018-07-05 Martin Sebor <msebor@redhat.com>
104 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
105 than its domain to compute its the upper bound of a char array.
107 2018-07-03 Carl Love <cel@us.ibm.com>
110 2018-07-03 Carl Love <cel@us.ibm.com>
112 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
113 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
114 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
115 VSX_BUILTIN_DOUBLEL_V4SF.
117 2018-07-04 Luis Machado <luis.machado@linaro.org>
119 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
121 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
123 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
126 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
129 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
130 ieee128_float_type_node to long_double_type_node unless
131 TARGET_LONG_DOUBLE_128 is set.
133 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
135 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
136 -fno-omit-frame-pointer when not optimizing.
138 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
141 * dwarf2out.c (loc_list_has_views): Adjust comments.
142 (dw_loc_list): Split single cross-partition range with
145 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
147 Backport from mainline
148 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
150 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
151 behavior of vec_pack (vector double, vector double) to match
152 behavior of vec_float2 (vector double, vector double).
154 2018-06-29 Jakub Jelinek <jakub@redhat.com>
156 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
158 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
160 Backport from mainline
161 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
163 * doc/sourcebuild.texi (Endianness): New subsubsection.
165 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
167 Backport from mainline
168 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
170 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
171 behavior of vec_packsu (vector unsigned long long, vector unsigned
172 long long) to match behavior of vec_packs with same signature.
174 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
177 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
179 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
180 unaligned vsx for 16B memset.
182 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
185 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
188 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
191 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
194 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
196 PR rtl-optimization/85645
197 * regrename.c (build_def_use): Also kill the chains that include the
198 destination of a REG_CFA_REGISTER note.
200 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
203 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
205 PR rtl-optimization/85645
206 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
207 insn that has a REG_CFA_REGISTER note.
209 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
211 * config/s390/s390.h (enum processor_flags): Do not use
212 default tune parameter when -march was specified.
214 2018-06-26 Jakub Jelinek <jakub@redhat.com>
217 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
218 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
221 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
224 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
226 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
227 added IEEE/IBM long double multilib support on PowerPC little
228 endian Linux systems.
229 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
230 (MULTILIB_DEFAULTS): Likewise.
231 * config/rs6000/rs6000.c (rs6000_option_override_internal):
233 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
234 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
235 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
237 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
240 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
242 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
244 2018-05-02 Jan Hubicka <jh@suse.cz>
246 Backport from mainline
247 2018-04-30 Jan Hubicka <jh@suse.cz>
249 * lto-wrapper.c (ltrans_priorities): New static var.
251 (run_gcc): Read priorities and if doing parallel build order
252 the Makefile by them.
254 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
257 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
260 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
261 ieee128 argument takes up only one (vector) register, not two (floating
264 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
266 PR tree-optimization/85989
267 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
269 (backprop::intersect_uses): Check it when deciding whether this
270 is a backedge reference.
271 (backprop::process_block): Add each phi to m_visited_phis
272 after visiting it, then clear it at the end.
274 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
276 Backport from mainline
278 2018-06-19 Tony Reix <tony.reix@atos.com>
279 Damien Bergamini <damien.bergamini@atos.com>
280 David Edelsohn <dje.gcc@gmail.com>
282 * collect2.c (static_obj): New variable.
283 (static_libs): New variable.
284 (is_in_list): Uncomment declaration.
285 (main): Track AIX libraries linked statically.
286 (is_in_list): Uncomment definition.
287 (scan_prog_file): Don't add AIX shared libraries initializer
288 to constructor list if linking statically.
290 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
293 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
295 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
296 conversion insn that shows up when pr85657-3.c is compiled using
297 IEEE 128-bit long double.
300 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
302 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
303 multiply and divide external functions from being created more
307 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
309 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
310 mode to check whether the mode is IBM extended.
312 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
315 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
316 floating point modes, so that IFmode is numerically greater than
317 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
318 to declare the ordering. This prevents IFmode from being
319 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
320 machine. Include rs6000-modes.h to share the fractional values
321 between genmodes* and the rest of the compiler.
325 * config/rs6000/rs6000-modes.h: New file.
326 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
327 meaning of rs6000_long_double_size so that 126..128 selects an
328 appropriate 128-bit floating point type.
329 (rs6000_option_override_internal): Likewise.
330 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
331 (TARGET_LONG_DOUBLE_128): Change the meaning of
332 rs6000_long_double_size so that 126..128 selects an appropriate
333 128-bit floating point type.
334 (LONG_DOUBLE_TYPE_SIZE): Update comment.
335 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
336 source and destination to match the standard usage.
337 (truncifkf2): Likewise.
338 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
339 ISA 2.07 to use an explicit clobber, instead of passing in a
341 (copysign<mode>3_soft): Likewise.
344 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
346 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
350 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
352 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
353 track if we pass or return IEEE 128-bit floating point.
354 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
355 C++ mangling that is compatible with GCC 8.1.
356 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
357 (init_cumulative_args): Note if we pass or return IEEE 128-bit
358 floating point types.
359 (rs6000_function_arg_advance_1): Likewise.
360 (rs6000_mangle_type): Optionally generate mangled names that match
361 what GCC 8.1 generated for IEEE 128-bit floating point types.
362 (rs6000_globalize_decl_name): If we have an external function that
363 passes or returns IEEE 128-bit types, generate a weak reference
364 from the mangled name used in GCC 8.1 to the current mangled
366 (rs6000_init_builtins): Make __ibm128 use the long double type if
367 long double is IBM extended double. Make __float128 use the long
368 double type if long double is IEEE 128-bit.
371 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
372 macro for __ibm128 built-in functions.
373 (PACK_IF): Add __ibm128 pack/unpack functions.
374 (UNPACK_IF): Likewise.
375 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
376 enable long double built-in functions if long double is IEEE
377 128-bit floating point.
378 (rs6000_invalid_builtin): Update long double built-in function
380 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
381 functions, adjust the built-in function to use the long double
382 built-in function if __ibm128 and long double are the same type.
383 * doc/extend.texi (PowerPC builtins): Update documention for
384 __builtin_{,un}pack_longdouble. Add documentation for
385 __builtin_{,un}pack_ibm128.
388 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
390 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
391 the 128-bit floating point types. Fix function comment.
393 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
396 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
399 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
400 define __ibm128 as long double.
401 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
402 as a distinct type with IEEE 128-bit floating point is supported.
403 (init_float128_ieee): Fix up conversions between IFmode and IEEE
404 128-bit types to use the correct functions.
405 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
406 convert between 128-bit floating point types that have different
407 modes but the same representation, instead of using gen_lowpart to
409 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
411 (IFKF_reg): New attributes to give the register constraints for
413 (extend<mode>tf2_internal): New insns to mark an explicit
414 conversion between 128-bit floating point types that have a
415 different mode but share the same representation.
417 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
420 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
421 -x assembler-with-cpp.
423 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
425 Backported from mainline
426 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
428 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
429 multilibs for *-*-rtems*.
430 * config/riscv/t-rtems: New file.
432 2018-06-20 Jakub Jelinek <jakub@redhat.com>
434 Backported from mainline
435 2018-06-16 Jakub Jelinek <jakub@redhat.com>
437 PR rtl-optimization/86108
438 * bb-reorder.c (create_forwarder_block): Renamed to ...
439 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
440 jump from new landing pad to the second part.
441 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
444 2018-06-20 Jakub Jelinek <jakub@redhat.com>
447 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
450 PR tree-optimization/86231
451 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
452 anti-range don't overwrite *vr0min before using it to compute *vr0max.
454 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
456 Backport from mainline
457 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
459 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
461 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
463 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
465 * gimplify.c (gimplify_init_constructor): Really never clear for an
466 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
468 2018-06-18 Martin Sebor <msebor@redhat.com>
471 * calls.c (alloc_max_size): Correct a logic error/typo.
472 Treat excessive arguments as infinite. Warn for invalid arguments.
473 * doc/invoke.texi (-Walloc-size-larger-than): Update.
475 2018-06-16 Jakub Jelinek <jakub@redhat.com>
478 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
479 documented as preserved for backward compatibility only.
480 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
482 2018-06-15 Jakub Jelinek <jakub@redhat.com>
485 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
486 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
487 Only call store_expr for halves if the mode is the same.
490 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
491 Fix up comment formatting.
493 2018-06-14 Jakub Jelinek <jakub@redhat.com>
496 * match.pd ((A +- CST1) +- CST2): Punt if last resort
497 unsigned_type_for returns NULL.
500 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
501 subregs of multi-word pseudos unless the float mode has word size.
503 2018-06-14 Richard Biener <rguenther@suse.de>
506 * tree-vect-generic.c (build_word_mode_vector_type): Remove
507 duplicate and harmful type_hash_canon.
508 * tree.c (type_hash_canon): Assert we didn't find ourselves.
510 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
512 Backport from mainline
513 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
515 * config/rtems.h (STDINT_LONG32): Define.
517 2018-05-05 Roland McGrath <mcgrathr@google.com>
520 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
521 any section for which we don't know a specific type it should have,
522 regardless of name. Previously this was done only for the exact
523 names ".init_array", ".fini_array", and ".preinit_array".
524 (default_elf_asm_named_section): Add comment about
525 relationship with default_section_type_flags and SECTION_NOTYPE.
526 (get_section): Don't consider it a type conflict if one side has
527 SECTION_NOTYPE and the other doesn't, as long as neither has the
528 SECTION_BSS et al used in the default_section_type_flags logic.
530 2018-06-13 Martin Sebor <msebor@redhat.com>
532 PR tree-optimization/86114
533 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
535 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
537 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
540 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
541 offsets for register save directives. Emit a second batch of save
542 directives, if need be, when the function accesses prior frames.
544 2018-06-11 Jason Merrill <jason@redhat.com>
546 PR c++/80485 - inline function non-zero address.
547 * symtab.c (nonzero_address): Check DECL_COMDAT.
549 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
551 Backport from mainline
552 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
555 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
558 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
561 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
564 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
565 on the correct operand.
566 (*movdi_internal64): Ditto.
568 2018-06-08 Carl Love <cel@us.ibm.com>
570 Backport from mainline
571 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
572 vextract_fp_from_shortl): Add BE support.
574 2018-06-08 Carl Love <cel@us.ibm.com>
576 Backport from mainline
577 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
578 Calculate index using natureal element order.
579 (first_match_or_eos_index_<mode>):
580 Calculate index using natural element order.
581 (first_match_index_<mode>):
582 Calculate index using natural element order.
583 (first_match_or_eos_index_<mode>):
584 Calculate index using natural order.
585 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
587 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
588 P9V_BUILTIN_VCLZLSBB_V16QI.
589 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
592 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
594 Backport from mainline
595 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
598 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
599 qualifier_void_pointer and qualifier_const_void_pointer.
600 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
601 (arm_init_builtins): Handle the above.
602 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
603 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
604 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
607 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
610 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
611 Generate SImode target register for null target.
612 <case IX86_BUILTIN_XGETBV>: Ditto.
613 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
614 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
616 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
618 Backport from mainline
619 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
622 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
623 Don't handle -mcpu=power8 if -mpower9-vector is also used.
625 2018-06-06 Richard Biener <rguenther@suse.de>
627 Backport from mainline
628 2018-06-04 Richard Biener <rguenther@suse.de>
630 PR tree-optimization/86038
631 * tracer.c (find_best_successor): Check probability for
632 being initialized, bail out if not.
634 2018-05-30 Richard Biener <rguenther@suse.de>
636 PR tree-optimization/85964
637 * tracer.c (better_p): Drop initialized count check, we only
638 call the function with initialized counts now.
639 (find_best_successor): Do find a best edge if one
640 has uninitialized count.
641 (find_best_predecessor): Likewise. Do BB frequency check only
642 if count is initialized.
644 2018-05-28 Richard Biener <rguenther@suse.de>
646 PR tree-optimization/85934
647 * tree-vect-generic.c (expand_vector_operations_1): Hoist
648 vector boolean check before scalar optimization.
650 2018-05-22 Richard Biener <rguenther@suse.de>
652 PR tree-optimization/85863
653 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
654 comparisons when vectype is specified.
655 (vectorizable_condition): Do not specify vectype for
656 vect_is_simple_cond when SLP vectorizing.
658 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
660 Backport from mainline
661 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
663 * config/s390/s390-builtin-types.def: Add void function type.
664 * config/s390/s390-builtins.def: Use the function type for the
667 2018-06-04 Martin Sebor <msebor@redhat.com>
670 * calls.c (maybe_warn_nonstring_arg): Use string length to set
671 or ajust the presumed bound on an operation to avoid unnecessary
674 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
677 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
678 of bits to ignore when comparing architectures.
680 2018-06-04 Jakub Jelinek <jakub@redhat.com>
683 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
685 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
687 PR tree-optimization/85712
688 Backport from mainline:
689 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
691 PR tree-optimization/85712
692 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
694 (alloc_cand_and_find_basis): Initialize first_interp field.
695 (slsr_process_mul): Modify first_interp field.
696 (slsr_process_add): Likewise.
697 (slsr_process_cast): Modify first_interp field for each new
699 (slsr_process_copy): Likewise.
700 (dump_candidate): Dump first_interp field.
701 (replace_mult_candidate): Process all interpretations, not just
703 (replace_rhs_if_not_dup): Likewise.
704 (replace_one_candidate): Likewise.
706 Backport from mainline:
707 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
709 PR tree-optimization/85712
710 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
711 this candidate has already been replaced in-situ by a copy.
713 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
716 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
717 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
719 (sse4_1_round<mode>2): Use nonimmediate_operand
720 for operand 1 predicate.
722 2018-05-31 Jakub Jelinek <jakub@redhat.com>
725 * bb-reorder.c (pass_partition_blocks::gate): Return false for
726 functions with naked attribute.
728 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
731 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
732 and movx for Haswell.
734 2018-05-30 Jakub Jelinek <jakub@redhat.com>
736 Backported from mainline
737 2018-05-14 Jakub Jelinek <jakub@redhat.com>
740 * config/i386/i386.md: Disallow non-commutative arithmetics in
741 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
742 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
743 in the peephole2 before it.
745 2018-05-08 Jakub Jelinek <jakub@redhat.com>
748 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
749 after cmpelim optimization.
751 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
753 Backport from mainline
754 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
758 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
760 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
764 * varasm.c (assemble_alias): Check ifunc_resolver only on
767 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
770 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
772 (cgraph_node::create_alias): Likewise.
773 (cgraph_node::get_availability): Check ifunc_resolver instead
774 of looking up ifunc attribute.
775 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
776 * varasm.c (do_assemble_alias): Likewise.
777 (assemble_alias): Likewise.
778 (default_binds_local_p_3): Likewise.
779 * cgraph.h (cgraph_node): Add ifunc_resolver.
780 (cgraph_node::only_called_directly_or_aliased_p): Return false
782 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
784 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
785 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
786 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
787 instead of looking up ifunc attribute.
789 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
791 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
793 2018-05-25 Richard Biener <rguenther@suse.de>
796 * tree-dump.c (dequeue_and_dump): Remove access to removed
797 operand 2 of a SWITCH_EXPR.
799 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
801 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
802 Add {q} suffix to insn mnemonic.
804 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
807 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
808 when memory input operand is handled.
810 2018-05-23 Bin Cheng <bin.cheng@arm.com>
812 Backport from mainline
813 2018-05-17 Bin Cheng <bin.cheng@arm.com>
814 Richard Biener <rguenther@suse.de>
816 PR tree-optimization/85793
817 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
818 for VMAT_ELEMENTWISE.
820 2018-05-23 Richard Biener <rguenther@suse.de>
823 * tree-data-ref.c (create_runtime_alias_checks): Defer
824 and ignore overflow warnings.
826 2018-05-23 Martin Liska <mliska@suse.cz>
828 Backport from mainline
829 2018-05-23 Yury Gribov <tetra2005@gmail.com>
831 PR tree-optimization/85822
832 * tree-vrp.c (is_masked_range_test): Fix handling of negative
835 2018-05-23 Martin Liska <mliska@suse.cz>
837 Backport from mainline
838 2018-05-11 Martin Liska <mliska@suse.cz>
841 * doc/extend.texi: Document LLVM style format for no_sanitize
844 2018-05-22 Martin Sebor <msebor@redhat.com>
847 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
849 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
851 PR tree-optimization/85814
852 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
853 a null return from get_strinfo when unsharing the next
854 strinfo in the chain.
856 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
858 Backport from mainline
859 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
860 Segher Boessenkool <segher@kernel.crashing.org>
863 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
865 2018-05-18 Toon Moene <toon@moene.org>
867 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
868 directly after that of -floop-interchange. Indicate that both
869 options are enabled by default when specifying -O3.
871 2018-05-11 Martin Jambor <mjambor@suse.cz>
874 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
877 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
880 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
882 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
885 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
887 (cortex-m0): Use armv6s-m isa.
888 (cortex-m0plus): Likewise.
889 (cortex-m1): Likewise.
890 (cortex-m0.small-multiply): Likewise.
891 (cortex-m0plus.small-multiply): Likewise.
892 (cortex-m1.small-multiply): Likewise.
894 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
896 PR rtl-optimization/85638
897 * bb-reorder.c: Include common/common-target.h.
898 (create_forwarder_block): New function extracted from...
899 (fix_up_crossing_landing_pad): ...here. Rename into...
900 (dw2_fix_up_crossing_landing_pad): ...this.
901 (sjlj_fix_up_crossing_landing_pad): New function.
902 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
903 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
904 from both partitions and exit the loop after one iteration.
906 2018-05-08 Richard Biener <rguenther@suse.de>
908 Backport from mainline
909 2018-05-04 Richard Biener <rguenther@suse.de>
912 * fold-const.c (negate_expr_p): Restrict negation of operand
913 zero of a division to when we know that can happen without
915 (fold_negate_expr_1): Likewise.
917 2018-05-03 Richard Biener <rguenther@suse.de>
919 PR tree-optimization/85615
920 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
921 to loops not nested in BBs loop father to avoid creating multi-entry
924 2018-05-02 Richard Biener <rguenther@suse.de>
927 * gimplify.c (gimplify_save_expr): When in SSA form allow
928 SAVE_EXPRs to compute to SSA vars.
930 2018-05-02 Richard Biener <rguenther@suse.de>
932 PR tree-optimization/85597
933 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
934 do not use split vect_get_vec_defs call but call vect_get_slp_defs
937 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
940 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
941 (check_arch): Likewise.
942 (check_fpu): Return the result rather than printing it.
943 (end arch): Fix operator precedence.
945 (END): Print the result from check_fpu.
947 2018-05-08 Richard Biener <rguenther@suse.de>
949 Backport from mainline
950 2018-05-04 Richard Biener <rguenther@suse.de>
952 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
953 * gimple-ssa-store-merging.c
954 (imm_store_chain_info::output_merged_store): Remove redundant create,
955 release split_store vector contents on failure.
956 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
957 scalar stmt vector on cache hit.
959 2018-05-04 Richard Biener <rguenther@suse.de>
962 * tree-complex.c (tree_lower_complex): Handle removed blocks.
964 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
966 Backport from mainline.
967 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
969 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
970 GET_MODE_MASK before any checking.
971 (nds32_can_use_bset_p): Likewise.
972 (nds32_can_use_btgl_p): Likewise.
974 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
976 Backport from mainline.
977 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
979 * config/nds32/nds32-doubleword.md: New define_split pattern for
980 illegal register number.
982 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
984 Backport from mainline.
985 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
987 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
989 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
991 Backport from mainline.
992 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
994 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
996 2018-05-06 Jakub Jelinek <jakub@redhat.com>
999 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1000 the type is addressable. Don't force op into register if it has
1003 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1005 Backport from trunk.
1006 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1009 * config/aarch64/constraints.md (Usg): Limit to 31.
1012 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1014 * BASE-VER: Set to 8.1.1.
1016 2018-05-02 Release Manager
1018 * GCC 8.1.0 released.
1020 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1022 PR tree-optimization/85586
1023 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1024 exit early for statements in the same group if the accesses are
1027 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1030 * doc/install.texi2html: Replace _002d with - and _002a with * in
1031 generated html files using sed.
1033 2018-04-27 Martin Jambor <mjambor@suse.cz>
1036 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1037 the jump function allows for passing through aggregate values.
1039 2018-04-27 Jeff Law <law@redhat.com>
1041 * config.gcc: Mark tile* targets as deprecated/obsolete.
1043 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1045 PR tree-optimization/85529
1046 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1047 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1048 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1049 zero extension or masking of the MSB bit.
1050 (optimize_range_tests): Add FIRST_BB argument, pass it through
1051 to optimize_range_tests_var_bound.
1052 (maybe_optimize_range_tests, reassociate_bb): Adjust
1053 optimize_range_tests callers.
1055 2018-04-26 Richard Biener <rguenther@suse.de>
1056 Jakub Jelinek <jakub@redhat.com>
1058 * cgraph.h (symbol_table): Just declare debug method here.
1059 * symtab.c (symbol_table::debug): Define.
1061 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1063 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1065 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
1067 Backport from mainline
1068 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1071 * config/i386/i386.c (ix86_expand_builtin): Change memory
1072 operand to XI, extend p0 to Pmode.
1073 * config/i386/i386.md: Change unspec volatile and operand
1074 1 mode to XI, change operand 0 mode to P.
1076 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1078 * DEV-PHASE: Set to prerelease.
1080 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1082 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1083 in __abskf2 and __powikf2.
1085 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1088 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1089 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1090 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1091 Use the above on operand 2. Reindent.
1092 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1094 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1097 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1099 * config/i386/i386.opt (mcet): Removed.
1100 * doc/install.texi: Remove -mcet documentation.
1101 * doc/invoke.texi: Likewise.
1103 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1106 * doc/install.texi: Remove -mcet from bootstrap-cet.
1108 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1111 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1112 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1116 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1117 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1118 containing a CONST_VECTOR.
1120 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1122 * doc/install.texi: Update newlib dependency for nvptx.
1124 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1127 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1128 instead of INTVAL when shifting x left.
1130 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1132 PR tree-optimization/85478
1133 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1134 vect_grouped_store_supported for single element vectors.
1136 2018-04-24 Richard Biener <rguenther@suse.de>
1139 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1140 load cost increase to the case of non-constant step.
1142 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1145 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1146 destination if any_malformed_asm.
1148 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1151 * expr.c (store_field): In the bitfield case, if the value comes from
1152 a function call and is returned in registers by means of a PARALLEL,
1153 do not change the mode of the temporary unless BLKmode and VOIDmode.
1155 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1157 PR rtl-optimization/85423
1158 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1159 dependencies to debug insns when the previous insn is non-debug.
1161 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1163 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1164 enums into a single definition.
1165 (fls): Fix predicates and printing.
1168 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1170 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1171 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1172 and short u6 immediate.
1173 (check_if_valid_sleep_operand): Remove.
1174 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1176 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1178 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1179 flag_always_save_lp condition.
1180 * config/nds32/nds32.opt (malways-save-lp): New option.
1182 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1184 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1185 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1186 * config/nds32/nds32.h
1187 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1188 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1190 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1192 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1193 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1195 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1196 Chung-Ju Wu <jasonwucj@gmail.com>
1198 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1200 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1201 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1203 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1205 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1207 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1209 * config/nds32/nds32-protos.h (nds32_data_alignment,
1210 nds32_local_alignment): Declare.
1211 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1212 nds32_local_alignment): New functions.
1213 (TARGET_CONSTANT_ALIGNMENT): Define.
1214 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1216 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1218 * config/nds32/nds32.c
1219 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1220 (TARGET_MODES_TIEABLE_P): Likewise.
1222 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1224 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1227 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1228 Chung-Ju Wu <jasonwucj@gmail.com>
1230 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1231 for unaligned access.
1232 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1234 * config/nds32/nds32-intrinsic.md: Likewise.
1235 * config/nds32/nds32_intrinsic.h: Likewise.
1236 * config/nds32/nds32.h (nds32_builtins): Likewise.
1237 * config/nds32/nds32.opt (munaligned-access): New option.
1238 * config/nds32/nds32.c (nds32_asm_file_start): Display
1239 flag_unaligned_access status.
1241 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1243 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1244 -mno-relax is present.
1245 * config/riscv/linux.h (LINK_SPEC): Ditto.
1247 2018-04-20 Martin Sebor <msebor@redhat.com>
1250 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1252 (gimple_fold_builtin_stxcpy_chk): Same.
1253 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1255 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1258 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1259 __powikf2 when long double is IEEE 128-bit.
1261 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1263 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1264 step to make sure stack always aligned.
1266 2018-04-20 Carl Love <cel@us.ibm.com>
1269 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1270 size check for arg0.
1272 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1273 Tom de Vries <tom@codesourcery.com>
1276 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1277 Emit insns for calls too.
1278 (nvptx_find_par): Always look for worker-level predecessor insn.
1279 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1281 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1282 (nvptx_process_pars): Propagate frames for calls.
1284 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1287 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1289 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1290 (ix86_handle_option): Don't handle OPT_mibt.
1291 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1293 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1295 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1296 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1297 (ix86_target_macros): Define __CET__ with flag_cf_protection
1298 for -fcf-protection.
1299 * config/i386/i386.c (isa2_opts): Remove -mibt.
1300 * config/i386/i386.h (TARGET_IBT): Removed.
1301 (TARGET_IBT_P): Likewise.
1302 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1303 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1304 * config/i386/i386.opt (mcet): Update help message.
1307 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1308 -mcet as an alias for -mshstk.
1310 2018-04-20 Richard Biener <rguenther@suse.de>
1313 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1314 complexity by forcing a single use of the multiply operand.
1316 2018-04-20 Martin Jambor <mjambor@suse.cz>
1319 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1320 recursion dependency to only apply to non-clones.
1322 2018-04-20 Martin Jambor <mjambor@suse.cz>
1325 * ipa-cp.c (create_specialized_node): Check that clones of
1326 self-recursive edges exist during IPA-CP.
1328 2018-04-19 Toon Moene <toon@moene.org>
1330 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1333 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1335 PR tree-optimization/85467
1336 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1337 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1338 VECTOR_CST element to type.
1340 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1343 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1344 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1345 (builtin_longjmp): Likewise.
1346 (save_stack_nonlocal): New pattern.
1347 (restore_stack_nonlocal): Likewise.
1349 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1352 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1353 Replace ASM_OUTPUT_LABEL with fprintf.
1355 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1358 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1359 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1360 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1361 define __IBT__ and __SHSTK__ for -fcf-protection.
1362 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1364 (ix86_trampoline_init): Likewise.
1365 (x86_output_mi_thunk): Likewise.
1366 (ix86_notrack_prefixed_insn_p): Likewise.
1367 (ix86_option_override_internal): Don't disallow -fcf-protection.
1368 * config/i386/i386.md (rdssp<mode>): Also enable for
1370 (incssp<mode>): Likewise.
1371 (nop_endbr): Likewise.
1372 * config/i386/i386.opt (mcet): Change help message to built-in
1376 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1377 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1378 enable CET built-in functions.
1380 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1382 * common/config/i386/i386-common.c
1383 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1384 OPTION_MASK_ISA_MOVDIRI_UNSET,
1385 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1386 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1387 * config.gcc (movdirintrin.h): New header.
1388 * config/i386/cpuid.h (bit_MOVDIRI,
1389 bit_MOVDIR64B): New bits.
1390 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1392 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1393 (VOID, PVOID, PCVOID)): New function types.
1394 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1395 __builtin_ia32_directstoreu_u64,
1396 __builtin_ia32_movdir64b): New builtins.
1397 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1398 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1400 (ix86_valid_target_attribute_inner_p): Ditto.
1401 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1402 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1403 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1404 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1405 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1406 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1407 (movdiri<mode>, movdir64b_<mode>): New.
1408 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1409 * config/i386/immintrin.h: Include movdirintrin.h.
1410 * config/i386/movdirintrin.h: New file.
1411 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1413 2018-04-19 Richard Biener <rguenther@suse.de>
1416 * cfg.c (clear_bb_flags): When loop state says we have
1417 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1419 2018-04-19 Richard Biener <rguenther@suse.de>
1421 PR tree-optimization/84737
1422 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1423 copying restrict info.
1424 (vect_setup_realignment): Use it.
1425 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1426 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1427 the first DR to all generated stores.
1428 (vectorizable_load): Likewise for loads.
1430 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1432 PR tree-optimization/85446
1433 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1434 the integral and pointer types to have the same precision.
1436 * doc/install.texi: Document --disable-cet being the default and
1439 2018-04-18 Martin Liska <mliska@suse.cz>
1441 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1444 2018-04-18 Martin Liska <mliska@suse.cz>
1447 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1450 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1451 arguments if they are comparable.
1453 2018-04-18 Martin Liska <mliska@suse.cz>
1456 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1459 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1462 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1465 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1466 ENDBR after calling __morestack.
1468 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1471 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1472 by using gcc_base_ver to generate a gcc_driver_version, and use
1473 it when generating GCC_DRIVER_NAME.
1474 * configure: Regenerate.
1476 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1479 * config.gcc: Obsolete powerpc*-*-*spe*.
1481 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1484 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1485 (stabstr_D): Change type of unum from unsigned int to
1486 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1489 2018-04-17 Jim Wilson <jimw@sifive.com>
1492 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1493 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1494 Set arg_pointer_offset after using pretend_args_size.
1496 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1498 PR rtl-optimization/85431
1499 * dse.c (record_store): Ignore zero width stores.
1502 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1503 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1504 __builtin_stack_restore rather than after it.
1505 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1506 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1507 argument instead of virtual_dynamic_stack_rtx.
1509 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1511 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1513 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1514 Add note to error message to explain internal mapping of overloaded
1515 built-in function name to non-overloaded built-in function name.
1516 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1519 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1522 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1523 where the inputs overlap with the output.
1525 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1527 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1528 (=v, v) alternative and explicit "memory" attribute.
1529 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1530 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1532 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1533 "sselog1" type instead of "sselog".
1534 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1535 "sselog". Remove explicit "memory" attribute.
1536 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1537 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1539 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1540 "sselog1" type instead of "sselog". Remove explicit "memory"
1542 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1543 use "sselog1" type instead of "sselog". Remove explicit "memory"
1545 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1546 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1548 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1549 "sselog1" type instead of "sselog". Remove explicit "memory"
1551 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1552 use "sselog1" type instead of "sselog". Remove explicit "memory"
1556 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1559 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1562 2018-04-17 Martin Jambor <mjambor@suse.cz>
1565 * ipa-cp.c (create_specialized_node): Call
1566 expand_all_artificial_thunks if necessary.
1568 2018-04-17 Martin Liska <mliska@suse.cz>
1571 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1572 in message, remote space in between '_G' and '('.
1574 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1577 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1578 avx512f_vmcmp<mode>3<round_saeonly_name>,
1579 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1580 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1581 avx512f_rndscale<mode><round_saeonly_name>,
1582 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1583 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1584 Use %<iptr>2 instead of %2 for -masm=intel.
1585 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1586 avx512f_vcvttss2usi<round_saeonly_name>,
1587 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1589 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1590 avx512f_vcvttsd2usi<round_saeonly_name>,
1591 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1592 Use %q1 instead of %1 for -masm=intel.
1593 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1594 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1595 of %3 for -masm=intel.
1596 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1598 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1600 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1602 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1603 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1605 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1607 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1608 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1609 %g1 and one with %0 and %1.
1610 (avx512er_vmrcp28<mode><round_saeonly_name>,
1611 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1613 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1614 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1615 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1616 of %0 and %{%4%} for -masm=intel.
1617 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1618 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1619 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1620 order of %0 and %{%5%}%{z%} for -masm=intel.
1622 2018-04-17 Jan Hubicka <jh@suse.cz>
1625 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1627 2018-04-17 Martin Liska <mliska@suse.cz>
1630 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1631 for target_clone error message. Make default implementation
1632 clone to be a local declaration.
1633 (separate_attrs): Add new argument and check for an empty
1635 (expand_target_clones): Handle it.
1636 (ipa_target_clone): Make redirection just for target_clones
1639 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1640 Tom de Vries <tom@codesourcery.com>
1643 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1644 tiled basic blocks without omp continue statements.
1646 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1649 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1650 vec_extract expression as having side effects to make sure it gets
1653 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1656 * config/i386/i386.c (get_builtin_code_for_version): Check
1659 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1662 * gcc/config.gcc: Support "skylake".
1663 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1665 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1666 (processor_target_table): Add "skylake".
1667 (ix86_option_override_internal): Add "skylake".
1668 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1669 PROCESSOR_CANNONLAKE.
1670 (get_builtin_code_for_version): Fix priority for
1671 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1672 PROCESSOR_SKYLAKE-AVX512.
1673 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1674 (processor_type): Add PROCESSOR_SKYLAKE.
1676 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1677 Jason Merrill <jason@redhat.com>
1680 * convert.c (convert_to_integer_1): Use direct recursion for
1681 enumeral types and types with a precision less than the number
1682 of bits in their mode.
1684 2018-04-16 Julia Koval <julia.koval@intel.com>
1687 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1688 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1690 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1693 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1694 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1695 and -mno-direct-move.
1697 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1700 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1701 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1702 Ensure negative shifts result in {0}.
1704 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1706 PR rtl-optimization/79916
1707 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1708 regs (if any) to define how to gnerate SD moves when LRA is in
1711 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1713 PR rtl-optimization/85393
1714 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1715 * except.c (expand_dw2_landing_pad_for_region): Make static.
1716 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1717 a label and unconditional jump to old_bb, rather than
1718 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1721 PR rtl-optimization/85376
1722 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1723 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1724 instead of a specific value.
1726 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1727 Bin Cheng <bin.cheng@arm.com>
1729 PR tree-optimization/82965
1730 PR tree-optimization/83991
1731 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1732 by_profile_only parameter.
1733 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1734 information if the loop was predicted to iterate too many times.
1735 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1737 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1740 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1743 2018-04-13 Martin Liska <mliska@suse.cz>
1744 Jakub Jelinek <jakub@redhat.com>
1747 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1748 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1749 * builtins.c (expand_builtin_memory_copy_args): Use
1750 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1751 handle dest_addr == pc_rtx.
1753 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1756 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1757 asked to not generate direct moves.
1758 (fix_trunc<mode>si2_stfiwx): Similar.
1759 (fix_trunc<mode>si2_internal): Similar.
1761 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1764 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1765 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1766 lookup if dest in some wider mode is known to be const0_rtx and
1767 if so, record permanent equivalence for it to be ZERO_EXTEND of
1768 the narrower mode destination.
1770 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1772 * lto-streamer-out.c (output_function): Revert 259346.
1773 * omp-expand.c (expand_oacc_for): Likewise.
1775 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1777 PR rtl-optimization/85354
1778 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1779 * sel-sched.c (sel_global_init): ... here.
1781 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1784 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1785 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1786 mode for PE-COFF targets.
1787 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1788 (i386_pe_asm_lto_end): Likewise.
1789 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1790 (TARGET_ASM_LTO_END): Likewise.
1791 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1792 (i386_pe_asm_lto_start): New function.
1793 (i386_pe_asm_lto_end): Likewise.
1795 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1796 Richard Biener <rguenther@suse.de>
1799 * lto-streamer-out.c (output_function): Fix CFG loop state before
1801 * omp-expand.c (expand_oacc_for): Handle calls to internal
1802 functions like regular functions.
1804 2018-04-12 Richard Biener <rguenther@suse.de>
1807 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1808 for the early LTO debug to properly generate references to it
1809 during DIE emission. Do not re-use that for the skeleton for
1811 (dwarf2out_early_finish): Likewise.
1813 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1816 * config/i386/sse.md
1817 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1818 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1819 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1820 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1821 and output is a reg, avoid creating invalid lowpart subreg, but
1822 instead split into a 512-bit move. Don't split if not AVX512VL,
1823 input is xmm16+ reg and output is a mem.
1824 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1825 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1826 xmm16+ reg and output is a mem.
1828 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1830 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1831 also for flag_dwarf2_cfi_asm.
1833 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1835 PR rtl-optimization/85342
1836 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1837 a bool scalar var inside of the loop instead. Don't try to update
1838 recog_data.operand after failed apply_change_group.
1840 2018-04-12 Tom de Vries <tom@codesourcery.com>
1843 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1844 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1845 array with flexible array member as array without given dimension.
1846 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1847 argument for undefined param to true.
1849 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1852 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1853 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1854 from PowerPC section.
1855 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1856 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1857 help text that is too long.
1858 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1859 help text that is too long.
1860 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1861 help text that is too long.
1863 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1865 * config/alpha/alpha.md (stack_probe_internal): Rename
1866 from "probe_stack". Update all callers.
1868 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1870 PR rtl-optimization/84566
1871 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1872 sched_macro_fuse_insns.
1874 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1877 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1878 (compute_block_dependences): ... from here.
1880 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1882 PR tree-optimization/85331
1883 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1884 from int to HOST_WIDE_INT.
1886 2018-04-11 Martin Jambor <mjambor@suse.cz>
1889 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1890 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1891 not the same as the source val.
1892 (cgraph_edge_brings_value_p): New parameter.
1893 (gather_edges_for_value): Pass destination value to
1894 cgraph_edge_brings_value_p.
1895 (perhaps_add_new_callers): Likewise.
1896 (get_info_about_necessary_edges): Likewise and exclude values brought
1897 only by self-recursive edges.
1898 (create_specialized_node): Redirect only clones of self-calling edges.
1899 (+self_recursive_pass_through_p): New function.
1900 (find_more_scalar_values_for_callers_subset): Use it.
1901 (find_aggregate_values_for_callers_subset): Likewise.
1902 (known_aggs_to_agg_replacement_list): Removed.
1903 (decide_whether_version_node): Re-calculate known constants for all
1904 remaining context clones.
1906 2018-04-11 Richard Biener <rguenther@suse.de>
1909 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1910 from early DWARF output.
1911 (dwarf2out_early_finish): Output line info unconditionally into
1912 early DWARF and add reference to it.
1914 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1917 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1918 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1919 other than V2DFmode using iptr mode attribute.
1920 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1922 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1924 PR rtl-optimization/84659
1925 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1927 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1930 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1932 (output_loc_list): Pass address of a dummy size variable even in the
1933 locview handling loop.
1934 (index_location_lists): Add comment on why skip_loc_list_entry can't
1937 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1940 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1943 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1946 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1947 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1948 and -mstring-compare-inline-limit.
1950 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1953 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1954 for stack clash protection in a register whenever we need it to be in
1957 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1959 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1960 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1962 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1965 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1967 (mlong-double-): Ditto.
1968 * config/rs6000/sysv4.opt (msdata=): Ditto.
1969 (mtls-size=): Ditto.
1971 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1973 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1974 erroneous entries for
1975 "vector int vec_ldl (int, long int *)", and
1976 "vector unsigned int vec_ldl (int, unsigned long int *)".
1977 Add comments and entries for
1978 "vector bool char vec_ldl (int, bool char *)",
1979 "vector bool short vec_ldl (int, bool short *)",
1980 "vector bool int vec_ldl (int, bool int *)",
1981 "vector bool long long vec_ldl (int, bool long long *)",
1982 "vector pixel vec_ldl (int, pixel *)",
1983 "vector long long vec_ldl (int, long long *)",
1984 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1985 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1986 type tree bool_long_long_type_node and correct definition of
1987 bool_V2DI_type_node to make reference to this new type tree.
1988 (rs6000_mangle_type): Replace erroneous reference to
1989 bool_long_type_node with bool_long_long_type_node.
1990 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1991 comments to emphasize sign distinctions for char and int types and
1992 replace RS6000_BTI_bool_long constant with
1993 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1994 use of RS6000_BTI_pixel.
1995 (bool_long_type_node): Remove this macro definition.
1996 (bool_long_long_type_node): New macro definition
1998 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2000 PR rtl-optimization/85300
2001 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2002 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2003 simplify_unary_operation fails.
2005 2018-04-10 Martin Liska <mliska@suse.cz>
2007 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2008 cgraph_edge and ipa_ref.
2010 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2014 * config/i386/sse.md
2015 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2016 computation of the VEC_MERGE selector from mask.
2017 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2018 Fix decoding of the VEC_MERGE selector into mask.
2020 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2022 PR tree-optimization/85286
2023 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2025 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2027 * final.c (final_1): Set insn_last_address as well as
2028 insn_current_address.
2030 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2033 * explow.c (emit_stack_probe): Call validize_mem on memory location
2034 before passing it to gen_probe_stack. Create address operand and
2035 legitimize it for the probe_stack_address case.
2037 2018-04-09 Jan Hubicka <jh@suse.cz>
2040 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2041 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2042 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2044 (free_lang_data): Rebuild type inheritance graph.
2046 2018-04-09 Martin Sebor <msebor@redhat.com>
2048 * invoke.texi (-finline-small-functions): Mention other optimization
2050 (-findirect-inlining, -fpartial-inlining): Same.
2051 (-finline-functions-called-once): Same.
2052 (-freorder-blocks-and-partition): Same.
2054 2018-04-09 Jan Hubicka <jh@suse.cz>
2057 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2058 jumps; choose last target that matches the criteria (i.e.
2059 no partition changes for non-crossing jumps).
2060 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2061 support for redirecting crossing jumps to non-crossing.
2063 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2065 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2066 also for naked functions.
2068 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2070 * config/arc/arc.md (add_shift): New pattern.
2071 (add_shift2): Likewise.
2072 (sub_shift): Likewise.
2073 (sub_shift_cmp0_noout): Likewise.
2074 (compare_si_ashiftsi): Likewise.
2075 (xbfu_cmp0_noout): New combine pattern.
2076 (xbfu_cmp0"): Likewise.
2077 (movsi_set_cc_insn): Place the predicable variant first.
2078 (commutative_binary_cmp0_noout): Remove clobber.
2079 (commutative_binary_cmp0): New pattern.
2080 (noncommutative_binary_cmp0): Likewise.
2081 (noncommutative_binary_cmp0_noout): Likewise.
2082 (noncommutative_binary_comparison_result_used): Removed.
2083 (rsub_cmp0): New pattern.
2084 (rsub_cmp0_noout): Likewise.
2085 (extzvsi): Changed, keep only meaningful variants.
2086 (SQH, SEZ): New iterators.
2087 (SQH_postfix): New mode attribute.
2088 (SEZ_prefix): New code attribute.
2089 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2090 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2091 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2093 (noncommutative_operator): Check the availability of barrel
2096 2018-04-09 Richard Biener <rguenther@suse.de>
2098 PR tree-optimization/85284
2099 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2100 Only use the niter constraining form of simple_iv when the exit
2103 2018-04-09 Tom de Vries <tom@codesourcery.com>
2106 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2107 (define_expand "*memory_barrier"): New define_expand.
2108 (define_insn "memory_barrier"): New insn.
2110 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2112 PR rtl-optimization/80463
2113 PR rtl-optimization/83972
2114 PR rtl-optimization/83480
2116 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2117 correct producer for the insn.
2118 (tidy_control_flow): Fixup seqnos in case of debug insns.
2120 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2122 PR rtl-optimization/83913
2124 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2125 different sched-times when merging exprs.
2127 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2129 PR rtl-optimization/83962
2131 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2132 tidy_fallthru_edge and tidy_control_flow.
2134 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2136 PR rtl-optimization/83530
2138 * sel-sched.c (force_next_insn): New global variable.
2139 (remove_insn_for_debug): When force_next_insn is true, also leave only
2140 next insn in the ready list.
2141 (sel_sched_region): When the region wasn't scheduled, make another pass
2142 over it with force_next_insn set to 1.
2144 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2146 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2148 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2149 for interrupt control.
2150 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2151 functions for interrupt control.
2152 * config/nds32/nds32-intrinsic.md: Likewise.
2153 * config/nds32/nds32_intrinsic.h: Likewise.
2154 * config/nds32/nds32.h (nds32_builtins): Likewise.
2156 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2158 * config/nds32/nds32.c (nds32_init_machine_status,
2159 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2160 strict_aligned_p field.
2161 (nds32_expand_to_rtl_hook): New function.
2162 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2163 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2165 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2166 Chung-Ju Wu <jasonwucj@gmail.com>
2168 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2169 * config/nds32/nds32-n7.md: New file.
2170 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2171 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2173 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2174 * config/nds32/nds32.md (pipeline_model): Add n7.
2175 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2176 * config/nds32/pipelines.md: Include n7 settings.
2178 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2179 Chung-Ju Wu <jasonwucj@gmail.com>
2181 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2182 * config/nds32/nds32-e8.md: New file.
2183 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2184 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2186 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2187 * config/nds32/nds32.md (pipeline_model): Add e8.
2188 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2189 * config/nds32/pipelines.md: Include e8 settings.
2191 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2192 Chung-Ju Wu <jasonwucj@gmail.com>
2194 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2195 * config/nds32/nds32-n8.md: New file.
2196 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2197 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2199 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2200 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2201 * config/nds32/nds32.md (pipeline_model): Add n8.
2202 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2203 * config/nds32/pipelines.md: Include n8 settings.
2205 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2206 Chung-Ju Wu <jasonwucj@gmail.com>
2208 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2209 * config/nds32/nds32-n9-2r1w.md: New file.
2210 * config/nds32/nds32-n9-3r2w.md: New file.
2211 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2212 nds32_register_ports): New or modify for cpu n9.
2213 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2215 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2216 * config/nds32/nds32-utils.c: New file.
2217 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2218 TARGET_MUL_SLOW): Define.
2219 * config/nds32/nds32.md (pipeline_model): New attribute.
2220 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2221 New options that support cpu n9.
2222 * config/nds32/pipelines.md: Include n9 settings.
2223 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2225 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2227 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2228 information if necessary.
2229 (output_cond_branch_compare_zero): Likewise.
2230 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2231 (nds32_target_alignment): Refine for alignment.
2232 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2233 (FUNCTION_BOUNDARY): Modify.
2234 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2236 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2238 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2240 * config/nds32/constants.md (unspec_volatile_element): Add values for
2241 TLB operation and data prefetch.
2242 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2243 functions for TLB operation and data prefetch.
2244 * config/nds32/nds32-intrinsic.md: Likewise.
2245 * config/nds32/nds32_intrinsic.h: Likewise.
2246 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2247 (nds32_print_operand): Likewise.
2248 * config/nds32/nds32.h (nds32_builtins): Likewise.
2250 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2251 Andrew Pinski <pinsika@gcc.gnu.org>
2254 * match.pd: Use constant_boolean_node of correct type instead of
2255 boolean_true_node or boolean_false_node for simplifying
2256 pointer comparisons to zero.
2258 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2260 PR tree-optimization/80021
2261 * tree.c (verify_type_variant): Make error call in verify_variant_match
2262 translatable and remove final full stop.
2264 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2266 * config/nds32/constants.md (unspec_volatile_element): Add
2267 UNSPEC_VOLATILE_EH_RETURN.
2268 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2269 nds32_output_stack_pop): Support dwarf exception handling process.
2270 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2271 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2272 exception handling process.
2273 (nds32_compute_stack_frame): Likewise.
2274 (nds32_return_addr_rtx): Likewise.
2275 (nds32_initial_elimination_offset): Likewise.
2276 (nds32_expand_prologue): Likewise.
2277 (nds32_expand_epilogue): Likewise.
2278 (nds32_dynamic_chain_address): New function.
2279 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2281 (DYNAMIC_CHAIN_ADDRESS): Define.
2282 (EH_RETURN_DATA_REGNO): Define.
2283 (EH_RETURN_STACKADJ_RTX): Define.
2284 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2285 patterns for dwarf exception handling.
2287 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2289 * config/nds32/nds32.h: Clean up obsolete macros.
2291 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2293 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2294 Add enum values for particular instructions.
2295 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2296 particular intrinsic functions.
2297 * config/nds32/nds32-intrinsic.md: Likewise.
2298 * config/nds32/nds32_intrinsic.h: Likewise.
2299 * config/nds32/nds32.h (nds32_builtins): Likewise.
2300 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2301 (btst, ave): New patterns for particular instructions.
2303 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2305 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2306 Add enum values for atomic load/store and memory sync.
2307 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2309 * config/nds32/nds32-intrinsic.md: Likewise.
2310 * config/nds32/nds32_intrinsic.h: Likewise.
2311 * config/nds32/nds32.h (nds32_builtins): Likewise.
2313 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2315 PR tree-optimization/85257
2316 * fold-const.c (native_encode_vector): If not all elts could fit
2317 and off is -1, return 0 rather than offset.
2318 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2319 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2320 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2321 adjust buffer in native_interpret_expr call.
2323 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2325 * config/nds32/constants.md (unspec_volatile_element): Add cache
2326 control enum values.
2327 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2328 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2329 * config/nds32/nds32.c (nds32_cctl_names): New.
2330 (nds32_print_operand): Handle cache control register names.
2331 * config/nds32/nds32.h (nds32_builtins): New enum values.
2332 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2334 * config/nds32/nds32.md (type): Add mmu.
2335 * config/nds32/pipelines.md (simple_insn): Add mmu.
2337 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2339 * config/nds32/nds32.md (type): Remove call.
2340 * config/nds32/pipelines.md (simple_insn): Likewise.
2342 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2344 * config/nds32/constants.md (unspec_volatile_element): Add
2345 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2346 UNSPEC_VOLATILE_FMFCFG.
2347 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2348 description for fmfcfg and fmfcsr.
2349 (bdesc_1arg): Add fmtcsr.
2350 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2351 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2352 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2353 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2354 unspec_fmfcfg): New patterns.
2355 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2356 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2357 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2358 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2359 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2360 __nds32__fmfcfg): Define.
2362 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2364 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2365 intrinsic register names.
2366 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2367 intrinsic register enum values and macros.
2369 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2371 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2372 for load/store addressing form.
2373 (nds32_print_operand_address): Likewise.
2375 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2378 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2379 based on LABEL_REF. Remove useless assertion.
2380 (pic_address_needs_scratch): Fix formatting.
2381 (sparc_legitimize_pic_address): Minor tweaks.
2382 (sparc_delegitimize_address): Adjust assertion accordingly.
2383 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2384 into symbolic_operand.
2385 (movsi_high_pic_label_ref): Likewise.
2386 (movsi_lo_sum_pic_label_ref): Likewise.
2387 (movdi_pic_label_ref): Likewise.
2388 (movdi_high_pic_label_ref): Likewise.
2389 (movdi_lo_sum_pic_label_ref): Likewise.
2391 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2393 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2394 custom LIB_SPEC setup.
2396 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2397 Kito Cheng <kito.cheng@gmail.com>
2399 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2400 * config/riscv/freebsd.h: New.
2402 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2404 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2405 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2408 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2409 Kito Cheng <kito.cheng@gmail.com>
2411 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2412 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2413 * config/nds32/nds32-protos.h (nds32_output_call,
2414 nds32_output_return): Declare.
2415 * config/nds32/nds32.md: Refine all the call and return patterns.
2417 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2420 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2421 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2423 PR rtl-optimization/84872
2424 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2425 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2428 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2430 * expr.c (copy_blkmode_to_reg): Revert 254862.
2431 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2433 2018-04-06 Richard Biener <rguenther@suse.de>
2436 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2437 after seeing a component reference with an adjacent field. Treat
2438 refs to arrays at struct end of external decls similar to
2439 refs to unconstrained commons.
2441 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2444 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2445 look through SAVE_EXPRs with non-side-effects argument. Adjust
2447 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2450 2018-04-06 Richard Biener <rguenther@suse.de>
2453 * alias.c (find_base_term): New wrapper around find_base_term
2454 unwinding CSELIB_VAL_PTR changes.
2455 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2458 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2460 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2462 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2463 constant definitions.
2464 ("nop"): lr 0,0 -> nopr r0
2465 ("nop_lr0", "nop_lr1"): New insn definitions.
2467 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2469 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2470 NDS32_V3PUSH_AVAILABLE_P macro.
2472 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2473 Chung-Ju Wu <jasonwucj@gmail.com>
2475 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2476 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2477 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2478 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2479 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2480 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2481 * config/nds32/constraints.md: New constraints and checking for hard
2482 float configuration.
2483 * config/nds32/iterators.md: New mode iterator and attribute for hard
2484 float configuration.
2485 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2487 * config/nds32/nds32-fpu.md: New file.
2488 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2489 deal with hard float code generation.
2490 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2492 (abi_type, float_reg_number): New enum type.
2493 * config/nds32/nds32-predicates.c: New predicates for hard float.
2494 * config/nds32/nds32-protos.h: Declare functions for hard float.
2495 * config/nds32/nds32.c: Implementation for hard float configuration.
2496 * config/nds32/nds32.h: Definitions for hard float configuration.
2497 * config/nds32/nds32.md: Include hard float machine description and
2498 modify patterns for hard float configuration.
2499 * config/nds32/nds32.opt: New options for hard float configuration.
2500 * config/nds32/predicates.md: New predicates for hard float
2503 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2505 * common/config/nds32/nds32-common.c
2506 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2508 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2511 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2512 CONSTRUCTOR_ELT (ctor, ...)->value.
2514 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2517 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2519 2018-04-05 Tom de Vries <tom@codesourcery.com>
2522 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2525 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2526 Kito Cheng <kito.cheng@gmail.com>
2528 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2529 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2530 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2531 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2533 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2534 Kito Cheng <kito.cheng@gmail.com>
2536 * config/nds32/constraints.md (Ufe): New memory constraint.
2537 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2538 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2539 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2541 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2542 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2546 * config/nds32/nds32.md: Use optimize_size in the condition for
2547 alu-shift instructions.
2549 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2551 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2553 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2555 * config/nds32/nds32.md (negsi2): Refine pattern.
2557 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2558 Chung-Ju Wu <jasonwucj@gmail.com>
2560 * config/nds32/iterators.md (shift_rotate): New code iterator.
2561 (shift): New code attribute.
2562 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2563 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2564 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2565 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2566 bit-wise operations.
2567 (andsi3, *andsi3): Ditto.
2568 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2569 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2570 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2571 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2572 nds32_ior_operand, nds32_xor_operand): New predicates.
2574 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2576 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2577 (addsi3, subsi3): ... this.
2579 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2581 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2583 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2585 * config/nds32/nds32.md: Adjust indention.
2587 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2589 * config/nds32/nds32.md (feature): New attribute.
2591 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config/nds32/nds32.md (subtype): New attribute.
2595 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2598 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2599 expansion to perform a bitwise AND of the argument followed by a
2600 boolean negation of the result.
2602 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2604 PR rtl-optimization/84878
2605 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2606 the basic block. Assert the use reference is not artificial and that
2607 it has an associated insn.
2609 2018-04-04 Michael Matz <matz@suse.de>
2611 * builtins.c (compute_objsize): Pass correct operand
2612 to array_at_struct_end_p.
2614 2018-04-04 Richard Biener <rguenther@suse.de>
2617 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2618 from contexts for DINFO_LEVEL_TERSE and below.
2620 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2622 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2623 resiter_operand condition.
2624 * config/nds32/nds32.md (*move<mode>): Ditto.
2626 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2627 Monk Chiang <sh.chiang04@gmail.com>
2629 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2633 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2635 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2636 Kito Cheng <kito.cheng@gmail.com>
2638 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2639 nds32_cond_code_str, output_cond_branch,
2640 output_cond_branch_compare_zero, nds32_expand_cbranch,
2641 nds32_expand_cstore, nds32_expand_movcc,
2642 nds32_output_cbranchsi4_equality_zero,
2643 nds32_output_cbranchsi4_equality_reg,
2644 nds32_output_cbranchsi4_equality_reg_or_const_int,
2645 nds32_output_cbranchsi4_greater_less_zero: New functions.
2646 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2647 nds32_expand_cstore, nds32_expand_movcc,
2648 nds32_output_cbranchsi4_equality_zero,
2649 nds32_output_cbranchsi4_equality_reg,
2650 nds32_output_cbranchsi4_equality_reg_or_const_int,
2651 nds32_output_cbranchsi4_greater_less_zero): Declare.
2652 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2653 nds32_rimm11s_operand): New predicates.
2654 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2655 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2658 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2660 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2661 * config/nds32/nds32.md: Ditto.
2662 * config/nds32/pipelines.md: Ditto.
2664 2018-04-04 Richard Biener <rguenther@suse.de>
2666 PR tree-optimization/85168
2667 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2668 propagating abnormals.
2670 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2672 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2674 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2675 Kito Cheng <kito.cheng@gmail.com>
2677 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2678 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2679 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2680 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2681 * config/nds32/nds32.md (sibcall_internal): New.
2682 (sibcall_register): Remove.
2683 (sibcall_immediate): Remove.
2684 (sibcall_value_internal): New.
2685 (sibcall_value_register): Remove.
2686 (sibcall_value_immediate): Remove.
2687 * config/nds32/predicates.md (nds32_general_register_operand): New.
2688 (nds32_call_address_operand): New.
2690 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2692 PR rtl-optimization/85167
2693 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2694 bb_defs if *split_p, instead preinitialize it to NULL.
2696 PR tree-optimization/85156
2697 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2698 evaluating the argument multiple times.
2700 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2702 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2704 (_mm_cvtpd_ps): Likewise.
2705 (_mm_cvttpd_epi32): Likewise.
2706 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2707 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2708 vector, pixel, and bool following altivec.h include.
2710 2018-04-03 Martin Sebor <msebor@redhat.com>
2712 * doc/extend.texi (Common Function Attributes): Clarify.
2713 (const attribute): Likewise.
2714 (pure attribute): Likewise.
2716 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2719 * config/i386/i386.c (ix86_expand_vector_set): Use
2720 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2722 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2724 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2725 instructions when changing rounding bits to preserve precision bits
2726 in the x87 control word.
2728 2018-04-03 Martin Liska <mliska@suse.cz>
2730 PR tree-optimization/82491
2731 * rtl.h (strip_offset_and_add): Replace += suboffset with
2732 poly_uint64 () + suboffset.
2734 2018-03-29 Martin Liska <mliska@suse.cz>
2735 Martin Jambor <mjambor@suse.cz>
2738 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2739 param_type is not an integral or pointer type.
2741 2018-04-03 Richard Biener <rguenther@suse.de>
2743 * sese.h (recompute_all_dominators): Remove.
2745 2018-04-02 Martin Sebor <msebor@redhat.com>
2747 * doc/invoke.texi (-Wrestrict): Fix typos.
2749 2018-04-02 Jim Wilson <jimw@sifive.com>
2751 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2752 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2753 (<optab>di3, <optab>si3_extend): Likewise.
2754 (<optab>si3_mask, <optab>si3_mask_1): New.
2755 (<optab>di3_mask, <optab>di3_mask_1): New.
2756 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2757 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2758 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2760 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2762 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2765 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2767 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2768 (nds32_canonicalize_comparison): New function.
2770 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2771 Kito Cheng <kito.cheng@gmail.com>
2772 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2774 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2775 * config/nds32/constants.md (unspec_volatile_element): Add
2776 UNSPEC_VOLATILE_RELAX_GROUP.
2777 * config/nds32/nds32-relax-opt.c: New file.
2778 * config/nds32/nds32-predicates.c
2779 (nds32_symbol_load_store_p): New function.
2780 * config/nds32/nds32-protos.h
2781 (nds32_symbol_load_store_p): Declare function.
2782 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2783 * config/nds32/nds32.c
2784 (nds32_register_pass): New function to register pass.
2785 (nds32_register_passes): New function to register passes.
2786 * config/nds32/nds32.md (relax_group): New pattern.
2787 * config/nds32/nds32.opt (mrelax-hint): New option.
2788 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2790 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2792 * config/nds32/t-nds32: Modify files dependency.
2794 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2796 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2797 (PROFILE_HOOK): Define its implementation.
2799 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2801 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2802 type and 32-bit size.
2804 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2807 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2808 (V_128_256): New mode iterator.
2809 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2810 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2811 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2813 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2814 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2816 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2819 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2820 NaN inputs correctly.
2822 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2825 * config/rs6000/vsx.md (??r): New mode attribute.
2826 (*vsx_mov<mode>_64bit): Use it.
2827 (*vsx_mov<mode>_32bit): Likewise.
2829 2018-03-30 Martin Sebor <msebor@redhat.com>
2831 PR tree-optimization/84818
2832 * builtins.c (check_access): Use warning_n.
2834 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2837 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2839 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2842 2018-03-30 Julia Koval <julia.koval@intel.com>
2845 * x86-tune.def (movx, partial_reg_dependency): Enable for
2848 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2851 * lra-constraints.c (process_alt_operands): Move setting
2852 this_alternative_matches below.
2854 2018-03-29 Martin Liska <mliska@suse.cz>
2857 * doc/invoke.texi: Document how LTO works with debug info.
2858 Describe auto-load support of binutils. Mention 'x86-64'
2859 as valid option value of -march option.
2861 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2863 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2866 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2867 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2868 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2871 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2874 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2875 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2876 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2877 for RS6000_BTM_POWERPC64.
2878 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2879 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2880 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2885 2018-03-28 Carl Love <cel@us.ibm.com>
2888 2017-09-27 Carl Love <cel@us.ibm.com>
2890 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2891 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2892 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2895 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2897 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2898 instead of __vector bool.
2899 (_mm_max_pu8): Likewise.
2900 (_mm_min_pi16): Likewise.
2902 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2905 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2906 (DIVWEUO): Likewise.
2908 (DIVDEUO): Likewise.
2909 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2910 DIVWEUO and DIVDEUO.
2911 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2912 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2913 (div_extend): Likewise.
2914 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2916 (__builtin_divweuo): Likewise.
2917 (__builtin_divdeo): Likewise.
2918 (__builtin_divdeuo): Likewise.
2920 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2923 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2924 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2926 PR tree-optimization/82004
2927 * gimple-match-head.c (optimize_pow_to_exp): New function.
2928 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2929 Don't fold to exp if optimize_pow_to_exp is false.
2931 2018-03-28 Martin Liska <mliska@suse.cz>
2934 * calls.c (initialize_argument_information): Fix trailing space.
2935 * common.opt: Fix typo and provide better explanation for
2936 -fsanitize-coverage option.
2937 * config/i386/i386.opt: Fix typo.
2939 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2940 Martin Liska <mliska@suse.cz>
2943 * gimplify.c (asan_poison_variable): Don't do the check for
2944 gimplify_omp_ctxp here.
2945 (gimplify_decl_expr): Do it here.
2946 (gimplify_target_expr): Likewise.
2948 2018-03-28 Martin Liska <mliska@suse.cz>
2951 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2952 chkp_type_bounds_count if MPX is not enabled.
2954 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2956 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2958 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2961 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2962 function to create the function decl for complex long double
2963 multiply and divide for -mabi=ieeelongdouble.
2964 (init_float128_ieee): Call it.
2966 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2969 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2970 -fcf-protection=branch -mibt.
2971 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2973 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2976 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2978 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2981 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2982 extern array declarations.
2984 2018-03-27 Richard Biener <rguenther@suse.de>
2987 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2988 explicit single_use checks.
2990 2018-03-27 Richard Biener <rguenther@suse.de>
2992 PR tree-optimization/85082
2993 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2996 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2998 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2999 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3000 Turn on fasynchronous-unwind-tables and funwind-tables.
3002 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3005 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3006 (*bmi_blsr_<mode>_ccz): Ditto.
3008 2018-03-26 Tom de Vries <tom@codesourcery.com>
3010 PR tree-optimization/85063
3011 * omp-general.c (offloading_function_p): New function. Factor out
3013 * omp-offload.c (pass_omp_target_link::gate): ... here.
3014 * omp-general.h (offloading_function_p): Declare.
3015 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3016 with attribute omp declare target for offloading functions.
3018 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3020 PR tree-optimization/84005
3021 * tree-data-ref.h (get_base_for_alignment): Declare.
3022 * tree-data-ref.c (get_base_for_alignment_1): New function.
3023 (get_base_for_alignment): Likewise.
3024 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3025 get_base_for_alignment to find a suitable base object, instead
3026 of always using drb->base_address.
3028 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3031 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3032 known size by default.
3034 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3037 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3038 and non BLKmode operands.
3040 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3043 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3044 Clean up attributes.
3046 2018-03-23 Richard Biener <rguenther@suse.de>
3049 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3050 we are going to emit early debug for LTO.
3052 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3055 * function.c (match_asm_constraints_1): Don't optimize if input
3056 doesn't satisfy general_operand predicate for output's mode.
3059 * alias.c (write_dependence_p): Don't require for x_canonicalized
3060 non-VOIDmode if x has VOIDmode.
3063 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3064 just don't try to optimize it rather than assert it never happens.
3066 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3068 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3069 macro expansions for definition of ST_INTERNAL_<mode> and
3070 LD_INTERNAL_<mode> builtins.
3071 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3073 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3075 (altivec_expand_st_builtin): Likewise.
3076 (altivec_expand_builtin): Remove calls to deleted functions.
3077 (rs6000_address_for_altivec): Delete this function.
3078 * config/rs6000/vector.md: Remove expands for
3079 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3081 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3084 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3085 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3086 re-computing once computed.
3087 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3088 (arm_init_machine_status): Initialize
3089 machine->static_chain_stack_bytes.
3091 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3094 * doc/extend.texi: Add four new prototypes for vec_ld.
3095 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3096 definitions for more logical presentation.
3097 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3098 entries for V1TI variants of __builtin_altivec_ld builtin.
3099 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3100 handling of V1TI variant of LVX icode pattern.
3101 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3102 (rs6000_gimple_fold_builtin): Likewise.
3103 (altivec_init_builtins): Add code to define
3104 __builtin_altivec_lvx_v1ti function.
3106 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3109 * function.c (match_asm_constraints_1): Don't do the optimization
3110 if input isn't a REG, SUBREG, MEM or constant.
3112 2018-03-22 Tom de Vries <tom@codesourcery.com>
3114 PR tree-optimization/84956
3115 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3116 bb_has_abnormal_pred.
3118 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3121 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3122 DECL_INITIAL (decl) to decl at the end.
3123 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3126 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3128 * doc/extend.texi (__builtin_tgmath): Document when complex
3129 integer types are treated as _Complex _Float64.
3131 2018-03-21 Tom de Vries <tom@codesourcery.com>
3133 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3135 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3137 PR tree-optimization/84960
3138 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3139 if it is ENTRY block, move them into single succ of ENTRY in that case.
3141 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3143 PR tree-optimization/84811
3144 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3145 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3146 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3147 (known_subrange_p): Update accordingly. Cast each value involved
3148 in the size comparison, rather than casting the result of the
3151 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3153 PR tree-optimization/84982
3154 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3155 by flipping the least significant bit rather than all bits from
3156 bitpos to bitpos + bitsize - 1.
3158 2018-03-21 Nathan Sidwell <nathan@acm.org>
3160 * doc/extend.texi (Deprecated Features): Remove mention of
3161 long-deleted deprecations.
3163 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3166 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3167 * configure: Regenerate.
3169 2018-03-21 Tom de Vries <tom@codesourcery.com>
3171 PR tree-optimization/83126
3172 * tree-parloops.c (num_phis): New function.
3173 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3175 2018-03-21 Nathan Sidwell <nathan@acm.org>
3177 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3178 mention anon-struct/union members and trailing attributes.
3180 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3182 PR tree-optimization/84969
3183 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3184 builtin memset partitions if they set different rhs values.
3186 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3188 PR rtl-optimization/84989
3189 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3190 VEC_DUPLICATE with scalar result mode.
3192 2018-03-21 Martin Liska <mliska@suse.cz>
3195 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3196 not intended return statement.
3198 2018-03-21 Martin Liska <mliska@suse.cz>
3201 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3202 (chkp_find_bound_slots_1): Limit number of iterations.
3204 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3207 * Minor grammar fixes for x86 options.
3209 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3212 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3213 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3215 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3218 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3219 (altivec_lvx_<mode>_1op): Likewise.
3220 (altivec_stvx_<mode>_2op): Likewise.
3221 (altivec_stvx_<mode>_1op): Likewise.
3222 (altivec_lvx_<VM2:mode>): New define_expand.
3223 (altivec_stvx_<VM2:mode>): Likewise.
3224 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3225 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3226 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3227 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3228 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3229 (rs6000_gen_lvx): Likewise.
3230 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3231 (altivec_expand_stv_builtin): Likewise.
3232 (altivec_expand_builtin): Likewise.
3233 * config/rs6000/vector.md: Likewise.
3235 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3238 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3241 2018-03-20 Richard Biener <rguenther@suse.de>
3244 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3245 sign-conversions as zero, fall back to standard scalar_stmt
3248 2018-03-20 Martin Liska <mliska@suse.cz>
3251 * predict.c (rebuild_frequencies): Handle case when we have
3252 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3254 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3257 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3258 flag_section_anchors.
3259 * varasm.c (use_blocks_for_decl_p): Remove hack for
3260 dw2_force_const_mem.
3263 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3265 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3266 be created, use lowpart_subreg of operands[0] rather than operands[0]
3268 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3269 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3270 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3271 and n constraint instead of aarch64_shift_imm_di and Usd.
3272 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3273 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3275 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3278 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3279 to favor GPR over NEON registers.
3280 (<shift>di3_neon): Likewise.
3282 2018-03-20 Tom de Vries <tom@codesourcery.com>
3285 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3286 (nvptx_process_pars): Emit bar.sync asap and alap.
3288 2018-03-20 Tom de Vries <tom@codesourcery.com>
3291 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3292 seen_label if seen_label is already set.
3294 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3297 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3298 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3299 Use 1U instead of 1. Formatting fixes.
3302 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3303 instead of TREE_TYPE (s1) for the return value.
3305 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3307 PR tree-optimization/84946
3308 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3309 bitsize + bitsize in poly_uint64 rather than poly_int64.
3312 * dwarf2asm.c: Include fold-const.c.
3313 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3314 of decl rather than decl itself.
3316 PR rtl-optimization/84643
3317 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3319 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3322 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3323 calling assemble_variable.
3325 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3328 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3329 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3331 2018-03-19 Jim Wilson <jimw@sifive.com>
3334 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3335 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3336 (riscv_first_stack_step): Likewise.
3337 (riscv_option_override): Use STACK_BOUNDARY instead of
3339 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3341 (BIGGEST_ALIGNMENT): Set to 128.
3342 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3343 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3346 2018-03-19 Richard Biener <rguenther@suse.de>
3348 PR tree-optimization/84933
3349 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3350 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3352 2018-03-19 Richard Biener <rguenther@suse.de>
3354 PR tree-optimization/84859
3355 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3356 (cond_if_else_store_replacement): Perform sinking operation on
3357 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3358 Generalize what a BB with a single eligible store is.
3360 2018-03-19 Richard Biener <rguenther@suse.de>
3362 PR tree-optimization/84929
3363 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3364 chrec_is_positive against non-chrec arg.
3366 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3369 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3371 2018-03-18 Martin Liska <mliska@suse.cz>
3373 PR rtl-optimization/84635
3374 * regrename.c (build_def_use): Use matches_mode only when
3377 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3379 PR tree-optimization/84913
3380 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3381 vectorize chains of COND_EXPRs.
3383 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3387 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3389 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3391 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3393 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3395 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3396 Kito Cheng <kito.cheng@gmail.com>
3398 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3399 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3400 (nds32_adjust_reg_alloc_order): New function.
3401 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3403 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3405 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3406 nds32_print_operand, nds32_print_operand_address): Use
3407 HOST_WIDE_INT_PRINT_DEC instead.
3409 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3411 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3413 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3416 * config/i386/i386.c (initial_ix86_tune_features,
3417 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3419 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3420 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3421 rather than 1u << ix86_tune. Formatting fix.
3422 (ix86_option_override_internal): Change ix86_arch_mask from
3423 unsigned int to unsigned HOST_WIDE_INT, initialize to
3424 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3425 (ix86_function_specific_restore): Likewise.
3427 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3430 * postreload.c (reload_combine_recognize_pattern): Perform
3431 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3432 truncate_int_for_mode the result for the destination's mode.
3435 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3436 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3438 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3441 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3444 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3445 regno_allocno_class_array and sorted_pseudos.
3446 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3447 insns where regno is used.
3449 2018-03-16 Martin Liska <mliska@suse.cz>
3452 * multiple_target.c (create_dispatcher_calls): Redirect
3453 reference in the symbol table.
3455 2018-03-16 Martin Liska <mliska@suse.cz>
3458 * multiple_target.c (create_dispatcher_calls): Redirect also
3461 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3465 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3466 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3467 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3469 2018-03-16 Julia Koval <julia.koval@intel.com>
3471 * doc/invoke.texi (Skylake Server): Add CLWB.
3472 Cannonlake): Remove CLWB.
3474 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3476 PR tree-optimization/84841
3477 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3479 (FLOAT_ONE_CONST_TYPE): Define.
3480 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3481 (sort_by_operand_rank): Put entries with higher constant_type last
3482 rather than first to match comments.
3484 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3486 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3489 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3492 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3493 check and use address of target always.
3495 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3498 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3499 (indirect_thunk_bnd_needed): Likewise.
3500 (indirect_thunks_used): Likewise.
3501 (indirect_thunks_bnd_used): Likewise.
3502 (indirect_return_needed): New.
3503 (indirect_return_bnd_needed): Likewise.
3504 (output_indirect_thunk_function): Add a bool argument for
3506 (output_indirect_thunk_function): Don't generate alias for
3507 function return thunk.
3508 (ix86_code_end): Call output_indirect_thunk_function to generate
3509 function return thunks.
3510 (ix86_output_function_return): Set indirect_return_bnd_needed
3511 and indirect_return_needed instead of indirect_thunk_bnd_needed
3512 and indirect_thunk_needed.
3514 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3516 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3517 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3518 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3520 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3521 Paul Hua <paul.hua.gm@gmail.com>
3524 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3526 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3528 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3529 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3532 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3535 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3536 instead of GET_MODE_SIZE when comparing Units.
3538 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3541 * varasm.c (hash_section): Return an unchangeble hash value
3542 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3543 Return !aarch64_can_use_per_function_literal_pools_p ().
3545 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3548 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3549 rather than address of cmode as last argument to prepare_cmp_insn.
3551 2018-03-15 Julia Koval <julia.koval@intel.com>
3553 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3554 F_AVX512VNNI, F_AVX512BITALG): New.
3556 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3559 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3560 insn for floating-point loads and stores.
3562 2018-03-14 Carl Love <cel@us.ibm.com>
3564 * config/rs6000/rs6000-c.c: Add macro definitions for
3565 ALTIVEC_BUILTIN_VEC_PERMXOR.
3566 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3567 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3568 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3569 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3571 * config/doc/extend.texi: Add prototypes for vec_permxor.
3573 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3576 * diagnostic-show-locus.c (class layout_point): Convert m_line
3577 from int to linenum_type.
3578 (line_span::comparator): Use linenum "compare" function when
3579 comparing line numbers.
3580 (test_line_span): New function.
3581 (layout_range::contains_point): Convert param "row" from int to
3583 (layout_range::intersects_line_p): Likewise.
3584 (layout::will_show_line_p): Likewise.
3585 (layout::print_source_line): Likewise.
3586 (layout::should_print_annotation_line_p): Likewise.
3587 (layout::print_annotation_line): Likewise.
3588 (layout::print_leading_fixits): Likewise.
3589 (layout::annotation_line_showed_range_p): Likewise.
3590 (struct line_corrections): Likewise for field m_row.
3591 (line_corrections::line_corrections): Likewise for param "row".
3592 (layout::print_trailing_fixits): Likewise.
3593 (layout::get_state_at_point): Likewise.
3594 (layout::get_x_bound_for_row): Likewise.
3595 (layout::print_line): Likewise.
3596 (diagnostic_show_locus): Likewise for locals "last_line" and
3598 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3599 * input.c (selftest::test_linenum_comparisons): New function.
3600 (selftest::input_c_tests): Call it.
3601 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3602 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3603 * selftest.h (ASSERT_GT): New macro.
3604 (ASSERT_GT_AT): New macro.
3605 (ASSERT_LT): New macro.
3606 (ASSERT_LT_AT): New macro.
3608 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3610 PR rtl-optimization/84780
3611 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3613 2018-03-14 Martin Liska <mliska@suse.cz>
3615 * tree.c (record_node_allocation_statistics): Use
3616 get_stats_node_kind.
3617 (get_stats_node_kind): New function extracted from
3618 record_node_allocation_statistics.
3619 (free_node): Use get_stats_node_kind.
3621 2018-03-14 Richard Biener <rguenther@suse.de>
3623 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3624 that the value-set of ANTIC_IN doesn't grow.
3627 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3629 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3630 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3632 2018-03-14 Julia Koval <julia.koval@intel.com>
3634 * config.gcc (icelake-client, icelake-server): New.
3636 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3637 (initial_ix86_arch_features): Ditto.
3638 (PTA_SKYLAKE): Add SGX.
3639 (PTA_ICELAKE): Remove.
3640 (PTA_ICELAKE_CLIENT): New.
3641 (PTA_ICELAKE_SERVER): New.
3642 (ix86_option_override_internal): Split up icelake on icelake client and
3644 (get_builtin_code_for_version): Ditto.
3645 (fold_builtin_cpu): Ditto.
3646 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3647 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3648 * config/i386/i386.h (processor_type): Ditto.
3649 * doc/invoke.texi: Ditto.
3651 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3654 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3655 INTEGER_CST offset, add it together with bitpos / 8 and
3656 sign extend based on POINTER_SIZE.
3660 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3663 * config/i386/constraints.md (Yc): New register constraint.
3664 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3665 Use Yc constraint for alternative 2 of operand 0. Remove
3666 preferred_for_speed attribute.
3668 2018-03-14 Richard Biener <rguenther@suse.de>
3670 PR tree-optimization/84830
3671 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3672 with the old one to avoid oscillations.
3674 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3677 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3679 (assign_by_spills): Return a flag of reload assignment failure.
3680 Do not process the reload assignment failures. Do not spill other
3681 reload pseudos if they has the same reg class. Update n if
3683 (lra_assign): Add a return arg. Set up from the result of
3684 assign_by_spills call.
3685 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3686 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3687 usage_insns if it is not NULL.
3688 (spill_hard_reg_in_range): New function.
3689 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3690 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3691 function prototypes.
3692 (lra_assign): Change prototype.
3693 * lra.c (lra): Add code to deal with fails by splitting hard reg
3696 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3698 * config/riscv/riscv.opt (mrelax): New option.
3699 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3700 "norelax" when riscv_mrelax is disabled.
3701 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3703 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3706 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3707 reassociation for int modes.
3709 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3711 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3712 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3714 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3715 * config/aarch64/aarch64-sve.md
3716 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3717 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3718 (*extend<mode><Vwide>2): Rename to...
3719 (aarch64_sve_extend<mode><Vwide>2): ...this.
3720 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3721 renaming the old pattern to...
3722 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3724 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3725 define_expand, renaming the old pattern to...
3726 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3727 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3728 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3729 account when deciding which SVE instruction the optab should use.
3730 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3732 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3734 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3735 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3736 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3737 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3738 (tlsdesc_small_<mode>): Turn a define_expand and use
3739 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3740 (tlsdesc_small_advsimd_<mode>): ...this.
3741 (tlsdesc_small_sve_<mode>): New pattern.
3743 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3745 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3746 (UNSPEC_UMUL_HIGHPART): New constants.
3747 (MUL_HIGHPART): New int iteraor.
3748 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3749 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3751 (*<su>mul<mode>3_highpart): New define_insn.
3753 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3756 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3759 2018-03-13 Martin Liska <mliska@suse.cz>
3762 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3764 (sem_item_optimizer::~sem_item_optimizer): Release it.
3765 (sem_item_optimizer::merge_classes): Register variable aliases.
3766 (sem_item_optimizer::fixup_pt_set): New function.
3767 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3768 * ipa-icf.h: Declare new variables and functions.
3770 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3773 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3774 integer_pow2p@2 and test integer_pow2p in condition.
3775 (A < 0 ? C : 0): Similarly for @1.
3778 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3779 characters starting at p contain '\0' character, don't look beyond
3783 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3784 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3787 * reg-stack.c (change_stack): Change update_end var from int to
3788 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3789 also call set_block_for_insn on the newly added insns and rescan.
3792 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3793 on the last operand.
3796 * tree.c (stabilize_reference_1): Return save_expr (e) for
3797 STATEMENT_LIST even if it doesn't have side-effects.
3799 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3801 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3803 2018-03-12 Renlin Li <renlin.li@arm.com>
3805 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3806 aarch64_output_scalar_simd_mov_immediate.
3808 2018-03-12 Martin Sebor <msebor@redhat.com>
3810 PR tree-optimization/83456
3811 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3812 for perfectly overlapping calls to memcpy.
3813 (gimple_fold_builtin_memory_chk): Same.
3814 (gimple_fold_builtin_strcpy): Handle no-warning.
3815 (gimple_fold_builtin_stxcpy_chk): Same.
3816 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3818 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3820 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3821 parameter. Use it for SFmode.
3822 (rs6000_function_arg_advance_1): Adjust.
3823 (rs6000_function_arg): Adjust.
3824 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3826 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3828 PR rtl-optimization/84169
3829 PR rtl-optimization/84780
3830 * combine.c (can_combine_p): Check for a 2-insn combination whether
3831 the destination register is used between the two insns, too.
3833 2018-03-12 Richard Biener <rguenther@suse.de>
3835 PR tree-optimization/84803
3836 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3837 for refs DR analysis didn't process.
3839 2018-03-12 Richard Biener <rguenther@suse.de>
3841 PR tree-optimization/84777
3842 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3843 force-vectorize loops ignore whether we are optimizing for size.
3845 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3847 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3848 (TARGET_MD_ASM_ADJUST): Define.
3850 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3851 Kito Cheng <kito.cheng@gmail.com>
3852 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32.c (nds32_compute_stack_frame,
3855 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3856 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3857 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3858 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3859 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3860 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3861 * config/nds32/nds32.md (prologue, epilogue): Use macro
3862 NDS32_V3PUSH_AVAILABLE_P to do checking.
3864 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3867 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3868 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3869 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3870 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3871 addition of most attributes on !orig_type_die or the attribute not
3872 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3874 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3875 Chung-Ju Wu <jasonwucj@gmail.com>
3877 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3879 * config/nds32/nds32.opt (mvh): New option.
3881 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3882 Chung-Ju Wu <jasonwucj@gmail.com>
3884 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3886 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3887 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3890 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3891 Chung-Ju Wu <jasonwucj@gmail.com>
3893 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3895 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3896 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3898 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3899 Kito Cheng <kito.cheng@gmail.com>
3900 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3903 UNSPEC_FFMISM and UNSPEC_FLMISM.
3904 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3905 for ffb, ffmism and flmism.
3906 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3907 (unspec_ffmism): Ditto.
3908 (unspec_flmism): Ditto.
3909 (nds32_expand_builtin_impl): Check if string extension is available.
3910 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3911 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3913 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3916 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3919 * lra-assigns.c (assign_by_spills): Return a flag of reload
3920 assignment failure. Do not process the reload assignment
3921 failures. Do not spill other reload pseudos if they has the same
3923 (lra_assign): Add a return arg. Set up from the result of
3924 assign_by_spills call.
3925 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3926 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3927 usage_insns if it is not NULL.
3928 (spill_hard_reg_in_range): New function.
3929 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3930 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3931 function prototypes.
3932 (lra_assign): Change prototype.
3933 * lra.c (lra): Add code to deal with fails by splitting hard reg
3936 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3939 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3941 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3944 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3945 (dw_val_node): Add val_symbolic_view.
3946 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3947 (symview_upper_bound): New.
3948 (new_line_info_table): Initialize symviews_since_reset.
3949 (dwarf2out_source_line): Count symviews_since_reset and set
3950 symview_upper_bound.
3951 (dw_val_equal_p): Handle symview.
3952 (add_AT_symview): New.
3953 (print_dw_val): Handle symview.
3954 (attr_checksum, attr_checksum_ordered): Likewise.
3955 (same_dw_val_p, size_of_die): Likewise.
3956 (value_format, output_die): Likewise.
3957 (add_high_low_attributes): Use add_AT_symview for entry_view.
3958 (dwarf2out_finish): Reset symview_upper_bound, clear
3961 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3964 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3965 Add strict argument and use it.
3966 (rs6000_split_multireg_move): Update for new strict argument.
3967 (mem_operand_gpr): Disallow all non-offsettable addresses.
3968 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3970 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3973 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3974 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3975 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3978 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3979 decl, use remap_type if we want to use the type.
3981 2018-03-09 Martin Sebor <msebor@redhat.com>
3983 PR tree-optimization/84526
3984 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3986 (builtin_access::generic_overlap): Be prepared to handle non-array
3989 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3991 PR rtl-optimization/84682
3992 * lra-constraints.c (process_address_1): Check is_address flag
3993 for address constraints.
3994 (process_alt_operands): Likewise.
3995 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3996 preprocess_constraints.
3997 * recog.h (preprocess_constraints): Add oploc parameter.
3999 * recog.c (preprocess_constraints): Test address_operand for
4000 CT_ADDRESS constraints.
4002 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4005 * lra-assigns.c (assign_by_spills): Return a flag of reload
4006 assignment failure. Do not process the reload assignment
4007 failures. Do not spill other reload pseudos if they has the same
4009 (lra_assign): Add a return arg. Set up from the result of
4010 assign_by_spills call.
4011 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4012 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4013 usage_insns if it is not NULL.
4014 (spill_hard_reg_in_range): New function.
4015 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4016 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4017 function prototypes.
4018 (lra_assign): Change prototype.
4019 * lra.c (lra): Add code to deal with fails by splitting hard reg
4022 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4025 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4026 Accept complain bool parameter. Only emit errors if it is true.
4027 (arm_parse_cpu_option_name): Likewise.
4028 (arm_target_thumb_only): Adjust callers of the above.
4029 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4030 prototype to take a default true bool parameter.
4031 (arm_parse_arch_option_name): Likewise.
4033 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4034 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4038 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4039 * configure: Regenerate.
4040 * configure.ac ("linker --version-script option"): New.
4041 ("linker soname option"): New.
4043 2018-03-09 Richard Biener <rguenther@suse.de>
4045 PR tree-optimization/84775
4046 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4047 immediate uses of predicate stmts and mark them modified.
4050 PR tree-optimization/84178
4051 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4053 (version_loop_for_if_conversion): Delay update_ssa call.
4054 (tree_if_conversion): Delay update_ssa until after predicate
4057 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4060 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4061 when the function accesses prior frames.
4063 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4066 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4067 gen_llsym, otherwise call maybe_gen_llsym.
4070 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4071 has ',' character inside of it.
4073 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4076 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4077 as clobbering CC_REGNUM.
4079 2018-03-08 Richard Biener <rguenther@suse.de>
4082 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4083 (follow_copies_to_constant): Do not follow SSA names registered
4086 2018-03-08 Richard Biener <rguenther@suse.de>
4088 PR tree-optimization/84178
4089 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4091 (version_loop_for_if_conversion): Delay update_ssa call.
4092 (tree_if_conversion): Delay update_ssa until after predicate
4095 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4097 PR tree-optimization/84178
4098 * tree-if-conv.c (release_bb_predicate): Remove the
4099 the assertion that the stmts have NULL use_ops.
4100 Discard the statements, asserting that they haven't
4101 yet been added to a BB.
4103 2018-03-08 Richard Biener <rguenther@suse.de>
4105 PR tree-optimization/84746
4106 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4107 (phi_translate): Pass in destination ANTIC_OUT set.
4108 (phi_translate_1): Likewise. For a simplified result lookup
4109 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4110 (phi_translate_set): Adjust.
4111 (do_pre_regular_insertion): Likewise.
4112 (do_pre_partial_partial_insertion): Likewise.
4114 2018-03-08 Martin Liska <mliska@suse.cz>
4116 PR gcov-profile/84735
4117 * doc/gcov.texi: Document usage of profile files.
4118 * gcov-io.h: Document changes in the format.
4120 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4124 * dwarf2out.c (struct dw_line_info_table): Update comments for
4126 (FORCE_RESET_NEXT_VIEW): New.
4127 (FORCE_RESETTING_VIEW_P): New.
4128 (RESETTING_VIEW_P): Check for -1 too.
4129 (ZERO_VIEW_P): Likewise.
4130 (new_line_info_table): Force-reset next view.
4131 (dwarf2out_begin_function): Likewise.
4132 (dwarf2out_source_line): Simplify zero_view_p initialization.
4133 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4134 view directly. Omit view when omitting .loc at line 0.
4136 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4138 PR tree-optimization/84740
4139 * tree-switch-conversion.c (process_switch): Call build_constructors
4140 only if info.phi_count is non-zero.
4142 PR tree-optimization/84739
4143 * tree-tailcall.c (find_tail_calls): Check call arguments against
4144 DECL_ARGUMENTS (current_function_decl) rather than
4145 DECL_ARGUMENTS (func) when checking for tail recursion.
4147 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4149 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4150 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4151 Volker Reichelt's entry and add entries for people that perform
4152 GCC fuzzy testing and report numerous bugs.
4154 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4157 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4158 readonly data in sdata, if that is disabled.
4159 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4160 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4161 -mreadonly-in-sdata option.
4163 2018-03-07 Martin Sebor <msebor@redhat.com>
4165 PR tree-optimization/84468
4166 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4167 basic block when looking for nul assignment.
4169 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4172 * except.h (output_function_exception_table): Adjust prototype.
4173 * except.c (output_function_exception_table): Remove FNNAME parameter
4174 and add SECTION parameter. Ouput one part of the table at a time.
4175 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4176 the first part of the exception table and emit unwind directives.
4177 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4178 (i386_pe_seh_cold_init): Likewise.
4179 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4180 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4181 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4182 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4183 * config/i386/winnt.c: Include except.h.
4184 (struct seh_frame_state): Add reg_offset, after_prologue and
4185 in_cold_section fields.
4186 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4187 (i386_pe_seh_cold_init): New function.
4188 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4189 to seh->in_cold_section.
4190 (seh_emit_push): Record the offset of the push.
4191 (seh_emit_save): Record the offet of the save.
4192 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4193 Test seh->after_prologue to disregard the epilogue.
4194 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4195 (i386_pe_end_cold_function): New function.
4197 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4200 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4201 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4204 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4206 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4207 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4210 * multiple_target.c: Include tree-inline.h and intl.h.
4211 (expand_target_clones): Diagnose and fail if node->definition and
4212 !tree_versionable_function_p (node->decl).
4214 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4216 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4218 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4219 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4220 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4222 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4225 * combine.c (try_combine): Use reg_or_subregno instead of handling
4226 just paradoxical SUBREGs and REGs.
4228 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4230 * config/arc/arc.c (arc_finalize_pic): Remove function.
4231 (arc_must_save_register): We use single base PIC register, remove
4232 checks to save/restore the PIC register.
4233 (arc_expand_prologue): Likewise.
4234 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4236 (arc_verify_short): Likewise.
4237 (arc_attr_type): Likewise.
4238 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4239 (walk_stores): Likewise.
4240 (arc_address_cost): Make it static.
4241 (arc_verify_short): Likewise.
4242 (branch_dest): Likewise.
4243 (arc_attr_type): Likewise.
4244 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4245 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4246 (arc_final_prescan_insn): Remove inserting the nops due to
4247 hardware hazards. It is done in reorg step.
4248 (insn_length_variant_t): Remove.
4249 (insn_length_parameters_t): Likewise.
4250 (arc_insn_length_parameters): Likewise.
4251 (arc_get_insn_variants): Likewise.
4252 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4254 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4257 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4260 PR tree-optimization/84687
4261 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4263 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4265 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4267 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4268 Rename to ppc_speculation_barrier.
4269 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4270 __builtin_ppc_speculation_barrier.
4272 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4275 * combine.c (combine_simplify_rtx): Don't try to simplify if
4276 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4279 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4281 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4282 to 32 bytes when compiling for POWER9.
4284 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4287 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4288 regparm >= 3 with no arg reg available also for calls with
4289 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4292 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4294 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4296 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4299 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4301 2018-03-05 Richard Biener <rguenther@suse.de>
4303 PR tree-optimization/84486
4304 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4305 When inserting a __builtin_assume_aligned call set the LHS
4306 SSA name alignment info accordingly.
4308 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4310 PR tree-optimization/84114
4311 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4312 Avoid reassociation of FLOAT_MODE addition.
4314 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4316 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4317 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4318 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4319 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4320 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4321 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4322 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4324 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4325 __builtin_ia32_wbinvd): New builtins.
4326 (SPECIAL_ARGS2): New.
4327 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4328 (SPECIAL_ARGS2): New.
4329 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4330 (ix86_valid_target_attribute_inner_p): Ditto.
4331 (ix86_init_mmx_sse_builtins): Add special_args2.
4332 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4333 TARGET_WBNOINVD_P): New.
4334 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4335 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4336 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4337 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4338 * config/i386/pconfigintrin.h: New file.
4339 * config/i386/wbnoinvdintrin.h: Ditto.
4340 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4341 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4343 2018-03-05 Richard Biener <rguenther@suse.de>
4345 PR tree-optimization/84670
4346 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4348 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4349 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4350 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4351 doesn't grow if all successors (recursively) were visited at least
4354 2018-03-05 Richard Biener <rguenther@suse.de>
4356 PR tree-optimization/84650
4357 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4358 if executed in the loop pipeline.
4360 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4362 * doc/configfiles.texi (Configuration Files): Move info about
4363 conditionalizing $target-protos.h to...
4364 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4365 differs from $target-protos.h.
4367 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4368 Chung-Ju Wu <jasonwucj@gmail.com>
4370 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4371 * config/nds32/nds32-multiple.md (setmemsi): Define.
4372 * config/nds32/nds32-memory-manipulation.c
4373 (nds32_gen_dup_4_byte_to_word_value): New.
4374 (emit_setmem_word_loop): New.
4375 (emit_setmem_byte_loop): New.
4376 (nds32_expand_setmem_loop): New.
4377 (nds32_expand_setmem_loop_v3m): New.
4378 (nds32_expand_setmem_unroll): New.
4379 (nds32_expand_setmem): New.
4381 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4382 Chung-Ju Wu <jasonwucj@gmail.com>
4384 * config/nds32/nds32-memory-manipulation.c
4385 (nds32_emit_load_store): New.
4386 (nds32_emit_post_inc_load_store): New.
4387 (nds32_emit_mem_move): New.
4388 (nds32_emit_mem_move_block): New.
4389 (nds32_expand_movmemsi_loop_unknown_size): New.
4390 (nds32_expand_movmemsi_loop_known_size): New.
4391 (nds32_expand_movmemsi_loop): New.
4392 (nds32_expand_movmemsi_unroll): New.
4393 (nds32_expand_movmemqi): Rename ...
4394 (nds32_expand_movmemsi): ... to this.
4395 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4396 (movmemsi): ... to this.
4397 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4398 (nds32_expand_movmemsi): ... to this.
4400 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4401 Monk Chiang <sh.chiang04@gmail.com>
4402 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32-protos.h
4405 (nds32_expand_load_multiple): New arguments.
4406 (nds32_expand_store_multiple): Ditto.
4407 (nds32_valid_multiple_load_store): Rename ...
4408 (nds32_valid_multiple_load_store_p): ... to this.
4409 * config/nds32/nds32-memory-manipulation.c
4410 (nds32_expand_load_multiple): Refine implementation.
4411 (nds32_expand_store_multiple): Ditto.
4412 * config/nds32/nds32-multiple.md
4413 (load_multiple): Update nds32_expand_load_multiple interface.
4414 (store_multiple): Update nds32_expand_store_multiple interface.
4415 * config/nds32/nds32-predicates.c
4416 (nds32_valid_multiple_load_store): Rename ...
4417 (nds32_valid_multiple_load_store_p): ... to this and refine
4419 * config/nds32/predicates.md
4420 (nds32_load_multiple_and_update_address_operation): New predicate.
4421 (nds32_store_multiple_and_update_address_operation): New predicate.
4423 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4424 Chung-Ju Wu <jasonwucj@gmail.com>
4426 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4427 (combo): New attribute.
4428 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4430 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4432 * config/nds32/nds32.opt: Change -mcmodel= default value.
4434 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4435 Monk Chiang <sh.chiang04@gmail.com>
4436 Chung-Ju Wu <jasonwucj@gmail.com>
4438 * config/nds32/constants.md (unspec_element): New enum.
4439 * config/nds32/constraints.md (Umw): New constraint.
4440 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4441 * config/nds32/nds32-intrinsic.md: Likewise.
4442 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4443 (nds32_valid_smw_lwm_base_p): New.
4444 (nds32_output_smw_single_word): New.
4445 (nds32_output_lmw_single_word): New.
4446 (nds32_expand_unaligned_load): New.
4447 (nds32_expand_unaligned_store): New.
4448 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4449 (nds32_output_smw_single_word): Declare.
4450 (nds32_output_lmw_single_word): Declare.
4451 (nds32_expand_unaligned_load): Declare.
4452 (nds32_expand_unaligned_store): Declare.
4453 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4454 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4455 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4456 NDS32_BUILTIN_UASTORE_DW.
4457 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4460 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4461 Kito Cheng <kito.cheng@gmail.com>
4462 Chung-Ju Wu <jasonwucj@gmail.com>
4464 * config/nds32/nds32-intrinsic.c
4465 (nds32_expand_builtin_null_ftype_reg): Delete.
4466 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4467 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4468 (nds32_read_argument): New.
4469 (nds32_legitimize_target): Ditto.
4470 (nds32_legitimize_argument): Ditto.
4471 (nds32_check_constant_argument): Ditto.
4472 (nds32_expand_unop_builtin): Ditto.
4473 (nds32_expand_unopimm_builtin): Ditto.
4474 (nds32_expand_binop_builtin): Ditto.
4475 (nds32_builtin_decl_impl): Ditto.
4476 (builtin_description): Ditto.
4477 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4478 (nds32_init_builtins_impl): Ditto.
4479 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4480 (nds32_builtin_decl): New.
4481 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4482 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4484 2018-03-02 Jeff Law <law@redhat.com>
4486 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4487 (redundant_insn, fill_simple_delay_slots): Likewise.
4488 (fill_slots_from_thread): Likewise.
4489 * resource.c (mark_referenced_resources): Likewise.
4490 (mark_set_resources, find_dead_or_set_registers): Likewise.
4492 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4494 * substring-locations.h (format_warning_va): Formatting fix for
4496 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4498 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4499 * substring-locations.c: Include intl.h.
4500 (format_warning_va): Turned into small wrapper around
4501 format_warning_n_va, renamed to ...
4502 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4503 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4505 (format_warning_at_substring_n): New function.
4506 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4507 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4508 format_warning_at_substring with just a shorter name instead of
4509 const function pointer.
4510 (fmtwarn_n): New function.
4511 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4512 appropriate, get rid of all the fmtstr temporaries, move conditionals
4513 with G_() wrapped string literals directly into fmtwarn arguments,
4514 cast dir.len to (int), formatting fixes.
4516 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4518 * doc/invoke.texi: Remove "Cilk Plus" references.
4520 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4521 Richard Biener <rguenther@suse.de>
4524 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4525 for error or warning attributes if CALL_FROM_THUNK_P is set.
4528 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4531 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4532 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4535 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4536 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4538 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4539 instead of -1U in last predictors element's probability member.
4541 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4544 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4545 arguments if they are comparable.
4547 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4549 PR tree-optimization/84634
4550 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4551 masks and masked_loop_p with a single loop_masks, making sure it's
4552 null for bb vectorization.
4554 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4556 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4557 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4558 loop->force_vectorize to check whether there is no alias.
4560 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4563 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4565 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4567 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4568 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4569 instead of a loop around prev_real_insn.
4570 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4574 * config/i386/i386.c (ix86_print_operand): Use conditional
4575 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4578 2018-03-02 Richard Biener <rguenther@suse.de>
4580 PR tree-optimization/84427
4581 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4582 (bitmap_set_subtract_values): Rewrite to handle multiple
4585 (prune_clobbered_mems): Likewise.
4586 (phi_translate): Take edge instead of pred/phiblock.
4587 (phi_translate_1): Likewise.
4588 (phi_translate_set): Likewise. Insert all translated
4589 exprs for a value into the set, keeping possibly multiple
4590 expressions per value.
4591 (compute_antic_aux): Adjust for phi_translate changes.
4592 When intersecting union the expressions and prune those
4593 not in the final value set, keeping possibly multiple
4594 expressions per value. Do not use value-insertion
4595 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4596 all expressions. Add verification that the value-sets
4597 only shrink during iteration.
4598 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4599 (do_pre_regular_insertion): Likewise.
4600 (do_pre_partial_partial_insertion): Likewise.
4602 2018-03-02 Richard Biener <rguenther@suse.de>
4605 * config/darwin.c (saved_debug_info_level): New static global.
4606 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4607 (darwin_asm_lto_end): Restore debug info generation settings.
4609 2018-03-01 Martin Liska <mliska@suse.cz>
4612 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4615 2018-03-01 Richard Biener <rguenther@suse.de>
4618 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4619 type annotation with LTO when debug was disabled at compile-time.
4621 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4623 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4625 (mips_final_postscan_insn): Likewise.
4627 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4629 PR rtl-optimization/84528
4630 * alias.c (init_alias_target): Add commentary.
4631 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4632 a unique base value if the frame pointer is not eliminated
4633 to the stack pointer.
4635 2018-03-01 Tom de Vries <tom@codesourcery.com>
4637 PR rtl-optimization/83327
4638 * lra-int.h (hard_regs_spilled_into): Declare.
4639 * lra.c (hard_regs_spilled_into): Define.
4640 (init_reg_info): Init hard_regs_spilled_into.
4641 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4642 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4643 (process_bb_lives): Handle hard_regs_spilled_into.
4644 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4645 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4647 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4649 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4650 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4651 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4652 * config/rs6000/aix72.h: New file.
4654 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4656 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4657 instead of warning_at with conditional singular and plural messages
4661 * stor-layout.c (update_alignment_for_field): For
4662 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4663 && !DECL_PACKED (field), do the alignment update, just use
4664 only desired_align instead of MAX (type_align, desired_align)
4666 (place_field): Don't do known_align < desired_align handling
4667 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4668 is non-NULL, instead do it after rli->prev_field handling and
4669 only if not within a bitfield word. For DECL_PACKED (field)
4670 use type_align of BITS_PER_UNIT.
4672 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4674 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4675 superfluous parentheses and trailing spaces.
4677 2018-02-28 Richard Biener <rguenther@suse.de>
4679 PR tree-optimization/84584
4680 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4681 SCoPs with fake exit edge.
4683 2018-02-28 Martin Liska <mliska@suse.cz>
4686 * timevar.c (timer::print): Fix format to properly print 100%
4689 2018-02-28 Richard Biener <rguenther@suse.de>
4692 * genmatch.c (capture_info::walk_match): Do not mark
4693 captured expressions without operands as expr_p given
4694 they act more like predicates and should be subject to
4695 "lost tail" side-effect preserving.
4697 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4699 PR rtl-optimization/81611
4700 * auto-inc-dec.c (attempt_change): Move dead note from
4701 mem_insn if it's the next use of regno
4702 (find_address): Take address use of reg holding
4703 non-incremented value. Add parm to limit search to the named
4705 (merge_in_block): Attempt to use a mem insn that is the next
4706 use of the original regno.
4708 2018-02-27 Martin Sebor <msebor@redhat.com>
4711 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4712 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4714 2018-02-27 Martin Sebor <msebor@redhat.com>
4716 PR translation/84207
4717 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4718 n argument to unsigned HOST_WIDE_INT.
4719 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4720 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4721 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4722 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4724 2018-02-27 Richard Biener <rguenther@suse.de>
4726 PR tree-optimization/84512
4727 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4728 Do not use the estimate returned from record_stmt_cost for
4729 the scalar iteration cost but sum properly using add_stmt_cost.
4731 2018-02-27 Richard Biener <rguenther@suse.de>
4733 PR tree-optimization/84466
4734 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4735 Adjust last change to less strictly validate use operands.
4737 2018-02-27 Martin Liska <mliska@suse.cz>
4739 PR gcov-profile/84548
4740 * gcov.c (process_file): Allow partial overlap and consider it
4741 also as group functions.
4742 (output_lines): Properly calculate range of lines for a group.
4744 2018-02-27 Martin Liska <mliska@suse.cz>
4746 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4747 'ggc' suffixes. Change first column width.
4748 (timer::print): Fix formatting of the column.
4750 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4752 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4753 preserve inline entry blocks for the sake of debug inline
4754 entry point markers alone.
4755 (remove_unused_locals): Suggest in comments a better place to
4756 force the preservation of inline entry blocks that are
4757 otherwise unused, but do not preserve them.
4759 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4761 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4763 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4766 * config/i386/constraints.md (Bs): Replace
4767 ix86_indirect_branch_register with
4768 TARGET_INDIRECT_BRANCH_REGISTER.
4770 * config/i386/i386.md (indirect_jump): Likewise.
4771 (tablejump): Likewise.
4772 (*sibcall_memory): Likewise.
4773 (*sibcall_value_memory): Likewise.
4774 Peepholes of indirect call and jump via memory: Likewise.
4775 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4776 (*sibcall_value_GOT_32): Likewise.
4777 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4778 (GOT_memory_operand): Likewise.
4779 (call_insn_operand): Likewise.
4780 (sibcall_insn_operand): Likewise.
4781 (GOT32_symbol_operand): Likewise.
4782 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4784 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4786 PR rtl-optimization/83496
4787 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4788 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4789 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4790 redundant insn, if any.
4791 (relax_delay_slots): Likewise.
4792 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4794 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4796 PR tree-optimization/83965
4797 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4798 that grouped statements are part of a reduction chain. Return
4799 true if the statement is not marked as a reduction itself but
4801 (vect_recog_dot_prod_pattern): Don't check whether the statement
4802 is part of a group here.
4803 (vect_recog_sad_pattern): Likewise.
4804 (vect_recog_widen_sum_pattern): Likewise.
4806 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4809 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4811 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4814 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4816 (ix86_output_indirect_function_return): New prototype.
4817 (ix86_split_simple_return_pop_internal): Likewise.
4818 * config/i386/i386.c (indirect_return_via_cx): New.
4819 (indirect_return_via_cx_bnd): Likewise.
4820 (indirect_thunk_name): Handle return va CX_REG.
4821 (output_indirect_thunk_function): Create alias for
4822 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4823 (ix86_output_indirect_jmp): Remove the bool argument.
4824 (ix86_output_indirect_function_return): New function.
4825 (ix86_split_simple_return_pop_internal): Likewise.
4826 * config/i386/i386.md (*indirect_jump): Don't pass false
4827 to ix86_output_indirect_jmp.
4828 (*tablejump_1): Likewise.
4829 (simple_return_pop_internal): Change it to define_insn_and_split.
4830 Call ix86_split_simple_return_pop_internal to split it for
4832 (simple_return_indirect_internal): Call
4833 ix86_output_indirect_function_return instead of
4834 ix86_output_indirect_jmp.
4836 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4839 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4840 memset and value initialization afterwards.
4842 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4844 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4846 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4849 * common/config/aarch64/aarch64-common.c
4850 (aarch_option_optimization_table[]): Switch
4851 off fomit-frame-pointer
4853 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4854 Chung-Ju Wu <jasonwucj@gmail.com>
4856 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4858 (store_multiple): Ditto.
4860 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4862 * config.gcc: Add --with-cpu support for nds32 target.
4863 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4864 * config/nds32/nds32.opt: Add -mcpu= option.
4866 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4868 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4869 isel=yes): Warn for these deprecated options.
4871 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4873 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4874 ISA_2_5_MASKS_EMBEDDED.
4876 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4878 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4879 p->max as pointers rather than using iterative_hash_expr.
4881 2018-02-23 Carl Love <cel@us.ibm.com>
4883 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4884 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4886 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4887 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4888 P8V_BUILTIN_VEC_VUNSIGNED2.
4890 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4893 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4894 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4896 * lra-constraints.c (curr_insn_transform): Set up
4897 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4899 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4900 * lra-eliminations.c (spill_pseudos): Ditto.
4901 (process_insn_for_elimination): Ditto.
4902 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4903 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4904 LRA_NON_CLOBBERED_ALT.
4906 2018-02-22 Martin Sebor <msebor@redhat.com>
4908 PR tree-optimization/84480
4909 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4910 to maybe_diag_stxncpy_trunc. Call it.
4911 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4912 from gimple_fold_builtin_strcpy. Print inlining stack.
4913 (handle_builtin_stxncpy): Print inlining stack.
4914 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4916 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4919 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4920 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4921 and -fcheck-pointer-bounds are used together.
4922 (indirect_thunk_prefix): New enum.
4923 (indirect_thunk_need_prefix): New function.
4924 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4925 "_nt" instead of "_bnd" for NOTRACK prefix.
4926 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4927 (output_indirect_thunk_function): Likewise.
4929 (ix86_code_end): Update output_indirect_thunk_function calls.
4930 (ix86_output_indirect_branch_via_reg): Replace
4931 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4932 (ix86_output_indirect_branch_via_push): Likewise.
4933 (ix86_output_function_return): Likewise.
4934 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4935 incompatible with -fcf-protection=branch and
4936 -fcheck-pointer-bounds.
4938 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4941 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4942 Change gcc_assert call to output_operand_lossage.
4944 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4946 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4948 2018-02-22 DJ Delorie <dj@redhat.com>
4949 Sebastian Perta <sebastian.perta@renesas.com>
4950 Oleg Endo <olegendo@gcc.gnu.org>
4952 * config/rx/rx.c (rx_rtx_costs): New function.
4953 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4955 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4957 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4959 2018-02-22 Martin Liska <mliska@suse.cz>
4962 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4963 Add "native" as a possible value.
4965 2018-02-22 Martin Liska <mliska@suse.cz>
4968 * config/i386/i386.c (ix86_option_override_internal):
4969 Add "native" as a possible value for -march and -mtune.
4971 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4974 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4975 to all type variants.
4977 PR tree-optimization/84503
4978 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4979 width as info->bitpos + info->bitsize - start.
4980 (merged_store_group::merge_overlapping): Simplify width computation.
4981 (check_no_overlap): New function.
4982 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4983 start + width and last_order of the group, fail if check_no_overlap
4985 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4986 to group if check_no_overlap fails.
4988 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4990 * config/rs6000/altivec.md: Delete contraint arguments to
4991 define_expand, define_split, and define_peephole2, and in
4992 define_insn_and_split if always unused.
4993 * config/rs6000/darwin.md: Ditto.
4994 * config/rs6000/dfp.md: Ditto.
4995 * config/rs6000/rs6000.md: Ditto.
4996 * config/rs6000/sync.md: Ditto.
4997 * config/rs6000/vector.md: Ditto.
4998 * config/rs6000/vsx.md: Ditto.
5000 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5002 * config/rs6000/altivec.md: Write output control strings as braced
5003 blocks instead of double-quoted strings.
5004 * config/rs6000/darwin.md: Ditto.
5005 * config/rs6000/rs6000.md: Ditto.
5006 * config/rs6000/vector.md: Ditto.
5007 * config/rs6000/vsx.md: Ditto.
5009 2018-02-21 Jason Merrill <jason@redhat.com>
5011 PR c++/84314 - ICE with templates and fastcall attribute.
5012 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5014 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5016 * ipa-cp.c (determine_versionability): Fix comment typos.
5018 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5021 * ipa-cp.c (determine_versionability): Do not version functions caling
5024 2018-02-21 Martin Liska <mliska@suse.cz>
5027 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5028 Add "native" as a possible value.
5029 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5030 the macro when native cpu detection is available.
5032 2018-02-21 Martin Liska <mliska@suse.cz>
5035 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5036 Add "native" as a possible value.
5037 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5038 when native cpu detection is available.
5040 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5041 Martin Sebor <msebor@redhat.com>
5043 PR tree-optimization/84478
5044 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5046 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5047 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5048 support which is conservatively correct, for 2 only stay conservative
5049 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5050 argument to the 2 argument get_range_strlen, adjust 6 arg
5051 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5053 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5054 (gimple_fold_builtin_strlen): Pass true as last argument to
5057 2018-02-20 Martin Sebor <msebor@redhat.com>
5060 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5061 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5062 (builtin_memref::builtin_memref): Factor out parts into
5063 set_base_and_offset and call it.
5065 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5068 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5069 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5070 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5071 search at the associated MODE_INT.
5073 2018-02-20 Jeff Law <law@redhat.com>
5076 PR tree-optimization/81592
5078 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5079 for range data rather than using global data.
5080 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5081 range data rather than using global data.
5082 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5083 pass it to children as needed.
5084 (struct directive::fmtresult): Similarly.
5085 (struct directive::set_width): Similarly.
5086 (struct directive::set_precision): Similarly.
5087 (format_integer, format_directive, parse_directive): Similarly.
5088 (format_none): Accept unnamed vr_values parameter.
5089 (format_percent, format_floating, format_character): Similarly.
5090 (format_string, format_plain): Similarly.
5091 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5092 the EVRP range analyzer for range data rather than using global data.
5093 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5094 gimple-ssa-evrp-analyze.h
5095 (class sprintf_dom_walker): Add after_dom_children member function.
5096 Add evrp_range_analyzer member.
5097 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5098 range analyzer as needed.
5099 (sprintf_dom_walker::after_dom_children): New member function.
5100 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5102 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5103 (evrp_range_analyzer::pop_to_marker): Likewise.
5105 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5107 PR tree-optimization/84419
5108 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5109 with the required type if its current type is compatible but
5112 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5115 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5116 after vectorization.
5118 2018-02-20 Martin Liska <mliska@suse.cz>
5121 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5122 possible values if we don't have a hint.
5124 2018-02-20 Martin Liska <mliska@suse.cz>
5128 * final.c (shorten_branches): Build align_tab array with one
5130 * opts.c (finish_options): Add alignment option limit check.
5131 (MAX_CODE_ALIGN): Likewise.
5132 (MAX_CODE_ALIGN_VALUE): Likewise.
5133 * doc/invoke.texi: Document maximum allowed option value for
5134 all -falign-* options.
5136 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5139 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5140 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5141 * var-tracking.c (emit_note_insn_var_location): Remove all references
5142 to NOTE_INSN_CALL_ARG_LOCATION.
5143 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5144 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5145 Use copy_rtx_if_shared.
5146 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5147 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5148 (dwarf2out_var_location): Remove handling of
5149 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5151 * final.c (final_scan_insn): Remove all references to
5152 NOTE_INSN_CALL_ARG_LOCATION.
5153 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5154 before dumping final insns.
5155 * except.c (emit_note_eh_region_end): Remove all references to
5156 NOTE_INSN_CALL_ARG_LOCATION.
5157 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5158 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5159 * config/arc/arc.c (hwloop_optimize): Likewise.
5160 * config/arm/arm.c (create_fix_barrier): Likewise.
5161 * config/s390/s390.c (s390_chunkify_start): Likewise.
5162 * config/sh/sh.c (find_barrier): Likewise.
5163 * config/i386/i386.c (rest_of_insert_endbranch,
5164 ix86_seh_fixup_eh_fallthru): Likewise.
5165 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5166 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5167 * config/frv/frv.c (frv_function_prologue): Likewise.
5168 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5170 (note_outside_basic_block_p): Remove all references to
5171 NOTE_INSN_CALL_ARG_LOCATION.
5172 * gengtype.c (adjust_field_rtx_def): Likewise.
5173 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5175 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5176 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5179 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5182 PR tree-optimization/84452
5183 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5184 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5187 2018-02-19 Martin Liska <mliska@suse.cz>
5190 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5192 2018-02-19 Martin Liska <mliska@suse.cz>
5193 Richard Sandiford <richard.sandiford@linaro.org>
5195 PR tree-optimization/82491
5196 * gimple-fold.c (get_base_constructor): Make earlier bail out
5199 2018-02-19 Carl Love <cel@us.ibm.com>
5201 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5202 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5204 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5205 P8V_BUILTIN_VEC_NEG.
5207 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5209 * config/rl78/rl78.md (movdf): New define expand.
5211 2018-02-19 Martin Liska <mliska@suse.cz>
5214 * doc/invoke.texi: Fix typo.
5215 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5217 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5219 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5220 handle rs6000_single_float and rs6000_double_float specially for
5223 2018-02-16 Jeff Law <law@redhat.com>
5225 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5228 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5231 * ipa-inline.c (inline_small_functions): Fix a typo.
5233 2018-02-16 Nathan Sidwell <nathan@acm.org>
5235 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5237 2018-02-16 Carl Love <cel@us.ibm.com>
5239 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5240 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5241 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5242 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5243 expansion to P8V_BUILTIN_VEC_FLOAT2.
5245 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5247 PR rtl-optimization/70023
5248 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5249 src_regno into account.
5251 2018-02-16 Carl Love <cel@us.ibm.com>
5253 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5254 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5255 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5256 * config/rs6000/rs6000.c: Remove case statements for
5257 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5258 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5259 and P9V_BUILTIN_VEC_VINSERT4B.
5260 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5261 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5262 * config/rs6000/vsx.md:
5263 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5266 2018-02-16 Carl Love <cel@us.ibm.com>
5268 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5270 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5272 * config/rs6000/rs6000-c.c: Add the definitions for
5273 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5274 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5275 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5276 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5277 definition for insert4b and define insn *insert3b_internal.
5278 * doc/extend.texi: Add documentation for vec_extract4b.
5280 2018-02-16 Nathan Sidwell <nathan@acm.org>
5282 * doc/extend.texi (Backwards Compatibility): Mention friend
5283 injection. Note for-scope is deprecated.
5284 * doc/invoke.texi (-ffriend-injection): Deprecate.
5286 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5288 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5289 that moved to I2, also allow destinations that are a paradoxical
5290 subreg (instead of a normal reg).
5292 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5295 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5298 2018-02-16 Richard Biener <rguenther@suse.de>
5300 PR tree-optimization/84037
5301 PR tree-optimization/84016
5303 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5304 Adjust vec_construct for the fact we need additional higher latency
5305 128bit inserts for AVX256 and AVX512 vector builds.
5306 (ix86_add_stmt_cost): Scale vector construction cost for
5309 2018-02-16 Richard Biener <rguenther@suse.de>
5311 PR tree-optimization/84417
5312 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5313 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5314 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5316 2018-02-16 Martin Liska <mliska@suse.cz>
5319 * internal-fn.def (ASAN_CHECK): Set proper flags.
5320 (ASAN_MARK): Likewise.
5322 2018-02-16 Julia Koval <julia.koval@intel.com>
5324 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5325 from PTA_CANNONLAKE.
5327 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5330 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5331 Use ++iter rather than iter++ for std::list iterators.
5332 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5333 defer deleting them until all nodes in the forest are processed. Do
5334 free even leaf nodes. Change to_process into auto_vec.
5337 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5338 * vec.h (vec_default_construct): Use memset instead of placement new
5339 if BROKEN_VALUE_INITIALIZATION is defined.
5340 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5341 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5344 PR rtl-optimization/83723
5345 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5346 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5347 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5349 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5351 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5353 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5355 PR rtl-optimization/81443
5356 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5357 from inner REGs to paradoxical SUBREGs.
5359 2018-02-16 Richard Biener <rguenther@suse.de>
5361 PR tree-optimization/84399
5362 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5363 For operands we can analyze at their definition make sure we can
5364 analyze them at each use as well.
5366 2018-02-16 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84190
5369 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5370 volatile accesses if the decl isn't volatile.
5372 2018-02-15 Jason Merrill <jason@redhat.com>
5374 PR c++/84314 - ICE with templates and fastcall attribute.
5375 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5376 TYPE_CANONICAL on an existing type.
5378 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5380 PR tree-optimization/84383
5381 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5382 dstoff nor call operand_equal_p if dstbase is NULL.
5384 PR tree-optimization/84334
5385 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5386 also a CONSTANT_CLASS_P, punt.
5388 2018-02-14 Jim Wilson <jimw@sifive.com>
5390 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5391 first SMALL_OPERAND check. New local min_second_step. Move assert
5392 to where locals are set. Add TARGET_RVC support.
5393 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5395 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5397 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5399 2018-02-14 Martin Sebor <msebor@redhat.com>
5401 PR tree-optimization/83698
5402 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5403 arrays constrain the offset range to their bounds.
5404 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5405 (builtin_access::overlap): Avoid setting the size of overlap if it's
5407 (maybe_diag_overlap): Also consider arrays when deciding what values
5408 of offsets to include in diagnostics.
5410 2018-02-14 Martin Sebor <msebor@redhat.com>
5413 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5414 that correspond to the kind of a declaration.
5416 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5419 * config/pa/pa.md: Load address of PIC label using the linkage table
5420 if the label is nonlocal.
5422 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5424 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5425 warning message if user requests -maltivec=be.
5426 * doc/invoke.texi: Document deprecation of -maltivec=be.
5428 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5431 * config/rs6000/rs6000-c.c: Update definitions for
5432 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5433 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5435 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5438 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5439 add _get_ssp intrinsics. Remove argument from
5440 __builtin_ia32_rdssp[d|q].
5441 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5442 * config/i386/i386-builtin.def: Remove argument from
5443 __builtin_ia32_rdssp[d|q].
5444 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5445 ix86_expand_special_args_builtin for _rdssp[d|q].
5446 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5447 Clear register before usage.
5448 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5449 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5451 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5453 PR tree-optimization/84357
5454 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5455 operand 1 of an ARRAY_REF too.
5457 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5460 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5461 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5463 (set_of_reg): New struct.
5464 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5465 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5466 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5468 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5469 Split into bitclr, bitset, bitinvert patterns if appropriate.
5470 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5471 use rx_fuse_in_memory_bitop.
5472 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5473 to named insn, correct maximum insn length.
5475 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5478 * machmode.def: Define a complex mode for PARTIAL_INT.
5479 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5481 * doc/rtl.texi: Document CSPImode.
5482 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5484 (msp430_hard_regno_nregs_with_padding): Likewise.
5486 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5489 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5491 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5493 PR rtl-optimization/84169
5494 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5495 we generated a parallel as new i3 and we split that to new i2 and i3
5496 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5497 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5498 those to i2, not i1. Partially rewrite this scan code.
5500 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5503 * stor-layout.c (place_field): For variable length fields, adjust
5504 offset_align afterwards not just based on the field's alignment,
5505 but also on the size.
5508 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5509 of exps and logs in the use_exp2 case.
5511 2018-02-13 Jeff Law <law@redhat.com>
5513 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5516 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5519 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5523 * common.opt (gas-loc-support, gas-locview-support): New.
5524 (ginline-points, ginternal-reset-location-views): New.
5525 * doc/invoke.texi: Document them. Use @itemx where intended.
5526 (gvariable-location-views): Adjust.
5527 * target.def (reset_location_view): New.
5528 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5529 (TARGET_RESET_LOCATION_VIEW): New.
5530 * doc/tm.texi: Rebuilt.
5531 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5532 (dwarf2out_default_as_locview_support): New.
5533 (output_asm_line_debug_info): Use option variables.
5534 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5535 (output_loc_list): Likewise.
5536 (add_high_low_attributes): Check option variables.
5537 Don't output entry view attribute in strict mode.
5538 (gen_inlined_subroutine_die): Check option variables.
5539 (dwarf2out_inline_entry): Likewise.
5540 (init_sections_and_labels): Likewise.
5541 (dwarf2out_early_finish): Likewise.
5542 (maybe_reset_location_view): New, from...
5543 (dwarf2out_var_location): ... here. Call it.
5544 * debug.h (dwarf2out_default_as_loc_support): Declare.
5545 (dwarf2out_default_as_locview_support): Declare.
5546 * hooks.c (hook_int_rtx_insn_0): New.
5547 * hooks.h (hook_int_rtx_insn_0): Declare.
5548 * toplev.c (process_options): Take -gas-loc-support and
5549 -gas-locview-support from dwarf2out. Enable
5550 -gvariable-location-views by default only with locview
5551 assembler support. Enable -ginternal-reset-location-views by
5552 default only if the target defines the corresponding hook.
5553 Enable -ginline-points by default if location views are
5554 enabled; force it disabled if statement frontiers are
5556 * tree-inline.c (expand_call_inline): Check option variables.
5557 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5559 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5561 PR tree-optimization/84321
5562 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5563 handling. Also check whether the anti-range contains any values
5564 that satisfy the mask; switch to a VR_RANGE if not.
5566 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5569 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5571 2018-02-13 Martin Jambor <mjambor@suse.cz>
5574 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5575 of call statements, also set location of a load to a temporary.
5577 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5579 * config/rl78/rl78.c (add_vector_labels): New function.
5580 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5581 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5582 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5583 which checks that no arguments are passed.
5584 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5585 * doc/extend.texi: Documentation for the new attribute.
5587 2018-02-13 Andreas Schwab <schwab@suse.de>
5589 * config/riscv/linux.h (CPP_SPEC): Define.
5591 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5594 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5595 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5596 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5597 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5598 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5599 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5600 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5601 temporarily for AES and PCLMUL builtins.
5603 PR tree-optimization/84339
5604 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5605 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5609 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5610 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5611 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5613 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5615 * omp-simd-clone.h: New file.
5616 * omp-simd-clone.c: Include omp-simd-clone.h.
5617 (expand_simd_clones): No longer static.
5618 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5619 cgraph.h and omp-simd-clone.h.
5620 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5621 (vect_recog_widen_shift_pattern): Formatting fix.
5622 (vect_pattern_recog_1): Don't check optab for calls.
5625 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5626 operands[2] into a REG before using gen_lowpart on it.
5628 2018-02-12 Jeff Law <law@redhat.com>
5631 * config/sh/sh.c (find_barrier): Consider a sibling call
5634 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5635 successfully back substituting a reg.
5637 2018-02-12 Richard Biener <rguenther@suse.de>
5639 PR tree-optimization/84037
5640 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5641 parameter, move visited init to caller.
5642 (vect_slp_analyze_operations): Separate cost from validity
5643 check, initialize visited once for all instances.
5644 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5646 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5648 (vect_model_promotion_demotion_cost): Likewise.
5649 (vectorizable_bswap): Guard cost modeling with !slp_node
5650 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5652 (vectorizable_call): Likewise.
5653 (vectorizable_conversion): Likewise.
5654 (vectorizable_assignment): Likewise.
5655 (vectorizable_shift): Likewise.
5656 (vectorizable_operation): Likewise.
5657 (vectorizable_store): Likewise.
5658 (vectorizable_load): Likewise.
5659 (vectorizable_condition): Likewise.
5660 (vectorizable_comparison): Likewise.
5662 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5665 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5666 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5667 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5669 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5672 * params.def (inline-min-speedup): Increase from 8 to 15.
5673 (max-inline-insns-auto): Decrease from 40 to 30.
5674 * ipa-split.c (consider_split): Add some buffer for function to
5675 be considered inlining candidate.
5676 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5679 2018-02-12 Richard Biener <rguenther@suse.de>
5681 PR tree-optimization/84037
5682 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5683 matched stmts if we cannot swap the non-matched ones.
5685 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5687 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5688 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5689 _mm_maskz_scalef_round_ss): New intrinsics.
5690 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5691 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5692 __builtin_ia32_scalefss_round): Remove.
5693 (__builtin_ia32_scalefsd_mask_round,
5694 __builtin_ia32_scalefss_mask_round): New intrinsics.
5695 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5696 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5697 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5698 "<round_constraint>")): Changed to ...
5699 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5700 "<round_scalar_constraint>")): ... this.
5701 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5702 %0, %1, %2<round_op3>}"): Changed to ...
5703 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5704 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5705 %2<round_scalar_mask_op3>}"): ... this.
5706 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5708 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5710 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5711 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5712 (_mm_maskz_sqrt_round_ss): New intrinsics.
5713 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5714 (__builtin_ia32_sqrtsd_mask_round)
5715 (__builtin_ia32_sqrtss_mask_round): New builtins.
5716 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5717 (__builtin_ia32_sqrtss_round): Remove.
5718 (__builtin_ia32_sqrtsd_mask_round)
5719 (__builtin_ia32_sqrtss_mask_round): New builtins.
5720 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5721 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5722 ((match_operand:VF_128 1 "vector_operand"
5723 "xBm,<round_constraint>")): Changed to ...
5724 ((match_operand:VF_128 1 "vector_operand"
5725 "xBm,<round_scalar_constraint>")): ... this.
5726 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5727 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5728 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5729 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5730 %<iptr>1<round_scalar_mask_op3>}): ... this.
5731 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5732 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5734 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5737 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5738 Cast vec_cmpeq result to correct type.
5739 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5740 Cast vec_cmpgt result to correct type.
5742 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5744 * final.c (final_scan_insn_1): Renamed from...
5745 (final_scan_insn): ... this. New wrapper, to recover
5746 seen from the outermost call in recursive ones.
5747 * config/sparc/sparc.c (output_return): Drop seen from call.
5748 (output_sibcall): Likewise.
5749 * config/visium/visium.c (output_branch): Likewise.
5751 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5753 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5756 2018-02-10 Alan Modra <amodra@gmail.com>
5759 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5760 Specify LR as an input.
5762 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5765 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5766 remove_member_access_dummy_vars): New functions.
5767 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5768 lower_omp_1, execute_lower_omp): Use them.
5770 PR rtl-optimization/84308
5771 * shrink-wrap.c (spread_components): Release todo vector.
5773 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5775 PR rtl-optimization/57193
5776 * ira-color.c (struct allocno_color_data): Add member
5777 conflict_allocno_hard_prefs.
5778 (update_conflict_allocno_hard_prefs): New.
5779 (bucket_allocno_compare_func): Add a preference based on
5780 conflict_allocno_hard_prefs.
5781 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5782 (color_allocnos): Remove a dead code. Initiate
5783 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5785 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5788 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5789 constraint from =wa to wa. Avoid a subreg on the output operand,
5790 instead use a pseudo and subreg it in a move.
5791 (p9_xxbrd_<mode>): Changed to ...
5792 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5793 (p9_xxbrd_v2df): New expander.
5794 (p9_xxbrw_<mode>): Changed to ...
5795 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5796 (p9_xxbrw_v4sf): New expander.
5798 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5800 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5802 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5805 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5806 multiply in 32-bit mode.
5807 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5808 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5811 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5813 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5814 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5815 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5816 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5818 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5821 * dwarf2out.c (is_trivial_indirect_ref): New function.
5822 (dwarf2out_late_global_decl): Do not generate a location
5823 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5824 and that are not defined in the current unit.
5826 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5828 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5829 instead of a libcall for UNORDERED.
5831 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5834 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5835 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5837 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5840 * config/s390/s390.c (s390_set_current_function): Invoke
5841 s390_indirect_branch_settings also if fndecl didn't change.
5843 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5845 * config/rs6000/rs6000.md (blockage): Set length to zero.
5847 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5849 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5851 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5854 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5855 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5859 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5860 PARALLEL incoming that failed vt_get_decl_and_offset check.
5863 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5864 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5866 (get_variable_section): For decls in named .bss* sections pass true as
5867 second argument to bss_initializer_p.
5869 2018-02-09 Marek Polacek <polacek@redhat.com>
5870 Jakub Jelinek <jakub@redhat.com>
5873 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5874 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5875 Sync some changes from cxx_fold_indirect_ref.
5877 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5879 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5881 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5882 (BLOCK_INLINE_ENTRY_LABEL): New.
5883 (dwarf2out_var_location): Disregard inline entry markers.
5884 (inline_entry_data): New struct.
5885 (inline_entry_data_hasher): New hashtable type.
5886 (inline_entry_data_hasher::hash): New.
5887 (inline_entry_data_hasher::equal): New.
5888 (inline_entry_data_table): New variable.
5889 (add_high_low_attributes): Add DW_AT_entry_pc and
5890 DW_AT_GNU_entry_view attributes if a pending entry is found
5891 in inline_entry_data_table. Add old entry_pc attribute only
5892 if debug nonbinding markers are disabled.
5893 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5894 markers are enabled.
5895 (block_within_block_p, dwarf2out_inline_entry): New.
5896 (dwarf2out_finish): Check that no entries remained in
5897 inline_entry_data_table.
5898 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5899 (final_scan_insn, notice_source_line): Likewise.
5900 (rest_of_clean_state): Skip inline entry markers.
5901 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5903 * gimple.c (gimple_build_debug_inline_entry): New.
5904 * gimple.h (enum gimple_debug_subcode): Add
5905 GIMPLE_DEBUG_INLINE_ENTRY.
5906 (gimple_build_debug_inline_entry): Declare.
5907 (gimple_debug_inline_entry_p): New.
5908 (gimple_debug_nonbind_marker_p): Adjust.
5909 * insn-notes.def (INLINE_ENTRY): New.
5910 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5911 inline entry marker notes.
5912 (print_insn): Likewise.
5913 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5914 (INSN_DEBUG_MARKER_KIND): Likewise.
5915 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5916 * tree-inline.c (expand_call_inline): Build and insert
5917 debug_inline_entry stmt.
5918 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5919 inline entry blocks early, if nonbind markers are enabled.
5920 (dump_scope_block): Dump fragment info.
5921 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5922 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5923 (gimple_build_debug_inline_entry): New.
5924 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5925 Enable/disable inline entry points too.
5926 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5927 (DEBUG_INSN): Describe inline entry markers.
5929 * common.opt (gvariable-location-views): New.
5930 (gvariable-location-views=incompat5): New.
5931 * config.in: Rebuilt.
5932 * configure: Rebuilt.
5933 * configure.ac: Test assembler for view support.
5934 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5935 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5936 * dwarf2out.c (var_loc_view): New typedef.
5937 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5938 (dwarf2out_locviews_in_attribute): New.
5939 (dwarf2out_locviews_in_loclist): New.
5940 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5941 (enum dw_line_info_opcode): Add LI_adv_address.
5942 (struct dw_line_info_table): Add view.
5943 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5944 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5945 (zero_view_p): New variable.
5946 (ZERO_VIEW_P): New macro.
5947 (output_asm_line_debug_info): New.
5948 (struct var_loc_node): Add view.
5949 (add_AT_view_list, AT_loc_list): New.
5950 (add_var_loc_to_decl): Add view param. Test it against last.
5951 (new_loc_list): Add view params. Record them.
5952 (AT_loc_list_ptr): Handle loc and view lists.
5953 (view_list_to_loc_list_val_node): New.
5954 (print_dw_val): Handle dw_val_class_view_list.
5955 (size_of_die): Likewise.
5956 (value_format): Likewise.
5957 (loc_list_has_views): New.
5958 (gen_llsym): Set vl_symbol too.
5959 (maybe_gen_llsym, skip_loc_list_entry): New.
5960 (dwarf2out_maybe_output_loclist_view_pair): New.
5961 (output_loc_list): Output view list or entries too.
5962 (output_view_list_offset): New.
5963 (output_die): Handle dw_val_class_view_list.
5964 (output_dwarf_version): New.
5965 (output_compilation_unit_header): Use it.
5966 (output_skeleton_debug_sections): Likewise.
5967 (output_rnglists, output_line_info): Likewise.
5968 (output_pubnames, output_aranges): Update version comments.
5969 (output_one_line_info_table): Output view numbers in asm comments.
5970 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5971 Call maybe_gen_llsym.
5972 (loc_list_from_tree_1): Adjust.
5973 (add_AT_location_description): Create view list attribute if
5974 needed, check it's absent otherwise.
5975 (convert_cfa_to_fb_loc_list): Adjust.
5976 (maybe_emit_file): Call output_asm_line_debug_info for test.
5977 (dwarf2out_var_location): Reset views as needed. Precompute
5978 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5979 attribute. Set view.
5980 (new_line_info_table): Reset next view.
5981 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5982 (dwarf2out_source_line): Likewise. Output view resets and labels to
5983 the assembler, or select appropriate line info opcodes.
5984 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5985 (optimize_string_length): Catch it. Adjust.
5986 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5987 dw_val_class_view_list, and remove it if no longer needed.
5988 (hash_loc_list): Hash view numbers.
5989 (loc_list_hasher::equal): Compare them.
5990 (optimize_location_lists): Check whether a view list symbol is
5991 needed, and whether the locview attribute is present, and
5992 whether they match. Remove the locview attribute if no longer
5994 (index_location_lists): Call skip_loc_list_entry for test.
5995 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5996 Use output_dwarf_version.
5997 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5998 (struct dw_val_node): Add val_view_list.
5999 * final.c (SEEN_NEXT_VIEW): New.
6000 (set_next_view_needed): New.
6001 (clear_next_view_needed): New.
6002 (maybe_output_next_view): New.
6003 (final_start_function): Rename to...
6004 (final_start_function_1): ... this. Take pointer to FIRST,
6005 add SEEN parameter. Emit param bindings in the initial view.
6006 (final_start_function): Reintroduce SEEN-less interface.
6007 (final): Rename to...
6008 (final_1): ... this. Take SEEN parameter. Output final pending
6009 next view at the end.
6010 (final): Reintroduce seen-less interface.
6011 (final_scan_insn): Output pending next view before switching
6012 sections or ending a block. Mark the next view as needed when
6013 outputting variable locations. Notify debug backend of section
6014 changes, and of location view changes.
6015 (rest_of_handle_final): Adjust.
6016 * toplev.c (process_options): Autodetect value for debug variable
6017 location views option. Warn on incompat5 without -gdwarf-5.
6018 * doc/invoke.texi (gvariable-location-views): New.
6019 (gvariable-location-views=incompat5): New.
6020 (gno-variable-location-views): New.
6022 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6024 PR tree-optimization/84136
6025 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6026 that the result of find_edge is non-NULL.
6028 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6031 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6032 storing integer register in SImode. Fix cost of 256 and 512
6033 byte aligned SSE register store.
6035 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6037 * config/i386/i386.c (ix86_multiplication_cost): Fix
6038 multiplication cost for TARGET_AVX512DQ.
6040 2018-02-08 Marek Polacek <polacek@redhat.com>
6042 PR tree-optimization/84238
6043 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6046 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6048 PR tree-optimization/84265
6049 * tree-vect-stmts.c (vectorizable_store): Don't treat
6050 VMAT_CONTIGUOUS accesses as grouped.
6051 (vectorizable_load): Likewise.
6053 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6055 PR tree-optimization/81635
6056 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6057 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6058 (test_round_for_mask): New functions.
6059 (wide_int_cc_tests): Call test_round_for_mask.
6060 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6061 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6062 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6063 range returned by get_range_info.
6065 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6068 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6069 * symtab.c: Include builtins.h
6070 (symtab_node::output_to_lto_symbol_table_p): Move here
6071 from lto-streamer-out.c:output_symbol_p.
6072 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6073 (output_symbol_p): Move all logic to symtab.c
6074 (produce_symtab): Update.
6076 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6078 * config/s390/s390-opts.h (enum indirect_branch): Define.
6079 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6080 (s390_indirect_branch_via_thunk)
6081 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6082 (enum s390_indirect_branch_type): Define.
6083 * config/s390/s390.c (struct s390_frame_layout, struct
6084 machine_function): Remove.
6085 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6086 (indirect_branch_table_label_no, indirect_branch_table_name):
6088 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6089 (enum s390_indirect_branch_option): Define.
6090 (s390_return_addr_from_memory): New function.
6091 (s390_handle_string_attribute): New function.
6092 (s390_attribute_table): Add new attribute handler.
6093 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6094 (s390_indirect_branch_via_thunk): New function.
6095 (s390_indirect_branch_via_inline_thunk): New function.
6096 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6097 sibling call optimization for non z10 compiles.
6098 (s390_emit_call): Force indirect branch target to be a single
6099 register. Add r1 clobber for non-z10 compiles.
6100 (s390_emit_epilogue): Emit return jump via return_use expander.
6101 (s390_reorg): Handle JUMP_INSNs as execute targets.
6102 (s390_option_override_internal): Perform validity checks for the
6103 new command line options.
6104 (s390_indirect_branch_attrvalue): New function.
6105 (s390_indirect_branch_settings): New function.
6106 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6107 (s390_output_indirect_thunk_function): New function.
6108 (s390_code_end): Implement target hook.
6109 (s390_case_values_threshold): Implement target hook.
6110 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6112 * config/s390/s390.h (struct s390_frame_layout)
6113 (struct machine_function): Move here from s390.c.
6114 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6115 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6116 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6117 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6118 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6119 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6120 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6121 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6122 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6123 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6124 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6125 (mnemonic attribute): Add values which aren't recognized
6127 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6128 pattern for branch conversion. Fix mnemonic attribute.
6129 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6130 indirect branch via thunk if requested.
6131 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6132 ("*indirect_jump"): Disable for branch conversion using out of
6134 ("indirect_jump_via_thunk<mode>_z10")
6135 ("indirect_jump_via_thunk<mode>")
6136 ("indirect_jump_via_inlinethunk<mode>_z10")
6137 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6138 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6139 ("casesi_jump_via_inlinethunk<mode>_z10")
6140 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6141 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6142 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6143 ("*indirect2_jump"): Disable for branch conversion.
6144 ("casesi_jump"): Turn into expander and expand patterns for branch
6146 ("return_use"): New expander.
6147 ("*return"): Emit return via thunk and rename it to ...
6148 ("*return<mode>"): ... this one.
6149 * config/s390/s390.opt: Add new options and and enum for the
6152 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6154 * lra-constraints.c (match_reload): Unconditionally use
6155 gen_lowpart_SUBREG, rather than selecting between that
6156 and equivalent gen_rtx_SUBREG code.
6158 2018-02-08 Richard Biener <rguenther@suse.de>
6160 PR tree-optimization/84233
6161 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6162 changed flag instead of boguously re-using phi_inserted.
6164 2018-02-08 Martin Jambor <mjambor@suse.cz>
6166 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6167 static local variables.
6169 2018-02-08 Richard Biener <rguenther@suse.de>
6171 PR tree-optimization/84278
6172 * tree-vect-stmts.c (vectorizable_store): When looking for
6173 smaller vector types to perform grouped strided loads/stores
6174 make sure the mode is supported by the target.
6175 (vectorizable_load): Likewise.
6177 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6179 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6180 Increase LDP/STP opportunities by adding adjacent callee-saves.
6182 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6184 PR rtl-optimization/84068
6185 PR rtl-optimization/83459
6186 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6188 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6190 PR tree-optimization/84224
6191 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6192 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6195 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6198 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6199 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6200 count, remove one USE.
6202 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6204 * doc/install.texi (Configuration): Document the
6205 --with-long-double-format={ibm,ieee} PowerPC configuration
6209 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6210 Convert from define_expand to be define_insn_and_split. Rework
6211 float/double/_Float128 conversions to QI/HI/SImode to work with
6212 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6213 conversions to QI/HImode types did a store and then a load to
6214 truncate the value. For conversions to VSX registers, don't split
6215 the insn, instead emit the code directly. Use the code iterator
6216 any_fix to combine signed and unsigned conversions.
6217 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6218 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6219 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6220 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6221 (fix_<mode>di2_hw): Likewise.
6222 (fixuns_<mode>di2_hw): Likewise.
6223 (fix_<mode>si2_hw): Likewise.
6224 (fixuns_<mode>si2_hw): Likewise.
6225 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6226 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6227 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6228 fix<uns>_trunc<SFDF:mode>si2_p8.
6229 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6231 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6232 (fix<uns>_<mode>_mem): Likewise.
6233 (fctiw<u>z_<mode>_mem): Likewise.
6234 (fix<uns>_<mode>_mem): Likewise.
6235 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6236 the register allocator from doing a direct move to the GPRs to do
6237 a store, and instead use the ISA 3.0 store byte/half-word from
6238 vector register instruction. For IEEE 128-bit floating point,
6239 also optimize stores of 32-bit ints.
6240 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6242 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6244 * genextract.c (push_pathstr_operand): New function to support
6246 (walk_rtx): Call push_pathstr_operand.
6247 (print_path): Support [a-zA-Z].
6249 2018-02-07 Richard Biener <rguenther@suse.de>
6251 PR tree-optimization/84037
6252 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6253 (cse_and_gimplify_to_preheader): Declare.
6254 (vect_get_place_in_interleaving_chain): Likewise.
6255 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6257 (_loop_vec_info::~_loop_vec_info): Delete it.
6258 (cse_and_gimplify_to_preheader): New function.
6259 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6260 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6261 (vectorizable_load): Likewise. For grouped stores always base
6262 the IV on the first element.
6263 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6264 condition before gimplifying.
6266 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6268 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6269 *DIV_EXPR and *MOD_EXPR.
6271 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6274 * config/i386/i386.c (ix86_option_override_internal): Mask out
6275 the CF_SET bit when checking -fcf-protection.
6277 2018-02-07 Tom de Vries <tom@codesourcery.com>
6280 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6283 2018-02-07 Richard Biener <rguenther@suse.de>
6285 PR tree-optimization/84204
6286 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6289 PR tree-optimization/84205
6290 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6291 special-case isl_ast_op_zdiv_r.
6293 PR tree-optimization/84223
6294 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6295 Only add conditions from within the region.
6296 (gather_bbs::after_dom_children): Adjust.
6298 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6301 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6302 * config/avr/avr.md: Only post-reload split REG-REG moves if
6303 either register is GENERAL_REG_P.
6305 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6307 PR tree-optimization/84235
6308 * tree-ssa-scopedtables.c
6309 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6310 if the subtraction is performed in floating point type where NaNs are
6311 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6312 build 1. Formatting fix.
6314 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6317 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6318 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6319 and skip it regardless of bb boundaries. Use CALL_P macro,
6320 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6321 unnecessarily, formatting fix.
6323 2018-02-06 Michael Collison <michael.collison@arm.com>
6325 * config/arm/thumb2.md:
6326 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6327 (*thumb_mov_notscc): Ditto.
6329 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6332 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6333 unsigned_fix, not "s".
6335 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6337 * configure.ac (gcc_fn_eh_frame_ro): New function.
6338 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6339 correct .eh_frame permissions.
6340 * configure: Regenerate.
6342 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6344 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6347 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6349 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6350 Display warning message for -mno-speculate-indirect-jumps.
6352 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6354 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6356 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6358 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6360 PR tree-optimization/84225
6361 * tree-eh.c (find_trapping_overflow): Only call
6362 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6364 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6367 * config/i386/i386.c: Reimplement the check of possible options
6368 -mibt/-mshstk conbination. Change error messages.
6369 * doc/invoke.texi: Fix a typo: remove extra '='.
6371 2018-02-06 Marek Polacek <polacek@redhat.com>
6373 PR tree-optimization/84228
6374 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6376 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6379 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6380 emitted arch directives.
6381 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6382 __ARM_FEATURE_COPROC before changing architectures.
6384 2018-02-06 Richard Biener <rguenther@suse.de>
6386 * config/i386/i386.c (print_reg): Fix typo.
6387 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6389 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6391 * configure: Regenerate.
6393 2018-02-05 Martin Sebor <msebor@redhat.com>
6395 PR tree-optimization/83369
6396 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6399 2018-02-05 Martin Liska <mliska@suse.cz>
6401 * doc/invoke.texi: Cherry-pick upstream r323995.
6403 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6405 * ira.c (ira_init_register_move_cost): Adjust comment.
6407 2018-02-05 Martin Liska <mliska@suse.cz>
6409 PR gcov-profile/84137
6410 * doc/gcov.texi: Fix typo in documentation.
6412 2018-02-05 Martin Liska <mliska@suse.cz>
6414 PR gcov-profile/83879
6415 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6416 using dlopen functionality.
6418 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6420 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6421 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6422 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6423 _mm_maskz_range_round_ss): New intrinsics.
6424 (__builtin_ia32_rangesd128_round)
6425 (__builtin_ia32_rangess128_round): Remove.
6426 (__builtin_ia32_rangesd128_mask_round,
6427 __builtin_ia32_rangess128_mask_round): New builtins.
6428 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6429 __builtin_ia32_rangess128_round): Remove.
6430 (__builtin_ia32_rangesd128_mask_round,
6431 __builtin_ia32_rangess128_mask_round): New builtins.
6432 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6433 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6434 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6435 "<round_saeonly_constraint>")): Changed to ...
6436 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6437 "<round_saeonly_scalar_constraint>")): ... this.
6438 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6439 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6440 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6441 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6442 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6444 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6446 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6448 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6449 Remove all values except native, 8540 and 8548.
6451 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6453 * config/i386/i386.c (ix86_output_function_return): Pass
6454 INVALID_REGNUM, instead of -1, as invalid register number to
6455 indirect_thunk_name and output_indirect_thunk.
6457 2018-02-02 Julia Koval <julia.koval@intel.com>
6459 * config.gcc: Add -march=icelake.
6460 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6461 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6462 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6463 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6464 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6465 (processor_target_table): Add icelake.
6466 (ix86_option_override_internal): Handle new PTAs.
6467 (get_builtin_code_for_version): Handle icelake.
6468 (M_INTEL_COREI7_ICELAKE): New.
6469 (fold_builtin_cpu): Handle icelake.
6470 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6471 * doc/invoke.texi: Add -march=icelake.
6473 2018-02-02 Julia Koval <julia.koval@intel.com>
6475 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6476 to wide_int_bitmask.
6477 * wide-int-bitmask.h: New.
6479 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6482 * config/i386/i386.md: Replace Pmode with word_mode in
6483 builtin_setjmp_setup and builtin_longjmp to support x32.
6485 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6489 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6491 (rs6000_supported_cpu_names): New static variable.
6492 (linux_cpu_translation_table): Likewise.
6493 (elf_platform) <cpu>: Define new static variable and use it.
6494 Translate kernel AT_PLATFORM name to canonical name if needed.
6495 Error if platform name is unknown.
6497 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6500 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6502 2018-02-01 Jeff Law <law@redhat.com>
6505 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6506 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6507 the scratch if RELEASE_VIA_POP is false.
6508 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6509 If we have to save a temporary register, decrement SIZE appropriately.
6510 Pass new arguments to release_scratch_register_on_entry.
6511 (ix86_adjust_stack_and_probe): Likewise.
6512 (ix86_emit_probe_stack_range): Pass new arguments to
6513 release_scratch_register_on_entry.
6515 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6517 PR rtl-optimization/84157
6518 * combine.c (change_zero_ext): Use REG_P predicate in
6519 front of HARD_REGISTER_P predicate.
6521 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6523 * config/avr/avr.c (avr_option_override): Move disabling of
6524 -fdelete-null-pointer-checks to...
6525 * common/config/avr/avr-common.c (avr_option_optimization_table):
6528 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6530 PR tree-optimization/81635
6531 * tree-data-ref.c (split_constant_offset_1): For types that
6532 wrap on overflow, try to use range info to prove that wrapping
6535 2018-02-01 Renlin Li <renlin.li@arm.com>
6538 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6540 (aarch64_register_move_cost): Likewise.
6541 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6543 (REG_CLASS_NAMES): Likewise.
6544 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6545 TAILCALL_ADDR_REGS. Remove IP registers.
6546 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6548 2018-02-01 Richard Biener <rguenther@suse.de>
6550 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6551 for specifying RPO order and allow NULL for that.
6552 * domwalk.c (dom_walker::dom_walker): Likewise.
6553 (dom_walker::walk): Handle NULL RPO order.
6554 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6556 (rewrite_update_dom_walker): Likewise.
6557 (mark_def_dom_walker): Likewise.
6559 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6561 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6562 (aarch64_maybe_expand_sve_subreg_move): Declare.
6563 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6564 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6566 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6567 that are semantically a reverse operation.
6568 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6569 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6570 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6572 (aarch64_can_change_mode_class): For big-endian, forbid changes
6573 between two SVE modes if they have different element sizes.
6575 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6577 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6578 the TImode handling for big-endian targets.
6580 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6582 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6583 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6585 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6586 Remove BSWAP handing for big-endian targets and use the form of
6587 LD1RQ appropariate for the mode.
6589 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6591 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6592 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6595 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6598 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6599 check for operands that need to go through aarch64_sve_reload_be.
6601 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6603 PR tree-optimization/81661
6604 PR tree-optimization/84117
6605 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6606 * tree-eh.c: Include gimplify.h.
6607 (find_trapping_overflow, replace_trapping_overflow,
6608 rewrite_to_non_trapping_overflow): New functions.
6609 * tree-vect-loop.c: Include tree-eh.h.
6610 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6611 * tree-data-ref.c: Include tree-eh.h.
6612 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6614 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6616 PR rtl-optimization/84123
6617 * combine.c (change_zero_ext): Check if hard register satisfies
6618 can_change_dest_mode before calling gen_lowpart_SUBREG.
6620 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6623 * ira.c (ira_init_register_move_cost): Remove assert.
6625 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6627 PR rtl-optimization/84071
6628 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6629 * doc/tm.texi: Regenerate.
6631 2018-01-31 Richard Biener <rguenther@suse.de>
6633 PR tree-optimization/84132
6634 * tree-data-ref.c (analyze_miv_subscript): Properly
6635 check whether evolution_function_is_affine_multivariate_p
6636 before calling gcd_of_steps_may_divide_p.
6638 2018-01-31 Julia Koval <julia.koval@intel.com>
6641 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6642 * config/i386/i386.md (rdpid_rex64) New.
6643 (rdpid): Make 32bit only.
6645 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6648 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6649 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6651 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6654 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6656 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6658 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6660 PR rtl-optimization/84071
6661 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6662 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6664 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6666 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6667 (arc_attribute_table): Add 'aux' attribute.
6668 (arc_in_small_data_p): Consider aux like variables.
6669 (arc_is_aux_reg_p): New function.
6670 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6671 (arc_get_aux_arg): New function.
6672 (prepare_move_operands): Handle aux-register access.
6673 (arc_handle_aux_attribute): New function.
6674 * doc/extend.texi (ARC Variable attributes): Add subsection.
6676 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6678 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6679 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6680 (arc_attribute_table): Add 'uncached' attribute.
6681 (arc_print_operand): Print '.di' flag for uncached memory
6683 (arc_in_small_data_p): Do not consider for small data the uncached
6685 (arc_is_uncached_mem_p): New function.
6686 * config/arc/predicates.md (compact_store_memory_operand): Check
6687 for uncached memory accesses.
6688 (nonvol_nonimm_operand): Likewise.
6689 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6691 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6694 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6695 falign-loops=): Add Optimization flag.
6697 2018-01-30 Jeff Law <law@redhat.com>
6700 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6701 INT_REGISTERS_SAVED. Check it prior to calling
6702 get_scratch_register_on_entry.
6703 (ix86_adjust_stack_and_probe): Similarly.
6704 (ix86_emit_probe_stack_range): Similarly.
6705 (ix86_expand_prologue): Corresponding changes.
6707 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6710 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6711 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6713 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6716 * lra-constraints.c (curr_insn_transform): Process AND in the
6719 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6721 PR rtl-optimization/83986
6722 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6723 dependence against last_pending_memory_flush in addition to
6726 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6728 PR tree-optimization/81611
6729 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6732 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6735 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6738 2018-01-30 Richard Biener <rguenther@suse.de>
6739 Jakub Jelinek <jakub@redhat.com>
6741 PR tree-optimization/84111
6742 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6743 inner loops added during recursion, as they don't have up-to-date
6746 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6749 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6750 (can_inline_edge_by_limits_p): ... here.
6751 (can_early_inline_edge_p, check_callers,
6752 update_caller_keys, update_callee_keys, recursive_inlining,
6753 add_new_edges_to_heap, speculation_useful_p,
6754 inline_small_functions,
6755 inline_small_functions, flatten_function,
6756 inline_to_all_callers_1): Update.
6758 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6760 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6763 2018-01-30 Richard Biener <rguenther@suse.de>
6765 PR tree-optimization/83008
6766 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6767 invariant and constant vector uses in stmts when they need
6770 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6773 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6774 * configure: Regenerate.
6776 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6778 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6780 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6781 Use gen_rtx_REG rather than gen_lowpart.
6783 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6785 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6786 rather than 0 when creating partial subregs.
6788 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6790 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6793 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6796 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6797 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6798 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6799 flags. This restores the settings used before the 2017-07-24.
6800 Turning off pre increment/decrement/modify allows IVOPTS to
6801 optimize DF/SF loops where the index is an int.
6803 2018-01-29 Richard Biener <rguenther@suse.de>
6804 Kelvin Nilsen <kelvin@gcc.gnu.org>
6807 * tree-vect-stmts.c (vectorizable_call): Don't call
6808 targetm.vectorize_builtin_md_vectorized_function if callee is
6811 2018-01-22 Carl Love <cel@us.ibm.com>
6813 * doc/extend.tex: Fix typo in second arg in
6814 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6816 2018-01-29 Richard Biener <rguenther@suse.de>
6818 PR tree-optimization/84086
6819 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6820 (flush_ssaname_freelist): When SSA names were released reset
6821 the SCEV hash table.
6823 2018-01-29 Richard Biener <rguenther@suse.de>
6825 PR tree-optimization/84057
6826 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6827 removed paths when removing edges.
6829 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6831 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6832 -mfunction-return=@var{choice}.
6834 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6837 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6839 (layout::print_source_line): Likewise.
6840 (test_get_line_width_without_trailing_whitespace): Add test cases.
6842 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6845 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6848 2018-01-26 Jim Wilson <jimw@sifive.com>
6850 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6852 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6855 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6857 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6858 and CMP + SUB-immediate -> SUBS.
6860 2018-01-26 Martin Sebor <msebor@redhat.com>
6862 PR tree-optimization/83896
6863 * tree-ssa-strlen.c (get_string_len): Rename...
6864 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6865 Avoid assuming length is constant.
6866 (handle_char_store): Use HOST_WIDE_INT for string length.
6868 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6871 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6872 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6874 2018-01-26 Richard Biener <rguenther@suse.de>
6876 PR rtl-optimization/84003
6877 * dse.c (record_store): Only record redundant stores when
6878 the earlier store aliases at least all accesses the later one does.
6880 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6882 PR rtl-optimization/83985
6883 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6884 REG_CFA_RESTORE insns.
6885 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6886 REG_CFA_RESTORE insns here.
6889 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6890 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6892 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6894 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6895 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6896 (arc_init): Likewise.
6897 (arc_override_options): Likewise.
6898 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6900 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6902 * config/arc/arc.h (TARGET_DBNZ): Define.
6903 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6904 properly set the tune attribute.
6905 (dbnz): Use TARGET_DBNZ guard.
6906 * config/arc/arc.opt (mtune): Add core3 option.
6908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6910 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6911 recognize new pic like addresses.
6912 (arc_delegitimize_address): Clean up.
6914 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6916 * config/arc/arc-arches.def: Option mrf16 valid for all
6918 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6919 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6920 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6921 * config/arc/arc-tables.opt: Regenerate.
6922 * config/arc/arc.c (arc_conditional_register_usage): Handle
6923 reduced register file case.
6924 (arc_file_start): Set must have build attributes.
6925 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6927 * config/arc/arc.opt (mrf16): Add new option.
6928 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6929 * config/arc/genmultilib.awk: Handle new mrf16 option.
6930 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6931 * config/arc/t-multilib: Regenerate.
6932 * doc/invoke.texi (ARC Options): Document mrf16 option.
6934 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6936 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6937 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6938 (arc_attribute_table): Add 'secure_call' attribute.
6939 (arc_print_operand): Print secure call operand.
6940 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6942 (arc_is_secure_call_p): New function. * config/arc/arc.md
6943 (call_i): Add support for sjli instruction.
6944 (call_value_i): Likewise.
6945 * config/arc/constraints.md (Csc): New constraint.
6947 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6948 John Eric Martin <John.Martin@emmicro-us.com>
6950 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6951 * config/arc/arc.c (_arc_jli_section): New struct.
6952 (arc_jli_section): New type.
6953 (rc_jli_sections): New static variable.
6954 (arc_handle_jli_attribute): New function.
6955 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6956 (arc_file_end): New function.
6957 (TARGET_ASM_FILE_END): Define.
6958 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6959 (arc_add_jli_section): New function.
6960 (jli_call_scan): Likewise.
6961 (arc_reorg): Call jli_call_scan.
6962 (arc_output_addsi): Remove 'S' from printing asm operand.
6963 (arc_is_jli_call_p): New function.
6964 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6966 (movhi_insn): Likewise.
6967 (movsi_insn): Likewise.
6968 (movsi_set_cc_insn): Likewise.
6969 (loadqi_update): Likewise.
6970 (load_zeroextendqisi_update): Likewise.
6971 (load_signextendqisi_update): Likewise.
6972 (loadhi_update): Likewise.
6973 (load_zeroextendhisi_update): Likewise.
6974 (load_signextendhisi_update): Likewise.
6975 (loadsi_update): Likewise.
6976 (loadsf_update): Likewise.
6977 (movsicc_insn): Likewise.
6978 (bset_insn): Likewise.
6979 (bxor_insn): Likewise.
6980 (bclr_insn): Likewise.
6981 (bmsk_insn): Likewise.
6982 (bicsi3_insn): Likewise.
6983 (cmpsi_cc_c_insn): Likewise.
6984 (movsi_ne): Likewise.
6985 (movsi_cond_exec): Likewise.
6986 (clrsbsi2): Likewise.
6997 (call_i): Remove 'S' asm letter, add jli instruction.
6998 (call_value_i): Likewise.
6999 * config/arc/arc.op (mjli-always): New option.
7000 * config/arc/constraints.md (Cji): New constraint.
7001 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7003 (subsf3_fpx): Likewise.
7004 (mulsf3_fpx): Likewise.
7005 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7007 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7009 * doc/invoke.texi (ARC): Document mjli-always option.
7011 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7013 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7014 avoid addition with 0 and use incw and decw where possible.
7016 2018-01-26 Richard Biener <rguenther@suse.de>
7018 PR tree-optimization/81082
7019 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7020 association if it requires casting to unsigned.
7021 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7022 from fold_plusminus_mult_expr to catch important cases late when
7023 range info is available.
7025 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7027 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7028 * configure.ac (hidden_linkonce): New test.
7029 * configure: Regenerate.
7030 * config.in: Regenerate.
7032 2018-01-26 Julia Koval <julia.koval@intel.com>
7034 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7035 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7036 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7037 _mm_mask_bitshuffle_epi64_mask): Fix type.
7038 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7039 USI_FTYPE_V4DI_V4DI_USI): Remove.
7040 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7041 __builtin_ia32_vpshufbitqmb256_mask,
7042 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7043 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7044 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7046 2018-01-26 Alan Modra <amodra@gmail.com>
7049 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7050 UNSPEC_VBPERMQ. Sort other unspecs.
7052 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7054 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7056 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7059 * predict.c (drop_profile): Do not push/pop cfun; update also
7061 (handle_missing_profiles): Fix logic looking for zero profiles.
7063 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7066 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7067 on functions with #pragma omp declare simd or functions with simd
7069 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7070 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7071 Remove trailing \n from warning_at calls.
7073 2018-01-25 Tom de Vries <tom@codesourcery.com>
7076 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7077 for neutered workers.
7079 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7082 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7083 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7085 2018-01-24 Jeff Law <law@redhat.com>
7088 * i386.c (get_probe_interval): Move to earlier point.
7089 (ix86_compute_frame_layout): If -fstack-clash-protection and
7090 the frame is larger than the probe interval, then use pushes
7091 to save registers rather than reg->mem moves.
7092 (ix86_expand_prologue): Remove conditional for int_registers_saved
7095 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7098 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7099 min/max for never referenced object.
7101 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7104 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7106 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7107 attributes from DECL_ATTRIBUTES (decl) without affecting
7108 DECL_ATTRIBUTES (current_function_decl).
7109 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7110 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7112 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7114 PR tree-optimization/83979
7115 * fold-const.c (fold_comparison): Use constant_boolean_node
7116 instead of boolean_{true,false}_node.
7118 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7120 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7123 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7125 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7126 Simplify the clause that sets the length attribute.
7127 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7128 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7129 clause that sets the length attribute.
7130 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7132 2018-01-24 Tom de Vries <tom@codesourcery.com>
7135 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7136 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7137 Add strict parameter.
7138 (prevent_branch_around_nothing): Insert dummy insn between branch to
7139 label and label with no ptx insn inbetween.
7140 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7142 2018-01-24 Tom de Vries <tom@codesourcery.com>
7145 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7146 for neutered threads in warp.
7147 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7149 2018-01-24 Richard Biener <rguenther@suse.de>
7151 PR tree-optimization/83176
7152 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7155 2018-01-24 Richard Biener <rguenther@suse.de>
7157 PR tree-optimization/82819
7158 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7159 code generating pluses that are no-ops in the target precision.
7161 2018-01-24 Richard Biener <rguenther@suse.de>
7164 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7166 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7168 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7169 to merge probabilities.
7170 * predict.c (probably_never_executed): Also mark as cold functions
7171 with global 0 profile and guessed local profile.
7172 * profile-count.c (profile_probability::combine_with_count): New
7174 * profile-count.h (profile_probability::operator*,
7175 profile_probability::operator*=, profile_probability::operator/,
7176 profile_probability::operator/=): Reduce precision to adjusted
7177 and set value to guessed on contradictory divisions.
7178 (profile_probability::combine_with_freq): Remove.
7179 (profile_probability::combine_wiht_count): Declare.
7180 (profile_count::force_nonzero):: Set to adjusted.
7181 (profile_count::probability_in):: Set quality to adjusted.
7182 * tree-ssa-tail-merge.c (replace_block_by): Use
7185 2018-01-23 Andrew Waterman <andrew@sifive.com>
7186 Jim Wilson <jimw@sifive.com>
7188 * config/riscv/riscv.c (riscv_stack_boundary): New.
7189 (riscv_option_override): Set riscv_stack_boundary. Handle
7190 riscv_preferred_stack_boundary_arg.
7191 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7192 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7193 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7194 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7195 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7196 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7198 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7201 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7202 of struct ix86_frame.
7203 (ix86_expand_epilogue): Likewise. Add a local variable for
7204 the reg_save_offset field in struct ix86_frame.
7206 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7208 PR tree-optimization/82604
7209 * tree-loop-distribution.c (enum partition_kind): New enum item
7210 PKIND_PARTIAL_MEMSET.
7211 (partition_builtin_p): Support above new enum item.
7212 (generate_code_for_partition): Ditto.
7213 (compute_access_range): Differentiate cases that equality can be
7214 proven at all loops, the innermost loops or no loops.
7215 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7216 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7217 (finalize_partitions, distribute_loop): Don't fuse partition of
7218 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7219 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7222 2018-01-23 Martin Liska <mliska@suse.cz>
7224 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7225 order to ignore the predictor.
7226 (PRED_POLYMORPHIC_CALL): Likewise.
7227 (PRED_RECURSIVE_CALL): Likewise.
7229 2018-01-23 Martin Liska <mliska@suse.cz>
7231 * tree-profile.c (tree_profiling): Print function header to
7232 aware reader which function we are working on.
7233 * value-prof.c (gimple_find_values_to_profile): Do not print
7234 not interesting value histograms.
7236 2018-01-23 Martin Liska <mliska@suse.cz>
7238 * profile-count.h (enum profile_quality): Add
7239 profile_uninitialized as the first value. Do not number values
7240 as they are zero based.
7241 (profile_count::verify): Update sanity check.
7242 (profile_probability::verify): Likewise.
7244 2018-01-23 Nathan Sidwell <nathan@acm.org>
7246 * doc/invoke.texi (ffor-scope): Deprecate.
7248 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7250 PR tree-optimization/83510
7251 * domwalk.c (set_all_edges_as_executable): New function.
7252 (dom_walker::dom_walker): Convert bool param
7253 "skip_unreachable_blocks" to enum reachability. Move setup of
7254 edge flags to set_all_edges_as_executable and only do it when
7255 reachability is REACHABLE_BLOCKS.
7256 * domwalk.h (enum dom_walker::reachability): New enum.
7257 (dom_walker::dom_walker): Convert bool param
7258 "skip_unreachable_blocks" to enum reachability.
7259 (set_all_edges_as_executable): New decl.
7260 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7261 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7263 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7264 but converting true to REACHABLE_BLOCKS.
7265 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7267 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7268 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7269 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7271 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7272 if check_all_array_refs will be called.
7274 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7276 * tree.c (selftest::test_location_wrappers): Add more test
7279 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7281 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7282 (selftest::test_bit_in_range): Likewise.
7284 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7287 * doc/sourcebuild.texi (vect_float): Say that the selector
7288 only describes the situation when -funsafe-math-optimizations is on.
7289 (vect_float_strict): Document.
7291 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7293 PR tree-optimization/83965
7294 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7295 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7296 instead of checking only for a reduction.
7297 (vect_recog_widen_sum_pattern): Likewise.
7299 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7301 * predict.c (probably_never_executed): Only use precise profile info.
7302 (compute_function_frequency): Skip after inlining hack since we now
7303 have quality checking.
7305 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7307 * profile-count.h (profile_probability::very_unlikely,
7308 profile_probability::unlikely, profile_probability::even): Set
7309 precision to guessed.
7311 2018-01-23 Richard Biener <rguenther@suse.de>
7313 PR tree-optimization/83963
7314 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7315 Properly terminate dominator walk when crossing the exit edge not
7316 when visiting its source block.
7318 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7321 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7322 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7324 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7326 PR tree-optimization/83957
7327 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7328 semicolon after for body surrounded by braces.
7330 PR tree-optimization/83081
7331 * profile-count.h (profile_probability::split): New method.
7332 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7333 Use profile_probability::split.
7334 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7335 when splitting a single conditional jump into 2.
7337 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7339 PR tree-optimization/69452
7340 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7345 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7346 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7347 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7349 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7351 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7352 * config/rl78/rl78.md (movdi): New define_expand.
7353 * config/rl78/rl78.c (rl78_split_movdi): New function.
7355 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7358 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7360 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7361 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7362 128-bit to produce an UNSPEC move to get the double word with the
7363 signbit and then a shift directly to do signbit.
7364 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7365 implementation with a new version that just does either a direct
7366 move or a regular move. Move memory interface to separate insns.
7367 Move insns so they are next to the expander.
7368 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7369 with signbit move. Split big and little endian case.
7370 (signbit<mode>2_dm_mem_le): Likewise.
7371 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7372 (signbit<mode>2_dm2): Likewise.
7374 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7376 * config/rl78/rl78.md (anddi3): New define_expand.
7378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7380 * config/rl78/rl78.md (umindi3): New define_expand.
7382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7384 * config/rl78/rl78.md (smindi3): New define_expand.
7386 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7388 * config/rl78/rl78.md (smaxdi3): New define_expand.
7390 2018-01-22 Carl Love <cel@us.ibm.com>
7392 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7393 LVX_V1TI): Add macro expansion.
7394 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7395 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7396 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7397 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7398 Change check to determine if the instruction is a byte reversing
7399 entry. Fix typo in comment.
7400 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7401 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7402 Add def_builtin calls for new builtins.
7403 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7404 Add define_insn expansion.
7406 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7408 * config/rl78/rl78.md (umaxdi3): New define_expand.
7410 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7412 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7413 for non-QImode registers.
7415 2018-01-22 Richard Biener <rguenther@suse.de>
7417 PR tree-optimization/83963
7418 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7419 including the loop exit block.
7420 (scop_detection::merge_sese): Likewise.
7421 (scop_detection::add_scop): Do it here instead.
7423 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7425 * doc/sourcebuild.texi (arm_softfloat): Document.
7427 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7430 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7431 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7432 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7434 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7435 David Edelsohn <dje.gcc@gmail.com>
7438 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7439 Change "crset eq" to "crset 2".
7440 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7441 (*call_indirect_aix<mode>_nospec): Likewise.
7442 (*call_value_indirect_aix<mode>_nospec): Likewise.
7443 (*call_indirect_elfv2<mode>_nospec): Likewise.
7444 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7445 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7446 change assembly output from . to $.
7447 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7448 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7449 (*tablejump<mode>_internal1_nospec): Likewise.
7451 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7454 * config/sh/sh_optimize_sett_clrt.cc:
7455 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7457 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7459 PR tree-optimization/83940
7460 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7461 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7462 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7463 use it to pass back the definition type.
7464 (vect_check_store_rhs): Likewise rhs_dt_out.
7465 (vect_build_gather_load_calls): Add a mask_dt argument and use
7466 it instead of a call to vect_is_simple_use.
7467 (vectorizable_store): Update calls to vect_check_load_store_mask
7468 and vect_check_store_rhs. Use the dt returned by the latter instead
7469 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7470 instead of calls to vect_is_simple_use. Pass the scalar rather
7471 than the vector operand to vect_is_simple_use when handling
7472 second and subsequent copies of an rhs value.
7473 (vectorizable_load): Update calls to vect_check_load_store_mask
7474 and vect_build_gather_load_calls. Use the cached mask_dt and
7475 gs_info.offset_dt instead of calls to vect_is_simple_use.
7477 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7480 * tree-emutls.c: Include gimplify.h.
7481 (lower_emutls_2): New function.
7482 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7483 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7484 it before further processing.
7487 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7488 UINTVAL (trueop1) instead of INTVAL (op1).
7490 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7494 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7495 INCOMING_FRAME_SP_OFFSET if not defined.
7496 (scan_trace): Add ENTRY argument. If true and
7497 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7498 emit a note to adjust the CFA offset.
7499 (create_cfi_notes): Adjust scan_trace callers.
7500 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7501 INCOMING_FRAME_SP_OFFSET in the CIE.
7502 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7503 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7505 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7506 * doc/tm.texi: Regenerated.
7508 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7510 PR rtl-optimization/83147
7511 * lra-constraints.c (remove_inheritance_pseudos): Use
7512 lra_substitute_pseudo_within_insn.
7514 2018-01-19 Tom de Vries <tom@codesourcery.com>
7515 Cesar Philippidis <cesar@codesourcery.com>
7518 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7520 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7523 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7524 spaces for function labels.
7526 2018-01-19 Martin Liska <mliska@suse.cz>
7528 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7529 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7530 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7531 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7532 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7533 (PRED_CONST_RETURN): Change from 69 to 65.
7534 (PRED_NULL_RETURN): Change from 91 to 71.
7535 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7536 (PRED_LOOP_GUARD): Change from 66 to 73.
7538 2018-01-19 Martin Liska <mliska@suse.cz>
7540 * predict.c (predict_insn_def): Add new assert.
7541 (struct branch_predictor): Change type to signed integer.
7542 (test_prediction_value_range): Amend test to cover
7544 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7545 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7546 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7547 (PRED_LOOP_IV_COMPARE): Likewise.
7548 * predict.h (PROB_UNINITIALIZED): Define new constant.
7550 2018-01-19 Martin Liska <mliska@suse.cz>
7552 * predict.c (dump_prediction): Add new format for
7553 analyze_brprob.py script which is enabled with -details
7555 * profile-count.h (precise_p): New function.
7557 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7559 PR tree-optimization/83922
7560 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7561 there are no statements that need masking.
7562 (vect_active_double_reduction_p): New function.
7563 (vect_analyze_loop_operations): Use it when handling phis that
7564 are not in the loop header.
7566 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7568 PR tree-optimization/83914
7569 * tree-vect-loop.c (vectorizable_induction): Don't convert
7570 init_expr or apply the peeling adjustment for inductions
7571 that are nested within the vectorized loop.
7573 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7575 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7578 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7582 * function.h (gimplify_parameters): Add gimple_seq * argument.
7583 * function.c: Include gimple.h and options.h.
7584 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7585 for the added local temporaries if needed.
7586 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7587 if there are any parameter cleanups, wrap whole body into a
7588 try/finally with the cleanups.
7590 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7593 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7594 Use GET_MODE_CLASS for scalar floating point.
7596 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7600 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7601 Fix call of call_cgraph_insertion_hooks.
7603 2018-01-18 Martin Sebor <msebor@redhat.com>
7605 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7607 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7610 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7613 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7616 * common.opt: (-ffile-prefix-map): New option.
7617 * opts.c (common_handle_option): Defer it.
7618 * opts-global.c (handle_common_deferred_options): Handle it.
7619 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7620 * file-prefix-map.h: New file.
7621 (remap_debug_filename, add_debug_prefix_map): ...here.
7622 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7623 * final.c (debug_prefix_map, add_debug_prefix_map
7624 remap_debug_filename): Move to...
7625 * file-prefix-map.c: New file.
7626 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7627 generalize, get rid of alloca(), use strrchr() instead of strchr().
7628 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7629 Implement in terms of add_prefix_map().
7630 (remap_macro_filename, remap_debug_filename): Implement in term of
7632 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7633 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7634 * dbxout.c: Include file-prefix-map.h.
7635 * varasm.c: Likewise.
7636 * vmsdbgout.c: Likewise.
7637 * xcoffout.c: Likewise.
7638 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7639 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7640 * doc/invoke.texi (-ffile-prefix-map): Document.
7641 (-fdebug-prefix-map): Update description.
7643 2018-01-18 Martin Liska <mliska@suse.cz>
7645 * config/i386/i386.c (indirect_thunk_name): Document that also
7647 (output_indirect_thunk): Document why both instructions
7648 (pause and lfence) are generated.
7650 2018-01-18 Richard Biener <rguenther@suse.de>
7652 PR tree-optimization/83887
7653 * graphite-scop-detection.c
7654 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7655 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7656 (scop_detection::merge_sese): Re-implement with a flood-fill
7657 algorithm that properly finds a SESE region if it exists.
7659 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7662 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7663 pointer_diff optimizations use view_convert instead of convert.
7665 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7667 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7668 Generate different code for -mno-speculate-indirect-jumps.
7669 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7670 (*call_indirect_aix<mode>): Disable for
7671 -mno-speculate-indirect-jumps.
7672 (*call_indirect_aix<mode>_nospec): New define_insn.
7673 (*call_value_indirect_aix<mode>): Disable for
7674 -mno-speculate-indirect-jumps.
7675 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7676 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7677 -mno-speculate-indirect-jumps.
7678 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7680 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7682 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7683 long double type, set the flags for noting the default long double
7684 type, even if we don't pass or return a long double type.
7686 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7689 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7692 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7694 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7695 support for merge[hl].
7696 (fold_mergehl_helper): New helper function.
7697 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7698 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7699 (altivec_vmrglw_direct): Add xxmrglw insn.
7701 2018-01-17 Andrew Waterman <andrew@sifive.com>
7703 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7704 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7706 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7709 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7710 call the lto_location_cache before reading the
7711 DECL_SOURCE_LOCATION of the types.
7713 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7714 Richard Sandiford <richard.sandiford@linaro.org>
7716 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7717 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7718 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7719 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7720 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7722 * config/aarch64/constraints.md (aarch64_movti_operand):
7724 * config/aarch64/predicates.md (Uti): Add new constraint.
7726 2018-01-17 Carl Love <cel@us.ibm.com>
7728 * config/rs6000/vsx.md (define_expand xl_len_r,
7729 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7730 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7732 (define_expand, define_insn): Move the shift left from the
7733 define_insn to the define_expand for lxvl and stxvl instructions.
7734 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7735 and XL_LEN_R definitions to PURE.
7737 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7739 * config/i386/i386.c (indirect_thunk_name): Declare regno
7740 as unsigned int. Compare regno with INVALID_REGNUM.
7741 (output_indirect_thunk): Ditto.
7742 (output_indirect_thunk_function): Ditto.
7743 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7744 in the call to output_indirect_thunk_function.
7746 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7749 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7750 rather than the size of inner_type to determine the stack slot size
7751 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7753 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7756 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7759 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7761 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7762 endian Linux systems to optionally enable multilibs for selecting
7763 the long double type if the user configured an explicit type.
7764 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7765 have no long double multilibs if not defined.
7766 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7767 warn if the user used -mabi={ieee,ibm}longdouble and we built
7768 multilibs for long double.
7769 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7770 appropriate multilib option.
7771 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7773 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7774 for building long double multilibs.
7775 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7777 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7779 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7782 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7784 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7787 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7790 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7793 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7795 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7796 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7798 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7800 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7801 different rtl trees depending on TARGET_64BIT.
7802 (rs6000_gen_lvx): Likewise.
7804 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7806 * config/visium/visium.md (nop): Tweak comment.
7807 (hazard_nop): Likewise.
7809 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7811 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7812 -mspeculate-indirect-jumps.
7813 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7814 for -mno-speculate-indirect-jumps.
7815 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7816 (*call_value_indirect_elfv2<mode>): Disable for
7817 -mno-speculate-indirect-jumps.
7818 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7819 (indirect_jump): Emit different RTL for
7820 -mno-speculate-indirect-jumps.
7821 (*indirect_jump<mode>): Disable for
7822 -mno-speculate-indirect-jumps.
7823 (*indirect_jump<mode>_nospec): New define_insn.
7824 (tablejump): Emit different RTL for
7825 -mno-speculate-indirect-jumps.
7826 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7827 (tablejumpsi_nospec): New define_expand.
7828 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7829 (tablejumpdi_nospec): New define_expand.
7830 (*tablejump<mode>_internal1): Disable for
7831 -mno-speculate-indirect-jumps.
7832 (*tablejump<mode>_internal1_nospec): New define_insn.
7833 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7836 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7838 * caller-save.c (insert_save): Drop unnecessary parameter. All
7841 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7842 Richard Biener <rguenth@suse.de>
7845 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7846 return early, inline manually is_gimple_sizepos. Make sure if we
7847 call gimplify_expr we don't end up with a gimple constant.
7848 * tree.c (variably_modified_type_p): Don't return true for
7849 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7850 * gimplify.h (is_gimple_sizepos): Remove.
7852 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7854 PR tree-optimization/83857
7855 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7856 vectorizable_live_operation for pure SLP statements.
7857 (vectorizable_live_operation): Handle PHIs.
7859 2018-01-16 Richard Biener <rguenther@suse.de>
7861 PR tree-optimization/83867
7862 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7863 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7865 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7868 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7869 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7870 If off is not INTEGER_CST, issue a may not be aligned warning
7871 rather than isn't aligned. Use isn%'t rather than isn't.
7872 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7874 <case MULT_EXPR>: Improve the case when bottom and one of the
7875 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7876 operand, in that case check if the other operand is multiple of
7877 bottom divided by the INTEGER_CST operand.
7879 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7882 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7883 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7884 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7885 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7886 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7887 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7888 (pa_function_arg_size): New function.
7890 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7892 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7893 in a separate statement.
7895 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7897 PR tree-optimization/83847
7898 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7899 group gathers and scatters.
7901 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7903 PR rtl-optimization/86620
7904 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7906 PR rtl-optimization/83213
7907 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7908 to last if both are JUMP_INSNs.
7910 PR tree-optimization/83843
7911 * gimple-ssa-store-merging.c
7912 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7913 store_immediate_info for bswap/nop orig_stores.
7915 2018-01-15 Andrew Waterman <andrew@sifive.com>
7917 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7919 <UDIV>: Increase cost if !TARGET_DIV.
7921 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7923 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7924 (define_attr "cr_logical_3op"): New.
7925 (cceq_ior_compare): Adjust.
7926 (cceq_ior_compare_complement): Adjust.
7927 (*cceq_rev_compare): Adjust.
7928 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7929 (is_cracked_insn): Adjust.
7930 (insn_must_be_first_in_group): Adjust.
7931 * config/rs6000/40x.md: Adjust.
7932 * config/rs6000/440.md: Adjust.
7933 * config/rs6000/476.md: Adjust.
7934 * config/rs6000/601.md: Adjust.
7935 * config/rs6000/603.md: Adjust.
7936 * config/rs6000/6xx.md: Adjust.
7937 * config/rs6000/7450.md: Adjust.
7938 * config/rs6000/7xx.md: Adjust.
7939 * config/rs6000/8540.md: Adjust.
7940 * config/rs6000/cell.md: Adjust.
7941 * config/rs6000/e300c2c3.md: Adjust.
7942 * config/rs6000/e500mc.md: Adjust.
7943 * config/rs6000/e500mc64.md: Adjust.
7944 * config/rs6000/e5500.md: Adjust.
7945 * config/rs6000/e6500.md: Adjust.
7946 * config/rs6000/mpc.md: Adjust.
7947 * config/rs6000/power4.md: Adjust.
7948 * config/rs6000/power5.md: Adjust.
7949 * config/rs6000/power6.md: Adjust.
7950 * config/rs6000/power7.md: Adjust.
7951 * config/rs6000/power8.md: Adjust.
7952 * config/rs6000/power9.md: Adjust.
7953 * config/rs6000/rs64.md: Adjust.
7954 * config/rs6000/titan.md: Adjust.
7956 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7958 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7959 ix86_indirect_branch_register logic.
7961 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7963 * config/i386/constraints.md (Bs): Update
7964 ix86_indirect_branch_register check. Don't check
7965 ix86_indirect_branch_register with GOT_memory_operand.
7967 * config/i386/predicates.md (GOT_memory_operand): Don't check
7968 ix86_indirect_branch_register here.
7969 (GOT32_symbol_operand): Likewise.
7971 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7973 * config/i386/predicates.md (constant_call_address_operand):
7974 Rewrite ix86_indirect_branch_register logic.
7975 (sibcall_insn_operand): Likewise.
7977 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7979 * config/i386/constraints.md (Bs): Replace
7980 ix86_indirect_branch_thunk_register with
7981 ix86_indirect_branch_register.
7983 * config/i386/i386.md (indirect_jump): Likewise.
7984 (tablejump): Likewise.
7985 (*sibcall_memory): Likewise.
7986 (*sibcall_value_memory): Likewise.
7987 Peepholes of indirect call and jump via memory: Likewise.
7988 * config/i386/i386.opt: Likewise.
7989 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7990 (GOT_memory_operand): Likewise.
7991 (call_insn_operand): Likewise.
7992 (sibcall_insn_operand): Likewise.
7993 (GOT32_symbol_operand): Likewise.
7995 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7998 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7999 type rather than type addr's type points to.
8000 (expand_omp_atomic_mutex): Likewise.
8001 (expand_omp_atomic): Likewise.
8003 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8006 * config/i386/i386.c (output_indirect_thunk_function): Use
8007 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8008 for __x86_return_thunk.
8010 2018-01-15 Richard Biener <rguenther@suse.de>
8013 * expmed.c (extract_bit_field_1): Fix typo.
8015 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8018 * config/arm/iterators.md (VF): New mode iterator.
8019 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8020 Remove integer-related logic from pattern.
8021 (neon_vabd<mode>_3): Likewise.
8023 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8026 * common.opt (fstrict-overflow): No longer an alias.
8027 (fwrapv-pointer): New option.
8028 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8029 also for pointer types based on flag_wrapv_pointer.
8030 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8031 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8032 opts->x_flag_wrapv got set.
8033 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8034 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8035 POINTER_TYPE_OVERFLOW_UNDEFINED.
8036 * match.pd: Likewise in address comparison pattern.
8037 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8039 2018-01-15 Richard Biener <rguenther@suse.de>
8042 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8043 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8044 Reset type names to their identifier if their TYPE_DECL doesn't
8045 have linkage (and thus is used for ODR and devirt).
8046 (save_debug_info_for_decl): Remove.
8047 (save_debug_info_for_type): Likewise.
8048 (add_tree_to_fld_list): Adjust.
8049 * tree-pretty-print.c (dump_generic_node): Make dumping of
8050 type names more robust.
8052 2018-01-15 Richard Biener <rguenther@suse.de>
8054 * BASE-VER: Bump to 8.0.1.
8056 2018-01-14 Martin Sebor <msebor@redhat.com>
8059 * builtins.c (check_access): Avoid warning when the no-warning bit
8062 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8064 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8065 * ira-color (allocno_hard_regs_compare): Likewise.
8067 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8070 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8071 Use .pushsection/.popsection.
8073 2018-01-14 Martin Sebor <msebor@redhat.com>
8076 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8078 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8080 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8081 entry from extra_headers.
8082 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8083 extra_headers, make the list bitwise identical to the i?86-*-* one.
8085 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8087 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8088 -mcmodel=large with -mindirect-branch=thunk,
8089 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8090 -mfunction-return=thunk-extern.
8091 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8092 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8093 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8095 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8097 * config/i386/i386.c (print_reg): Print the name of the full
8098 integer register without '%'.
8099 (ix86_print_operand): Handle 'V'.
8100 * doc/extend.texi: Document 'V' modifier.
8102 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8104 * config/i386/constraints.md (Bs): Disallow memory operand for
8105 -mindirect-branch-register.
8107 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8108 (GOT_memory_operand): Likewise.
8109 (call_insn_operand): Likewise.
8110 (sibcall_insn_operand): Likewise.
8111 (GOT32_symbol_operand): Likewise.
8112 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8113 for -mindirect-branch-register.
8114 (tablejump): Likewise.
8115 (*sibcall_memory): Likewise.
8116 (*sibcall_value_memory): Likewise.
8117 Disallow peepholes of indirect call and jump via memory for
8118 -mindirect-branch-register.
8119 (*call_pop): Replace m with Bw.
8120 (*call_value_pop): Likewise.
8121 (*sibcall_pop_memory): Replace m with Bs.
8122 * config/i386/i386.opt (mindirect-branch-register): New option.
8123 * doc/invoke.texi: Document -mindirect-branch-register option.
8125 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8127 * config/i386/i386-protos.h (ix86_output_function_return): New.
8128 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8129 set function_return_type.
8130 (indirect_thunk_name): Add ret_p to indicate thunk for function
8132 (output_indirect_thunk_function): Pass false to
8133 indirect_thunk_name.
8134 (ix86_output_indirect_branch_via_reg): Likewise.
8135 (ix86_output_indirect_branch_via_push): Likewise.
8136 (output_indirect_thunk_function): Create alias for function
8137 return thunk if regno < 0.
8138 (ix86_output_function_return): New function.
8139 (ix86_handle_fndecl_attribute): Handle function_return.
8140 (ix86_attribute_table): Add function_return.
8141 * config/i386/i386.h (machine_function): Add
8142 function_return_type.
8143 * config/i386/i386.md (simple_return_internal): Use
8144 ix86_output_function_return.
8145 (simple_return_internal_long): Likewise.
8146 * config/i386/i386.opt (mfunction-return=): New option.
8147 (indirect_branch): Mention -mfunction-return=.
8148 * doc/extend.texi: Document function_return function attribute.
8149 * doc/invoke.texi: Document -mfunction-return= option.
8151 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8153 * config/i386/i386-opts.h (indirect_branch): New.
8154 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8155 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8156 with local indirect jump when converting indirect call and jump.
8157 (ix86_set_indirect_branch_type): New.
8158 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8159 (indirectlabelno): New.
8160 (indirect_thunk_needed): Likewise.
8161 (indirect_thunk_bnd_needed): Likewise.
8162 (indirect_thunks_used): Likewise.
8163 (indirect_thunks_bnd_used): Likewise.
8164 (INDIRECT_LABEL): Likewise.
8165 (indirect_thunk_name): Likewise.
8166 (output_indirect_thunk): Likewise.
8167 (output_indirect_thunk_function): Likewise.
8168 (ix86_output_indirect_branch_via_reg): Likewise.
8169 (ix86_output_indirect_branch_via_push): Likewise.
8170 (ix86_output_indirect_branch): Likewise.
8171 (ix86_output_indirect_jmp): Likewise.
8172 (ix86_code_end): Call output_indirect_thunk_function if needed.
8173 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8175 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8176 (ix86_attribute_table): Add indirect_branch.
8177 * config/i386/i386.h (machine_function): Add indirect_branch_type
8178 and has_local_indirect_jump.
8179 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8181 (tablejump): Likewise.
8182 (*indirect_jump): Use ix86_output_indirect_jmp.
8183 (*tablejump_1): Likewise.
8184 (simple_return_indirect_internal): Likewise.
8185 * config/i386/i386.opt (mindirect-branch=): New option.
8186 (indirect_branch): New.
8189 (thunk-inline): Likewise.
8190 (thunk-extern): Likewise.
8191 * doc/extend.texi: Document indirect_branch function attribute.
8192 * doc/invoke.texi: Document -mindirect-branch= option.
8194 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8197 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8199 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8201 * ipa-inline.c (want_inline_small_function_p): Return false if
8202 inlining has already failed with CIF_FINAL_ERROR.
8203 (update_caller_keys): Call want_inline_small_function_p before
8205 (update_callee_keys): Likewise.
8207 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8209 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8211 (rs6000_quadword_masked_address_p): Likewise.
8212 (quad_aligned_load_p): Likewise.
8213 (quad_aligned_store_p): Likewise.
8214 (const_load_sequence_p): Add comment to describe the outer-most loop.
8215 (mimic_memory_attributes_and_flags): New function.
8216 (rs6000_gen_stvx): Likewise.
8217 (replace_swapped_aligned_store): Likewise.
8218 (rs6000_gen_lvx): Likewise.
8219 (replace_swapped_aligned_load): Likewise.
8220 (replace_swapped_load_constant): Capitalize argument name in
8221 comment describing this function.
8222 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8223 and stores that access quad-word aligned addresses and replace
8224 with stvx or lvx instructions when appropriate.
8225 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8226 New function prototype.
8227 (rs6000_quadword_masked_address_p): Likewise.
8228 (rs6000_gen_lvx): Likewise.
8229 (rs6000_gen_stvx): Likewise.
8230 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8231 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8232 when memory address is aligned.
8233 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8234 this split to select lvx instruction when memory address is aligned.
8235 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8236 instruction when memory address is aligned.
8237 (*vsx_le_perm_load_v16qi): Likewise.
8238 (four unnamed splitters): Modify to select the stvx instruction
8239 when memory is aligned.
8241 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8243 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8244 which appears in the queue multiple times.
8246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8247 Alan Hayward <alan.hayward@arm.com>
8248 David Sherwood <david.sherwood@arm.com>
8250 * tree-vectorizer.h (vec_lower_bound): New structure.
8251 (_loop_vec_info): Add check_nonzero and lower_bounds.
8252 (LOOP_VINFO_CHECK_NONZERO): New macro.
8253 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8254 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8255 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8256 fields. Make seg_len the distance travelled, not including the
8258 (dr_direction_indicator): Declare.
8259 (dr_zero_step_indicator): Likewise.
8260 (dr_known_forward_stride_p): Likewise.
8261 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8263 (runtime_alias_check_p): Allow runtime alias checks with
8265 (operator ==): Compare access_size and align.
8266 (prune_runtime_alias_test_list): Rework for new distinction between
8267 the access_size and seg_len.
8268 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8270 (get_segment_min_max): New function.
8271 (create_intersect_range_checks): Use it.
8272 (dr_step_indicator): New function.
8273 (dr_direction_indicator): Likewise.
8274 (dr_zero_step_indicator): Likewise.
8275 (dr_known_forward_stride_p): Likewise.
8276 * tree-loop-distribution.c (data_ref_segment_size): Return
8277 DR_STEP * (niters - 1).
8278 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8280 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8281 (vect_preserves_scalar_order_p): New function, split out from...
8282 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8283 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8284 (vect_vfa_access_size): New function.
8285 (vect_vfa_align): Likewise.
8286 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8287 (dump_lower_bound): New function.
8288 (vect_check_lower_bound): Likewise.
8289 (vect_small_gap_p): Likewise.
8290 (vectorizable_with_step_bound_p): Likewise.
8291 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8292 depencies if the vectorization factor is 1. Convert the checks
8293 for nonzero steps into checks on the bounds of DR_STEP. Try using
8294 a bunds check for variable steps if the minimum required step is
8295 relatively small. Update calls to the dr_with_seg_len
8296 constructor and to vect_compile_time_alias.
8297 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8299 (vect_loop_versioning): Call it.
8300 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8302 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8305 Alan Hayward <alan.hayward@arm.com>
8306 David Sherwood <david.sherwood@arm.com>
8308 * doc/sourcebuild.texi (vect_scatter_store): Document.
8309 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8311 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8313 * genopinit.c (main): Add supports_vec_scatter_store and
8314 supports_vec_scatter_store_cached to target_optabs.
8315 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8316 IFN_MASK_SCATTER_STORE.
8317 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8319 * internal-fn.h (internal_store_fn_p): Declare.
8320 (internal_fn_stored_value_index): Likewise.
8321 * internal-fn.c (scatter_store_direct): New macro.
8322 (expand_scatter_store_optab_fn): New function.
8323 (direct_scatter_store_optab_supported_p): New macro.
8324 (internal_store_fn_p): New function.
8325 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8326 IFN_MASK_SCATTER_STORE.
8327 (internal_fn_mask_index): Likewise.
8328 (internal_fn_stored_value_index): New function.
8329 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8331 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8332 * optabs-query.c (supports_vec_scatter_store_p): New function.
8333 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8334 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8335 true for scatter stores.
8336 (vect_gather_scatter_fn_p): Handle scatter stores too.
8337 (vect_check_gather_scatter): Consider using scatter stores if
8338 supports_vec_scatter_store_p.
8339 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8341 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8342 internal_fn_stored_value_index.
8343 (check_load_store_masking): Handle scatter stores too.
8344 (vect_get_store_rhs): Make public.
8345 (vectorizable_call): Use internal_store_fn_p.
8346 (vectorizable_store): Handle scatter store internal functions.
8347 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8348 when deciding whether the end of the group has been reached.
8349 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8350 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8351 (mask_scatter_store<mode>): New insns.
8353 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8354 Alan Hayward <alan.hayward@arm.com>
8355 David Sherwood <david.sherwood@arm.com>
8357 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8358 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8359 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8361 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8362 Use vect_truncate_gather_scatter_offset if we can't treat the
8363 operation as a normal gather load or scatter store.
8364 (get_group_load_store_type): Take the gather_scatter_info
8365 as argument. Try using a gather load or scatter store for
8366 single-element groups.
8367 (get_load_store_type): Update calls to get_group_load_store_type
8368 and vect_use_strided_gather_scatters_p.
8370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8371 Alan Hayward <alan.hayward@arm.com>
8372 David Sherwood <david.sherwood@arm.com>
8374 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8375 optional tree argument.
8376 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8378 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8379 but continue to use the current value as a fallback.
8380 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8381 to compare the updates.
8382 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8383 (get_load_store_type): Use it when handling a strided access.
8384 (vect_get_strided_load_store_ops): New function.
8385 (vect_get_data_ptr_increment): Likewise.
8386 (vectorizable_load): Handle strided gather loads. Always pass
8387 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
8394 (mask_gather_load@var{m}): Likewise.
8395 * genopinit.c (main): Add supports_vec_gather_load and
8396 supports_vec_gather_load_cached to target_optabs.
8397 * optabs-tree.c (init_tree_optimization_optabs): Use
8398 ggc_cleared_alloc to allocate target_optabs.
8399 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8400 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8402 * internal-fn.h (internal_load_fn_p): Declare.
8403 (internal_gather_scatter_fn_p): Likewise.
8404 (internal_fn_mask_index): Likewise.
8405 (internal_gather_scatter_fn_supported_p): Likewise.
8406 * internal-fn.c (gather_load_direct): New macro.
8407 (expand_gather_load_optab_fn): New function.
8408 (direct_gather_load_optab_supported_p): New macro.
8409 (direct_internal_fn_optab): New function.
8410 (internal_load_fn_p): Likewise.
8411 (internal_gather_scatter_fn_p): Likewise.
8412 (internal_fn_mask_index): Likewise.
8413 (internal_gather_scatter_fn_supported_p): Likewise.
8414 * optabs-query.c (supports_at_least_one_mode_p): New function.
8415 (supports_vec_gather_load_p): Likewise.
8416 * optabs-query.h (supports_vec_gather_load_p): Declare.
8417 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8418 and memory_type field.
8419 (NUM_PATTERNS): Bump to 15.
8420 * tree-vect-data-refs.c: Include internal-fn.h.
8421 (vect_gather_scatter_fn_p): New function.
8422 (vect_describe_gather_scatter_call): Likewise.
8423 (vect_check_gather_scatter): Try using internal functions for
8424 gather loads. Recognize existing calls to a gather load function.
8425 (vect_analyze_data_refs): Consider using gather loads if
8426 supports_vec_gather_load_p.
8427 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8428 (vect_get_gather_scatter_offset_type): Likewise.
8429 (vect_convert_mask_for_vectype): Likewise.
8430 (vect_add_conversion_to_patterm): Likewise.
8431 (vect_try_gather_scatter_pattern): Likewise.
8432 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8433 (vect_vect_recog_func_ptrs): Add it.
8434 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8435 internal_fn_mask_index and internal_gather_scatter_fn_p.
8436 (check_load_store_masking): Take the gather_scatter_info as an
8437 argument and handle gather loads.
8438 (vect_get_gather_scatter_ops): New function.
8439 (vectorizable_call): Check internal_load_fn_p.
8440 (vectorizable_load): Likewise. Handle gather load internal
8442 (vectorizable_store): Update call to check_load_store_masking.
8443 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8444 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8445 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8446 (aarch64_gather_scale_operand_d): New predicates.
8447 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8448 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
8455 * doc/md.texi (fold_left_plus_@var{m}): Document.
8456 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8457 * internal-fn.c (fold_left_direct): Define.
8458 (expand_fold_left_optab_fn): Likewise.
8459 (direct_fold_left_optab_supported_p): Likewise.
8460 * fold-const-call.c (fold_const_fold_left): New function.
8461 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8462 * tree-parloops.c (valid_reduction_p): New function.
8463 (gather_scalar_reductions): Use it.
8464 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8465 (vect_finish_replace_stmt): Declare.
8466 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8467 (needs_fold_left_reduction_p): New function, split out from...
8468 (vect_is_simple_reduction): ...here. Accept reductions that
8469 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8470 (vect_force_simple_reduction): Also store the reduction type in
8471 the assignment's STMT_VINFO_REDUC_TYPE.
8472 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8473 (merge_with_identity): New function.
8474 (vect_expand_fold_left): Likewise.
8475 (vectorize_fold_left_reduction): Likewise.
8476 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8477 scalar phi in place for it. Check for target support and reject
8478 cases that would reassociate the operation. Defer the transform
8479 phase to vectorize_fold_left_reduction.
8480 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8481 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8482 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8486 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8487 call to ifc_temp_var.
8489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8490 Alan Hayward <alan.hayward@arm.com>
8491 David Sherwood <david.sherwood@arm.com>
8493 * target.def (legitimize_address_displacement): Take the original
8494 offset as a poly_int.
8495 * targhooks.h (default_legitimize_address_displacement): Update
8497 * targhooks.c (default_legitimize_address_displacement): Likewise.
8498 * doc/tm.texi: Regenerate.
8499 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8500 as an argument, moving assert of ad->disp == ad->disp_term to...
8501 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8502 Try calling targetm.legitimize_address_displacement before expanding
8503 the address rather than afterwards, and adjust for the new interface.
8504 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8505 Match the new hook interface. Handle SVE addresses.
8506 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8509 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8511 * Makefile.in (OBJS): Add early-remat.o.
8512 * target.def (select_early_remat_modes): New hook.
8513 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8514 * doc/tm.texi: Regenerate.
8515 * targhooks.h (default_select_early_remat_modes): Declare.
8516 * targhooks.c (default_select_early_remat_modes): New function.
8517 * timevar.def (TV_EARLY_REMAT): New timevar.
8518 * passes.def (pass_early_remat): New pass.
8519 * tree-pass.h (make_pass_early_remat): Declare.
8520 * early-remat.c: New file.
8521 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8523 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8526 Alan Hayward <alan.hayward@arm.com>
8527 David Sherwood <david.sherwood@arm.com>
8529 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8530 vfm1 with a bound_epilog parameter.
8531 (vect_do_peeling): Update calls accordingly, and move the prologue
8532 call earlier in the function. Treat the base bound_epilog as 0 for
8533 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8534 this base when peeling for gaps.
8535 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8536 with fully-masked loops.
8537 (vect_estimate_min_profitable_iters): Handle the single peeled
8538 iteration in that case.
8540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8541 Alan Hayward <alan.hayward@arm.com>
8542 David Sherwood <david.sherwood@arm.com>
8544 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8545 single-element interleaving even if the size is not a power of 2.
8546 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8547 accesses for single-element interleaving if the group size is
8550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8551 Alan Hayward <alan.hayward@arm.com>
8552 David Sherwood <david.sherwood@arm.com>
8554 * doc/md.texi (fold_extract_last_@var{m}): Document.
8555 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8556 * optabs.def (fold_extract_last_optab): New optab.
8557 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8558 * internal-fn.c (fold_extract_direct): New macro.
8559 (expand_fold_extract_optab_fn): Likewise.
8560 (direct_fold_extract_optab_supported_p): Likewise.
8561 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8562 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8563 EXTRACT_LAST_REDUCTION.
8564 (get_initial_def_for_reduction): Do not create an initial vector
8565 for EXTRACT_LAST_REDUCTION reductions.
8566 (vectorizable_reduction): Leave the scalar phi in place for
8567 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8568 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8569 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8570 transform phase to vectorizable_condition.
8571 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8573 (vect_finish_stmt_generation): ...here.
8574 (vect_finish_replace_stmt): New function.
8575 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8576 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8578 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8581 Alan Hayward <alan.hayward@arm.com>
8582 David Sherwood <david.sherwood@arm.com>
8584 * doc/md.texi (extract_last_@var{m}): Document.
8585 * optabs.def (extract_last_optab): New optab.
8586 * internal-fn.def (EXTRACT_LAST): New internal function.
8587 * internal-fn.c (cond_unary_direct): New macro.
8588 (expand_cond_unary_optab_fn): Likewise.
8589 (direct_cond_unary_optab_supported_p): Likewise.
8590 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8591 loops using EXTRACT_LAST.
8592 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8593 (extract_last_<mode>): ...this optab.
8594 (vec_extract<mode><Vel>): Update accordingly.
8596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8597 Alan Hayward <alan.hayward@arm.com>
8598 David Sherwood <david.sherwood@arm.com>
8600 * target.def (empty_mask_is_expensive): New hook.
8601 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8602 * doc/tm.texi: Regenerate.
8603 * targhooks.h (default_empty_mask_is_expensive): Declare.
8604 * targhooks.c (default_empty_mask_is_expensive): New function.
8605 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8606 if the target says that empty masks are expensive.
8607 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8609 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8612 Alan Hayward <alan.hayward@arm.com>
8613 David Sherwood <david.sherwood@arm.com>
8615 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8616 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8617 (vect_use_loop_mask_for_alignment_p): New function.
8618 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8619 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8620 niters_skip argument. Make sure that the first niters_skip elements
8621 of the first iteration are inactive.
8622 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8623 Update call to vect_set_loop_masks_directly.
8624 (get_misalign_in_elems): New function, split out from...
8625 (vect_gen_prolog_loop_niters): ...here.
8626 (vect_update_init_of_dr): Take a code argument that specifies whether
8627 the adjustment should be added or subtracted.
8628 (vect_update_init_of_drs): Likewise.
8629 (vect_prepare_for_masked_peels): New function.
8630 (vect_do_peeling): Skip prologue peeling if we're using a mask
8631 instead. Update call to vect_update_inits_of_drs.
8632 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8634 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8635 alignment. Do not include the number of peeled iterations in
8636 the minimum threshold in that case.
8637 (vectorizable_induction): Adjust the start value down by
8638 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8639 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8640 Take the number of skipped iterations into account when calculating
8642 * tree-vect-stmts.c (vect_gen_while_not): New function.
8644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8645 Alan Hayward <alan.hayward@arm.com>
8646 David Sherwood <david.sherwood@arm.com>
8648 * doc/sourcebuild.texi (vect_fully_masked): Document.
8649 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8651 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8653 (vect_analyze_loop_2): ...here. Don't check the vectorization
8654 factor against the number of loop iterations if the loop is
8657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8658 Alan Hayward <alan.hayward@arm.com>
8659 David Sherwood <david.sherwood@arm.com>
8661 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8662 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8663 (dump_groups): Update accordingly.
8664 (iv_use::mem_type): New member variable.
8665 (address_p): New function.
8666 (record_use): Add a mem_type argument and initialize the new
8668 (record_group_use): Add a mem_type argument. Use address_p.
8669 Remove obsolete null checks of base_object. Update call to record_use.
8670 (find_interesting_uses_op): Update call to record_group_use.
8671 (find_interesting_uses_cond): Likewise.
8672 (find_interesting_uses_address): Likewise.
8673 (get_mem_type_for_internal_fn): New function.
8674 (find_address_like_use): Likewise.
8675 (find_interesting_uses_stmt): Try find_address_like_use before
8676 calling find_interesting_uses_op.
8677 (addr_offset_valid_p): Use the iv mem_type field as the type
8678 of the addressed memory.
8679 (add_autoinc_candidates): Likewise.
8680 (get_address_cost): Likewise.
8681 (split_small_address_groups_p): Use address_p.
8682 (split_address_groups): Likewise.
8683 (add_iv_candidate_for_use): Likewise.
8684 (autoinc_possible_for_pair): Likewise.
8685 (rewrite_groups): Likewise.
8686 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8687 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8688 (get_alias_ptr_type_for_ptr_address): New function.
8689 (rewrite_use_address): Rewrite address uses in calls that were
8690 identified by find_address_like_use.
8692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8693 Alan Hayward <alan.hayward@arm.com>
8694 David Sherwood <david.sherwood@arm.com>
8696 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8698 * gimple-expr.h (is_gimple_addressable: Likewise.
8699 * gimple-expr.c (is_gimple_address): Likewise.
8700 * internal-fn.c (expand_call_mem_ref): New function.
8701 (expand_mask_load_optab_fn): Use it.
8702 (expand_mask_store_optab_fn): Likewise.
8704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8705 Alan Hayward <alan.hayward@arm.com>
8706 David Sherwood <david.sherwood@arm.com>
8708 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8709 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8710 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8711 (cond_umax@var{mode}): Document.
8712 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8713 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8714 (cond_umin_optab, cond_umax_optab): New optabs.
8715 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8716 (COND_IOR, COND_XOR): New internal functions.
8717 * internal-fn.h (get_conditional_internal_fn): Declare.
8718 * internal-fn.c (cond_binary_direct): New macro.
8719 (expand_cond_binary_optab_fn): Likewise.
8720 (direct_cond_binary_optab_supported_p): Likewise.
8721 (get_conditional_internal_fn): New function.
8722 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8723 Cope with reduction statements that are vectorized as calls rather
8725 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8726 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8727 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8728 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8729 (UNSPEC_COND_EOR): New unspecs.
8730 (optab): Add mappings for them.
8731 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8732 (sve_int_op, sve_fp_op): New int attributes.
8734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8735 Alan Hayward <alan.hayward@arm.com>
8736 David Sherwood <david.sherwood@arm.com>
8738 * optabs.def (while_ult_optab): New optab.
8739 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8740 * internal-fn.def (WHILE_ULT): New internal function.
8741 * internal-fn.h (direct_internal_fn_supported_p): New override
8742 that takes two types as argument.
8743 * internal-fn.c (while_direct): New macro.
8744 (expand_while_optab_fn): New function.
8745 (convert_optab_supported_p): Likewise.
8746 (direct_while_optab_supported_p): New macro.
8747 * wide-int.h (wi::udiv_ceil): New function.
8748 * tree-vectorizer.h (rgroup_masks): New structure.
8749 (vec_loop_masks): New typedef.
8750 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8752 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8753 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8754 (vect_max_vf): New function.
8755 (slpeel_make_loop_iterate_ntimes): Delete.
8756 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8757 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8758 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8759 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8760 internal-fn.h, stor-layout.h and optabs-query.h.
8761 (vect_set_loop_mask): New function.
8762 (add_preheader_seq): Likewise.
8763 (add_header_seq): Likewise.
8764 (interleave_supported_p): Likewise.
8765 (vect_maybe_permute_loop_masks): Likewise.
8766 (vect_set_loop_masks_directly): Likewise.
8767 (vect_set_loop_condition_masked): Likewise.
8768 (vect_set_loop_condition_unmasked): New function, split out from
8769 slpeel_make_loop_iterate_ntimes.
8770 (slpeel_make_loop_iterate_ntimes): Rename to..
8771 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8772 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8773 (vect_do_peeling): Update call accordingly.
8774 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8776 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8777 mask_compare_type, can_fully_mask_p and fully_masked_p.
8778 (release_vec_loop_masks): New function.
8779 (_loop_vec_info): Use it to free the loop masks.
8780 (can_produce_all_loop_masks_p): New function.
8781 (vect_get_max_nscalars_per_iter): Likewise.
8782 (vect_verify_full_masking): Likewise.
8783 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8784 retries, and free the mask rgroups before retrying. Check loop-wide
8785 reasons for disallowing fully-masked loops. Make the final decision
8786 about whether use a fully-masked loop or not.
8787 (vect_estimate_min_profitable_iters): Do not assume that peeling
8788 for the number of iterations will be needed for fully-masked loops.
8789 (vectorizable_reduction): Disable fully-masked loops.
8790 (vectorizable_live_operation): Likewise.
8791 (vect_halve_mask_nunits): New function.
8792 (vect_double_mask_nunits): Likewise.
8793 (vect_record_loop_mask): Likewise.
8794 (vect_get_loop_mask): Likewise.
8795 (vect_transform_loop): Handle the case in which the final loop
8796 iteration might handle a partial vector. Call vect_set_loop_condition
8797 instead of slpeel_make_loop_iterate_ntimes.
8798 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8799 (check_load_store_masking): New function.
8800 (prepare_load_store_mask): Likewise.
8801 (vectorizable_store): Handle fully-masked loops.
8802 (vectorizable_load): Likewise.
8803 (supportable_widening_operation): Use vect_halve_mask_nunits for
8805 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8806 (vect_gen_while): New function.
8807 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8808 (aarch64_uqdec<mode>): New insn.
8810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8811 Alan Hayward <alan.hayward@arm.com>
8812 David Sherwood <david.sherwood@arm.com>
8814 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8815 (reduc_xor_scal_optab): New optabs.
8816 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8817 (reduc_xor_scal_@var{m}): Document.
8818 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8819 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8821 * fold-const-call.c (fold_const_call): Handle them.
8822 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8823 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8824 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8825 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8826 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8827 (UNSPEC_XORV): New unspecs.
8828 (optab): Add entries for them.
8829 (BITWISEV): New int iterator.
8830 (bit_reduc_op): New int attributes.
8832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8833 Alan Hayward <alan.hayward@arm.com>
8834 David Sherwood <david.sherwood@arm.com>
8836 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8837 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8838 * optabs.def (vec_shl_insert_optab): New optab.
8839 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8840 (duplicate_and_interleave): Likewise.
8841 * tree-vect-loop.c: Include internal-fn.h.
8842 (neutral_op_for_slp_reduction): New function, split out from
8843 get_initial_defs_for_reduction.
8844 (get_initial_def_for_reduction): Handle option 2 for variable-length
8845 vectors by loading the neutral value into a vector and then shifting
8846 the initial value into element 0.
8847 (get_initial_defs_for_reduction): Replace the code argument with
8848 the neutral value calculated by neutral_op_for_slp_reduction.
8849 Use gimple_build_vector for constant-length vectors.
8850 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8851 but the first group_size elements have a neutral value.
8852 Use duplicate_and_interleave otherwise.
8853 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8854 Update call to get_initial_defs_for_reduction. Handle SLP
8855 reductions for variable-length vectors by creating one vector
8856 result for each scalar result, with the elements associated
8857 with other scalar results stubbed out with the neutral value.
8858 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8859 Require IFN_VEC_SHL_INSERT for double reductions on
8860 variable-length vectors, or SLP reductions that have
8861 a neutral value. Require can_duplicate_and_interleave_p
8862 support for variable-length unchained SLP reductions if there
8863 is no neutral value, such as for MIN/MAX reductions. Also require
8864 the number of vector elements to be a multiple of the number of
8865 SLP statements when doing variable-length unchained SLP reductions.
8866 Update call to vect_create_epilog_for_reduction.
8867 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8868 and remove initial values.
8869 (duplicate_and_interleave): Make public.
8870 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8871 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8874 Alan Hayward <alan.hayward@arm.com>
8875 David Sherwood <david.sherwood@arm.com>
8877 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8878 (can_duplicate_and_interleave_p): New function.
8879 (vect_get_and_check_slp_defs): Take the vector of statements
8880 rather than just the current one. Remove excess parentheses.
8881 Restriction rejectinon of vect_constant_def and vect_external_def
8882 for variable-length vectors to boolean types, or types for which
8883 can_duplicate_and_interleave_p is false.
8884 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8885 (duplicate_and_interleave): New function.
8886 (vect_get_constant_vectors): Use gimple_build_vector for
8887 constant-length vectors and suitable variable-length constant
8888 vectors. Use duplicate_and_interleave for other variable-length
8889 vectors. Don't defer the update when inserting new statements.
8891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8892 Alan Hayward <alan.hayward@arm.com>
8893 David Sherwood <david.sherwood@arm.com>
8895 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8896 min_profitable_iters doesn't go negative.
8898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8903 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8904 * optabs.def (vec_mask_load_lanes_optab): New optab.
8905 (vec_mask_store_lanes_optab): Likewise.
8906 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8907 (MASK_STORE_LANES): Likewise.
8908 * internal-fn.c (mask_load_lanes_direct): New macro.
8909 (mask_store_lanes_direct): Likewise.
8910 (expand_mask_load_optab_fn): Handle masked operations.
8911 (expand_mask_load_lanes_optab_fn): New macro.
8912 (expand_mask_store_optab_fn): Handle masked operations.
8913 (expand_mask_store_lanes_optab_fn): New macro.
8914 (direct_mask_load_lanes_optab_supported_p): Likewise.
8915 (direct_mask_store_lanes_optab_supported_p): Likewise.
8916 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8918 (vect_load_lanes_supported): Likewise.
8919 * tree-vect-data-refs.c (strip_conversion): New function.
8920 (can_group_stmts_p): Likewise.
8921 (vect_analyze_data_ref_accesses): Use it instead of checking
8922 for a pair of assignments.
8923 (vect_store_lanes_supported): Take a masked_p parameter.
8924 (vect_load_lanes_supported): Likewise.
8925 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8926 vect_store_lanes_supported and vect_load_lanes_supported.
8927 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8928 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8929 parameter. Don't allow gaps for masked accesses.
8930 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8931 and vect_load_lanes_supported.
8932 (get_load_store_type): Take a masked_p parameter and update
8933 call to get_group_load_store_type.
8934 (vectorizable_store): Update call to get_load_store_type.
8935 Handle IFN_MASK_STORE_LANES.
8936 (vectorizable_load): Update call to get_load_store_type.
8937 Handle IFN_MASK_LOAD_LANES.
8939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8940 Alan Hayward <alan.hayward@arm.com>
8941 David Sherwood <david.sherwood@arm.com>
8943 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8945 * config/aarch64/aarch64-protos.h
8946 (aarch64_sve_struct_memory_operand_p): Declare.
8947 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8948 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8949 (VPRED, vpred): Handle SVE structure modes.
8950 * config/aarch64/constraints.md (Utx): New constraint.
8951 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8952 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8953 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8954 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8955 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8956 structure modes. Split into pieces after RA.
8957 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8958 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8960 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8961 SVE structure modes.
8962 (aarch64_classify_address): Likewise.
8963 (sizetochar): Move earlier in file.
8964 (aarch64_print_operand): Handle SVE register lists.
8965 (aarch64_array_mode): New function.
8966 (aarch64_sve_struct_memory_operand_p): Likewise.
8967 (TARGET_ARRAY_MODE): Redefine.
8969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8970 Alan Hayward <alan.hayward@arm.com>
8971 David Sherwood <david.sherwood@arm.com>
8973 * target.def (array_mode): New target hook.
8974 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8975 * doc/tm.texi: Regenerate.
8976 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8977 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8978 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8980 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8984 Alan Hayward <alan.hayward@arm.com>
8985 David Sherwood <david.sherwood@arm.com>
8987 * fold-const.c (fold_binary_loc): Check the argument types
8988 rather than the result type when testing for a vector operation.
8990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8992 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8993 * doc/tm.texi: Regenerate.
8995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9000 (sve): Document new AArch64 extension.
9001 * doc/md.texi (w): Extend the description of the AArch64
9002 constraint to include SVE vectors.
9003 (Upl, Upa): Document new AArch64 predicate constraints.
9004 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9006 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9007 (msve-vector-bits=): New option.
9008 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9009 SVE when these are disabled.
9010 (sve): New extension.
9011 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9012 modes. Adjust their number of units based on aarch64_sve_vg.
9013 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9014 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9015 aarch64_addr_query_type.
9016 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9017 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9018 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9019 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9020 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9021 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9022 (aarch64_simd_imm_zero_p): Delete.
9023 (aarch64_check_zero_based_sve_index_immediate): Declare.
9024 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9025 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9026 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9027 (aarch64_sve_float_mul_immediate_p): Likewise.
9028 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9030 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9031 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9032 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9033 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9034 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9035 (aarch64_regmode_natural_size): Likewise.
9036 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9037 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9039 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9040 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9041 for VG and the SVE predicate registers.
9042 (V_ALIASES): Add a "z"-prefixed alias.
9043 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9044 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9045 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9046 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9047 (REG_CLASS_NAMES): Add entries for them.
9048 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9049 and the predicate registers.
9050 (aarch64_sve_vg): Declare.
9051 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9052 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9053 (REGMODE_NATURAL_SIZE): Define.
9054 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9056 * config/aarch64/aarch64.c: Include cfgrtl.h.
9057 (simd_immediate_info): Add a constructor for series vectors,
9058 and an associated step field.
9059 (aarch64_sve_vg): New variable.
9060 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9061 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9062 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9063 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9064 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9065 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9066 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9067 (aarch64_get_mask_mode): New functions.
9068 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9069 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9070 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9071 predicate modes and predicate registers. Explicitly restrict
9072 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9073 to store a vector mode if it is recognized by
9074 aarch64_classify_vector_mode.
9075 (aarch64_regmode_natural_size): New function.
9076 (aarch64_hard_regno_caller_save_mode): Return the original mode
9078 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9079 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9080 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9081 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9083 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9084 does not overlap dest if the function is frame-related. Handle
9086 (aarch64_split_add_offset): New function.
9087 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9088 them aarch64_add_offset.
9089 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9090 and update call to aarch64_sub_sp.
9091 (aarch64_add_cfa_expression): New function.
9092 (aarch64_expand_prologue): Pass extra temporary registers to the
9093 functions above. Handle the case in which we need to emit new
9094 DW_CFA_expressions for registers that were originally saved
9095 relative to the stack pointer, but now have to be expressed
9096 relative to the frame pointer.
9097 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9099 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9100 IP0 and IP1 values for SVE frames.
9101 (aarch64_expand_vec_series): New function.
9102 (aarch64_expand_sve_widened_duplicate): Likewise.
9103 (aarch64_expand_sve_const_vector): Likewise.
9104 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9105 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9106 into the register, rather than emitting a SET directly.
9107 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9108 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9109 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9110 (offset_9bit_signed_scaled_p): New functions.
9111 (aarch64_replicate_bitmask_imm): New function.
9112 (aarch64_bitmask_imm): Use it.
9113 (aarch64_cannot_force_const_mem): Reject expressions involving
9114 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9115 (aarch64_classify_index): Handle SVE indices, by requiring
9116 a plain register index with a scale that matches the element size.
9117 (aarch64_classify_address): Handle SVE addresses. Assert that
9118 the mode of the address is VOIDmode or an integer mode.
9119 Update call to aarch64_classify_symbol.
9120 (aarch64_classify_symbolic_expression): Update call to
9121 aarch64_classify_symbol.
9122 (aarch64_const_vec_all_in_range_p): New function.
9123 (aarch64_print_vector_float_operand): Likewise.
9124 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9125 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9126 and the FP immediates 1.0 and 0.5.
9127 (aarch64_print_address_internal): Handle SVE addresses.
9128 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9129 (aarch64_regno_regclass): Handle predicate registers.
9130 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9132 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9133 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9134 (aarch64_convert_sve_vector_bits): New function.
9135 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9136 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9138 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9139 Handle SVE vector and predicate modes. Accept VL-based constants
9140 that need only one temporary register, and VL offsets that require
9141 no temporary registers.
9142 (aarch64_conditional_register_usage): Mark the predicate registers
9143 as fixed if SVE isn't available.
9144 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9145 Return true for SVE vector and predicate modes.
9146 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9147 rather than an unsigned int. Handle SVE modes.
9148 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9150 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9152 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9153 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9154 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9155 (aarch64_sve_float_mul_immediate_p): New functions.
9156 (aarch64_sve_valid_immediate): New function.
9157 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9158 Explicitly reject structure modes. Check for INDEX constants.
9159 Handle PTRUE and PFALSE constants.
9160 (aarch64_check_zero_based_sve_index_immediate): New function.
9161 (aarch64_simd_imm_zero_p): Delete.
9162 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9163 vector modes. Accept constants in the range of CNT[BHWD].
9164 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9165 ask for an Advanced SIMD mode.
9166 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9167 (aarch64_simd_vector_alignment): Handle SVE predicates.
9168 (aarch64_vectorize_preferred_vector_alignment): New function.
9169 (aarch64_simd_vector_alignment_reachable): Use it instead of
9171 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9172 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9174 (MAX_VECT_LEN): Delete.
9175 (expand_vec_perm_d): Add a vec_flags field.
9176 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9177 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9178 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9180 (aarch64_evpc_rev): Rename to...
9181 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9182 (aarch64_evpc_rev_global): New function.
9183 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9184 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9186 (aarch64_evpc_sve_tbl): New function.
9187 (aarch64_expand_vec_perm_const_1): Update after rename of
9188 aarch64_evpc_rev. Handle SVE permutes too, trying
9189 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9190 than aarch64_evpc_tbl.
9191 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9192 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9193 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9194 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9195 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9196 (aarch64_expand_sve_vcond): New functions.
9197 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9198 of aarch64_vector_mode_p.
9199 (aarch64_dwarf_poly_indeterminate_value): New function.
9200 (aarch64_compute_pressure_classes): Likewise.
9201 (aarch64_can_change_mode_class): Likewise.
9202 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9203 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9204 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9205 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9206 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9207 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9208 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9209 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9211 (Dn, Dl, Dr): Accept const as well as const_vector.
9212 (Dz): Likewise. Compare against CONST0_RTX.
9213 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9214 of "vector" where appropriate.
9215 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9216 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9217 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9218 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9219 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9220 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9221 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9222 (v_int_equiv): Extend to SVE modes.
9223 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9225 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9226 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9227 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9228 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9229 (SVE_COND_FP_CMP): New int iterators.
9230 (perm_hilo): Handle the new unpack unspecs.
9231 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9233 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9234 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9235 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9236 (aarch64_equality_operator, aarch64_constant_vector_operand)
9237 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9238 (aarch64_sve_nonimmediate_operand): Likewise.
9239 (aarch64_sve_general_operand): Likewise.
9240 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9241 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9242 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9243 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9244 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9245 (aarch64_sve_float_arith_immediate): Likewise.
9246 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9247 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9248 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9249 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9250 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9251 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9252 (aarch64_sve_float_arith_operand): Likewise.
9253 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9254 (aarch64_sve_float_mul_operand): Likewise.
9255 (aarch64_sve_vec_perm_operand): Likewise.
9256 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9257 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9258 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9259 as well as const_vector.
9260 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9261 in file. Use CONST0_RTX and CONSTM1_RTX.
9262 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9263 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9264 Use aarch64_simd_imm_zero.
9265 * config/aarch64/aarch64-sve.md: New file.
9266 * config/aarch64/aarch64.md: Include it.
9267 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9268 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9269 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9270 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9271 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9272 (sve): New attribute.
9273 (enabled): Disable instructions with the sve attribute unless
9275 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9276 aarch64_expand_mov_immediate.
9277 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9278 CNT[BHSD] immediates.
9279 (movti): Split CONST_POLY_INT moves into two halves.
9280 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9281 Split additions that need a temporary here if the destination
9282 is the stack pointer.
9283 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9284 (*add<mode>3_poly_1): New instruction.
9285 (set_clobber_cc): New expander.
9287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9289 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9290 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9291 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9292 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9293 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9294 Change innermode from fixed_mode_size to machine_mode.
9295 (simplify_subreg): Update call accordingly. Handle a constant-sized
9296 subreg of a variable-length CONST_VECTOR.
9298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9299 Alan Hayward <alan.hayward@arm.com>
9300 David Sherwood <david.sherwood@arm.com>
9302 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9303 (add_offset_to_base): New function, split out from...
9304 (create_mem_ref): ...here. When handling a scale other than 1,
9305 check first whether the address is valid without the offset.
9306 Add it into the base if so, leaving the index and scale as-is.
9308 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9311 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9312 fold_for_warn before checking if arg2 is INTEGER_CST.
9314 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9316 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9317 (store_multiple_operation): Delete.
9318 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9319 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9320 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9321 guarded by TARGET_STRING.
9322 (rs6000_output_load_multiple): Delete.
9323 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9324 OPTION_MASK_STRING / TARGET_STRING handling.
9325 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9326 (const rs6000_opt_masks) <"string">: Change mask to 0.
9327 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9328 (MASK_STRING): Delete.
9329 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9331 (load_multiple): Delete.
9338 (store_multiple): Delete.
9345 (movmemsi_8reg): Delete.
9346 (corresponding unnamed define_insn): Delete.
9347 (movmemsi_6reg): Delete.
9348 (corresponding unnamed define_insn): Delete.
9349 (movmemsi_4reg): Delete.
9350 (corresponding unnamed define_insn): Delete.
9351 (movmemsi_2reg): Delete.
9352 (corresponding unnamed define_insn): Delete.
9353 (movmemsi_1reg): Delete.
9354 (corresponding unnamed define_insn): Delete.
9355 * config/rs6000/rs6000.opt (mno-string): New.
9356 (mstring): Replace by deprecation warning stub.
9357 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9359 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9361 * regrename.c (regrename_do_replace): If replacing the same
9362 reg multiple times, try to reuse last created gen_raw_REG.
9365 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9366 main to workaround a bug in GDB.
9368 2018-01-12 Tom de Vries <tom@codesourcery.com>
9371 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9373 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9375 PR rtl-optimization/80481
9376 * ira-color.c (get_cap_member): New function.
9377 (allocnos_conflict_by_live_ranges_p): Use it.
9378 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9379 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9381 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9384 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9385 (*saddl_se_1): Ditto.
9387 (*ssubl_se_1): Ditto.
9389 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9391 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9392 rather than wi::to_widest for DR_INITs.
9393 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9394 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9395 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9397 (vect_analyze_group_access_1): Note that here.
9399 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9401 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9402 polynomial type sizes.
9404 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9406 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9407 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9408 (gimple_add_tmp_var): Likewise.
9410 2018-01-12 Martin Liska <mliska@suse.cz>
9412 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9413 (gimple_alloc_sizes): Likewise.
9414 (dump_gimple_statistics): Use PRIu64 in printf format.
9415 * gimple.h: Change uint64_t to int.
9417 2018-01-12 Martin Liska <mliska@suse.cz>
9419 * tree-core.h: Use uint64_t instead of int.
9420 * tree.c (tree_node_counts): Likewise.
9421 (tree_node_sizes): Likewise.
9422 (dump_tree_statistics): Use PRIu64 in printf format.
9424 2018-01-12 Martin Liska <mliska@suse.cz>
9426 * Makefile.in: As qsort_chk is implemented in vec.c, add
9427 vec.o to linkage of gencfn-macros.
9428 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9429 passing the info to record_node_allocation_statistics.
9430 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9432 * ggc-common.c (struct ggc_usage): Add operator== and use
9433 it in operator< and compare function.
9434 * mem-stats.h (struct mem_usage): Likewise.
9435 * vec.c (struct vec_usage): Remove operator< and compare
9436 function. Can be simply inherited.
9438 2018-01-12 Martin Jambor <mjambor@suse.cz>
9441 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9442 * tree-ssa-math-opts.c: Include domwalk.h.
9443 (convert_mult_to_fma_1): New function.
9444 (fma_transformation_info): New type.
9445 (fma_deferring_state): Likewise.
9446 (cancel_fma_deferring): New function.
9447 (result_of_phi): Likewise.
9448 (last_fma_candidate_feeds_initial_phi): Likewise.
9449 (convert_mult_to_fma): Added deferring logic, split actual
9450 transformation to convert_mult_to_fma_1.
9451 (math_opts_dom_walker): New type.
9452 (math_opts_dom_walker::after_dom_children): New method, body moved
9453 here from pass_optimize_widening_mul::execute, added deferring logic
9455 (pass_optimize_widening_mul::execute): Moved most of code to
9456 math_opts_dom_walker::after_dom_children.
9457 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9458 * config/i386/i386.c (ix86_option_override_internal): Added
9459 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9461 2018-01-12 Richard Biener <rguenther@suse.de>
9464 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9465 inline instance vars.
9467 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9470 * config/rx/rx.c (rx_is_restricted_memory_address):
9473 2018-01-12 Richard Biener <rguenther@suse.de>
9475 PR tree-optimization/80846
9476 * target.def (split_reduction): New target hook.
9477 * targhooks.c (default_split_reduction): New function.
9478 * targhooks.h (default_split_reduction): Declare.
9479 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9480 target requests first reduce vectors by combining low and high
9482 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9483 (get_vectype_for_scalar_type_and_size): Export.
9484 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9485 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9486 * doc/tm.texi: Regenerate.
9487 * config/i386/i386.c (ix86_split_reduction): Implement
9488 TARGET_VECTORIZE_SPLIT_REDUCTION.
9490 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9493 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9494 in PIC mode except for TARGET_VXWORKS_RTP.
9495 * config/sparc/sparc.c: Include cfgrtl.h.
9496 (TARGET_INIT_PIC_REG): Define.
9497 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9498 (sparc_pic_register_p): New predicate.
9499 (sparc_legitimate_address_p): Use it.
9500 (sparc_legitimize_pic_address): Likewise.
9501 (sparc_delegitimize_address): Likewise.
9502 (sparc_mode_dependent_address_p): Likewise.
9503 (gen_load_pcrel_sym): Remove 4th parameter.
9504 (load_got_register): Adjust call to above. Remove obsolete stuff.
9505 (sparc_expand_prologue): Do not call load_got_register here.
9506 (sparc_flat_expand_prologue): Likewise.
9507 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9508 (sparc_use_pseudo_pic_reg): New function.
9509 (sparc_init_pic_reg): Likewise.
9510 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9511 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9513 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9515 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9516 Add item for branch_cost.
9518 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9520 PR rtl-optimization/83565
9521 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9522 not extend the result to a larger mode for rotate operations.
9523 (num_sign_bit_copies1): Likewise.
9525 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9528 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9530 Use values-Xc.o for -pedantic.
9531 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9533 2018-01-12 Martin Liska <mliska@suse.cz>
9536 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9538 (possible_polymorphic_call_targets): Use it.
9539 (ipa_devirt): Likewise.
9541 2018-01-12 Martin Liska <mliska@suse.cz>
9543 * profile-count.h (enum profile_quality): Use 0 as invalid
9544 enum value of profile_quality.
9546 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9548 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9549 -mext-string options.
9551 2018-01-12 Richard Biener <rguenther@suse.de>
9553 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9554 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9555 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9557 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9559 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9561 * configure.ac (--with-long-double-format): Add support for the
9562 configuration option to change the default long double format on
9564 * config.gcc (powerpc*-linux*-*): Likewise.
9565 * configure: Regenerate.
9566 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9567 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9568 used without modification.
9570 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9572 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9573 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9574 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9575 MISC_BUILTIN_SPEC_BARRIER.
9576 (rs6000_init_builtins): Likewise.
9577 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9579 (speculation_barrier): New define_insn.
9580 * doc/extend.texi: Document __builtin_speculation_barrier.
9582 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9585 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9586 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9587 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9589 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9590 integral modes instead of "ss" and "sd".
9591 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9592 vectors with 32-bit and 64-bit elements.
9593 (vecdupssescalarmodesuffix): New mode attribute.
9594 (vec_dup<mode>): Use it.
9596 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9599 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9600 frame if argument is passed on stack.
9602 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9605 * ree.c (combine_reaching_defs): Optimize also
9606 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9607 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9609 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9612 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9614 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9617 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9618 after they are computed.
9620 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9622 PR tree-optimization/83695
9623 * gimple-loop-linterchange.cc
9624 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9625 reset cached scev information after interchange.
9626 (pass_linterchange::execute): Remove call to scev_reset_htab.
9628 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9630 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9631 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9632 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9633 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9634 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9635 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9636 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9637 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9638 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9639 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9640 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9641 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9642 (V_lane_reg): Likewise.
9643 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9645 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9646 (vfmal_lane_low<mode>_intrinsic,
9647 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9648 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9649 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9650 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9651 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9652 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9654 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9656 * config/arm/arm-cpus.in (fp16fml): New feature.
9657 (ALL_SIMD): Add fp16fml.
9658 (armv8.2-a): Add fp16fml as an option.
9659 (armv8.3-a): Likewise.
9660 (armv8.4-a): Add fp16fml as part of fp16.
9661 * config/arm/arm.h (TARGET_FP16FML): Define.
9662 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9664 * config/arm/arm-modes.def (V2HF): Define.
9665 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9666 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9667 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9668 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9669 vfmsl_low, vfmsl_high): New set of builtins.
9670 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9671 (vfml_op): New code attribute.
9672 (VFMLHALVES): New int iterator.
9673 (VFML, VFMLSEL): New mode attributes.
9674 (V_reg): Define mapping for V2HF.
9675 (V_hi, V_lo): New mode attributes.
9676 (VF_constraint): Likewise.
9677 (vfml_half, vfml_half_selector): New int attributes.
9678 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9680 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9681 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9683 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9684 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9685 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9686 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9688 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9689 Document new effective target and option set.
9691 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9693 * config/arm/arm-cpus.in (armv8_4): New feature.
9694 (ARMv8_4a): New fgroup.
9695 (armv8.4-a): New arch.
9696 * config/arm/arm-tables.opt: Regenerate.
9697 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9698 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9699 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9700 Add matching rules for -march=armv8.4-a and extensions.
9701 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9703 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9706 * config/rx/rx.md (BW): New mode attribute.
9707 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9709 2018-01-11 Richard Biener <rguenther@suse.de>
9711 PR tree-optimization/83435
9712 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9713 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9714 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9716 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9717 Alan Hayward <alan.hayward@arm.com>
9718 David Sherwood <david.sherwood@arm.com>
9720 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9722 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9723 (aarch64_print_address_internal): Use it to check for a zero offset.
9725 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9726 Alan Hayward <alan.hayward@arm.com>
9727 David Sherwood <david.sherwood@arm.com>
9729 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9730 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9731 Return a poly_int64 rather than a HOST_WIDE_INT.
9732 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9733 rather than a HOST_WIDE_INT.
9734 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9735 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9736 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9737 final_offset from HOST_WIDE_INT to poly_int64.
9738 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9739 to_constant when getting the number of units in an Advanced SIMD
9741 (aarch64_builtin_vectorized_function): Check for a constant number
9743 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9745 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9746 attribute instead of GET_MODE_NUNITS.
9747 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9748 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9749 GET_MODE_SIZE for fixed-size registers.
9750 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9751 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9752 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9753 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9754 (aarch64_print_operand, aarch64_print_address_internal)
9755 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9756 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9757 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9758 Handle polynomial GET_MODE_SIZE.
9759 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9760 wider than SImode without modification.
9761 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9762 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9763 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9764 passing and returning SVE modes.
9765 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9766 rather than GEN_INT.
9767 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9768 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9769 (aarch64_allocate_and_probe_stack_space): Likewise.
9770 (aarch64_layout_frame): Cope with polynomial offsets.
9771 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9772 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9774 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9775 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9776 poly_int64 rather than a HOST_WIDE_INT.
9777 (aarch64_get_separate_components, aarch64_process_components)
9778 (aarch64_expand_prologue, aarch64_expand_epilogue)
9779 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9780 (aarch64_anchor_offset): New function, split out from...
9781 (aarch64_legitimize_address): ...here.
9782 (aarch64_builtin_vectorization_cost): Handle polynomial
9783 TYPE_VECTOR_SUBPARTS.
9784 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9786 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9787 number of elements from the PARALLEL rather than the mode.
9788 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9789 rather than GET_MODE_BITSIZE.
9790 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9791 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9792 (aarch64_expand_vec_perm_const_1): Handle polynomial
9793 d->perm.length () and d->perm elements.
9794 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9795 Apply to_constant to d->perm elements.
9796 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9797 polynomial CONST_VECTOR_NUNITS.
9798 (aarch64_move_pointer): Take amount as a poly_int64 rather
9800 (aarch64_progress_pointer): Avoid temporary variable.
9801 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9802 the mode attribute instead of GET_MODE.
9804 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9805 Alan Hayward <alan.hayward@arm.com>
9806 David Sherwood <david.sherwood@arm.com>
9808 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9809 x exists before using it.
9810 (aarch64_add_constant_internal): Rename to...
9811 (aarch64_add_offset_1): ...this. Replace regnum with separate
9812 src and dest rtxes. Handle the case in which they're different,
9813 including when the offset is zero. Replace scratchreg with an rtx.
9814 Use 2 additions if there is no spare register into which we can
9815 move a 16-bit constant.
9816 (aarch64_add_constant): Delete.
9817 (aarch64_add_offset): Replace reg with separate src and dest
9818 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9819 Use aarch64_add_offset_1.
9820 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9821 an rtx rather than an int. Take the delta as a poly_int64
9822 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9823 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9824 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9825 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9826 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9828 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9829 aarch64_add_constant.
9831 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9833 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9834 Use scalar_float_mode.
9836 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9838 * config/aarch64/aarch64-simd.md
9839 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9840 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9841 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9842 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9843 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9844 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9845 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9846 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9847 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9848 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9850 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9853 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9854 targ_options->x_arm_arch_string is non NULL.
9856 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9858 * config/aarch64/aarch64.h
9859 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9861 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9864 * expmed.c (emit_store_flag_force): Swap if const op0
9865 and change VOIDmode to mode of op0.
9867 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9869 PR rtl-optimization/83761
9870 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9871 than bytes to mode_for_size.
9873 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9876 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9877 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9880 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9883 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9884 when in layout mode.
9885 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9886 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9889 2018-01-10 Michael Collison <michael.collison@arm.com>
9891 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9892 * config/aarch64/aarch64-option-extension.def: Add
9893 AARCH64_OPT_EXTENSION of 'fp16fml'.
9894 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9895 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9896 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9897 * config/aarch64/constraints.md (Ui7): New constraint.
9898 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9899 (VFMLA_SEL_W): Ditto.
9902 (VFMLA16_LOW): New int iterator.
9903 (VFMLA16_HIGH): Ditto.
9904 (UNSPEC_FMLAL): New unspec.
9905 (UNSPEC_FMLSL): Ditto.
9906 (UNSPEC_FMLAL2): Ditto.
9907 (UNSPEC_FMLSL2): Ditto.
9908 (f16mac): New code attribute.
9909 * config/aarch64/aarch64-simd-builtins.def
9910 (aarch64_fmlal_lowv2sf): Ditto.
9911 (aarch64_fmlsl_lowv2sf): Ditto.
9912 (aarch64_fmlalq_lowv4sf): Ditto.
9913 (aarch64_fmlslq_lowv4sf): Ditto.
9914 (aarch64_fmlal_highv2sf): Ditto.
9915 (aarch64_fmlsl_highv2sf): Ditto.
9916 (aarch64_fmlalq_highv4sf): Ditto.
9917 (aarch64_fmlslq_highv4sf): Ditto.
9918 (aarch64_fmlal_lane_lowv2sf): Ditto.
9919 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9920 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9921 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9922 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9923 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9924 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9925 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9926 (aarch64_fmlal_lane_highv2sf): Ditto.
9927 (aarch64_fmlsl_lane_highv2sf): Ditto.
9928 (aarch64_fmlal_laneq_highv2sf): Ditto.
9929 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9930 (aarch64_fmlalq_lane_highv4sf): Ditto.
9931 (aarch64_fmlsl_lane_highv4sf): Ditto.
9932 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9933 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9934 * config/aarch64/aarch64-simd.md:
9935 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9936 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9937 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9938 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9939 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9940 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9941 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9942 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9943 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9944 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9945 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9946 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9947 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9948 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9949 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9950 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9951 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9952 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9953 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9954 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9955 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9956 (vfmlsl_low_u32): Ditto.
9957 (vfmlalq_low_u32): Ditto.
9958 (vfmlslq_low_u32): Ditto.
9959 (vfmlal_high_u32): Ditto.
9960 (vfmlsl_high_u32): Ditto.
9961 (vfmlalq_high_u32): Ditto.
9962 (vfmlslq_high_u32): Ditto.
9963 (vfmlal_lane_low_u32): Ditto.
9964 (vfmlsl_lane_low_u32): Ditto.
9965 (vfmlal_laneq_low_u32): Ditto.
9966 (vfmlsl_laneq_low_u32): Ditto.
9967 (vfmlalq_lane_low_u32): Ditto.
9968 (vfmlslq_lane_low_u32): Ditto.
9969 (vfmlalq_laneq_low_u32): Ditto.
9970 (vfmlslq_laneq_low_u32): Ditto.
9971 (vfmlal_lane_high_u32): Ditto.
9972 (vfmlsl_lane_high_u32): Ditto.
9973 (vfmlal_laneq_high_u32): Ditto.
9974 (vfmlsl_laneq_high_u32): Ditto.
9975 (vfmlalq_lane_high_u32): Ditto.
9976 (vfmlslq_lane_high_u32): Ditto.
9977 (vfmlalq_laneq_high_u32): Ditto.
9978 (vfmlslq_laneq_high_u32): Ditto.
9979 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9980 (AARCH64_FL_FOR_ARCH8_4): New.
9981 (AARCH64_ISA_F16FML): New ISA flag.
9982 (TARGET_F16FML): New feature flag for fp16fml.
9983 (doc/invoke.texi): Document new fp16fml option.
9985 2018-01-10 Michael Collison <michael.collison@arm.com>
9987 * config/aarch64/aarch64-builtins.c:
9988 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9989 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9990 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9991 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9992 (AARCH64_ISA_SHA3): New ISA flag.
9993 (TARGET_SHA3): New feature flag for sha3.
9994 * config/aarch64/iterators.md (sha512_op): New int attribute.
9995 (CRYPTO_SHA512): New int iterator.
9996 (UNSPEC_SHA512H): New unspec.
9997 (UNSPEC_SHA512H2): Ditto.
9998 (UNSPEC_SHA512SU0): Ditto.
9999 (UNSPEC_SHA512SU1): Ditto.
10000 * config/aarch64/aarch64-simd-builtins.def
10001 (aarch64_crypto_sha512hqv2di): New builtin.
10002 (aarch64_crypto_sha512h2qv2di): Ditto.
10003 (aarch64_crypto_sha512su0qv2di): Ditto.
10004 (aarch64_crypto_sha512su1qv2di): Ditto.
10005 (aarch64_eor3qv8hi): Ditto.
10006 (aarch64_rax1qv2di): Ditto.
10007 (aarch64_xarqv2di): Ditto.
10008 (aarch64_bcaxqv8hi): Ditto.
10009 * config/aarch64/aarch64-simd.md:
10010 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10011 (aarch64_crypto_sha512su0qv2di): Ditto.
10012 (aarch64_crypto_sha512su1qv2di): Ditto.
10013 (aarch64_eor3qv8hi): Ditto.
10014 (aarch64_rax1qv2di): Ditto.
10015 (aarch64_xarqv2di): Ditto.
10016 (aarch64_bcaxqv8hi): Ditto.
10017 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10018 (vsha512h2q_u64): Ditto.
10019 (vsha512su0q_u64): Ditto.
10020 (vsha512su1q_u64): Ditto.
10021 (veor3q_u16): Ditto.
10022 (vrax1q_u64): Ditto.
10023 (vxarq_u64): Ditto.
10024 (vbcaxq_u16): Ditto.
10025 * config/arm/types.md (crypto_sha512): New type attribute.
10026 (crypto_sha3): Ditto.
10027 (doc/invoke.texi): Document new sha3 option.
10029 2018-01-10 Michael Collison <michael.collison@arm.com>
10031 * config/aarch64/aarch64-builtins.c:
10032 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10033 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10034 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10035 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10036 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10037 (AARCH64_ISA_SM4): New ISA flag.
10038 (TARGET_SM4): New feature flag for sm4.
10039 * config/aarch64/aarch64-simd-builtins.def
10040 (aarch64_sm3ss1qv4si): Ditto.
10041 (aarch64_sm3tt1aq4si): Ditto.
10042 (aarch64_sm3tt1bq4si): Ditto.
10043 (aarch64_sm3tt2aq4si): Ditto.
10044 (aarch64_sm3tt2bq4si): Ditto.
10045 (aarch64_sm3partw1qv4si): Ditto.
10046 (aarch64_sm3partw2qv4si): Ditto.
10047 (aarch64_sm4eqv4si): Ditto.
10048 (aarch64_sm4ekeyqv4si): Ditto.
10049 * config/aarch64/aarch64-simd.md:
10050 (aarch64_sm3ss1qv4si): Ditto.
10051 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10052 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10053 (aarch64_sm4eqv4si): Ditto.
10054 (aarch64_sm4ekeyqv4si): Ditto.
10055 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10056 (sm3part_op): Ditto.
10057 (CRYPTO_SM3TT): Ditto.
10058 (CRYPTO_SM3PART): Ditto.
10059 (UNSPEC_SM3SS1): New unspec.
10060 (UNSPEC_SM3TT1A): Ditto.
10061 (UNSPEC_SM3TT1B): Ditto.
10062 (UNSPEC_SM3TT2A): Ditto.
10063 (UNSPEC_SM3TT2B): Ditto.
10064 (UNSPEC_SM3PARTW1): Ditto.
10065 (UNSPEC_SM3PARTW2): Ditto.
10066 (UNSPEC_SM4E): Ditto.
10067 (UNSPEC_SM4EKEY): Ditto.
10068 * config/aarch64/constraints.md (Ui2): New constraint.
10069 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10070 * config/arm/types.md (crypto_sm3): New type attribute.
10071 (crypto_sm4): Ditto.
10072 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10073 (vsm3tt1aq_u32): Ditto.
10074 (vsm3tt1bq_u32): Ditto.
10075 (vsm3tt2aq_u32): Ditto.
10076 (vsm3tt2bq_u32): Ditto.
10077 (vsm3partw1q_u32): Ditto.
10078 (vsm3partw2q_u32): Ditto.
10079 (vsm4eq_u32): Ditto.
10080 (vsm4ekeyq_u32): Ditto.
10081 (doc/invoke.texi): Document new sm4 option.
10083 2018-01-10 Michael Collison <michael.collison@arm.com>
10085 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10086 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10087 (AARCH64_FL_FOR_ARCH8_4): New.
10088 (AARCH64_FL_V8_4): New flag.
10089 (doc/invoke.texi): Document new armv8.4-a option.
10091 2018-01-10 Michael Collison <michael.collison@arm.com>
10093 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10094 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10095 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10096 * config/aarch64/aarch64-option-extension.def: Add
10097 AARCH64_OPT_EXTENSION of 'sha2'.
10098 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10099 (crypto): Disable sha2 and aes if crypto disabled.
10100 (crypto): Enable aes and sha2 if enabled.
10101 (simd): Disable sha2 and aes if simd disabled.
10102 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10104 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10105 (TARGET_SHA2): New feature flag for sha2.
10106 (TARGET_AES): New feature flag for aes.
10107 * config/aarch64/aarch64-simd.md:
10108 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10109 conditional on TARGET_AES.
10110 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10111 (aarch64_crypto_sha1hsi): Make pattern conditional
10113 (aarch64_crypto_sha1hv4si): Ditto.
10114 (aarch64_be_crypto_sha1hv4si): Ditto.
10115 (aarch64_crypto_sha1su1v4si): Ditto.
10116 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10117 (aarch64_crypto_sha1su0v4si): Ditto.
10118 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10119 (aarch64_crypto_sha256su0v4si): Ditto.
10120 (aarch64_crypto_sha256su1v4si): Ditto.
10121 (doc/invoke.texi): Document new aes and sha2 options.
10123 2018-01-10 Martin Sebor <msebor@redhat.com>
10125 PR tree-optimization/83781
10126 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10129 2018-01-11 Martin Sebor <msebor@gmail.com>
10130 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10132 PR tree-optimization/83501
10133 PR tree-optimization/81703
10135 * tree-ssa-strlen.c (get_string_cst): Rename...
10136 (get_string_len): ...to this. Handle global constants.
10137 (handle_char_store): Adjust.
10139 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10140 Jim Wilson <jimw@sifive.com>
10142 * config/riscv/riscv-protos.h (riscv_output_return): New.
10143 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10144 (riscv_attribute_table, riscv_output_return),
10145 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10146 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10147 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10149 (riscv_expand_prologue): Add early return for naked function.
10150 (riscv_expand_epilogue): Likewise.
10151 (riscv_function_ok_for_sibcall): Return false for naked function.
10152 (riscv_set_current_function): New.
10153 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10154 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10155 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10156 * doc/extend.texi (RISC-V Function Attributes): New.
10158 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10160 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10161 check for 128-bit long double before checking TCmode.
10162 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10163 128-bit long doubles before checking TFmode or TCmode.
10164 (FLOAT128_IBM_P): Likewise.
10166 2018-01-10 Martin Sebor <msebor@redhat.com>
10168 PR tree-optimization/83671
10169 * builtins.c (c_strlen): Unconditionally return zero for the empty
10171 Use -Warray-bounds for warnings.
10172 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10173 for non-constant array indices with COMPONENT_REF, arrays of
10174 arrays, and pointers to arrays.
10175 (gimple_fold_builtin_strlen): Determine and set length range for
10176 non-constant character arrays.
10178 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10180 PR middle-end/81897
10181 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10184 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10186 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10188 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10191 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10192 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10193 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10194 indexed_or_indirect_operand predicate.
10195 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10196 (*vsx_le_perm_load_v8hi): Likewise.
10197 (*vsx_le_perm_load_v16qi): Likewise.
10198 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10199 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10200 (*vsx_le_perm_store_v8hi): Likewise.
10201 (*vsx_le_perm_store_v16qi): Likewise.
10202 (eight unnamed splitters): Likewise.
10204 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10206 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10207 * config/rs6000/emmintrin.h: Likewise.
10208 * config/rs6000/mmintrin.h: Likewise.
10209 * config/rs6000/xmmintrin.h: Likewise.
10211 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10214 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10216 * tree.c (tree_nop_conversion): Return true for location wrapper
10218 (maybe_wrap_with_location): New function.
10219 (selftest::check_strip_nops): New function.
10220 (selftest::test_location_wrappers): New function.
10221 (selftest::tree_c_tests): Call it.
10222 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10223 (maybe_wrap_with_location): New decl.
10224 (EXPR_LOCATION_WRAPPER_P): New macro.
10225 (location_wrapper_p): New inline function.
10226 (tree_strip_any_location_wrapper): New inline function.
10228 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10231 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10232 stack_realign_offset for the largest alignment of stack slot
10234 (ix86_find_max_used_stack_alignment): New function.
10235 (ix86_finalize_stack_frame_flags): Use it. Set
10236 max_used_stack_alignment if we don't realign stack.
10237 * config/i386/i386.h (machine_function): Add
10238 max_used_stack_alignment.
10240 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10242 * config/arm/arm.opt (-mbranch-cost): New option.
10243 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10246 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10249 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10250 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10252 2018-01-10 Richard Biener <rguenther@suse.de>
10255 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10256 early out so it also covers the case where we have a non-NULL
10259 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10261 PR tree-optimization/83753
10262 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10263 for non-strided grouped accesses if the number of elements is 1.
10265 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10268 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10269 * i386.h (TARGET_USE_GATHER): Define.
10270 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10272 2018-01-10 Martin Liska <mliska@suse.cz>
10275 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10276 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10278 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10279 CLEANUP_NO_PARTITIONING is not set.
10281 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10283 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10284 for vectors, as a partial revert of r254296.
10285 * rtl.h (const_vec_p): Delete.
10286 (const_vec_duplicate_p): Don't test for vector CONSTs.
10287 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10288 * expmed.c (make_tree): Likewise.
10291 * common.md (E, F): Use CONSTANT_P instead of checking for
10293 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10294 checking for CONST_VECTOR.
10296 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10298 PR middle-end/83575
10299 * predict.c (force_edge_cold): Handle in more sane way edges
10300 with no prediction.
10302 2018-01-09 Carl Love <cel@us.ibm.com>
10304 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10306 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10307 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10308 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10309 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10310 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10311 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10312 * config/rs6000/rs6000-protos.h: Add extern defition for
10313 rs6000_generate_float2_double_code.
10314 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10316 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10317 (float2_v2df): Add define_expand.
10319 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10322 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10323 op_mode in the force_to_mode call.
10325 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10327 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10328 instead of checking each element individually.
10329 (aarch64_evpc_uzp): Likewise.
10330 (aarch64_evpc_zip): Likewise.
10331 (aarch64_evpc_ext): Likewise.
10332 (aarch64_evpc_rev): Likewise.
10333 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10334 instead of checking each element individually. Return true without
10336 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10337 whether all selected elements come from the same input, instead of
10338 checking each element individually. Remove calls to gen_rtx_REG,
10339 start_sequence and end_sequence and instead assert that no rtl is
10342 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10344 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10345 order of HIGH and CONST checks.
10347 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10349 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10350 if the destination isn't an SSA_NAME.
10352 2018-01-09 Richard Biener <rguenther@suse.de>
10354 PR tree-optimization/83668
10355 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10357 (canonicalize_loop_form): ... here, renamed from ...
10358 (canonicalize_loop_closed_ssa_form): ... this and amended to
10359 swap successor edges for loop exit blocks to make us use
10360 the RPO order we need for initial schedule generation.
10362 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10364 PR tree-optimization/64811
10365 * match.pd: When optimizing comparisons with Inf, avoid
10366 introducing or losing exceptions from comparisons with NaN.
10368 2018-01-09 Martin Liska <mliska@suse.cz>
10371 * asan.c (shadow_mem_size): Add gcc_assert.
10373 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10375 Don't save registers in main().
10378 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10379 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10380 * config/avr/avr.c (avr_set_current_function): Don't error if
10381 naked, OS_task or OS_main are specified at the same time.
10382 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10384 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10386 * common/config/avr/avr-common.c (avr_option_optimization_table):
10387 Switch on -mmain-is-OS_task for optimizing compilations.
10389 2018-01-09 Richard Biener <rguenther@suse.de>
10391 PR tree-optimization/83572
10392 * graphite.c: Include cfganal.h.
10393 (graphite_transform_loops): Connect infinite loops to exit
10394 and remove fake edges at the end.
10396 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10398 * ipa-inline.c (edge_badness): Revert accidental checkin.
10400 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10403 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10404 symbols; not inline clones.
10406 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10409 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10410 hard registers. Formatting fixes.
10412 PR preprocessor/83722
10413 * gcc.c (try_generate_repro): Pass
10414 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10415 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10418 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10419 Kito Cheng <kito.cheng@gmail.com>
10421 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10422 (riscv_leaf_function_p): Delete.
10423 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10425 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10427 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10429 (do_ifelse): New function.
10430 (do_isel): New function.
10431 (do_sub3): New function.
10432 (do_add3): New function.
10433 (do_load_mask_compare): New function.
10434 (do_overlap_load_compare): New function.
10435 (expand_compare_loop): New function.
10436 (expand_block_compare): Call expand_compare_loop() when appropriate.
10437 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10438 option description.
10439 (-mblock-compare-inline-loop-limit): New option.
10441 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10444 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10445 Reverse order of second and third operands in first alternative.
10446 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10447 of first and second elements in UNSPEC_VPERMR vector.
10448 (altivec_expand_vec_perm_le): Likewise.
10450 2018-01-08 Jeff Law <law@redhat.com>
10452 PR rtl-optimizatin/81308
10453 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10454 (process_switch): If group_case_labels makes a change, then set
10456 (pass_convert_switch::execute): If a switch is converted, then
10457 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10459 PR rtl-optimization/81308
10460 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10463 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10465 PR target/83663 - Revert r255946
10466 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10467 generation for cases where splatting a value is not useful.
10468 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10469 across a vec_duplicate and a paradoxical subreg forming a vector
10470 mode to a vec_concat.
10472 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10474 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10475 -march=armv8.3-a variants.
10476 * config/arm/t-multilib: Likewise.
10477 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10479 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10481 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10483 (cceq_ior_compare_complement): Give it a name so I can use it, and
10484 change boolean_or_operator predicate to boolean_operator so it can
10485 be used to generate a crand.
10486 (eqne): New code iterator.
10487 (bd/bd_neg): New code_attrs.
10488 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10489 a single define_insn.
10490 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10491 decrement (bdnzt/bdnzf/bdzt/bdzf).
10492 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10493 with the new names of the branch decrement patterns, and added the
10494 names of the branch decrement conditional patterns.
10496 2018-01-08 Richard Biener <rguenther@suse.de>
10498 PR tree-optimization/83563
10499 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10502 2018-01-08 Richard Biener <rguenther@suse.de>
10504 PR middle-end/83713
10505 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10507 2018-01-08 Richard Biener <rguenther@suse.de>
10509 PR tree-optimization/83685
10510 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10511 references to abnormals.
10513 2018-01-08 Richard Biener <rguenther@suse.de>
10516 * dwarf2out.c (output_indirect_strings): Handle empty
10517 skeleton_debug_str_hash.
10518 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10520 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10522 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10523 (emit_store_direct): Likewise.
10524 (arc_trampoline_adjust_address): Likewise.
10525 (arc_asm_trampoline_template): New function.
10526 (arc_initialize_trampoline): Use asm_trampoline_template.
10527 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10528 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10529 * config/arc/arc.md (flush_icache): Delete pattern.
10531 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10533 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10534 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10537 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10540 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10541 by not USED_FOR_TARGET.
10542 (make_pass_resolve_sw_modes): Likewise.
10544 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10546 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10549 2018-01-08 Richard Biener <rguenther@suse.de>
10551 PR middle-end/83580
10552 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10554 2018-01-08 Richard Biener <rguenther@suse.de>
10556 PR middle-end/83517
10557 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10559 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10561 PR middle-end/81897
10562 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10563 basic blocks with a small number of successors.
10564 (convert_control_dep_chain_into_preds): Improve handling of
10566 (dump_predicates): Split apart into...
10567 (dump_pred_chain): ...here...
10568 (dump_pred_info): ...and here.
10569 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10570 (can_chain_union_be_invalidated_p): Improve check for invalidation
10572 (uninit_uses_cannot_happen): Avoid unnecessary if
10573 convert_control_dep_chain_into_preds yielded nothing.
10575 2018-01-06 Martin Sebor <msebor@redhat.com>
10577 PR tree-optimization/83640
10578 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10579 subtracting negative offset from size.
10580 (builtin_access::overlap): Adjust offset bounds of the access to fall
10581 within the size of the object if possible.
10583 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10585 PR rtl-optimization/83699
10586 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10587 extract_bit_field_as_subreg to cases in which the extracted
10588 value is also a vector.
10590 * lra-constraints.c (process_alt_operands): Test for the equivalence
10591 substitutions when detecting a possible reload cycle.
10593 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10596 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10597 by default if flag_selective_schedling{,2}. Formatting fixes.
10599 PR rtl-optimization/83682
10600 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10601 if it has non-VECTOR_MODE element mode.
10602 (vec_duplicate_p): Likewise.
10604 PR middle-end/83694
10605 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10606 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10608 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10611 * config/i386/i386-builtin.def
10612 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10613 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10614 Require also OPTION_MASK_ISA_AVX512F in addition to
10615 OPTION_MASK_ISA_GFNI.
10616 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10617 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10618 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10619 to OPTION_MASK_ISA_GFNI.
10620 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10621 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10622 OPTION_MASK_ISA_AVX512BW.
10623 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10624 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10625 addition to OPTION_MASK_ISA_GFNI.
10626 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10627 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10628 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10629 to OPTION_MASK_ISA_GFNI.
10630 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10631 a requirement for all ISAs rather than any of them with a few
10633 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10635 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10636 bitmasks to be enabled with 3 exceptions, instead of requiring any
10637 enabled ISA with lots of exceptions.
10638 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10639 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10640 Change avx512bw in isa attribute to avx512f.
10641 * config/i386/sgxintrin.h: Add license boilerplate.
10642 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10643 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10644 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10645 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10647 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10648 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10649 temporarily sse2 rather than sse if not enabled already.
10652 * config/i386/sse.md (VI248_VLBW): Rename to ...
10653 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10654 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10655 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10656 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10657 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10658 mode iterator instead of VI248_VLBW.
10660 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10662 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10663 (record_modified): Skip clobbers; add debug output.
10664 (param_change_prob): Use sreal frequencies.
10666 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10668 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10669 punt for user-aligned variables.
10671 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10673 * tree-chrec.c (chrec_contains_symbols): Return true for
10676 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10679 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10680 of (x|y) == x for BICS pattern.
10682 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10684 PR tree-optimization/83605
10685 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10686 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10689 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10691 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10692 * config/epiphany/rtems.h: New file.
10694 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10695 Uros Bizjak <ubizjak@gmail.com>
10698 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10699 QIreg_operand instead of register_operand predicate.
10700 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10701 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10702 comments instead of -fmitigate[-_]rop.
10704 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10707 * cgraphunit.c (symbol_table::compile): Switch to text_section
10708 before calling assembly_start debug hook.
10709 * run-rtl-passes.c (run_rtl_passes): Likewise.
10712 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10714 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10715 range_int_cst_p rather than !symbolic_range_p before calling
10716 extract_range_from_multiplicative_op_1.
10718 2018-01-04 Jeff Law <law@redhat.com>
10720 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10721 redundant test in assertion.
10723 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10725 * doc/rtl.texi: Document machine_mode wrapper classes.
10727 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10729 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10730 using tree_to_uhwi.
10732 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10734 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10735 the VEC_PERM_EXPR fold to fail.
10737 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10740 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10741 to switched_sections.
10743 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10746 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10747 test for d.testing.
10749 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10752 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10753 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10755 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10758 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10759 is BLKmode and bitpos not zero or mode change is needed.
10761 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10764 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10767 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10770 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10771 instead of MULT rtx. Update all corresponding splitters.
10772 (*saddl_se): Ditto.
10773 (*ssub<modesuffix>): Ditto.
10774 (*ssubl_se): Ditto.
10775 (*cmp_sadd_di): Update split patterns.
10776 (*cmp_sadd_si): Ditto.
10777 (*cmp_sadd_sidi): Ditto.
10778 (*cmp_ssub_di): Ditto.
10779 (*cmp_ssub_si): Ditto.
10780 (*cmp_ssub_sidi): Ditto.
10781 * config/alpha/predicates.md (const23_operand): New predicate.
10782 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10783 Look for ASHIFT, not MULT inner operand.
10784 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10786 2018-01-04 Martin Liska <mliska@suse.cz>
10788 PR gcov-profile/83669
10789 * gcov.c (output_intermediate_file): Add version to intermediate
10791 * doc/gcov.texi: Document new field 'version' in intermediate
10792 file format. Fix location of '-k' option of gcov command.
10794 2018-01-04 Martin Liska <mliska@suse.cz>
10797 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10799 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10801 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10803 2018-01-03 Martin Sebor <msebor@redhat.com>
10805 PR tree-optimization/83655
10806 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10807 checking calls with invalid arguments.
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10811 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10812 (vectorizable_mask_load_store): Delete.
10813 (vectorizable_call): Return false for masked loads and stores.
10814 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10815 instead of gimple_assign_rhs1.
10816 (vectorizable_load): Handle IFN_MASK_LOAD.
10817 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10821 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10823 (vectorizable_mask_load_store): ...here.
10824 (vectorizable_load): ...and here.
10826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10828 * tree-vect-stmts.c (vect_build_all_ones_mask)
10829 (vect_build_zero_merge_argument): New functions, split out from...
10830 (vectorizable_load): ...here.
10832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10834 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10836 (vectorizable_mask_load_store): ...here.
10837 (vectorizable_store): ...and here.
10839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10841 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10843 (vectorizable_mask_load_store): ...here.
10845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10847 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10848 (vect_model_store_cost): Take a vec_load_store_type instead of a
10850 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10851 (vect_model_store_cost): Take a vec_load_store_type instead of a
10853 (vectorizable_mask_load_store): Update accordingly.
10854 (vectorizable_store): Likewise.
10855 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10859 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10860 IFN_MASK_LOAD calls here rather than...
10861 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10864 Alan Hayward <alan.hayward@arm.com>
10865 David Sherwood <david.sherwood@arm.com>
10867 * expmed.c (extract_bit_field_1): For vector extracts,
10868 fall back to extract_bit_field_as_subreg if vec_extract
10871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10876 they are variable or constant sized.
10877 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10878 slots for constant-sized data.
10880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10881 Alan Hayward <alan.hayward@arm.com>
10882 David Sherwood <david.sherwood@arm.com>
10884 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10885 handling COND_EXPRs with boolean comparisons, try to find a better
10886 basis for the mask type than the boolean itself.
10888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10890 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10891 is calculated and how it can be overridden.
10892 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10893 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10895 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10899 Alan Hayward <alan.hayward@arm.com>
10900 David Sherwood <david.sherwood@arm.com>
10902 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10903 Remove the mode argument.
10904 (aarch64_simd_valid_immediate): Remove the mode and inverse
10906 * config/aarch64/iterators.md (bitsize): New iterator.
10907 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10908 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10909 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10910 aarch64_simd_valid_immediate.
10911 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10912 (aarch64_reg_or_bic_imm): Likewise.
10913 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10914 with an insn_type enum and msl with a modifier_type enum.
10915 Replace element_width with a scalar_mode. Change the shift
10916 to unsigned int. Add constructors for scalar_float_mode and
10917 scalar_int_mode elements.
10918 (aarch64_vect_float_const_representable_p): Delete.
10919 (aarch64_can_const_movi_rtx_p)
10920 (aarch64_simd_scalar_immediate_valid_for_move)
10921 (aarch64_simd_make_constant): Update call to
10922 aarch64_simd_valid_immediate.
10923 (aarch64_advsimd_valid_immediate_hs): New function.
10924 (aarch64_advsimd_valid_immediate): Likewise.
10925 (aarch64_simd_valid_immediate): Remove mode and inverse
10926 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10927 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10928 and aarch64_float_const_representable_p on the result.
10929 (aarch64_output_simd_mov_immediate): Remove mode argument.
10930 Update call to aarch64_simd_valid_immediate and use of
10931 simd_immediate_info.
10932 (aarch64_output_scalar_simd_mov_immediate): Update call
10935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10936 Alan Hayward <alan.hayward@arm.com>
10937 David Sherwood <david.sherwood@arm.com>
10939 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10940 (mode_nunits): Likewise CONST_MODE_NUNITS.
10941 * machmode.def (ADJUST_NUNITS): Document.
10942 * genmodes.c (mode_data::need_nunits_adj): New field.
10943 (blank_mode): Update accordingly.
10944 (adj_nunits): New variable.
10945 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10947 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10948 listed in adj_nunits.
10949 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10950 listed in adj_nunits. Don't emit case statements for such modes.
10951 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10952 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10953 nothing if adj_nunits is nonnull.
10954 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10955 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10956 (emit_mode_fbit): Update use of print_maybe_const_decl.
10957 (emit_move_size): Likewise. Treat the array as non-const
10959 (emit_mode_adjustments): Handle adj_nunits.
10961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10963 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10964 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10965 (VECTOR_MODES): Use it.
10966 (make_vector_modes): Take the prefix as an argument.
10968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10969 Alan Hayward <alan.hayward@arm.com>
10970 David Sherwood <david.sherwood@arm.com>
10972 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10973 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10974 for MODE_VECTOR_BOOL.
10975 * machmode.def (VECTOR_BOOL_MODE): Document.
10976 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10977 (make_vector_bool_mode): New function.
10978 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10980 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10981 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10983 * stor-layout.c (int_mode_for_mode): Likewise.
10984 * tree.c (build_vector_type_for_mode): Likewise.
10985 * varasm.c (output_constant_pool_2): Likewise.
10986 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10987 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10988 for MODE_VECTOR_BOOL.
10989 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10990 of mode class checks.
10991 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10992 instead of a list of mode class checks.
10993 (expand_vector_scalar_condition): Likewise.
10994 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10997 Alan Hayward <alan.hayward@arm.com>
10998 David Sherwood <david.sherwood@arm.com>
11000 * machmode.h (mode_size): Change from unsigned short to
11002 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11003 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11004 or if measurement_type is not polynomial.
11005 (fixed_size_mode::includes_p): Check for constant-sized modes.
11006 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11007 return a poly_uint16 rather than an unsigned short.
11008 (emit_mode_size): Change the type of mode_size from unsigned short
11009 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11010 (emit_mode_adjustments): Cope with polynomial vector sizes.
11011 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11013 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11015 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11016 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11017 * caller-save.c (setup_save_areas): Likewise.
11018 (replace_reg_with_saved_mem): Likewise.
11019 * calls.c (emit_library_call_value_1): Likewise.
11020 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11021 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11022 (gen_lowpart_for_combine): Likewise.
11023 * convert.c (convert_to_integer_1): Likewise.
11024 * cse.c (equiv_constant, cse_insn): Likewise.
11025 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11026 (cselib_subst_to_values): Likewise.
11027 * dce.c (word_dce_process_block): Likewise.
11028 * df-problems.c (df_word_lr_mark_ref): Likewise.
11029 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11030 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11031 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11032 (rtl_for_decl_location): Likewise.
11033 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11034 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11035 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11036 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11037 (expand_expr_real_1): Likewise.
11038 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11039 (pad_below): Likewise.
11040 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11041 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11042 * ira.c (get_subreg_tracking_sizes): Likewise.
11043 * ira-build.c (ira_create_allocno_objects): Likewise.
11044 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11045 (ira_sort_regnos_for_alter_reg): Likewise.
11046 * ira-costs.c (record_operand_costs): Likewise.
11047 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11048 (resolve_simple_move): Likewise.
11049 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11050 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11051 (lra_constraints): Likewise.
11052 (CONST_POOL_OK_P): Reject variable-sized modes.
11053 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11054 (add_pseudo_to_slot, lra_spill): Likewise.
11055 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11056 * optabs-query.c (get_best_extraction_insn): Likewise.
11057 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11058 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11059 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11060 * recog.c (offsettable_address_addr_space_p): Likewise.
11061 * regcprop.c (maybe_mode_change): Likewise.
11062 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11063 * regrename.c (build_def_use): Likewise.
11064 * regstat.c (dump_reg_info): Likewise.
11065 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11066 (find_reloads, find_reloads_subreg_address): Likewise.
11067 * reload1.c (eliminate_regs_1): Likewise.
11068 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11069 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11070 (simplify_binary_operation_1, simplify_subreg): Likewise.
11071 * targhooks.c (default_function_arg_padding): Likewise.
11072 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11073 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11074 (verify_gimple_assign_ternary): Likewise.
11075 * tree-inline.c (estimate_move_cost): Likewise.
11076 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11077 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11078 (get_address_cost_ainc): Likewise.
11079 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11080 (vect_supportable_dr_alignment): Likewise.
11081 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11082 (vectorizable_reduction): Likewise.
11083 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11084 (vectorizable_operation, vectorizable_load): Likewise.
11085 * tree.c (build_same_sized_truth_vector_type): Likewise.
11086 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11087 * var-tracking.c (emit_note_insn_var_location): Likewise.
11088 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11089 (ADDR_VEC_ALIGN): Likewise.
11091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11092 Alan Hayward <alan.hayward@arm.com>
11093 David Sherwood <david.sherwood@arm.com>
11095 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11097 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11098 or if measurement_type is polynomial.
11099 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11100 * combine.c (make_extraction): Likewise.
11101 * dse.c (find_shift_sequence): Likewise.
11102 * dwarf2out.c (mem_loc_descriptor): Likewise.
11103 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11104 (extract_bit_field, extract_low_bits): Likewise.
11105 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11106 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11107 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11108 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11109 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11110 * reload.c (find_reloads): Likewise.
11111 * reload1.c (alter_reg): Likewise.
11112 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11113 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11114 * tree-if-conv.c (predicate_mem_writes): Likewise.
11115 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11116 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11117 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11118 * valtrack.c (dead_debug_insert_temp): Likewise.
11119 * varasm.c (mergeable_constant_section): Likewise.
11120 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11123 Alan Hayward <alan.hayward@arm.com>
11124 David Sherwood <david.sherwood@arm.com>
11126 * expr.c (expand_assignment): Cope with polynomial mode sizes
11127 when assigning to a CONCAT.
11129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 Alan Hayward <alan.hayward@arm.com>
11131 David Sherwood <david.sherwood@arm.com>
11133 * machmode.h (mode_precision): Change from unsigned short to
11135 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11137 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11138 or if measurement_type is not polynomial.
11139 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11140 in which the mode is already known to be a scalar_int_mode.
11141 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11142 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11144 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11145 for GET_MODE_PRECISION.
11146 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11147 for GET_MODE_PRECISION.
11148 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11150 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11151 (expand_field_assignment, make_extraction): Likewise.
11152 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11153 (get_last_value): Likewise.
11154 * convert.c (convert_to_integer_1): Likewise.
11155 * cse.c (cse_insn): Likewise.
11156 * expr.c (expand_expr_real_1): Likewise.
11157 * lra-constraints.c (simplify_operand_subreg): Likewise.
11158 * optabs-query.c (can_atomic_load_p): Likewise.
11159 * optabs.c (expand_atomic_load): Likewise.
11160 (expand_atomic_store): Likewise.
11161 * ree.c (combine_reaching_defs): Likewise.
11162 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11163 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11164 * tree.h (type_has_mode_precision_p): Likewise.
11165 * ubsan.c (instrument_si_overflow): Likewise.
11167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11168 Alan Hayward <alan.hayward@arm.com>
11169 David Sherwood <david.sherwood@arm.com>
11171 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11172 polynomial numbers of units.
11173 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11174 (valid_vector_subparts_p): New function.
11175 (build_vector_type): Remove temporary shim and take the number
11176 of units as a poly_uint64 rather than an int.
11177 (build_opaque_vector_type): Take the number of units as a
11178 poly_uint64 rather than an int.
11179 * tree.c (build_vector_from_ctor): Handle polynomial
11180 TYPE_VECTOR_SUBPARTS.
11181 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11182 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11183 (build_vector_from_val): If the number of units is variable,
11184 use build_vec_duplicate_cst for constant operands and
11185 VEC_DUPLICATE_EXPR otherwise.
11186 (make_vector_type): Remove temporary is_constant ().
11187 (build_vector_type, build_opaque_vector_type): Take the number of
11188 units as a poly_uint64 rather than an int.
11189 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11191 * cfgexpand.c (expand_debug_expr): Likewise.
11192 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11193 (store_constructor, expand_expr_real_1): Likewise.
11194 (const_scalar_mask_from_tree): Likewise.
11195 * fold-const-call.c (fold_const_reduction): Likewise.
11196 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11197 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11198 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11199 (fold_relational_const): Likewise.
11200 (native_interpret_vector): Likewise. Change the size from an
11201 int to an unsigned int.
11202 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11203 TYPE_VECTOR_SUBPARTS.
11204 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11205 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11206 duplicating a non-constant operand into a variable-length vector.
11207 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11208 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11209 * ipa-icf.c (sem_variable::equals): Likewise.
11210 * match.pd: Likewise.
11211 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11212 * print-tree.c (print_node): Likewise.
11213 * stor-layout.c (layout_type): Likewise.
11214 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11215 * tree-cfg.c (verify_gimple_comparison): Likewise.
11216 (verify_gimple_assign_binary): Likewise.
11217 (verify_gimple_assign_ternary): Likewise.
11218 (verify_gimple_assign_single): Likewise.
11219 * tree-pretty-print.c (dump_generic_node): Likewise.
11220 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11221 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11222 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11223 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11224 (vect_shift_permute_load_chain): Likewise.
11225 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11226 (expand_vector_condition, optimize_vector_constructor): Likewise.
11227 (lower_vec_perm, get_compute_type): Likewise.
11228 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11229 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11230 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11231 (vect_recog_mask_conversion_pattern): Likewise.
11232 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11233 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11234 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11235 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11236 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11237 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11238 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11239 (supportable_widening_operation): Likewise.
11240 (supportable_narrowing_operation): Likewise.
11241 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11243 * varasm.c (output_constant): Likewise.
11245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11246 Alan Hayward <alan.hayward@arm.com>
11247 David Sherwood <david.sherwood@arm.com>
11249 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11250 so that both the length == 3 and length != 3 cases set up their
11251 own permute vectors. Add comments explaining why we know the
11252 number of elements is constant.
11253 (vect_permute_load_chain): Likewise.
11255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11256 Alan Hayward <alan.hayward@arm.com>
11257 David Sherwood <david.sherwood@arm.com>
11259 * machmode.h (mode_nunits): Change from unsigned char to
11261 (ONLY_FIXED_SIZE_MODES): New macro.
11262 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11263 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11264 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11266 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11267 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11268 or if measurement_type is not polynomial.
11269 * genmodes.c (ZERO_COEFFS): New macro.
11270 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11272 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11273 Use ZERO_COEFFS when emitting initializers.
11274 * data-streamer.h (bp_pack_poly_value): New function.
11275 (bp_unpack_poly_value): Likewise.
11276 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11277 for GET_MODE_NUNITS.
11278 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11279 for GET_MODE_NUNITS.
11280 * tree.c (make_vector_type): Remove temporary shim and make
11281 the real function take the number of units as a poly_uint64
11282 rather than an int.
11283 (build_vector_type_for_mode): Handle polynomial nunits.
11284 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11285 * emit-rtl.c (const_vec_series_p_1): Likewise.
11286 (gen_rtx_CONST_VECTOR): Likewise.
11287 * fold-const.c (test_vec_duplicate_folding): Likewise.
11288 * genrecog.c (validate_pattern): Likewise.
11289 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11290 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11291 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11292 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11293 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11294 * rtlanal.c (subreg_get_info): Likewise.
11295 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11296 (vect_grouped_load_supported): Likewise.
11297 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11298 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11299 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11300 (simplify_const_unary_operation, simplify_binary_operation_1)
11301 (simplify_const_binary_operation, simplify_ternary_operation)
11302 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11303 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11304 instead of CONST_VECTOR_NUNITS.
11305 * varasm.c (output_constant_pool_2): Likewise.
11306 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11307 explicit-encoded elements in the XVEC for variable-length vectors.
11309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11311 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11314 Alan Hayward <alan.hayward@arm.com>
11315 David Sherwood <david.sherwood@arm.com>
11317 * coretypes.h (fixed_size_mode): Declare.
11318 (fixed_size_mode_pod): New typedef.
11319 * builtins.h (target_builtins::x_apply_args_mode)
11320 (target_builtins::x_apply_result_mode): Change type to
11321 fixed_size_mode_pod.
11322 * builtins.c (apply_args_size, apply_result_size, result_vector)
11323 (expand_builtin_apply_args_1, expand_builtin_apply)
11324 (expand_builtin_return): Update accordingly.
11326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11328 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11329 * cselib.c (cselib_hash_rtx): Likewise.
11330 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11331 CONST_VECTOR encoding.
11333 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11334 Jeff Law <law@redhat.com>
11337 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11338 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11339 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11340 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11343 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11344 explicitly probe *sp in a noreturn function if there were any callee
11345 register saves or frame pointer is needed.
11347 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11350 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11351 BLKmode for ternary, binary or unary expressions.
11354 * var-tracking.c (delete_vta_debug_insn): New inline function.
11355 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11356 insns from get_insns () to NULL instead of each bb separately.
11357 Use delete_vta_debug_insn. No longer static.
11358 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11359 delete_vta_debug_insns callers.
11360 * rtl.h (delete_vta_debug_insns): Declare.
11361 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11362 instead of variable_tracking_main.
11364 2018-01-03 Martin Sebor <msebor@redhat.com>
11366 PR tree-optimization/83603
11367 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11368 arguments past the endof the argument list in functions declared
11369 without a prototype.
11370 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11371 Avoid checking when arguments are null.
11373 2018-01-03 Martin Sebor <msebor@redhat.com>
11376 * doc/extend.texi (attribute const): Fix a typo.
11377 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11378 issuing -Wsuggest-attribute for void functions.
11380 2018-01-03 Martin Sebor <msebor@redhat.com>
11382 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11383 offset_int::from instead of wide_int::to_shwi.
11384 (maybe_diag_overlap): Remove assertion.
11385 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11386 * gimple-ssa-sprintf.c (format_directive): Same.
11387 (parse_directive): Same.
11388 (sprintf_dom_walker::compute_format_length): Same.
11389 (try_substitute_return_value): Same.
11391 2018-01-03 Jeff Law <law@redhat.com>
11393 PR middle-end/83654
11394 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11395 non-constant residual for zero at runtime and avoid probing in
11396 that case. Reorganize code for trailing problem to mirror handling
11399 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11401 PR tree-optimization/83501
11402 * tree-ssa-strlen.c (get_string_cst): New.
11403 (handle_char_store): Call get_string_cst.
11405 2018-01-03 Martin Liska <mliska@suse.cz>
11407 PR tree-optimization/83593
11408 * tree-ssa-strlen.c: Include tree-cfg.h.
11409 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11410 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11411 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11413 (strlen_dom_walker::before_dom_children): Call
11414 gimple_purge_dead_eh_edges. Dump tranformation with details
11416 (strlen_dom_walker::before_dom_children): Update call by adding
11417 new argument cleanup_eh.
11418 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11420 2018-01-03 Martin Liska <mliska@suse.cz>
11423 * cif-code.def (VARIADIC_THUNK): New enum value.
11424 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11427 2018-01-03 Jan Beulich <jbeulich@suse.com>
11429 * sse.md (mov<mode>_internal): Tighten condition for when to use
11430 vmovdqu<ssescalarsize> for TI and OI modes.
11432 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11434 Update copyright years.
11436 2018-01-03 Martin Liska <mliska@suse.cz>
11439 * ipa-visibility.c (function_and_variable_visibility): Skip
11440 functions with noipa attribure.
11442 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11444 * gcc.c (process_command): Update copyright notice dates.
11445 * gcov-dump.c (print_version): Ditto.
11446 * gcov.c (print_version): Ditto.
11447 * gcov-tool.c (print_version): Ditto.
11448 * gengtype.c (create_file): Ditto.
11449 * doc/cpp.texi: Bump @copying's copyright year.
11450 * doc/cppinternals.texi: Ditto.
11451 * doc/gcc.texi: Ditto.
11452 * doc/gccint.texi: Ditto.
11453 * doc/gcov.texi: Ditto.
11454 * doc/install.texi: Ditto.
11455 * doc/invoke.texi: Ditto.
11457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11459 * vector-builder.h (vector_builder::m_full_nelts): Change from
11460 unsigned int to poly_uint64.
11461 (vector_builder::full_nelts): Update prototype accordingly.
11462 (vector_builder::new_vector): Likewise.
11463 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11464 (vector_builder::operator ==): Likewise.
11465 (vector_builder::finalize): Likewise.
11466 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11467 Take the number of elements as a poly_uint64 rather than an
11469 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11470 from unsigned int to poly_uint64.
11471 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11472 (vec_perm_indices::new_vector): Likewise.
11473 (vec_perm_indices::length): Likewise.
11474 (vec_perm_indices::nelts_per_input): Likewise.
11475 (vec_perm_indices::input_nelts): Likewise.
11476 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11477 number of elements per input as a poly_uint64 rather than an
11478 unsigned int. Use the original encoding for variable-length
11479 vectors, rather than clamping each individual element.
11480 For the second and subsequent elements in each pattern,
11481 clamp the step and base before clamping their sum.
11482 (vec_perm_indices::series_p): Handle polynomial element counts.
11483 (vec_perm_indices::all_in_range_p): Likewise.
11484 (vec_perm_indices_to_tree): Likewise.
11485 (vec_perm_indices_to_rtx): Likewise.
11486 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11487 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11488 (tree_vector_builder::new_binary_operation): Handle polynomial
11489 element counts. Return false if we need to know the number
11490 of elements at compile time.
11491 * fold-const.c (fold_vec_perm): Punt if the number of elements
11492 isn't known at compile time.
11494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11496 * vec-perm-indices.h (vec_perm_builder): Change element type
11497 from HOST_WIDE_INT to poly_int64.
11498 (vec_perm_indices::element_type): Update accordingly.
11499 (vec_perm_indices::clamp): Handle polynomial element_types.
11500 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11501 (vec_perm_indices::all_in_range_p): Likewise.
11502 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11504 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11505 polynomial vec_perm_indices element types.
11506 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11507 * fold-const.c (fold_vec_perm): Likewise.
11508 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11509 * tree-vect-generic.c (lower_vec_perm): Likewise.
11510 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11511 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11512 element type to HOST_WIDE_INT.
11514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11515 Alan Hayward <alan.hayward@arm.com>
11516 David Sherwood <david.sherwood@arm.com>
11518 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11519 rather than an int. Use plus_constant.
11520 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11521 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11524 Alan Hayward <alan.hayward@arm.com>
11525 David Sherwood <david.sherwood@arm.com>
11527 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11528 a HOST_WIDE_INT to a poly_int64.
11530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11531 Alan Hayward <alan.hayward@arm.com>
11532 David Sherwood <david.sherwood@arm.com>
11534 * calls.c (load_register_parameters): Cope with polynomial
11535 mode sizes. Require a constant size for BLKmode parameters
11536 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11537 forces a parameter to be padded at the lsb end in order to
11538 fill a complete number of words, require the parameter size
11539 to be ordered wrt UNITS_PER_WORD.
11541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11542 Alan Hayward <alan.hayward@arm.com>
11543 David Sherwood <david.sherwood@arm.com>
11545 * reload1.c (spill_stack_slot_width): Change element type
11546 from unsigned int to poly_uint64_pod.
11547 (alter_reg): Treat mode sizes as polynomial.
11549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11550 Alan Hayward <alan.hayward@arm.com>
11551 David Sherwood <david.sherwood@arm.com>
11553 * reload.c (complex_word_subreg_p): New function.
11554 (reload_inner_reg_of_subreg, push_reload): Use it.
11556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11557 Alan Hayward <alan.hayward@arm.com>
11558 David Sherwood <david.sherwood@arm.com>
11560 * lra-constraints.c (process_alt_operands): Reject matched
11561 operands whose sizes aren't ordered.
11562 (match_reload): Refer to this check here.
11564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11565 Alan Hayward <alan.hayward@arm.com>
11566 David Sherwood <david.sherwood@arm.com>
11568 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11569 that the mode size is in the set {1, 2, 4, 8, 16}.
11571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11572 Alan Hayward <alan.hayward@arm.com>
11573 David Sherwood <david.sherwood@arm.com>
11575 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11576 Use plus_constant instead of gen_rtx_PLUS.
11578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11579 Alan Hayward <alan.hayward@arm.com>
11580 David Sherwood <david.sherwood@arm.com>
11582 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11583 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11584 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11585 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11586 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11587 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11588 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11589 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11590 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11591 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11593 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11594 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11595 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11596 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11597 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11598 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11599 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11600 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11601 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11602 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11604 * expr.c (emit_move_resolve_push): Treat the input and result
11605 of PUSH_ROUNDING as a poly_int64.
11606 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11607 (emit_push_insn): Likewise.
11608 * lra-eliminations.c (mark_not_eliminable): Likewise.
11609 * recog.c (push_operand): Likewise.
11610 * reload1.c (elimination_effects): Likewise.
11611 * rtlanal.c (nonzero_bits1): Likewise.
11612 * calls.c (store_one_arg): Likewise. Require the padding to be
11613 known at compile time.
11615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11616 Alan Hayward <alan.hayward@arm.com>
11617 David Sherwood <david.sherwood@arm.com>
11619 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11620 Use plus_constant instead of gen_rtx_PLUS.
11622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11623 Alan Hayward <alan.hayward@arm.com>
11624 David Sherwood <david.sherwood@arm.com>
11626 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11627 rather than an int.
11629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11630 Alan Hayward <alan.hayward@arm.com>
11631 David Sherwood <david.sherwood@arm.com>
11633 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11634 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11635 via stack temporaries. Treat the mode size as polynomial too.
11637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11638 Alan Hayward <alan.hayward@arm.com>
11639 David Sherwood <david.sherwood@arm.com>
11641 * expr.c (expand_expr_real_2): When handling conversions involving
11642 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11643 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11644 as a poly_uint64 too.
11646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11647 Alan Hayward <alan.hayward@arm.com>
11648 David Sherwood <david.sherwood@arm.com>
11650 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11653 Alan Hayward <alan.hayward@arm.com>
11654 David Sherwood <david.sherwood@arm.com>
11656 * combine.c (can_change_dest_mode): Handle polynomial
11657 REGMODE_NATURAL_SIZE.
11658 * expmed.c (store_bit_field_1): Likewise.
11659 * expr.c (store_constructor): Likewise.
11660 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11661 and polynomial REGMODE_NATURAL_SIZE.
11662 (gen_lowpart_common): Likewise.
11663 * reginfo.c (record_subregs_of_mode): Likewise.
11664 * rtlanal.c (read_modify_subreg_p): Likewise.
11666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11667 Alan Hayward <alan.hayward@arm.com>
11668 David Sherwood <david.sherwood@arm.com>
11670 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11671 numbers of elements.
11673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11674 Alan Hayward <alan.hayward@arm.com>
11675 David Sherwood <david.sherwood@arm.com>
11677 * match.pd: Cope with polynomial numbers of vector elements.
11679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11680 Alan Hayward <alan.hayward@arm.com>
11681 David Sherwood <david.sherwood@arm.com>
11683 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11684 in a POINTER_PLUS_EXPR.
11686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11687 Alan Hayward <alan.hayward@arm.com>
11688 David Sherwood <david.sherwood@arm.com>
11690 * omp-simd-clone.c (simd_clone_subparts): New function.
11691 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11692 (ipa_simd_modify_function_body): Likewise.
11694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11695 Alan Hayward <alan.hayward@arm.com>
11696 David Sherwood <david.sherwood@arm.com>
11698 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11699 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11700 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11701 (expand_vector_condition, vector_element): Likewise.
11702 (subparts_gt): New function.
11703 (get_compute_type): Use subparts_gt.
11704 (count_type_subparts): Delete.
11705 (expand_vector_operations_1): Use subparts_gt instead of
11706 count_type_subparts.
11708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11709 Alan Hayward <alan.hayward@arm.com>
11710 David Sherwood <david.sherwood@arm.com>
11712 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11713 (vect_compile_time_alias): ...this new function. Do the calculation
11714 on poly_ints rather than trees.
11715 (vect_prune_runtime_alias_test_list): Update call accordingly.
11717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11718 Alan Hayward <alan.hayward@arm.com>
11719 David Sherwood <david.sherwood@arm.com>
11721 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11723 (vect_schedule_slp_instance): Likewise.
11725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11726 Alan Hayward <alan.hayward@arm.com>
11727 David Sherwood <david.sherwood@arm.com>
11729 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11730 constant and extern definitions for variable-length vectors.
11731 (vect_get_constant_vectors): Note that the number of units
11732 is known to be constant.
11734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11735 Alan Hayward <alan.hayward@arm.com>
11736 David Sherwood <david.sherwood@arm.com>
11738 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11739 of units as polynomial. Choose between WIDE and NARROW based
11742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11743 Alan Hayward <alan.hayward@arm.com>
11744 David Sherwood <david.sherwood@arm.com>
11746 * tree-vect-stmts.c (simd_clone_subparts): New function.
11747 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11750 Alan Hayward <alan.hayward@arm.com>
11751 David Sherwood <david.sherwood@arm.com>
11753 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11754 vectors as polynomial. Use build_index_vector for
11755 IFN_GOMP_SIMD_LANE.
11757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11758 Alan Hayward <alan.hayward@arm.com>
11759 David Sherwood <david.sherwood@arm.com>
11761 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11762 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11763 for variable-length vectors.
11764 (vectorizable_mask_load_store): Treat the number of units as
11765 polynomial, asserting that it is constant if the condition has
11766 already been enforced.
11767 (vectorizable_store, vectorizable_load): 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-vect-loop.c (vectorizable_live_operation): Treat the number
11774 of units as polynomial. Punt if we can't tell at compile time
11775 which vector contains the final result.
11777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11778 Alan Hayward <alan.hayward@arm.com>
11779 David Sherwood <david.sherwood@arm.com>
11781 * tree-vect-loop.c (vectorizable_induction): Treat the number
11782 of units as polynomial. Punt on SLP inductions. Use an integer
11783 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11784 cast of such a series for variable-length floating-point
11787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11788 Alan Hayward <alan.hayward@arm.com>
11789 David Sherwood <david.sherwood@arm.com>
11791 * tree.h (build_index_vector): Declare.
11792 * tree.c (build_index_vector): New function.
11793 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11794 of units as polynomial, forcibly converting it to a constant if
11795 vectorizable_reduction has already enforced the condition.
11796 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11797 to create a {1,2,3,...} vector.
11798 (vectorizable_reduction): Treat the number of units as polynomial.
11799 Choose vectype_in based on the largest scalar element size rather
11800 than the smallest number of units. Enforce the restrictions
11803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11804 Alan Hayward <alan.hayward@arm.com>
11805 David Sherwood <david.sherwood@arm.com>
11807 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11808 number of units as polynomial.
11810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11811 Alan Hayward <alan.hayward@arm.com>
11812 David Sherwood <david.sherwood@arm.com>
11814 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11815 * target.def (autovectorize_vector_sizes): Return the vector sizes
11816 by pointer, using vector_sizes rather than a bitmask.
11817 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11818 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11819 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11821 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11822 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11823 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11824 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11825 * omp-general.c (omp_max_vf): Likewise.
11826 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11827 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11828 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11829 * tree-vect-slp.c (vect_slp_bb): Likewise.
11830 * doc/tm.texi: Regenerate.
11831 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11833 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11834 the vector size as a poly_uint64 rather than an unsigned int.
11835 (current_vector_size): Change from an unsigned int to a poly_uint64.
11836 (get_vectype_for_scalar_type): Update accordingly.
11837 * tree.h (build_truth_vector_type): Take the size and number of
11838 units as a poly_uint64 rather than an unsigned int.
11839 (build_vector_type): Add a temporary overload that takes
11840 the number of units as a poly_uint64 rather than an unsigned int.
11841 * tree.c (make_vector_type): Likewise.
11842 (build_truth_vector_type): Take the number of units as a poly_uint64
11843 rather than an unsigned int.
11845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11846 Alan Hayward <alan.hayward@arm.com>
11847 David Sherwood <david.sherwood@arm.com>
11849 * target.def (get_mask_mode): Take the number of units and length
11850 as poly_uint64s rather than unsigned ints.
11851 * targhooks.h (default_get_mask_mode): Update accordingly.
11852 * targhooks.c (default_get_mask_mode): Likewise.
11853 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11854 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11861 * omp-general.c (omp_max_vf): Likewise.
11862 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11863 (expand_omp_simd): Handle polynomial safelen.
11864 * omp-low.c (omplow_simd_context): Add a default constructor.
11865 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11866 (lower_rec_simd_input_clauses): Update accordingly.
11867 (lower_rec_input_clauses): Likewise.
11869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11870 Alan Hayward <alan.hayward@arm.com>
11871 David Sherwood <david.sherwood@arm.com>
11873 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11874 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11875 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11876 (vect_analyze_slp_cost): Likewise.
11877 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11878 (vect_model_load_cost): Likewise.
11880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11881 Alan Hayward <alan.hayward@arm.com>
11882 David Sherwood <david.sherwood@arm.com>
11884 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11885 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11886 from an unsigned int * to a poly_uint64_pod *.
11887 (calculate_unrolling_factor): New function.
11888 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11891 Alan Hayward <alan.hayward@arm.com>
11892 David Sherwood <david.sherwood@arm.com>
11894 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11895 from an unsigned int to a poly_uint64.
11896 (_loop_vec_info::slp_unrolling_factor): Likewise.
11897 (_loop_vec_info::vectorization_factor): Change from an int
11899 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11900 (vect_get_num_vectors): New function.
11901 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11902 (vect_get_num_copies): Use vect_get_num_vectors.
11903 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11904 to an unsigned int *.
11905 (vect_analyze_data_refs): Change min_vf from an int * to a
11907 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11908 than an unsigned HOST_WIDE_INT.
11909 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11910 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11911 to an unsigned int *.
11912 (vect_analyze_data_ref_dependences): Likewise.
11913 (vect_compute_data_ref_alignment): Handle polynomial vf.
11914 (vect_enhance_data_refs_alignment): Likewise.
11915 (vect_prune_runtime_alias_test_list): Likewise.
11916 (vect_shift_permute_load_chain): Likewise.
11917 (vect_supportable_dr_alignment): Likewise.
11918 (dependence_distance_ge_vf): Take the vectorization factor as a
11919 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11920 (vect_analyze_data_refs): Change min_vf from an int * to a
11922 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11923 vfm1 as a poly_uint64 rather than an int. Make the same change
11924 for the returned bound_scalar.
11925 (vect_gen_vector_loop_niters): Handle polynomial vf.
11926 (vect_do_peeling): Likewise. Update call to
11927 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11928 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11930 * tree-vect-loop.c (vect_determine_vectorization_factor)
11931 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11932 (vect_get_known_peeling_cost): Likewise.
11933 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11934 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11935 (vect_transform_loop): Likewise. Use the lowest possible VF when
11936 updating the upper bounds of the loop.
11937 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11938 rather than an int.
11939 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11940 polynomial unroll factors.
11941 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11942 (vect_make_slp_decision): Likewise.
11943 (vect_supported_load_permutation_p): Likewise, and polynomial
11945 (vect_analyze_slp_cost): Handle polynomial vf.
11946 (vect_slp_analyze_node_operations): Likewise.
11947 (vect_slp_analyze_bb_1): Likewise.
11948 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11949 than an unsigned HOST_WIDE_INT.
11950 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11951 (vectorizable_load): Handle polynomial vf.
11952 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11954 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11957 Alan Hayward <alan.hayward@arm.com>
11958 David Sherwood <david.sherwood@arm.com>
11960 * match.pd: Handle bit operations involving three constants
11961 and try to fold one pair.
11963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11965 * tree-vect-loop-manip.c: Include gimple-fold.h.
11966 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11967 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11968 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11969 Add a path that uses a step of VF instead of 1, but disable it
11971 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11972 and niters_no_overflow parameters. Update calls to
11973 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11974 Create a new SSA name if the latter choses to use a ste other
11975 than zero, and return it via niters_vector_mult_vf_var.
11976 * tree-vect-loop.c (vect_transform_loop): Update calls to
11977 vect_do_peeling, vect_gen_vector_loop_niters and
11978 slpeel_make_loop_iterate_ntimes.
11979 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11980 (vect_gen_vector_loop_niters): Update declarations after above changes.
11982 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11984 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11985 128-bit round to integer instructions.
11986 (ceil<mode>2): Likewise.
11987 (btrunc<mode>2): Likewise.
11988 (round<mode>2): Likewise.
11990 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11992 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11993 unaligned VSX load/store on P8/P9.
11994 (expand_block_clear): Allow the use of unaligned VSX
11995 load/store on P8/P9.
11997 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11999 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12001 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12002 swap associated with both a load and a store.
12004 2018-01-02 Andrew Waterman <andrew@sifive.com>
12006 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12007 * config/riscv/riscv.md (clear_cache): Use it.
12009 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12011 * web.c: Remove out-of-date comment.
12013 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12015 * expr.c (fixup_args_size_notes): Check that any existing
12016 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12017 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12018 (emit_single_push_insn): ...here.
12020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12022 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12023 (const_vector_encoded_nelts): New function.
12024 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12025 (const_vector_int_elt, const_vector_elt): Declare.
12026 * emit-rtl.c (const_vector_int_elt_1): New function.
12027 (const_vector_elt): Likewise.
12028 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12029 of CONST_VECTOR_ELT.
12031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12033 * expr.c: Include rtx-vector-builder.h.
12034 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12035 directly on the tree encoding.
12036 (const_vector_from_tree): Likewise.
12037 * optabs.c: Include rtx-vector-builder.h.
12038 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12039 sequence of "u" values.
12040 * vec-perm-indices.c: Include rtx-vector-builder.h.
12041 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12042 directly on the vec_perm_indices encoding.
12044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12046 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12047 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12048 * rtx-vector-builder.h: New file.
12049 * rtx-vector-builder.c: Likewise.
12050 * rtl.h (rtx_def::u2): Add a const_vector field.
12051 (CONST_VECTOR_NPATTERNS): New macro.
12052 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12053 (CONST_VECTOR_DUPLICATE_P): Likewise.
12054 (CONST_VECTOR_STEPPED_P): Likewise.
12055 (CONST_VECTOR_ENCODED_ELT): Likewise.
12056 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12057 (unwrap_const_vec_duplicate): Likewise.
12058 (const_vec_series_p): Check for a non-duplicated vector encoding.
12059 Say that the function only returns true for integer vectors.
12060 * emit-rtl.c: Include rtx-vector-builder.h.
12061 (gen_const_vec_duplicate_1): Delete.
12062 (gen_const_vector): Call gen_const_vec_duplicate instead of
12063 gen_const_vec_duplicate_1.
12064 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12065 (gen_const_vec_duplicate): Use rtx_vector_builder.
12066 (gen_const_vec_series): Likewise.
12067 (gen_rtx_CONST_VECTOR): Likewise.
12068 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12069 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12070 Build a new vector rather than modifying a CONST_VECTOR in-place.
12071 (handle_special_swappables): Update call accordingly.
12072 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12073 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12074 Build a new vector rather than modifying a CONST_VECTOR in-place.
12075 (handle_special_swappables): Update call accordingly.
12077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12079 * simplify-rtx.c (simplify_const_binary_operation): Use
12080 CONST_VECTOR_ELT instead of XVECEXP.
12082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12084 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12085 the selector elements to be different from the data elements
12086 if the selector is a VECTOR_CST.
12087 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12088 ssizetype for the selector.
12090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12092 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12093 before testing each element individually.
12094 * tree-vect-generic.c (lower_vec_perm): Likewise.
12096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12098 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12099 * selftest-run-tests.c (selftest::run_tests): Call it.
12100 * vector-builder.h (vector_builder::operator ==): New function.
12101 (vector_builder::operator !=): Likewise.
12102 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12103 (vec_perm_indices::all_from_input_p): New function.
12104 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12105 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12106 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12107 instead of reading the VECTOR_CST directly. Detect whether both
12108 vector inputs are the same before constructing the vec_perm_indices,
12109 and update the number of inputs argument accordingly. Use the
12110 utility functions added above. Only construct sel2 if we need to.
12112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12114 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12115 the broadcast of the low byte.
12116 (expand_mult_highpart): Use an explicit encoding for the permutes.
12117 * optabs-query.c (can_mult_highpart_p): Likewise.
12118 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12119 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12120 (vectorizable_bswap): Likewise.
12121 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12122 explicit encoding for the power-of-2 permutes.
12123 (vect_permute_store_chain): Likewise.
12124 (vect_grouped_load_supported): Likewise.
12125 (vect_permute_load_chain): Likewise.
12127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12129 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12130 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12131 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12132 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12133 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12134 (vect_gen_perm_mask_any): Likewise.
12136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12138 * int-vector-builder.h: New file.
12139 * vec-perm-indices.h: Include int-vector-builder.h.
12140 (vec_perm_indices): Redefine as an int_vector_builder.
12141 (auto_vec_perm_indices): Delete.
12142 (vec_perm_builder): Redefine as a stand-alone class.
12143 (vec_perm_indices::vec_perm_indices): New function.
12144 (vec_perm_indices::clamp): Likewise.
12145 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12146 (vec_perm_indices::new_vector): New function.
12147 (vec_perm_indices::new_expanded_vector): Update for new
12148 vec_perm_indices class.
12149 (vec_perm_indices::rotate_inputs): New function.
12150 (vec_perm_indices::all_in_range_p): Operate directly on the
12151 encoded form, without computing elided elements.
12152 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12153 encoding. Update for new vec_perm_indices class.
12154 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12155 the given vec_perm_builder.
12156 (expand_vec_perm_var): Update vec_perm_builder constructor.
12157 (expand_mult_highpart): Use vec_perm_builder instead of
12158 auto_vec_perm_indices.
12159 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12160 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12161 or double series encoding as appropriate.
12162 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12163 vec_perm_indices instead of auto_vec_perm_indices.
12164 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12165 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12166 (vect_permute_store_chain): Likewise.
12167 (vect_grouped_load_supported): Likewise.
12168 (vect_permute_load_chain): Likewise.
12169 (vect_shift_permute_load_chain): Likewise.
12170 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12171 (vect_transform_slp_perm_load): Likewise.
12172 (vect_schedule_slp_instance): Likewise.
12173 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12174 (vectorizable_mask_load_store): Likewise.
12175 (vectorizable_bswap): Likewise.
12176 (vectorizable_store): Likewise.
12177 (vectorizable_load): Likewise.
12178 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12179 vec_perm_indices instead of auto_vec_perm_indices. Use
12180 tree_to_vec_perm_builder to read the vector from a tree.
12181 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12182 vec_perm_builder instead of a vec_perm_indices.
12183 (have_whole_vector_shift): Use vec_perm_builder and
12184 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12185 truncation to calc_vec_perm_mask_for_shift.
12186 (vect_create_epilog_for_reduction): Likewise.
12187 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12188 from auto_vec_perm_indices to vec_perm_indices.
12189 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12190 instead of changing individual elements.
12191 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12192 the vector in d.perm.
12193 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12194 from auto_vec_perm_indices to vec_perm_indices.
12195 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12196 instead of changing individual elements.
12197 (arm_vectorize_vec_perm_const): Use new_vector to install
12198 the vector in d.perm.
12199 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12200 Update vec_perm_builder constructor.
12201 (rs6000_expand_interleave): Likewise.
12202 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12203 (rs6000_expand_interleave): Likewise.
12205 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12207 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12208 to qimode could truncate the indices.
12209 * optabs.c (expand_vec_perm_var): Likewise.
12211 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12213 * Makefile.in (OBJS): Add vec-perm-indices.o.
12214 * vec-perm-indices.h: New file.
12215 * vec-perm-indices.c: Likewise.
12216 * target.h (vec_perm_indices): Replace with a forward class
12218 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12219 * optabs.h: Include vec-perm-indices.h.
12220 (expand_vec_perm): Delete.
12221 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12222 (expand_vec_perm_const): Declare.
12223 * target.def (vec_perm_const_ok): Replace with...
12224 (vec_perm_const): ...this new hook.
12225 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12226 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12227 * doc/tm.texi: Regenerate.
12228 * optabs.def (vec_perm_const): Delete.
12229 * doc/md.texi (vec_perm_const): Likewise.
12230 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12231 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12232 expand_vec_perm for constant permutation vectors. Assert that
12233 the mode of variable permutation vectors is the integer equivalent
12234 of the mode that is being permuted.
12235 * optabs-query.h (selector_fits_mode_p): Declare.
12236 * optabs-query.c: Include vec-perm-indices.h.
12237 (selector_fits_mode_p): New function.
12238 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12239 is defined, instead of checking whether the vec_perm_const_optab
12240 exists. Use targetm.vectorize.vec_perm_const instead of
12241 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12242 fit in the vector mode before using a variable permute.
12243 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12244 vec_perm_indices instead of an rtx.
12245 (expand_vec_perm): Replace with...
12246 (expand_vec_perm_const): ...this new function. Take the selector
12247 as a vec_perm_indices rather than an rtx. Also take the mode of
12248 the selector. Update call to shift_amt_for_vec_perm_mask.
12249 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12250 Use vec_perm_indices::new_expanded_vector to expand the original
12251 selector into bytes. Check whether the indices fit in the vector
12252 mode before using a variable permute.
12253 (expand_vec_perm_var): Make global.
12254 (expand_mult_highpart): Use expand_vec_perm_const.
12255 * fold-const.c: Includes vec-perm-indices.h.
12256 * tree-ssa-forwprop.c: Likewise.
12257 * tree-vect-data-refs.c: Likewise.
12258 * tree-vect-generic.c: Likewise.
12259 * tree-vect-loop.c: Likewise.
12260 * tree-vect-slp.c: Likewise.
12261 * tree-vect-stmts.c: Likewise.
12262 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12264 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12265 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12266 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12267 (aarch64_vectorize_vec_perm_const): ...this new function.
12268 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12269 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12270 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12271 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12272 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12273 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12274 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12276 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12277 check for NEON modes.
12278 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12279 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12280 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12281 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12283 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12284 the old VEC_PERM_CONST conditions.
12285 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12286 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12287 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12288 (ia64_vectorize_vec_perm_const_ok): Merge into...
12289 (ia64_vectorize_vec_perm_const): ...this new function.
12290 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12291 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12292 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12293 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12294 * config/mips/mips.c (mips_expand_vec_perm_const)
12295 (mips_vectorize_vec_perm_const_ok): Merge into...
12296 (mips_vectorize_vec_perm_const): ...this new function.
12297 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12298 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12299 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12300 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12301 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12302 (rs6000_expand_vec_perm_const): Delete.
12303 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12305 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12306 (altivec_expand_vec_perm_const_le): Take each operand individually.
12307 Operate on constant selectors rather than rtxes.
12308 (altivec_expand_vec_perm_const): Likewise. Update call to
12309 altivec_expand_vec_perm_const_le.
12310 (rs6000_expand_vec_perm_const): Delete.
12311 (rs6000_vectorize_vec_perm_const_ok): Delete.
12312 (rs6000_vectorize_vec_perm_const): New function.
12313 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12314 an element count and rtx array.
12315 (rs6000_expand_extract_even): Update call accordingly.
12316 (rs6000_expand_interleave): Likewise.
12317 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12318 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12319 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12320 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12321 (rs6000_expand_vec_perm_const): Delete.
12322 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12323 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12324 (altivec_expand_vec_perm_const_le): Take each operand individually.
12325 Operate on constant selectors rather than rtxes.
12326 (altivec_expand_vec_perm_const): Likewise. Update call to
12327 altivec_expand_vec_perm_const_le.
12328 (rs6000_expand_vec_perm_const): Delete.
12329 (rs6000_vectorize_vec_perm_const_ok): Delete.
12330 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12331 reference to the SPE evmerge intructions.
12332 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12333 an element count and rtx array.
12334 (rs6000_expand_extract_even): Update call accordingly.
12335 (rs6000_expand_interleave): Likewise.
12336 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12337 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12339 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12343 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12344 vector mode and that that mode matches the mode of the data
12346 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12347 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12348 directly using expand_vec_perm_1 when forcing selectors into
12350 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12354 * optabs-query.h (can_vec_perm_p): Delete.
12355 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12356 * optabs-query.c (can_vec_perm_p): Split into...
12357 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12358 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12359 particular selector is valid.
12360 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12361 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12362 (vect_grouped_load_supported): Likewise.
12363 (vect_shift_permute_load_chain): Likewise.
12364 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12365 (vect_transform_slp_perm_load): Likewise.
12366 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12367 (vectorizable_bswap): Likewise.
12368 (vect_gen_perm_mask_checked): Likewise.
12369 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12370 implementations of variable permutation vectors into account
12371 when deciding which selector to use.
12372 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12373 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12374 with a false third argument.
12375 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12376 to test whether the constant selector is valid and can_vec_perm_var_p
12377 to test whether a variable selector is valid.
12379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12381 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12382 * optabs-query.c (can_vec_perm_p): Likewise.
12383 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12384 instead of vec_perm_indices.
12385 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12386 (vect_gen_perm_mask_checked): Likewise,
12387 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12388 (vect_gen_perm_mask_checked): Likewise,
12390 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12392 * optabs-query.h (qimode_for_vec_perm): Declare.
12393 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12394 (qimode_for_vec_perm): ...this new function.
12395 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12397 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12399 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12400 does not have a conditional at the top.
12402 2018-01-02 Richard Biener <rguenther@suse.de>
12404 * ipa-inline.c (big_speedup_p): Fix expression.
12406 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12409 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12412 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12416 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12417 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12418 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12419 cond_taken_branch_cost 3->4.
12421 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12423 PR tree-optimization/83581
12424 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12425 TODO_cleanup_cfg if any changes have been made.
12427 PR middle-end/83608
12428 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12429 convert_modes if target mode has the right side, but different mode
12432 PR middle-end/83609
12433 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12434 last argument when extracting from CONCAT. If either from_real or
12435 from_imag is NULL, use expansion through memory. If result is not
12436 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12437 the parts directly to inner mode, if even that fails, use expansion
12440 PR middle-end/83623
12441 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12442 check for bswap in mode rather than HImode and use that in expand_unop
12445 Copyright (C) 2018 Free Software Foundation, Inc.
12447 Copying and distribution of this file, with or without modification,
12448 are permitted in any medium without royalty provided the copyright
12449 notice and this notice are preserved.