| blob | fcf80cdff21bd4af52706b79817e313f32d09052 |
1 2018-07-14 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/86274
4 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
5 precondition.
6 (format_floating): Correct handling of infinities and NaNs.
8 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
10 Backport from mainline
11 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
12 Sunil K Pandey <sunil.k.pandey@intel.com>
14 PR target/84413
15 * config/i386/i386.c (m_CORE_AVX512): New.
16 (m_CORE_AVX2): Likewise.
17 (m_CORE_ALL): Add m_CORE_AVX2.
18 * config/i386/x86-tune.def: Replace m_HASWELL with m_CORE_AVX2.
19 Replace m_SKYLAKE_AVX512 with m_CORE_AVX512 on avx256_optimal
20 and remove the rest of m_SKYLAKE_AVX512.
22 2018-07-13 Marek Polacek <polacek@redhat.com>
24 Backport from trunk
25 2018-07-03 Marek Polacek <polacek@redhat.com>
27 PR middle-end/86202
28 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
30 2018-07-12 Richard Biener <rguenther@suse.de>
32 PR target/84829
33 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
34 Remove -mieee-fp handling.
36 2018-07-12 Jakub Jelinek <jakub@redhat.com>
38 PR tree-optimization/86492
39 * gimple-ssa-store-merging.c
40 (imm_store_chain_info::coalesce_immediate_stores): Call
41 check_no_overlap even for the merge_overlapping case.
43 2018-07-11 Jakub Jelinek <jakub@redhat.com>
45 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
46 Use __mmask64 type instead of __mmask8 for __M argument.
47 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
48 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
49 __U argument.
50 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
51 __mmask16 for __M argument.
52 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
53 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
54 to __mmask16 instead of __mmask8.
55 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
56 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
57 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
58 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
59 instead of __mmask16 for __U argument.
60 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
61 __mmask16 instead of __mmask8 for __U argument.
62 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
63 __U argument.
64 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
65 __mmask16.
66 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
67 argument.
68 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
69 __U argument.
70 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
71 __mmask16.
72 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
73 of __mmask16.
74 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
75 __U argument.
76 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
77 __U argument.
78 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
79 __U argument.
80 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
81 __U argument.
82 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
83 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
84 return type as well as __M argument type and all casts from __mmask8
85 to __mmask32.
86 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
87 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
88 return type as well as __M argument type and all casts from __mmask8
89 to __mmask16.
90 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
91 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
92 return type as well as __M argument type and all casts from __mmask8
93 to __mmask32.
94 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
95 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
96 return type as well as __M argument type and all casts from __mmask8
97 to __mmask16.
98 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
99 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
100 __mmask16.
102 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
104 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
105 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
106 for __U argument.
108 2018-07-05 Martin Sebor <msebor@redhat.com>
110 PR c++/86400
111 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
112 than its domain to compute its the upper bound of a char array.
114 2018-07-03 Carl Love <cel@us.ibm.com>
116 Backport from trunk
117 2018-07-03 Carl Love <cel@us.ibm.com>
119 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
120 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
121 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
122 VSX_BUILTIN_DOUBLEL_V4SF.
124 2018-07-04 Luis Machado <luis.machado@linaro.org>
126 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
127 global.
128 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
130 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
132 Backport from trunk
133 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
135 PR target/86285
136 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
137 ieee128_float_type_node to long_double_type_node unless
138 TARGET_LONG_DOUBLE_128 is set.
140 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
142 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
143 -fno-omit-frame-pointer when not optimizing.
145 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
147 PR debug/86064
148 * dwarf2out.c (loc_list_has_views): Adjust comments.
149 (dw_loc_list): Split single cross-partition range with
150 nonzero locview.
152 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
154 Backport from mainline
155 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
157 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
158 behavior of vec_pack (vector double, vector double) to match
159 behavior of vec_float2 (vector double, vector double).
161 2018-06-29 Jakub Jelinek <jakub@redhat.com>
163 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
165 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
167 Backport from mainline
168 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
170 * doc/sourcebuild.texi (Endianness): New subsubsection.
172 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
174 Backport from mainline
175 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
177 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
178 behavior of vec_packsu (vector unsigned long long, vector unsigned
179 long long) to match behavior of vec_packs with same signature.
181 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
183 Backport from trunk
184 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
186 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
187 unaligned vsx for 16B memset.
189 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
191 Backport from trunk
192 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
194 PR target/86222
195 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
196 correctly.
198 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
200 Backport from trunk
201 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
203 PR rtl-optimization/85645
204 * regrename.c (build_def_use): Also kill the chains that include the
205 destination of a REG_CFA_REGISTER note.
207 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
209 Backport from trunk
210 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
212 PR rtl-optimization/85645
213 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
214 insn that has a REG_CFA_REGISTER note.
216 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
218 * config/s390/s390.h (enum processor_flags): Do not use
219 default tune parameter when -march was specified.
221 2018-06-26 Jakub Jelinek <jakub@redhat.com>
223 PR target/86314
224 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
225 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
226 operands.
228 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
230 Back port from trunk
231 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
233 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
234 added IEEE/IBM long double multilib support on PowerPC little
235 endian Linux systems.
236 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
237 (MULTILIB_DEFAULTS): Likewise.
238 * config/rs6000/rs6000.c (rs6000_option_override_internal):
239 Likewise.
240 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
241 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
242 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
244 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
246 Backport from trunk
247 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
249 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
251 2018-05-02 Jan Hubicka <jh@suse.cz>
253 Backport from mainline
254 2018-04-30 Jan Hubicka <jh@suse.cz>
256 * lto-wrapper.c (ltrans_priorities): New static var.
257 (cmp_priority): New.
258 (run_gcc): Read priorities and if doing parallel build order
259 the Makefile by them.
261 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
263 Backport from trunk
264 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
266 PR target/86197
267 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
268 ieee128 argument takes up only one (vector) register, not two (floating
269 point) registers.
271 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
273 PR tree-optimization/85989
274 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
275 variable.
276 (backprop::intersect_uses): Check it when deciding whether this
277 is a backedge reference.
278 (backprop::process_block): Add each phi to m_visited_phis
279 after visiting it, then clear it at the end.
281 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
283 Backport from mainline
285 2018-06-19 Tony Reix <tony.reix@atos.com>
286 Damien Bergamini <damien.bergamini@atos.com>
287 David Edelsohn <dje.gcc@gmail.com>
289 * collect2.c (static_obj): New variable.
290 (static_libs): New variable.
291 (is_in_list): Uncomment declaration.
292 (main): Track AIX libraries linked statically.
293 (is_in_list): Uncomment definition.
294 (scan_prog_file): Don't add AIX shared libraries initializer
295 to constructor list if linking statically.
297 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
299 Back port from trunk
300 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
302 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
303 conversion insn that shows up when pr85657-3.c is compiled using
304 IEEE 128-bit long double.
306 Back port from trunk
307 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
309 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
310 multiply and divide external functions from being created more
311 than once.
313 Back port from trunk
314 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
316 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
317 mode to check whether the mode is IBM extended.
319 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
321 PR target/85358
322 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
323 floating point modes, so that IFmode is numerically greater than
324 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
325 to declare the ordering. This prevents IFmode from being
326 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
327 machine. Include rs6000-modes.h to share the fractional values
328 between genmodes* and the rest of the compiler.
329 (IFmode): Likewise.
330 (KFmode): Likewise.
331 (TFmode): Likewise.
332 * config/rs6000/rs6000-modes.h: New file.
333 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
334 meaning of rs6000_long_double_size so that 126..128 selects an
335 appropriate 128-bit floating point type.
336 (rs6000_option_override_internal): Likewise.
337 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
338 (TARGET_LONG_DOUBLE_128): Change the meaning of
339 rs6000_long_double_size so that 126..128 selects an appropriate
340 128-bit floating point type.
341 (LONG_DOUBLE_TYPE_SIZE): Update comment.
342 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
343 source and destination to match the standard usage.
344 (truncifkf2): Likewise.
345 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
346 ISA 2.07 to use an explicit clobber, instead of passing in a
347 temporary.
348 (copysign<mode>3_soft): Likewise.
350 Back port from trunk
351 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
353 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
354 TARGET_ELF.
356 Back port from trunk
357 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
359 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
360 track if we pass or return IEEE 128-bit floating point.
361 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
362 C++ mangling that is compatible with GCC 8.1.
363 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
364 (init_cumulative_args): Note if we pass or return IEEE 128-bit
365 floating point types.
366 (rs6000_function_arg_advance_1): Likewise.
367 (rs6000_mangle_type): Optionally generate mangled names that match
368 what GCC 8.1 generated for IEEE 128-bit floating point types.
369 (rs6000_globalize_decl_name): If we have an external function that
370 passes or returns IEEE 128-bit types, generate a weak reference
371 from the mangled name used in GCC 8.1 to the current mangled
372 name.
373 (rs6000_init_builtins): Make __ibm128 use the long double type if
374 long double is IBM extended double. Make __float128 use the long
375 double type if long double is IEEE 128-bit.
377 PR target/85657
378 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
379 macro for __ibm128 built-in functions.
380 (PACK_IF): Add __ibm128 pack/unpack functions.
381 (UNPACK_IF): Likewise.
382 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
383 enable long double built-in functions if long double is IEEE
384 128-bit floating point.
385 (rs6000_invalid_builtin): Update long double built-in function
386 error message.
387 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
388 functions, adjust the built-in function to use the long double
389 built-in function if __ibm128 and long double are the same type.
390 * doc/extend.texi (PowerPC builtins): Update documention for
391 __builtin_{,un}pack_longdouble. Add documentation for
392 __builtin_{,un}pack_ibm128.
394 Back port from trunk
395 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
397 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
398 the 128-bit floating point types. Fix function comment.
400 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
402 Back port from trunk
403 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
405 PR target/85657
406 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
407 define __ibm128 as long double.
408 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
409 as a distinct type with IEEE 128-bit floating point is supported.
410 (init_float128_ieee): Fix up conversions between IFmode and IEEE
411 128-bit types to use the correct functions.
412 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
413 convert between 128-bit floating point types that have different
414 modes but the same representation, instead of using gen_lowpart to
415 makean alias.
416 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
417 KFmode.
418 (IFKF_reg): New attributes to give the register constraints for
419 IFmode and KFmode.
420 (extend<mode>tf2_internal): New insns to mark an explicit
421 conversion between 128-bit floating point types that have a
422 different mode but share the same representation.
424 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
426 PR target/85994
427 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
428 -x assembler-with-cpp.
430 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
432 Backported from mainline
433 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
435 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
436 multilibs for *-*-rtems*.
437 * config/riscv/t-rtems: New file.
439 2018-06-20 Jakub Jelinek <jakub@redhat.com>
441 Backported from mainline
442 2018-06-16 Jakub Jelinek <jakub@redhat.com>
444 PR rtl-optimization/86108
445 * bb-reorder.c (create_forwarder_block): Renamed to ...
446 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
447 jump from new landing pad to the second part.
448 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
449 Adjust callers.
451 2018-06-20 Jakub Jelinek <jakub@redhat.com>
453 PR debug/86194
454 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
455 be narrowed.
457 PR tree-optimization/86231
458 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
459 anti-range don't overwrite *vr0min before using it to compute *vr0max.
461 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
463 Backport from mainline
464 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
466 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
467 constant.
468 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
470 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
472 * gimplify.c (gimplify_init_constructor): Really never clear for an
473 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
475 2018-06-18 Martin Sebor <msebor@redhat.com>
477 PR c/82063
478 * calls.c (alloc_max_size): Correct a logic error/typo.
479 Treat excessive arguments as infinite. Warn for invalid arguments.
480 * doc/invoke.texi (-Walloc-size-larger-than): Update.
482 2018-06-16 Jakub Jelinek <jakub@redhat.com>
484 PR middle-end/86095
485 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
486 documented as preserved for backward compatibility only.
487 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
489 2018-06-15 Jakub Jelinek <jakub@redhat.com>
491 PR middle-end/85878
492 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
493 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
494 Only call store_expr for halves if the mode is the same.
496 PR middle-end/86123
497 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
498 Fix up comment formatting.
500 2018-06-14 Jakub Jelinek <jakub@redhat.com>
502 PR middle-end/86122
503 * match.pd ((A +- CST1) +- CST2): Punt if last resort
504 unsigned_type_for returns NULL.
506 PR target/85945
507 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
508 subregs of multi-word pseudos unless the float mode has word size.
510 2018-06-14 Richard Biener <rguenther@suse.de>
512 PR middle-end/86139
513 * tree-vect-generic.c (build_word_mode_vector_type): Remove
514 duplicate and harmful type_hash_canon.
515 * tree.c (type_hash_canon): Assert we didn't find ourselves.
517 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
519 Backport from mainline
520 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
522 * config/rtems.h (STDINT_LONG32): Define.
524 2018-05-05 Roland McGrath <mcgrathr@google.com>
526 PR other/77609
527 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
528 any section for which we don't know a specific type it should have,
529 regardless of name. Previously this was done only for the exact
530 names ".init_array", ".fini_array", and ".preinit_array".
531 (default_elf_asm_named_section): Add comment about
532 relationship with default_section_type_flags and SECTION_NOTYPE.
533 (get_section): Don't consider it a type conflict if one side has
534 SECTION_NOTYPE and the other doesn't, as long as neither has the
535 SECTION_BSS et al used in the default_section_type_flags logic.
537 2018-06-13 Martin Sebor <msebor@redhat.com>
539 PR tree-optimization/86114
540 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
541 of integer types.
542 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
544 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
546 PR target/86048
547 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
548 offsets for register save directives. Emit a second batch of save
549 directives, if need be, when the function accesses prior frames.
551 2018-06-11 Jason Merrill <jason@redhat.com>
553 PR c++/80485 - inline function non-zero address.
554 * symtab.c (nonzero_address): Check DECL_COMDAT.
556 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
558 Backport from mainline
559 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
561 PR target/85755
562 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
563 addresses.
565 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
567 Backport from trunk
568 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
570 PR target/85755
571 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
572 on the correct operand.
573 (*movdi_internal64): Ditto.
575 2018-06-08 Carl Love <cel@us.ibm.com>
577 Backport from mainline
578 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
579 vextract_fp_from_shortl): Add BE support.
581 2018-06-08 Carl Love <cel@us.ibm.com>
583 Backport from mainline
584 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
585 Calculate index using natureal element order.
586 (first_match_or_eos_index_<mode>):
587 Calculate index using natural element order.
588 (first_match_index_<mode>):
589 Calculate index using natural element order.
590 (first_match_or_eos_index_<mode>):
591 Calculate index using natural order.
592 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
593 for BE and LE modes.
594 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
595 P9V_BUILTIN_VCLZLSBB_V16QI.
596 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
597 specific.
599 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
601 Backport from mainline
602 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
604 PR target/81497
605 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
606 qualifier_void_pointer and qualifier_const_void_pointer.
607 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
608 (arm_init_builtins): Handle the above.
609 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
610 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
611 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
612 void intrinsics.
614 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
616 PR target/85684
617 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
618 Generate SImode target register for null target.
619 <case IX86_BUILTIN_XGETBV>: Ditto.
620 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
621 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
623 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
625 Backport from mainline
626 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
628 PR target/63177
629 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
630 Don't handle -mcpu=power8 if -mpower9-vector is also used.
632 2018-06-06 Richard Biener <rguenther@suse.de>
634 Backport from mainline
635 2018-06-04 Richard Biener <rguenther@suse.de>
637 PR tree-optimization/86038
638 * tracer.c (find_best_successor): Check probability for
639 being initialized, bail out if not.
641 2018-05-30 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/85964
644 * tracer.c (better_p): Drop initialized count check, we only
645 call the function with initialized counts now.
646 (find_best_successor): Do find a best edge if one
647 has uninitialized count.
648 (find_best_predecessor): Likewise. Do BB frequency check only
649 if count is initialized.
651 2018-05-28 Richard Biener <rguenther@suse.de>
653 PR tree-optimization/85934
654 * tree-vect-generic.c (expand_vector_operations_1): Hoist
655 vector boolean check before scalar optimization.
657 2018-05-22 Richard Biener <rguenther@suse.de>
659 PR tree-optimization/85863
660 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
661 comparisons when vectype is specified.
662 (vectorizable_condition): Do not specify vectype for
663 vect_is_simple_cond when SLP vectorizing.
665 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
667 Backport from mainline
668 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
670 * config/s390/s390-builtin-types.def: Add void function type.
671 * config/s390/s390-builtins.def: Use the function type for the
672 tbeginc builtin.
674 2018-06-04 Martin Sebor <msebor@redhat.com>
676 PR c/85623
677 * calls.c (maybe_warn_nonstring_arg): Use string length to set
678 or ajust the presumed bound on an operation to avoid unnecessary
679 warnings.
681 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
683 PR target/86003
684 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
685 of bits to ignore when comparing architectures.
687 2018-06-04 Jakub Jelinek <jakub@redhat.com>
689 PR c++/86025
690 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
692 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
694 PR tree-optimization/85712
695 Backport from mainline:
696 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
698 PR tree-optimization/85712
699 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
700 first_interp field.
701 (alloc_cand_and_find_basis): Initialize first_interp field.
702 (slsr_process_mul): Modify first_interp field.
703 (slsr_process_add): Likewise.
704 (slsr_process_cast): Modify first_interp field for each new
705 interpretation.
706 (slsr_process_copy): Likewise.
707 (dump_candidate): Dump first_interp field.
708 (replace_mult_candidate): Process all interpretations, not just
709 subsequent ones.
710 (replace_rhs_if_not_dup): Likewise.
711 (replace_one_candidate): Likewise.
713 Backport from mainline:
714 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
716 PR tree-optimization/85712
717 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
718 this candidate has already been replaced in-situ by a copy.
720 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
722 PR target/85950
723 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
724 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
725 sequence.
726 (sse4_1_round<mode>2): Use nonimmediate_operand
727 for operand 1 predicate.
729 2018-05-31 Jakub Jelinek <jakub@redhat.com>
731 PR target/85984
732 * bb-reorder.c (pass_partition_blocks::gate): Return false for
733 functions with naked attribute.
735 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
737 PR target/85829
738 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
739 and movx for Haswell.
741 2018-05-30 Jakub Jelinek <jakub@redhat.com>
743 Backported from mainline
744 2018-05-14 Jakub Jelinek <jakub@redhat.com>
746 PR target/85756
747 * config/i386/i386.md: Disallow non-commutative arithmetics in
748 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
749 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
750 in the peephole2 before it.
752 2018-05-08 Jakub Jelinek <jakub@redhat.com>
754 PR target/85683
755 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
756 after cmpelim optimization.
758 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
760 Backport from mainline
761 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
763 PR target/85900
764 PR target/85345
765 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
767 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
769 PR target/85900
770 PR target/85345
771 * varasm.c (assemble_alias): Check ifunc_resolver only on
772 FUNCTION_DECL.
774 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
776 PR target/85345
777 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
778 attribute.
779 (cgraph_node::create_alias): Likewise.
780 (cgraph_node::get_availability): Check ifunc_resolver instead
781 of looking up ifunc attribute.
782 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
783 * varasm.c (do_assemble_alias): Likewise.
784 (assemble_alias): Likewise.
785 (default_binds_local_p_3): Likewise.
786 * cgraph.h (cgraph_node): Add ifunc_resolver.
787 (cgraph_node::only_called_directly_or_aliased_p): Return false
788 for IFUNC resolver.
789 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
790 attribute.
791 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
792 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
793 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
794 instead of looking up ifunc attribute.
796 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
798 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
800 2018-05-25 Richard Biener <rguenther@suse.de>
802 PR c++/85912
803 * tree-dump.c (dequeue_and_dump): Remove access to removed
804 operand 2 of a SWITCH_EXPR.
806 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
808 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
809 Add {q} suffix to insn mnemonic.
811 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
813 PR target/85903
814 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
815 when memory input operand is handled.
817 2018-05-23 Bin Cheng <bin.cheng@arm.com>
819 Backport from mainline
820 2018-05-17 Bin Cheng <bin.cheng@arm.com>
821 Richard Biener <rguenther@suse.de>
823 PR tree-optimization/85793
824 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
825 for VMAT_ELEMENTWISE.
827 2018-05-23 Richard Biener <rguenther@suse.de>
829 PR middle-end/85874
830 * tree-data-ref.c (create_runtime_alias_checks): Defer
831 and ignore overflow warnings.
833 2018-05-23 Martin Liska <mliska@suse.cz>
835 Backport from mainline
836 2018-05-23 Yury Gribov <tetra2005@gmail.com>
838 PR tree-optimization/85822
839 * tree-vrp.c (is_masked_range_test): Fix handling of negative
840 constants.
842 2018-05-23 Martin Liska <mliska@suse.cz>
844 Backport from mainline
845 2018-05-11 Martin Liska <mliska@suse.cz>
847 PR sanitizer/85556
848 * doc/extend.texi: Document LLVM style format for no_sanitize
849 attribute.
851 2018-05-22 Martin Sebor <msebor@redhat.com>
853 PR middle-end/85643
854 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
856 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
858 PR tree-optimization/85814
859 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
860 a null return from get_strinfo when unsharing the next
861 strinfo in the chain.
863 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
865 Backport from mainline
866 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
867 Segher Boessenkool <segher@kernel.crashing.org>
869 PR target/85698
870 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
872 2018-05-18 Toon Moene <toon@moene.org>
874 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
875 directly after that of -floop-interchange. Indicate that both
876 options are enabled by default when specifying -O3.
878 2018-05-11 Martin Jambor <mjambor@suse.cz>
880 PR ipa/85655
881 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
882 single const.
884 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
886 PR target/85733
887 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
889 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
891 PR target/85606
892 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
893 equivalent.
894 (cortex-m0): Use armv6s-m isa.
895 (cortex-m0plus): Likewise.
896 (cortex-m1): Likewise.
897 (cortex-m0.small-multiply): Likewise.
898 (cortex-m0plus.small-multiply): Likewise.
899 (cortex-m1.small-multiply): Likewise.
901 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
903 PR rtl-optimization/85638
904 * bb-reorder.c: Include common/common-target.h.
905 (create_forwarder_block): New function extracted from...
906 (fix_up_crossing_landing_pad): ...here. Rename into...
907 (dw2_fix_up_crossing_landing_pad): ...this.
908 (sjlj_fix_up_crossing_landing_pad): New function.
909 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
910 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
911 from both partitions and exit the loop after one iteration.
913 2018-05-08 Richard Biener <rguenther@suse.de>
915 Backport from mainline
916 2018-05-04 Richard Biener <rguenther@suse.de>
918 PR middle-end/85588
919 * fold-const.c (negate_expr_p): Restrict negation of operand
920 zero of a division to when we know that can happen without
921 overflow.
922 (fold_negate_expr_1): Likewise.
924 2018-05-03 Richard Biener <rguenther@suse.de>
926 PR tree-optimization/85615
927 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
928 to loops not nested in BBs loop father to avoid creating multi-entry
929 loops.
931 2018-05-02 Richard Biener <rguenther@suse.de>
933 PR middle-end/85567
934 * gimplify.c (gimplify_save_expr): When in SSA form allow
935 SAVE_EXPRs to compute to SSA vars.
937 2018-05-02 Richard Biener <rguenther@suse.de>
939 PR tree-optimization/85597
940 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
941 do not use split vect_get_vec_defs call but call vect_get_slp_defs
942 directly.
944 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
946 PR target/85658
947 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
948 (check_arch): Likewise.
949 (check_fpu): Return the result rather than printing it.
950 (end arch): Fix operator precedence.
951 (end cpu): Likewise.
952 (END): Print the result from check_fpu.
954 2018-05-08 Richard Biener <rguenther@suse.de>
956 Backport from mainline
957 2018-05-04 Richard Biener <rguenther@suse.de>
959 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
960 * gimple-ssa-store-merging.c
961 (imm_store_chain_info::output_merged_store): Remove redundant create,
962 release split_store vector contents on failure.
963 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
964 scalar stmt vector on cache hit.
966 2018-05-04 Richard Biener <rguenther@suse.de>
968 PR middle-end/85627
969 * tree-complex.c (tree_lower_complex): Handle removed blocks.
971 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
973 Backport from mainline.
974 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
976 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
977 GET_MODE_MASK before any checking.
978 (nds32_can_use_bset_p): Likewise.
979 (nds32_can_use_btgl_p): Likewise.
981 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
983 Backport from mainline.
984 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
986 * config/nds32/nds32-doubleword.md: New define_split pattern for
987 illegal register number.
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.c (nds32_print_operand): Set op_value ealier.
996 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
998 Backport from mainline.
999 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1001 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1003 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1005 PR c++/85659
1006 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1007 the type is addressable. Don't force op into register if it has
1008 BLKmode.
1010 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1012 Backport from trunk.
1013 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1015 PR target/85512
1016 * config/aarch64/constraints.md (Usg): Limit to 31.
1017 (Usj): Limit to 63.
1019 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1021 * BASE-VER: Set to 8.1.1.
1023 2018-05-02 Release Manager
1025 * GCC 8.1.0 released.
1027 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1029 PR tree-optimization/85586
1030 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1031 exit early for statements in the same group if the accesses are
1032 not strided.
1034 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1036 PR web/85578
1037 * doc/install.texi2html: Replace _002d with - and _002a with * in
1038 generated html files using sed.
1040 2018-04-27 Martin Jambor <mjambor@suse.cz>
1042 PR ipa/85549
1043 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1044 the jump function allows for passing through aggregate values.
1046 2018-04-27 Jeff Law <law@redhat.com>
1048 * config.gcc: Mark tile* targets as deprecated/obsolete.
1050 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1052 PR tree-optimization/85529
1053 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1054 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1055 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1056 zero extension or masking of the MSB bit.
1057 (optimize_range_tests): Add FIRST_BB argument, pass it through
1058 to optimize_range_tests_var_bound.
1059 (maybe_optimize_range_tests, reassociate_bb): Adjust
1060 optimize_range_tests callers.
1062 2018-04-26 Richard Biener <rguenther@suse.de>
1063 Jakub Jelinek <jakub@redhat.com>
1065 * cgraph.h (symbol_table): Just declare debug method here.
1066 * symtab.c (symbol_table::debug): Define.
1068 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1070 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1072 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
1074 Backport from mainline
1075 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1077 PR target/85473
1078 * config/i386/i386.c (ix86_expand_builtin): Change memory
1079 operand to XI, extend p0 to Pmode.
1080 * config/i386/i386.md: Change unspec volatile and operand
1081 1 mode to XI, change operand 0 mode to P.
1083 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1085 * DEV-PHASE: Set to prerelease.
1087 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1089 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1090 in __abskf2 and __powikf2.
1092 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1094 PR target/85512
1095 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1096 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1097 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1098 Use the above on operand 2. Reindent.
1099 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1101 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1103 PR target/85485
1104 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1105 handle OPT_mcet.
1106 * config/i386/i386.opt (mcet): Removed.
1107 * doc/install.texi: Remove -mcet documentation.
1108 * doc/invoke.texi: Likewise.
1110 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1112 PR target/85485
1113 * doc/install.texi: Remove -mcet from bootstrap-cet.
1115 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1117 PR target/85511
1118 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1119 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1120 if TARGET_64BIT.
1122 PR target/85503
1123 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1124 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1125 containing a CONST_VECTOR.
1127 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1129 * doc/install.texi: Update newlib dependency for nvptx.
1131 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1133 PR target/85508
1134 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1135 instead of INTVAL when shifting x left.
1137 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1139 PR tree-optimization/85478
1140 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1141 vect_grouped_store_supported for single element vectors.
1143 2018-04-24 Richard Biener <rguenther@suse.de>
1145 PR target/85491
1146 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1147 load cost increase to the case of non-constant step.
1149 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1151 PR target/84828
1152 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1153 destination if any_malformed_asm.
1155 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1157 PR middle-end/85496
1158 * expr.c (store_field): In the bitfield case, if the value comes from
1159 a function call and is returned in registers by means of a PARALLEL,
1160 do not change the mode of the temporary unless BLKmode and VOIDmode.
1162 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1164 PR rtl-optimization/85423
1165 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1166 dependencies to debug insns when the previous insn is non-debug.
1168 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1170 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1171 enums into a single definition.
1172 (fls): Fix predicates and printing.
1173 (seti): Likewise.
1175 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1177 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1178 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1179 and short u6 immediate.
1180 (check_if_valid_sleep_operand): Remove.
1181 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1183 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1185 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1186 flag_always_save_lp condition.
1187 * config/nds32/nds32.opt (malways-save-lp): New option.
1189 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1191 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1192 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1193 * config/nds32/nds32.h
1194 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1195 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1197 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1199 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1200 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1202 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1203 Chung-Ju Wu <jasonwucj@gmail.com>
1205 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1206 Declare.
1207 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1208 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1210 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1212 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1214 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1216 * config/nds32/nds32-protos.h (nds32_data_alignment,
1217 nds32_local_alignment): Declare.
1218 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1219 nds32_local_alignment): New functions.
1220 (TARGET_CONSTANT_ALIGNMENT): Define.
1221 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1223 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1225 * config/nds32/nds32.c
1226 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1227 (TARGET_MODES_TIEABLE_P): Likewise.
1229 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1231 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1232 level Ofast and Og.
1234 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1235 Chung-Ju Wu <jasonwucj@gmail.com>
1237 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1238 for unaligned access.
1239 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1240 unaligned access.
1241 * config/nds32/nds32-intrinsic.md: Likewise.
1242 * config/nds32/nds32_intrinsic.h: Likewise.
1243 * config/nds32/nds32.h (nds32_builtins): Likewise.
1244 * config/nds32/nds32.opt (munaligned-access): New option.
1245 * config/nds32/nds32.c (nds32_asm_file_start): Display
1246 flag_unaligned_access status.
1248 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1250 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1251 -mno-relax is present.
1252 * config/riscv/linux.h (LINK_SPEC): Ditto.
1254 2018-04-20 Martin Sebor <msebor@redhat.com>
1256 PR c/85365
1257 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1258 for null pointers.
1259 (gimple_fold_builtin_stxcpy_chk): Same.
1260 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1262 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1264 PR target/85456
1265 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1266 __powikf2 when long double is IEEE 128-bit.
1268 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1270 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1271 step to make sure stack always aligned.
1273 2018-04-20 Carl Love <cel@us.ibm.com>
1275 PR target/83402
1276 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1277 size check for arg0.
1279 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1280 Tom de Vries <tom@codesourcery.com>
1282 PR target/85445
1283 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1284 Emit insns for calls too.
1285 (nvptx_find_par): Always look for worker-level predecessor insn.
1286 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1287 calls.
1288 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1289 (nvptx_process_pars): Propagate frames for calls.
1291 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1293 PR target/85469
1294 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1295 Removed.
1296 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1297 (ix86_handle_option): Don't handle OPT_mibt.
1298 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1299 __SHSTK__.
1300 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1301 has_ibt and ibt.
1302 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1303 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1304 (ix86_target_macros): Define __CET__ with flag_cf_protection
1305 for -fcf-protection.
1306 * config/i386/i386.c (isa2_opts): Remove -mibt.
1307 * config/i386/i386.h (TARGET_IBT): Removed.
1308 (TARGET_IBT_P): Likewise.
1309 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1310 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1311 * config/i386/i386.opt (mcet): Update help message.
1312 (mshstk): Likewise.
1313 (mibt): Removed.
1314 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1315 -mcet as an alias for -mshstk.
1317 2018-04-20 Richard Biener <rguenther@suse.de>
1319 PR middle-end/85475
1320 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1321 complexity by forcing a single use of the multiply operand.
1323 2018-04-20 Martin Jambor <mjambor@suse.cz>
1325 ipa/85449
1326 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1327 recursion dependency to only apply to non-clones.
1329 2018-04-20 Martin Jambor <mjambor@suse.cz>
1331 ipa/85447
1332 * ipa-cp.c (create_specialized_node): Check that clones of
1333 self-recursive edges exist during IPA-CP.
1335 2018-04-19 Toon Moene <toon@moene.org>
1337 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1338 by -O3.
1340 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1342 PR tree-optimization/85467
1343 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1344 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1345 VECTOR_CST element to type.
1347 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1349 PR target/85397
1350 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1351 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1352 (builtin_longjmp): Likewise.
1353 (save_stack_nonlocal): New pattern.
1354 (restore_stack_nonlocal): Likewise.
1356 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1358 PR target/85404
1359 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1360 Replace ASM_OUTPUT_LABEL with fprintf.
1362 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1364 PR target/85417
1365 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1366 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1367 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1368 define __IBT__ and __SHSTK__ for -fcf-protection.
1369 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1370 TARGET_IBT.
1371 (ix86_trampoline_init): Likewise.
1372 (x86_output_mi_thunk): Likewise.
1373 (ix86_notrack_prefixed_insn_p): Likewise.
1374 (ix86_option_override_internal): Don't disallow -fcf-protection.
1375 * config/i386/i386.md (rdssp<mode>): Also enable for
1376 -fcf-protection.
1377 (incssp<mode>): Likewise.
1378 (nop_endbr): Likewise.
1379 * config/i386/i386.opt (mcet): Change help message to built-in
1380 functions only.
1381 (mibt): Likewise.
1382 (mshstk): Likewise.
1383 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1384 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1385 enable CET built-in functions.
1387 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1389 * common/config/i386/i386-common.c
1390 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1391 OPTION_MASK_ISA_MOVDIRI_UNSET,
1392 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1393 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1394 * config.gcc (movdirintrin.h): New header.
1395 * config/i386/cpuid.h (bit_MOVDIRI,
1396 bit_MOVDIR64B): New bits.
1397 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1398 and -mmvodir64b.
1399 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1400 (VOID, PVOID, PCVOID)): New function types.
1401 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1402 __builtin_ia32_directstoreu_u64,
1403 __builtin_ia32_movdir64b): New builtins.
1404 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1405 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1406 and -mmovdiri.
1407 (ix86_valid_target_attribute_inner_p): Ditto.
1408 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1409 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1410 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1411 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1412 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1413 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1414 (movdiri<mode>, movdir64b_<mode>): New.
1415 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1416 * config/i386/immintrin.h: Include movdirintrin.h.
1417 * config/i386/movdirintrin.h: New file.
1418 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1420 2018-04-19 Richard Biener <rguenther@suse.de>
1422 PR middle-end/85455
1423 * cfg.c (clear_bb_flags): When loop state says we have
1424 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1426 2018-04-19 Richard Biener <rguenther@suse.de>
1428 PR tree-optimization/84737
1429 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1430 copying restrict info.
1431 (vect_setup_realignment): Use it.
1432 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1433 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1434 the first DR to all generated stores.
1435 (vectorizable_load): Likewise for loads.
1437 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1439 PR tree-optimization/85446
1440 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1441 the integral and pointer types to have the same precision.
1443 * doc/install.texi: Document --disable-cet being the default and
1444 --enable-cet=auto.
1446 2018-04-18 Martin Liska <mliska@suse.cz>
1448 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1449 style.
1451 2018-04-18 Martin Liska <mliska@suse.cz>
1453 Revert
1454 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1456 PR ipa/83983
1457 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1458 arguments if they are comparable.
1460 2018-04-18 Martin Liska <mliska@suse.cz>
1462 Revert
1463 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1465 PR lto/84805
1466 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1467 incomplete types.
1469 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1471 PR target/85388
1472 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1473 ENDBR after calling __morestack.
1475 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1477 PR jit/85384
1478 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1479 by using gcc_base_ver to generate a gcc_driver_version, and use
1480 it when generating GCC_DRIVER_NAME.
1481 * configure: Regenerate.
1483 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1485 PR target/81084
1486 * config.gcc: Obsolete powerpc*-*-*spe*.
1488 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1490 PR debug/84637
1491 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1492 (stabstr_D): Change type of unum from unsigned int to
1493 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1494 type.
1496 2018-04-17 Jim Wilson <jimw@sifive.com>
1498 PR 84856
1499 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1500 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1501 Set arg_pointer_offset after using pretend_args_size.
1503 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1505 PR rtl-optimization/85431
1506 * dse.c (record_store): Ignore zero width stores.
1508 PR sanitizer/85230
1509 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1510 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1511 __builtin_stack_restore rather than after it.
1512 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1513 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1514 argument instead of virtual_dynamic_stack_rtx.
1516 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1518 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1519 New prototype.
1520 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1521 Add note to error message to explain internal mapping of overloaded
1522 built-in function name to non-overloaded built-in function name.
1523 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1524 function.
1526 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1528 PR target/85424
1529 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1530 where the inputs overlap with the output.
1532 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1534 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1535 (=v, v) alternative and explicit "memory" attribute.
1536 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1537 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1538 attributes.
1539 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1540 "sselog1" type instead of "sselog".
1541 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1542 "sselog". Remove explicit "memory" attribute.
1543 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1544 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1545 attributes.
1546 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1547 "sselog1" type instead of "sselog". Remove explicit "memory"
1548 attribute.
1549 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1550 use "sselog1" type instead of "sselog". Remove explicit "memory"
1551 attribute.
1552 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1553 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1554 attributes.
1555 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1556 "sselog1" type instead of "sselog". Remove explicit "memory"
1557 attribute.
1558 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1559 use "sselog1" type instead of "sselog". Remove explicit "memory"
1560 attribute.
1562 PR target/85430
1563 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1565 PR middle-end/85414
1566 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1567 on a SUBREG.
1569 2018-04-17 Martin Jambor <mjambor@suse.cz>
1571 PR ipa/85421
1572 * ipa-cp.c (create_specialized_node): Call
1573 expand_all_artificial_thunks if necessary.
1575 2018-04-17 Martin Liska <mliska@suse.cz>
1577 PR lto/85405
1578 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1579 in message, remote space in between '_G' and '('.
1581 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1583 PR target/85281
1584 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1585 avx512f_vmcmp<mode>3<round_saeonly_name>,
1586 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1587 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1588 avx512f_rndscale<mode><round_saeonly_name>,
1589 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1590 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1591 Use %<iptr>2 instead of %2 for -masm=intel.
1592 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1593 avx512f_vcvttss2usi<round_saeonly_name>,
1594 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1595 -masm=intel.
1596 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1597 avx512f_vcvttsd2usi<round_saeonly_name>,
1598 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1599 Use %q1 instead of %1 for -masm=intel.
1600 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1601 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1602 of %3 for -masm=intel.
1603 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1604 -masm=intel.
1605 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1606 -masm=intel.
1607 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1608 -masm=intel.
1609 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1610 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1611 %g1.
1612 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1613 -masm=intel.
1614 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1615 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1616 %g1 and one with %0 and %1.
1617 (avx512er_vmrcp28<mode><round_saeonly_name>,
1618 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1619 %1 for -masm=intel.
1620 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1621 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1622 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1623 of %0 and %{%4%} for -masm=intel.
1624 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1625 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1626 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1627 order of %0 and %{%5%}%{z%} for -masm=intel.
1629 2018-04-17 Jan Hubicka <jh@suse.cz>
1631 PR lto/85405
1632 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1634 2018-04-17 Martin Liska <mliska@suse.cz>
1636 PR ipa/85329
1637 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1638 for target_clone error message. Make default implementation
1639 clone to be a local declaration.
1640 (separate_attrs): Add new argument and check for an empty
1641 string.
1642 (expand_target_clones): Handle it.
1643 (ipa_target_clone): Make redirection just for target_clones
1644 functions.
1646 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1647 Tom de Vries <tom@codesourcery.com>
1649 PR middle-end/84955
1650 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1651 tiled basic blocks without omp continue statements.
1653 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1655 PR target/83660
1656 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1657 vec_extract expression as having side effects to make sure it gets
1658 a cleanup point.
1660 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1662 PR target/85403
1663 * config/i386/i386.c (get_builtin_code_for_version): Check
1664 error_mark_node.
1666 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1668 PR target/84331
1669 * gcc/config.gcc: Support "skylake".
1670 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1671 PROCESSOR_SKYLAKE.
1672 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1673 (processor_target_table): Add "skylake".
1674 (ix86_option_override_internal): Add "skylake".
1675 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1676 PROCESSOR_CANNONLAKE.
1677 (get_builtin_code_for_version): Fix priority for
1678 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1679 PROCESSOR_SKYLAKE-AVX512.
1680 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1681 (processor_type): Add PROCESSOR_SKYLAKE.
1683 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1684 Jason Merrill <jason@redhat.com>
1686 PR c++/85112
1687 * convert.c (convert_to_integer_1): Use direct recursion for
1688 enumeral types and types with a precision less than the number
1689 of bits in their mode.
1691 2018-04-16 Julia Koval <julia.koval@intel.com>
1693 PR target/84413
1694 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1695 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1697 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1699 PR target/85293
1700 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1701 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1702 and -mno-direct-move.
1704 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1706 PR target/83402
1707 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1708 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1709 Ensure negative shifts result in {0}.
1711 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1713 PR rtl-optimization/79916
1714 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1715 regs (if any) to define how to gnerate SD moves when LRA is in
1716 progress.
1718 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1720 PR rtl-optimization/85393
1721 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1722 * except.c (expand_dw2_landing_pad_for_region): Make static.
1723 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1724 a label and unconditional jump to old_bb, rather than
1725 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1726 basic block.
1728 PR rtl-optimization/85376
1729 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1730 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1731 instead of a specific value.
1733 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1734 Bin Cheng <bin.cheng@arm.com>
1736 PR tree-optimization/82965
1737 PR tree-optimization/83991
1738 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1739 by_profile_only parameter.
1740 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1741 information if the loop was predicted to iterate too many times.
1742 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1744 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1746 PR lto/71991
1747 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1748 always inline.
1750 2018-04-13 Martin Liska <mliska@suse.cz>
1751 Jakub Jelinek <jakub@redhat.com>
1753 PR middle-end/81657
1754 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1755 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1756 * builtins.c (expand_builtin_memory_copy_args): Use
1757 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1758 handle dest_addr == pc_rtx.
1760 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1762 PR target/85291
1763 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1764 asked to not generate direct moves.
1765 (fix_trunc<mode>si2_stfiwx): Similar.
1766 (fix_trunc<mode>si2_internal): Similar.
1768 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1770 PR debug/83157
1771 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1772 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1773 lookup if dest in some wider mode is known to be const0_rtx and
1774 if so, record permanent equivalence for it to be ZERO_EXTEND of
1775 the narrower mode destination.
1777 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1779 * lto-streamer-out.c (output_function): Revert 259346.
1780 * omp-expand.c (expand_oacc_for): Likewise.
1782 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1784 PR rtl-optimization/85354
1785 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1786 * sel-sched.c (sel_global_init): ... here.
1788 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1790 PR target/85238
1791 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1792 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1793 mode for PE-COFF targets.
1794 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1795 (i386_pe_asm_lto_end): Likewise.
1796 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1797 (TARGET_ASM_LTO_END): Likewise.
1798 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1799 (i386_pe_asm_lto_start): New function.
1800 (i386_pe_asm_lto_end): Likewise.
1802 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1803 Richard Biener <rguenther@suse.de>
1805 PR middle-end/84955
1806 * lto-streamer-out.c (output_function): Fix CFG loop state before
1807 streaming out.
1808 * omp-expand.c (expand_oacc_for): Handle calls to internal
1809 functions like regular functions.
1811 2018-04-12 Richard Biener <rguenther@suse.de>
1813 PR lto/85371
1814 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1815 for the early LTO debug to properly generate references to it
1816 during DIE emission. Do not re-use that for the skeleton for
1817 split-dwarf.
1818 (dwarf2out_early_finish): Likewise.
1820 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1822 PR target/85328
1823 * config/i386/sse.md
1824 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1825 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1826 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1827 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1828 and output is a reg, avoid creating invalid lowpart subreg, but
1829 instead split into a 512-bit move. Don't split if not AVX512VL,
1830 input is xmm16+ reg and output is a mem.
1831 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1832 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1833 xmm16+ reg and output is a mem.
1835 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1837 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1838 also for flag_dwarf2_cfi_asm.
1840 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1842 PR rtl-optimization/85342
1843 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1844 a bool scalar var inside of the loop instead. Don't try to update
1845 recog_data.operand after failed apply_change_group.
1847 2018-04-12 Tom de Vries <tom@codesourcery.com>
1849 PR target/85296
1850 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1851 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1852 array with flexible array member as array without given dimension.
1853 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1854 argument for undefined param to true.
1856 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1858 PR target/85321
1859 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1860 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1861 from PowerPC section.
1862 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1863 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1864 help text that is too long.
1865 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1866 help text that is too long.
1867 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1868 help text that is too long.
1870 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1872 * config/alpha/alpha.md (stack_probe_internal): Rename
1873 from "probe_stack". Update all callers.
1875 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1877 PR rtl-optimization/84566
1878 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1879 sched_macro_fuse_insns.
1881 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1883 PR target/84301
1884 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1885 (compute_block_dependences): ... from here.
1887 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1889 PR tree-optimization/85331
1890 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1891 from int to HOST_WIDE_INT.
1893 2018-04-11 Martin Jambor <mjambor@suse.cz>
1895 PR ipa/84149
1896 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1897 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1898 not the same as the source val.
1899 (cgraph_edge_brings_value_p): New parameter.
1900 (gather_edges_for_value): Pass destination value to
1901 cgraph_edge_brings_value_p.
1902 (perhaps_add_new_callers): Likewise.
1903 (get_info_about_necessary_edges): Likewise and exclude values brought
1904 only by self-recursive edges.
1905 (create_specialized_node): Redirect only clones of self-calling edges.
1906 (+self_recursive_pass_through_p): New function.
1907 (find_more_scalar_values_for_callers_subset): Use it.
1908 (find_aggregate_values_for_callers_subset): Likewise.
1909 (known_aggs_to_agg_replacement_list): Removed.
1910 (decide_whether_version_node): Re-calculate known constants for all
1911 remaining context clones.
1913 2018-04-11 Richard Biener <rguenther@suse.de>
1915 PR lto/85339
1916 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1917 from early DWARF output.
1918 (dwarf2out_early_finish): Output line info unconditionally into
1919 early DWARF and add reference to it.
1921 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1923 PR target/85281
1924 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1925 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1926 other than V2DFmode using iptr mode attribute.
1927 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1929 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1931 PR rtl-optimization/84659
1932 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1934 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1936 PR debug/85302
1937 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1938 SIZEP is NULL.
1939 (output_loc_list): Pass address of a dummy size variable even in the
1940 locview handling loop.
1941 (index_location_lists): Add comment on why skip_loc_list_entry can't
1942 call size_of_locs.
1944 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1946 PR target/85261
1947 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1948 into register.
1950 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1952 PR target/85321
1953 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1954 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1955 and -mstring-compare-inline-limit.
1957 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1959 PR target/85287
1960 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1961 for stack clash protection in a register whenever we need it to be in
1962 a register.
1964 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1966 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1967 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1969 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1971 PR target/85321
1972 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1973 the help text.
1974 (mlong-double-): Ditto.
1975 * config/rs6000/sysv4.opt (msdata=): Ditto.
1976 (mtls-size=): Ditto.
1978 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1980 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1981 erroneous entries for
1982 "vector int vec_ldl (int, long int *)", and
1983 "vector unsigned int vec_ldl (int, unsigned long int *)".
1984 Add comments and entries for
1985 "vector bool char vec_ldl (int, bool char *)",
1986 "vector bool short vec_ldl (int, bool short *)",
1987 "vector bool int vec_ldl (int, bool int *)",
1988 "vector bool long long vec_ldl (int, bool long long *)",
1989 "vector pixel vec_ldl (int, pixel *)",
1990 "vector long long vec_ldl (int, long long *)",
1991 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1992 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1993 type tree bool_long_long_type_node and correct definition of
1994 bool_V2DI_type_node to make reference to this new type tree.
1995 (rs6000_mangle_type): Replace erroneous reference to
1996 bool_long_type_node with bool_long_long_type_node.
1997 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1998 comments to emphasize sign distinctions for char and int types and
1999 replace RS6000_BTI_bool_long constant with
2000 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2001 use of RS6000_BTI_pixel.
2002 (bool_long_type_node): Remove this macro definition.
2003 (bool_long_long_type_node): New macro definition
2005 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2007 PR rtl-optimization/85300
2008 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2009 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2010 simplify_unary_operation fails.
2012 2018-04-10 Martin Liska <mliska@suse.cz>
2014 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2015 cgraph_edge and ipa_ref.
2017 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2019 PR target/85177
2020 PR target/85255
2021 * config/i386/sse.md
2022 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2023 computation of the VEC_MERGE selector from mask.
2024 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2025 Fix decoding of the VEC_MERGE selector into mask.
2027 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2029 PR tree-optimization/85286
2030 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2032 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2034 * final.c (final_1): Set insn_last_address as well as
2035 insn_current_address.
2037 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2039 PR target/85173
2040 * explow.c (emit_stack_probe): Call validize_mem on memory location
2041 before passing it to gen_probe_stack. Create address operand and
2042 legitimize it for the probe_stack_address case.
2044 2018-04-09 Jan Hubicka <jh@suse.cz>
2046 PR lto/85078
2047 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2048 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2049 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2050 walk basetypes.
2051 (free_lang_data): Rebuild type inheritance graph.
2053 2018-04-09 Martin Sebor <msebor@redhat.com>
2055 * invoke.texi (-finline-small-functions): Mention other optimization
2056 options.
2057 (-findirect-inlining, -fpartial-inlining): Same.
2058 (-finline-functions-called-once): Same.
2059 (-freorder-blocks-and-partition): Same.
2061 2018-04-09 Jan Hubicka <jh@suse.cz>
2063 PR rtl/84058
2064 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2065 jumps; choose last target that matches the criteria (i.e.
2066 no partition changes for non-crossing jumps).
2067 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2068 support for redirecting crossing jumps to non-crossing.
2070 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2072 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2073 also for naked functions.
2075 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2077 * config/arc/arc.md (add_shift): New pattern.
2078 (add_shift2): Likewise.
2079 (sub_shift): Likewise.
2080 (sub_shift_cmp0_noout): Likewise.
2081 (compare_si_ashiftsi): Likewise.
2082 (xbfu_cmp0_noout): New combine pattern.
2083 (xbfu_cmp0"): Likewise.
2084 (movsi_set_cc_insn): Place the predicable variant first.
2085 (commutative_binary_cmp0_noout): Remove clobber.
2086 (commutative_binary_cmp0): New pattern.
2087 (noncommutative_binary_cmp0): Likewise.
2088 (noncommutative_binary_cmp0_noout): Likewise.
2089 (noncommutative_binary_comparison_result_used): Removed.
2090 (rsub_cmp0): New pattern.
2091 (rsub_cmp0_noout): Likewise.
2092 (extzvsi): Changed, keep only meaningful variants.
2093 (SQH, SEZ): New iterators.
2094 (SQH_postfix): New mode attribute.
2095 (SEZ_prefix): New code attribute.
2096 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2097 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2098 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2099 of numerical value.
2100 (noncommutative_operator): Check the availability of barrel
2101 shifter option.
2103 2018-04-09 Richard Biener <rguenther@suse.de>
2105 PR tree-optimization/85284
2106 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2107 Only use the niter constraining form of simple_iv when the exit
2108 is always executed.
2110 2018-04-09 Tom de Vries <tom@codesourcery.com>
2112 PR target/84041
2113 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2114 (define_expand "*memory_barrier"): New define_expand.
2115 (define_insn "memory_barrier"): New insn.
2117 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2119 PR rtl-optimization/80463
2120 PR rtl-optimization/83972
2121 PR rtl-optimization/83480
2123 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2124 correct producer for the insn.
2125 (tidy_control_flow): Fixup seqnos in case of debug insns.
2127 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2129 PR rtl-optimization/83913
2131 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2132 different sched-times when merging exprs.
2134 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2136 PR rtl-optimization/83962
2138 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2139 tidy_fallthru_edge and tidy_control_flow.
2141 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2143 PR rtl-optimization/83530
2145 * sel-sched.c (force_next_insn): New global variable.
2146 (remove_insn_for_debug): When force_next_insn is true, also leave only
2147 next insn in the ready list.
2148 (sel_sched_region): When the region wasn't scheduled, make another pass
2149 over it with force_next_insn set to 1.
2151 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2153 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2154 into tm_file.
2155 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2156 for interrupt control.
2157 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2158 functions for interrupt control.
2159 * config/nds32/nds32-intrinsic.md: Likewise.
2160 * config/nds32/nds32_intrinsic.h: Likewise.
2161 * config/nds32/nds32.h (nds32_builtins): Likewise.
2163 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2165 * config/nds32/nds32.c (nds32_init_machine_status,
2166 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2167 strict_aligned_p field.
2168 (nds32_expand_to_rtl_hook): New function.
2169 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2170 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2172 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2173 Chung-Ju Wu <jasonwucj@gmail.com>
2175 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2176 * config/nds32/nds32-n7.md: New file.
2177 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2178 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2179 pipeline.
2180 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2181 * config/nds32/nds32.md (pipeline_model): Add n7.
2182 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2183 * config/nds32/pipelines.md: Include n7 settings.
2185 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2186 Chung-Ju Wu <jasonwucj@gmail.com>
2188 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2189 * config/nds32/nds32-e8.md: New file.
2190 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2191 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2192 pipeline.
2193 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2194 * config/nds32/nds32.md (pipeline_model): Add e8.
2195 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2196 * config/nds32/pipelines.md: Include e8 settings.
2198 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2199 Chung-Ju Wu <jasonwucj@gmail.com>
2201 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2202 * config/nds32/nds32-n8.md: New file.
2203 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2204 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2205 pipeline.
2206 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2207 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2208 * config/nds32/nds32.md (pipeline_model): Add n8.
2209 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2210 * config/nds32/pipelines.md: Include n8 settings.
2212 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2213 Chung-Ju Wu <jasonwucj@gmail.com>
2215 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2216 * config/nds32/nds32-n9-2r1w.md: New file.
2217 * config/nds32/nds32-n9-3r2w.md: New file.
2218 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2219 nds32_register_ports): New or modify for cpu n9.
2220 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2221 pipeline.
2222 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2223 * config/nds32/nds32-utils.c: New file.
2224 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2225 TARGET_MUL_SLOW): Define.
2226 * config/nds32/nds32.md (pipeline_model): New attribute.
2227 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2228 New options that support cpu n9.
2229 * config/nds32/pipelines.md: Include n9 settings.
2230 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2232 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2234 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2235 information if necessary.
2236 (output_cond_branch_compare_zero): Likewise.
2237 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2238 (nds32_target_alignment): Refine for alignment.
2239 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2240 (FUNCTION_BOUNDARY): Modify.
2241 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2242 align case.
2243 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2245 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2247 * config/nds32/constants.md (unspec_volatile_element): Add values for
2248 TLB operation and data prefetch.
2249 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2250 functions for TLB operation and data prefetch.
2251 * config/nds32/nds32-intrinsic.md: Likewise.
2252 * config/nds32/nds32_intrinsic.h: Likewise.
2253 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2254 (nds32_print_operand): Likewise.
2255 * config/nds32/nds32.h (nds32_builtins): Likewise.
2257 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2258 Andrew Pinski <pinsika@gcc.gnu.org>
2260 PR middle-end/82976
2261 * match.pd: Use constant_boolean_node of correct type instead of
2262 boolean_true_node or boolean_false_node for simplifying
2263 pointer comparisons to zero.
2265 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2267 PR tree-optimization/80021
2268 * tree.c (verify_type_variant): Make error call in verify_variant_match
2269 translatable and remove final full stop.
2271 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2273 * config/nds32/constants.md (unspec_volatile_element): Add
2274 UNSPEC_VOLATILE_EH_RETURN.
2275 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2276 nds32_output_stack_pop): Support dwarf exception handling process.
2277 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2278 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2279 exception handling process.
2280 (nds32_compute_stack_frame): Likewise.
2281 (nds32_return_addr_rtx): Likewise.
2282 (nds32_initial_elimination_offset): Likewise.
2283 (nds32_expand_prologue): Likewise.
2284 (nds32_expand_epilogue): Likewise.
2285 (nds32_dynamic_chain_address): New function.
2286 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2287 exception handling.
2288 (DYNAMIC_CHAIN_ADDRESS): Define.
2289 (EH_RETURN_DATA_REGNO): Define.
2290 (EH_RETURN_STACKADJ_RTX): Define.
2291 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2292 patterns for dwarf exception handling.
2294 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2296 * config/nds32/nds32.h: Clean up obsolete macros.
2298 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2300 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2301 Add enum values for particular instructions.
2302 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2303 particular intrinsic functions.
2304 * config/nds32/nds32-intrinsic.md: Likewise.
2305 * config/nds32/nds32_intrinsic.h: Likewise.
2306 * config/nds32/nds32.h (nds32_builtins): Likewise.
2307 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2308 (btst, ave): New patterns for particular instructions.
2310 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2312 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2313 Add enum values for atomic load/store and memory sync.
2314 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2315 and memory sync.
2316 * config/nds32/nds32-intrinsic.md: Likewise.
2317 * config/nds32/nds32_intrinsic.h: Likewise.
2318 * config/nds32/nds32.h (nds32_builtins): Likewise.
2320 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2322 PR tree-optimization/85257
2323 * fold-const.c (native_encode_vector): If not all elts could fit
2324 and off is -1, return 0 rather than offset.
2325 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2326 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2327 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2328 adjust buffer in native_interpret_expr call.
2330 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2332 * config/nds32/constants.md (unspec_volatile_element): Add cache
2333 control enum values.
2334 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2335 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2336 * config/nds32/nds32.c (nds32_cctl_names): New.
2337 (nds32_print_operand): Handle cache control register names.
2338 * config/nds32/nds32.h (nds32_builtins): New enum values.
2339 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2340 macros.
2341 * config/nds32/nds32.md (type): Add mmu.
2342 * config/nds32/pipelines.md (simple_insn): Add mmu.
2344 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2346 * config/nds32/nds32.md (type): Remove call.
2347 * config/nds32/pipelines.md (simple_insn): Likewise.
2349 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2351 * config/nds32/constants.md (unspec_volatile_element): Add
2352 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2353 UNSPEC_VOLATILE_FMFCFG.
2354 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2355 description for fmfcfg and fmfcsr.
2356 (bdesc_1arg): Add fmtcsr.
2357 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2358 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2359 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2360 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2361 unspec_fmfcfg): New patterns.
2362 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2363 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2364 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2365 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2366 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2367 __nds32__fmfcfg): Define.
2369 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2371 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2372 intrinsic register names.
2373 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2374 intrinsic register enum values and macros.
2376 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2378 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2379 for load/store addressing form.
2380 (nds32_print_operand_address): Likewise.
2382 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2384 PR target/85196
2385 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2386 based on LABEL_REF. Remove useless assertion.
2387 (pic_address_needs_scratch): Fix formatting.
2388 (sparc_legitimize_pic_address): Minor tweaks.
2389 (sparc_delegitimize_address): Adjust assertion accordingly.
2390 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2391 into symbolic_operand.
2392 (movsi_high_pic_label_ref): Likewise.
2393 (movsi_lo_sum_pic_label_ref): Likewise.
2394 (movdi_pic_label_ref): Likewise.
2395 (movdi_high_pic_label_ref): Likewise.
2396 (movdi_lo_sum_pic_label_ref): Likewise.
2398 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2400 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2401 custom LIB_SPEC setup.
2403 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2404 Kito Cheng <kito.cheng@gmail.com>
2406 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2407 * config/riscv/freebsd.h: New.
2409 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2411 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2412 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2413 file.
2415 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2416 Kito Cheng <kito.cheng@gmail.com>
2418 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2419 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2420 * config/nds32/nds32-protos.h (nds32_output_call,
2421 nds32_output_return): Declare.
2422 * config/nds32/nds32.md: Refine all the call and return patterns.
2424 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2426 PR debug/85252
2427 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2428 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2430 PR rtl-optimization/84872
2431 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2432 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2433 EDGE_CROSSING edge.
2435 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2437 * expr.c (copy_blkmode_to_reg): Revert 254862.
2438 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2440 2018-04-06 Richard Biener <rguenther@suse.de>
2442 PR middle-end/85244
2443 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2444 after seeing a component reference with an adjacent field. Treat
2445 refs to arrays at struct end of external decls similar to
2446 refs to unconstrained commons.
2448 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2450 PR sanitizer/85213
2451 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2452 look through SAVE_EXPRs with non-side-effects argument. Adjust
2453 recursive calls.
2454 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2455 save_p here.
2457 2018-04-06 Richard Biener <rguenther@suse.de>
2459 PR middle-end/85180
2460 * alias.c (find_base_term): New wrapper around find_base_term
2461 unwinding CSELIB_VAL_PTR changes.
2462 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2463 recursion.
2465 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2467 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2468 instructions.
2469 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2470 constant definitions.
2471 ("nop"): lr 0,0 -> nopr r0
2472 ("nop_lr0", "nop_lr1"): New insn definitions.
2474 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2476 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2477 NDS32_V3PUSH_AVAILABLE_P macro.
2479 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2480 Chung-Ju Wu <jasonwucj@gmail.com>
2482 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2483 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2484 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2485 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2486 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2487 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2488 * config/nds32/constraints.md: New constraints and checking for hard
2489 float configuration.
2490 * config/nds32/iterators.md: New mode iterator and attribute for hard
2491 float configuration.
2492 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2493 patterns.
2494 * config/nds32/nds32-fpu.md: New file.
2495 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2496 deal with hard float code generation.
2497 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2498 ARCH_V3S.
2499 (abi_type, float_reg_number): New enum type.
2500 * config/nds32/nds32-predicates.c: New predicates for hard float.
2501 * config/nds32/nds32-protos.h: Declare functions for hard float.
2502 * config/nds32/nds32.c: Implementation for hard float configuration.
2503 * config/nds32/nds32.h: Definitions for hard float configuration.
2504 * config/nds32/nds32.md: Include hard float machine description and
2505 modify patterns for hard float configuration.
2506 * config/nds32/nds32.opt: New options for hard float configuration.
2507 * config/nds32/predicates.md: New predicates for hard float
2508 configuration.
2510 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2512 * common/config/nds32/nds32-common.c
2513 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2515 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2517 PR middle-end/85195
2518 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2519 CONSTRUCTOR_ELT (ctor, ...)->value.
2521 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2523 PR target/85193
2524 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2526 2018-04-05 Tom de Vries <tom@codesourcery.com>
2528 PR target/85204
2529 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2530 cond jump.
2532 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2533 Kito Cheng <kito.cheng@gmail.com>
2535 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2536 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2537 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2538 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2540 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2541 Kito Cheng <kito.cheng@gmail.com>
2543 * config/nds32/constraints.md (Ufe): New memory constraint.
2544 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2545 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2546 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2547 operands.
2548 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2549 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2551 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2553 * config/nds32/nds32.md: Use optimize_size in the condition for
2554 alu-shift instructions.
2556 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2558 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2560 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2562 * config/nds32/nds32.md (negsi2): Refine pattern.
2564 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2565 Chung-Ju Wu <jasonwucj@gmail.com>
2567 * config/nds32/iterators.md (shift_rotate): New code iterator.
2568 (shift): New code attribute.
2569 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2570 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2571 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2572 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2573 bit-wise operations.
2574 (andsi3, *andsi3): Ditto.
2575 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2576 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2577 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2578 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2579 nds32_ior_operand, nds32_xor_operand): New predicates.
2581 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2583 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2584 (addsi3, subsi3): ... this.
2586 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2588 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2590 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2592 * config/nds32/nds32.md: Adjust indention.
2594 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2596 * config/nds32/nds32.md (feature): New attribute.
2598 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2600 * config/nds32/nds32.md (subtype): New attribute.
2602 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2604 PR target/85203
2605 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2606 expansion to perform a bitwise AND of the argument followed by a
2607 boolean negation of the result.
2609 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2611 PR rtl-optimization/84878
2612 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2613 the basic block. Assert the use reference is not artificial and that
2614 it has an associated insn.
2616 2018-04-04 Michael Matz <matz@suse.de>
2618 * builtins.c (compute_objsize): Pass correct operand
2619 to array_at_struct_end_p.
2621 2018-04-04 Richard Biener <rguenther@suse.de>
2623 PR lto/85176
2624 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2625 from contexts for DINFO_LEVEL_TERSE and below.
2627 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2629 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2630 resiter_operand condition.
2631 * config/nds32/nds32.md (*move<mode>): Ditto.
2633 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2634 Monk Chiang <sh.chiang04@gmail.com>
2636 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2638 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2640 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2642 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2643 Kito Cheng <kito.cheng@gmail.com>
2645 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2646 nds32_cond_code_str, output_cond_branch,
2647 output_cond_branch_compare_zero, nds32_expand_cbranch,
2648 nds32_expand_cstore, nds32_expand_movcc,
2649 nds32_output_cbranchsi4_equality_zero,
2650 nds32_output_cbranchsi4_equality_reg,
2651 nds32_output_cbranchsi4_equality_reg_or_const_int,
2652 nds32_output_cbranchsi4_greater_less_zero: New functions.
2653 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2654 nds32_expand_cstore, nds32_expand_movcc,
2655 nds32_output_cbranchsi4_equality_zero,
2656 nds32_output_cbranchsi4_equality_reg,
2657 nds32_output_cbranchsi4_equality_reg_or_const_int,
2658 nds32_output_cbranchsi4_greater_less_zero): Declare.
2659 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2660 nds32_rimm11s_operand): New predicates.
2661 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2662 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2663 patterns.
2665 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2667 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2668 * config/nds32/nds32.md: Ditto.
2669 * config/nds32/pipelines.md: Ditto.
2671 2018-04-04 Richard Biener <rguenther@suse.de>
2673 PR tree-optimization/85168
2674 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2675 propagating abnormals.
2677 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2679 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2681 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2682 Kito Cheng <kito.cheng@gmail.com>
2684 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2685 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2686 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2687 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2688 * config/nds32/nds32.md (sibcall_internal): New.
2689 (sibcall_register): Remove.
2690 (sibcall_immediate): Remove.
2691 (sibcall_value_internal): New.
2692 (sibcall_value_register): Remove.
2693 (sibcall_value_immediate): Remove.
2694 * config/nds32/predicates.md (nds32_general_register_operand): New.
2695 (nds32_call_address_operand): New.
2697 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2699 PR rtl-optimization/85167
2700 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2701 bb_defs if *split_p, instead preinitialize it to NULL.
2703 PR tree-optimization/85156
2704 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2705 evaluating the argument multiple times.
2707 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2709 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2710 than vector.
2711 (_mm_cvtpd_ps): Likewise.
2712 (_mm_cvttpd_epi32): Likewise.
2713 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2714 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2715 vector, pixel, and bool following altivec.h include.
2717 2018-04-03 Martin Sebor <msebor@redhat.com>
2719 * doc/extend.texi (Common Function Attributes): Clarify.
2720 (const attribute): Likewise.
2721 (pure attribute): Likewise.
2723 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2725 PR target/85169
2726 * config/i386/i386.c (ix86_expand_vector_set): Use
2727 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2729 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2731 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2732 instructions when changing rounding bits to preserve precision bits
2733 in the x87 control word.
2735 2018-04-03 Martin Liska <mliska@suse.cz>
2737 PR tree-optimization/82491
2738 * rtl.h (strip_offset_and_add): Replace += suboffset with
2739 poly_uint64 () + suboffset.
2741 2018-03-29 Martin Liska <mliska@suse.cz>
2742 Martin Jambor <mjambor@suse.cz>
2744 PR ipa/84947
2745 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2746 param_type is not an integral or pointer type.
2748 2018-04-03 Richard Biener <rguenther@suse.de>
2750 * sese.h (recompute_all_dominators): Remove.
2752 2018-04-02 Martin Sebor <msebor@redhat.com>
2754 * doc/invoke.texi (-Wrestrict): Fix typos.
2756 2018-04-02 Jim Wilson <jimw@sifive.com>
2758 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2759 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2760 (<optab>di3, <optab>si3_extend): Likewise.
2761 (<optab>si3_mask, <optab>si3_mask_1): New.
2762 (<optab>di3_mask, <optab>di3_mask_1): New.
2763 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2764 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2765 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2767 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2769 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2770 example.
2772 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2774 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2775 (nds32_canonicalize_comparison): New function.
2777 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2778 Kito Cheng <kito.cheng@gmail.com>
2779 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2781 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2782 * config/nds32/constants.md (unspec_volatile_element): Add
2783 UNSPEC_VOLATILE_RELAX_GROUP.
2784 * config/nds32/nds32-relax-opt.c: New file.
2785 * config/nds32/nds32-predicates.c
2786 (nds32_symbol_load_store_p): New function.
2787 * config/nds32/nds32-protos.h
2788 (nds32_symbol_load_store_p): Declare function.
2789 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2790 * config/nds32/nds32.c
2791 (nds32_register_pass): New function to register pass.
2792 (nds32_register_passes): New function to register passes.
2793 * config/nds32/nds32.md (relax_group): New pattern.
2794 * config/nds32/nds32.opt (mrelax-hint): New option.
2795 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2797 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2799 * config/nds32/t-nds32: Modify files dependency.
2801 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2803 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2804 (PROFILE_HOOK): Define its implementation.
2806 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2808 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2809 type and 32-bit size.
2811 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2813 PR middle-end/85090
2814 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2815 (V_128_256): New mode iterator.
2816 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2817 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2818 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2819 of V.
2820 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2821 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2823 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2825 PR target/83315
2826 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2827 NaN inputs correctly.
2829 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2831 PR target/80546
2832 * config/rs6000/vsx.md (??r): New mode attribute.
2833 (*vsx_mov<mode>_64bit): Use it.
2834 (*vsx_mov<mode>_32bit): Likewise.
2836 2018-03-30 Martin Sebor <msebor@redhat.com>
2838 PR tree-optimization/84818
2839 * builtins.c (check_access): Use warning_n.
2841 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2843 PR target/83822
2844 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2845 condition.
2846 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2847 condition.
2849 2018-03-30 Julia Koval <julia.koval@intel.com>
2851 PR target/84413
2852 * x86-tune.def (movx, partial_reg_dependency): Enable for
2853 m_SKYLAKE_AVX512.
2855 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2857 PR inline-asm/84985
2858 * lra-constraints.c (process_alt_operands): Move setting
2859 this_alternative_matches below.
2861 2018-03-29 Martin Liska <mliska@suse.cz>
2863 PR lto/84995.
2864 * doc/invoke.texi: Document how LTO works with debug info.
2865 Describe auto-load support of binutils. Mention 'x86-64'
2866 as valid option value of -march option.
2868 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2870 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2872 PR c/85094
2873 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2874 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2875 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2876 checking.
2878 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2880 PR target/84912
2881 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2882 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2883 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2884 for RS6000_BTM_POWERPC64.
2885 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2886 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2887 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2888 definition.
2889 (DIVDE): Use it.
2890 (DIVDEU): Likewise.
2892 2018-03-28 Carl Love <cel@us.ibm.com>
2894 Revert
2895 2017-09-27 Carl Love <cel@us.ibm.com>
2897 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2898 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2899 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2900 fctiw instruction.
2902 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2904 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2905 instead of __vector bool.
2906 (_mm_max_pu8): Likewise.
2907 (_mm_min_pi16): Likewise.
2909 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2911 PR target/84912
2912 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2913 (DIVWEUO): Likewise.
2914 (DIVDEO): Likewise.
2915 (DIVDEUO): Likewise.
2916 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2917 DIVWEUO and DIVDEUO.
2918 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2919 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2920 (div_extend): Likewise.
2921 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2922 builtin function.
2923 (__builtin_divweuo): Likewise.
2924 (__builtin_divdeo): Likewise.
2925 (__builtin_divdeuo): Likewise.
2927 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2929 PR target/85095
2930 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2931 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2933 PR tree-optimization/82004
2934 * gimple-match-head.c (optimize_pow_to_exp): New function.
2935 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2936 Don't fold to exp if optimize_pow_to_exp is false.
2938 2018-03-28 Martin Liska <mliska@suse.cz>
2940 PR other/84819
2941 * calls.c (initialize_argument_information): Fix trailing space.
2942 * common.opt: Fix typo and provide better explanation for
2943 -fsanitize-coverage option.
2944 * config/i386/i386.opt: Fix typo.
2946 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2947 Martin Liska <mliska@suse.cz>
2949 PR sanitizer/85081
2950 * gimplify.c (asan_poison_variable): Don't do the check for
2951 gimplify_omp_ctxp here.
2952 (gimplify_decl_expr): Do it here.
2953 (gimplify_target_expr): Likewise.
2955 2018-03-28 Martin Liska <mliska@suse.cz>
2957 PR target/84988
2958 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2959 chkp_type_bounds_count if MPX is not enabled.
2961 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2963 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2965 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2967 PR target/84914
2968 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2969 function to create the function decl for complex long double
2970 multiply and divide for -mabi=ieeelongdouble.
2971 (init_float128_ieee): Call it.
2973 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2975 PR target/85044
2976 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2977 -fcf-protection=branch -mibt.
2978 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2980 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2982 PR target/81863
2983 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2985 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2987 PR target/85056
2988 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2989 extern array declarations.
2991 2018-03-27 Richard Biener <rguenther@suse.de>
2993 PR middle-end/84067
2994 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2995 explicit single_use checks.
2997 2018-03-27 Richard Biener <rguenther@suse.de>
2999 PR tree-optimization/85082
3000 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3001 Valueize the VUSE.
3003 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3005 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3006 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3007 Turn on fasynchronous-unwind-tables and funwind-tables.
3009 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3011 PR target/85073
3012 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3013 (*bmi_blsr_<mode>_ccz): Ditto.
3015 2018-03-26 Tom de Vries <tom@codesourcery.com>
3017 PR tree-optimization/85063
3018 * omp-general.c (offloading_function_p): New function. Factor out
3019 of ...
3020 * omp-offload.c (pass_omp_target_link::gate): ... here.
3021 * omp-general.h (offloading_function_p): Declare.
3022 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3023 with attribute omp declare target for offloading functions.
3025 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3027 PR tree-optimization/84005
3028 * tree-data-ref.h (get_base_for_alignment): Declare.
3029 * tree-data-ref.c (get_base_for_alignment_1): New function.
3030 (get_base_for_alignment): Likewise.
3031 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3032 get_base_for_alignment to find a suitable base object, instead
3033 of always using drb->base_address.
3035 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3037 PR inline-asm/85022
3038 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3039 known size by default.
3041 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3043 PR inline-asm/85030
3044 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3045 and non BLKmode operands.
3047 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3049 PR target/85026
3050 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3051 Clean up attributes.
3053 2018-03-23 Richard Biener <rguenther@suse.de>
3055 PR debug/85020
3056 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3057 we are going to emit early debug for LTO.
3059 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3061 PR inline-asm/85034
3062 * function.c (match_asm_constraints_1): Don't optimize if input
3063 doesn't satisfy general_operand predicate for output's mode.
3065 PR inline-asm/85022
3066 * alias.c (write_dependence_p): Don't require for x_canonicalized
3067 non-VOIDmode if x has VOIDmode.
3069 PR sanitizer/85029
3070 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3071 just don't try to optimize it rather than assert it never happens.
3073 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3075 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3076 macro expansions for definition of ST_INTERNAL_<mode> and
3077 LD_INTERNAL_<mode> builtins.
3078 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3079 Remove prototype.
3080 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3081 function.
3082 (altivec_expand_st_builtin): Likewise.
3083 (altivec_expand_builtin): Remove calls to deleted functions.
3084 (rs6000_address_for_altivec): Delete this function.
3085 * config/rs6000/vector.md: Remove expands for
3086 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3088 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3090 PR target/84826
3091 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3092 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3093 re-computing once computed.
3094 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3095 (arm_init_machine_status): Initialize
3096 machine->static_chain_stack_bytes.
3098 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3100 PR target/84760
3101 * doc/extend.texi: Add four new prototypes for vec_ld.
3102 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3103 definitions for more logical presentation.
3104 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3105 entries for V1TI variants of __builtin_altivec_ld builtin.
3106 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3107 handling of V1TI variant of LVX icode pattern.
3108 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3109 (rs6000_gimple_fold_builtin): Likewise.
3110 (altivec_init_builtins): Add code to define
3111 __builtin_altivec_lvx_v1ti function.
3113 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3115 PR inline-asm/84941
3116 * function.c (match_asm_constraints_1): Don't do the optimization
3117 if input isn't a REG, SUBREG, MEM or constant.
3119 2018-03-22 Tom de Vries <tom@codesourcery.com>
3121 PR tree-optimization/84956
3122 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3123 bb_has_abnormal_pred.
3125 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3127 PR sanitizer/85018
3128 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3129 DECL_INITIAL (decl) to decl at the end.
3130 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3131 adjust the comment.
3133 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3135 * doc/extend.texi (__builtin_tgmath): Document when complex
3136 integer types are treated as _Complex _Float64.
3138 2018-03-21 Tom de Vries <tom@codesourcery.com>
3140 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3142 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3144 PR tree-optimization/84960
3145 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3146 if it is ENTRY block, move them into single succ of ENTRY in that case.
3148 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3150 PR tree-optimization/84811
3151 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3152 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3153 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3154 (known_subrange_p): Update accordingly. Cast each value involved
3155 in the size comparison, rather than casting the result of the
3156 subtraction.
3158 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3160 PR tree-optimization/84982
3161 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3162 by flipping the least significant bit rather than all bits from
3163 bitpos to bitpos + bitsize - 1.
3165 2018-03-21 Nathan Sidwell <nathan@acm.org>
3167 * doc/extend.texi (Deprecated Features): Remove mention of
3168 long-deleted deprecations.
3170 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3172 PR jit/84288
3173 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3174 * configure: Regenerate.
3176 2018-03-21 Tom de Vries <tom@codesourcery.com>
3178 PR tree-optimization/83126
3179 * tree-parloops.c (num_phis): New function.
3180 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3182 2018-03-21 Nathan Sidwell <nathan@acm.org>
3184 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3185 mention anon-struct/union members and trailing attributes.
3187 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3189 PR tree-optimization/84969
3190 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3191 builtin memset partitions if they set different rhs values.
3193 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3195 PR rtl-optimization/84989
3196 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3197 VEC_DUPLICATE with scalar result mode.
3199 2018-03-21 Martin Liska <mliska@suse.cz>
3201 PR ipa/84963
3202 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3203 not intended return statement.
3205 2018-03-21 Martin Liska <mliska@suse.cz>
3207 PR target/84988
3208 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3209 (chkp_find_bound_slots_1): Limit number of iterations.
3211 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3213 PR target/84838
3214 * Minor grammar fixes for x86 options.
3216 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3218 PR debug/84875
3219 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3220 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3222 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3224 PR target/83789
3225 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3226 (altivec_lvx_<mode>_1op): Likewise.
3227 (altivec_stvx_<mode>_2op): Likewise.
3228 (altivec_stvx_<mode>_1op): Likewise.
3229 (altivec_lvx_<VM2:mode>): New define_expand.
3230 (altivec_stvx_<VM2:mode>): Likewise.
3231 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3232 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3233 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3234 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3235 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3236 (rs6000_gen_lvx): Likewise.
3237 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3238 (altivec_expand_stv_builtin): Likewise.
3239 (altivec_expand_builtin): Likewise.
3240 * config/rs6000/vector.md: Likewise.
3242 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3244 PR target/82518
3245 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3246 BYTES_BIG_ENDIAN.
3248 2018-03-20 Richard Biener <rguenther@suse.de>
3250 PR target/84986
3251 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3252 sign-conversions as zero, fall back to standard scalar_stmt
3253 cost for the rest.
3255 2018-03-20 Martin Liska <mliska@suse.cz>
3257 PR ipa/84825
3258 * predict.c (rebuild_frequencies): Handle case when we have
3259 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3261 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3263 PR target/84990
3264 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3265 flag_section_anchors.
3266 * varasm.c (use_blocks_for_decl_p): Remove hack for
3267 dw2_force_const_mem.
3269 PR target/84845
3270 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3271 to ...
3272 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3273 be created, use lowpart_subreg of operands[0] rather than operands[0]
3274 itself.
3275 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3276 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3277 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3278 and n constraint instead of aarch64_shift_imm_di and Usd.
3279 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3280 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3282 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3284 PR target/82989
3285 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3286 to favor GPR over NEON registers.
3287 (<shift>di3_neon): Likewise.
3289 2018-03-20 Tom de Vries <tom@codesourcery.com>
3291 PR target/84952
3292 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3293 (nvptx_process_pars): Emit bar.sync asap and alap.
3295 2018-03-20 Tom de Vries <tom@codesourcery.com>
3297 PR target/84954
3298 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3299 seen_label if seen_label is already set.
3301 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3303 PR target/84945
3304 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3305 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3306 Use 1U instead of 1. Formatting fixes.
3308 PR c/84953
3309 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3310 instead of TREE_TYPE (s1) for the return value.
3312 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3314 PR tree-optimization/84946
3315 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3316 bitsize + bitsize in poly_uint64 rather than poly_int64.
3318 PR sanitizer/78651
3319 * dwarf2asm.c: Include fold-const.c.
3320 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3321 of decl rather than decl itself.
3323 PR rtl-optimization/84643
3324 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3326 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3328 PR sanitizer/78651
3329 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3330 calling assemble_variable.
3332 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3334 PR target/81647
3335 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3336 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3338 2018-03-19 Jim Wilson <jimw@sifive.com>
3340 PR bootstrap/84856
3341 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3342 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3343 (riscv_first_stack_step): Likewise.
3344 (riscv_option_override): Use STACK_BOUNDARY instead of
3345 MIN_STACK_BOUNDARY.
3346 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3347 MIN_STACK_BOUNDARY.
3348 (BIGGEST_ALIGNMENT): Set to 128.
3349 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3350 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3351 STACK_BOUNDARY.
3353 2018-03-19 Richard Biener <rguenther@suse.de>
3355 PR tree-optimization/84933
3356 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3357 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3359 2018-03-19 Richard Biener <rguenther@suse.de>
3361 PR tree-optimization/84859
3362 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3363 (cond_if_else_store_replacement): Perform sinking operation on
3364 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3365 Generalize what a BB with a single eligible store is.
3367 2018-03-19 Richard Biener <rguenther@suse.de>
3369 PR tree-optimization/84929
3370 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3371 chrec_is_positive against non-chrec arg.
3373 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3375 PR target/84711
3376 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3378 2018-03-18 Martin Liska <mliska@suse.cz>
3380 PR rtl-optimization/84635
3381 * regrename.c (build_def_use): Use matches_mode only when
3382 matches >= 0.
3384 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3386 PR tree-optimization/84913
3387 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3388 vectorize chains of COND_EXPRs.
3390 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3392 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3394 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3396 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3398 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3400 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3402 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3403 Kito Cheng <kito.cheng@gmail.com>
3405 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3406 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3407 (nds32_adjust_reg_alloc_order): New function.
3408 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3410 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3412 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3413 nds32_print_operand, nds32_print_operand_address): Use
3414 HOST_WIDE_INT_PRINT_DEC instead.
3416 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3418 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3420 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3422 PR target/84902
3423 * config/i386/i386.c (initial_ix86_tune_features,
3424 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3425 unsigned long long.
3426 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3427 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3428 rather than 1u << ix86_tune. Formatting fix.
3429 (ix86_option_override_internal): Change ix86_arch_mask from
3430 unsigned int to unsigned HOST_WIDE_INT, initialize to
3431 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3432 (ix86_function_specific_restore): Likewise.
3434 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3436 PR target/84899
3437 * postreload.c (reload_combine_recognize_pattern): Perform
3438 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3439 truncate_int_for_mode the result for the destination's mode.
3441 PR c/84909
3442 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3443 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3444 Likewise.
3445 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3446 Likewise.
3448 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3450 PR target/84876
3451 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3452 regno_allocno_class_array and sorted_pseudos.
3453 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3454 insns where regno is used.
3456 2018-03-16 Martin Liska <mliska@suse.cz>
3458 PR ipa/84833
3459 * multiple_target.c (create_dispatcher_calls): Redirect
3460 reference in the symbol table.
3462 2018-03-16 Martin Liska <mliska@suse.cz>
3464 PR ipa/84722
3465 * multiple_target.c (create_dispatcher_calls): Redirect also
3466 an alias.
3468 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3470 PR c++/79937
3471 PR c++/82410
3472 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3473 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3474 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3476 2018-03-16 Julia Koval <julia.koval@intel.com>
3478 * doc/invoke.texi (Skylake Server): Add CLWB.
3479 Cannonlake): Remove CLWB.
3481 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3483 PR tree-optimization/84841
3484 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3485 1 << 3.
3486 (FLOAT_ONE_CONST_TYPE): Define.
3487 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3488 (sort_by_operand_rank): Put entries with higher constant_type last
3489 rather than first to match comments.
3491 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3493 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3494 split predicate.
3496 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3498 PR c++/79085
3499 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3500 check and use address of target always.
3502 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3504 PR target/84574
3505 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3506 (indirect_thunk_bnd_needed): Likewise.
3507 (indirect_thunks_used): Likewise.
3508 (indirect_thunks_bnd_used): Likewise.
3509 (indirect_return_needed): New.
3510 (indirect_return_bnd_needed): Likewise.
3511 (output_indirect_thunk_function): Add a bool argument for
3512 function return.
3513 (output_indirect_thunk_function): Don't generate alias for
3514 function return thunk.
3515 (ix86_code_end): Call output_indirect_thunk_function to generate
3516 function return thunks.
3517 (ix86_output_function_return): Set indirect_return_bnd_needed
3518 and indirect_return_needed instead of indirect_thunk_bnd_needed
3519 and indirect_thunk_needed.
3521 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3523 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3524 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3525 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3527 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3528 Paul Hua <paul.hua.gm@gmail.com>
3530 PR c/84852
3531 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3533 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3535 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3536 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3537 resp. SFmode cases.
3539 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3541 PR target/84711
3542 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3543 instead of GET_MODE_SIZE when comparing Units.
3545 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3547 PR target/68256
3548 * varasm.c (hash_section): Return an unchangeble hash value
3549 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3550 Return !aarch64_can_use_per_function_literal_pools_p ().
3552 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3554 PR target/84860
3555 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3556 rather than address of cmode as last argument to prepare_cmp_insn.
3558 2018-03-15 Julia Koval <julia.koval@intel.com>
3560 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3561 F_AVX512VNNI, F_AVX512BITALG): New.
3563 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3565 PR target/83451
3566 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3567 insn for floating-point loads and stores.
3569 2018-03-14 Carl Love <cel@us.ibm.com>
3571 * config/rs6000/rs6000-c.c: Add macro definitions for
3572 ALTIVEC_BUILTIN_VEC_PERMXOR.
3573 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3574 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3575 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3576 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3577 UNSPEC_VPERMXOR.
3578 * config/doc/extend.texi: Add prototypes for vec_permxor.
3580 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3582 PR c/84852
3583 * diagnostic-show-locus.c (class layout_point): Convert m_line
3584 from int to linenum_type.
3585 (line_span::comparator): Use linenum "compare" function when
3586 comparing line numbers.
3587 (test_line_span): New function.
3588 (layout_range::contains_point): Convert param "row" from int to
3589 linenum_type.
3590 (layout_range::intersects_line_p): Likewise.
3591 (layout::will_show_line_p): Likewise.
3592 (layout::print_source_line): Likewise.
3593 (layout::should_print_annotation_line_p): Likewise.
3594 (layout::print_annotation_line): Likewise.
3595 (layout::print_leading_fixits): Likewise.
3596 (layout::annotation_line_showed_range_p): Likewise.
3597 (struct line_corrections): Likewise for field m_row.
3598 (line_corrections::line_corrections): Likewise for param "row".
3599 (layout::print_trailing_fixits): Likewise.
3600 (layout::get_state_at_point): Likewise.
3601 (layout::get_x_bound_for_row): Likewise.
3602 (layout::print_line): Likewise.
3603 (diagnostic_show_locus): Likewise for locals "last_line" and
3604 "row".
3605 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3606 * input.c (selftest::test_linenum_comparisons): New function.
3607 (selftest::input_c_tests): Call it.
3608 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3609 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3610 * selftest.h (ASSERT_GT): New macro.
3611 (ASSERT_GT_AT): New macro.
3612 (ASSERT_LT): New macro.
3613 (ASSERT_LT_AT): New macro.
3615 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3617 PR rtl-optimization/84780
3618 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3620 2018-03-14 Martin Liska <mliska@suse.cz>
3622 * tree.c (record_node_allocation_statistics): Use
3623 get_stats_node_kind.
3624 (get_stats_node_kind): New function extracted from
3625 record_node_allocation_statistics.
3626 (free_node): Use get_stats_node_kind.
3628 2018-03-14 Richard Biener <rguenther@suse.de>
3630 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3631 that the value-set of ANTIC_IN doesn't grow.
3633 Revert
3634 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3635 member.
3636 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3637 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3639 2018-03-14 Julia Koval <julia.koval@intel.com>
3641 * config.gcc (icelake-client, icelake-server): New.
3642 (icelake): Remove.
3643 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3644 (initial_ix86_arch_features): Ditto.
3645 (PTA_SKYLAKE): Add SGX.
3646 (PTA_ICELAKE): Remove.
3647 (PTA_ICELAKE_CLIENT): New.
3648 (PTA_ICELAKE_SERVER): New.
3649 (ix86_option_override_internal): Split up icelake on icelake client and
3650 icelake server.
3651 (get_builtin_code_for_version): Ditto.
3652 (fold_builtin_cpu): Ditto.
3653 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3654 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3655 * config/i386/i386.h (processor_type): Ditto.
3656 * doc/invoke.texi: Ditto.
3658 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3660 PR sanitizer/83392
3661 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3662 INTEGER_CST offset, add it together with bitpos / 8 and
3663 sign extend based on POINTER_SIZE.
3665 PR target/84844
3666 Revert
3667 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3669 PR target/78090
3670 * config/i386/constraints.md (Yc): New register constraint.
3671 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3672 Use Yc constraint for alternative 2 of operand 0. Remove
3673 preferred_for_speed attribute.
3675 2018-03-14 Richard Biener <rguenther@suse.de>
3677 PR tree-optimization/84830
3678 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3679 with the old one to avoid oscillations.
3681 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3683 PR target/83712
3684 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3685 pseudos.
3686 (assign_by_spills): Return a flag of reload assignment failure.
3687 Do not process the reload assignment failures. Do not spill other
3688 reload pseudos if they has the same reg class. Update n if
3689 necessary.
3690 (lra_assign): Add a return arg. Set up from the result of
3691 assign_by_spills call.
3692 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3693 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3694 usage_insns if it is not NULL.
3695 (spill_hard_reg_in_range): New function.
3696 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3697 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3698 function prototypes.
3699 (lra_assign): Change prototype.
3700 * lra.c (lra): Add code to deal with fails by splitting hard reg
3701 live ranges.
3703 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3705 * config/riscv/riscv.opt (mrelax): New option.
3706 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3707 "norelax" when riscv_mrelax is disabled.
3708 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3710 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3712 PR target/84743
3713 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3714 reassociation for int modes.
3716 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3718 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3719 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3720 for big-endian.
3721 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3722 * config/aarch64/aarch64-sve.md
3723 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3724 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3725 (*extend<mode><Vwide>2): Rename to...
3726 (aarch64_sve_extend<mode><Vwide>2): ...this.
3727 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3728 renaming the old pattern to...
3729 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3730 unsigned packs.
3731 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3732 define_expand, renaming the old pattern to...
3733 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3734 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3735 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3736 account when deciding which SVE instruction the optab should use.
3737 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3739 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3741 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3742 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3743 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3744 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3745 (tlsdesc_small_<mode>): Turn a define_expand and use
3746 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3747 (tlsdesc_small_advsimd_<mode>): ...this.
3748 (tlsdesc_small_sve_<mode>): New pattern.
3750 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3752 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3753 (UNSPEC_UMUL_HIGHPART): New constants.
3754 (MUL_HIGHPART): New int iteraor.
3755 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3756 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3757 define_expand.
3758 (*<su>mul<mode>3_highpart): New define_insn.
3760 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3762 PR lto/84805
3763 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3764 incomplete types.
3766 2018-03-13 Martin Liska <mliska@suse.cz>
3768 PR ipa/84658.
3769 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3770 vector.
3771 (sem_item_optimizer::~sem_item_optimizer): Release it.
3772 (sem_item_optimizer::merge_classes): Register variable aliases.
3773 (sem_item_optimizer::fixup_pt_set): New function.
3774 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3775 * ipa-icf.h: Declare new variables and functions.
3777 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3779 PR middle-end/84834
3780 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3781 integer_pow2p@2 and test integer_pow2p in condition.
3782 (A < 0 ? C : 0): Similarly for @1.
3784 PR middle-end/84831
3785 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3786 characters starting at p contain '\0' character, don't look beyond
3787 that.
3789 PR target/84827
3790 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3791 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3793 PR target/84828
3794 * reg-stack.c (change_stack): Change update_end var from int to
3795 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3796 also call set_block_for_insn on the newly added insns and rescan.
3798 PR target/84786
3799 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3800 on the last operand.
3802 PR c++/84704
3803 * tree.c (stabilize_reference_1): Return save_expr (e) for
3804 STATEMENT_LIST even if it doesn't have side-effects.
3806 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3808 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3810 2018-03-12 Renlin Li <renlin.li@arm.com>
3812 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3813 aarch64_output_scalar_simd_mov_immediate.
3815 2018-03-12 Martin Sebor <msebor@redhat.com>
3817 PR tree-optimization/83456
3818 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3819 for perfectly overlapping calls to memcpy.
3820 (gimple_fold_builtin_memory_chk): Same.
3821 (gimple_fold_builtin_strcpy): Handle no-warning.
3822 (gimple_fold_builtin_stxcpy_chk): Same.
3823 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3825 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3827 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3828 parameter. Use it for SFmode.
3829 (rs6000_function_arg_advance_1): Adjust.
3830 (rs6000_function_arg): Adjust.
3831 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3833 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3835 PR rtl-optimization/84169
3836 PR rtl-optimization/84780
3837 * combine.c (can_combine_p): Check for a 2-insn combination whether
3838 the destination register is used between the two insns, too.
3840 2018-03-12 Richard Biener <rguenther@suse.de>
3842 PR tree-optimization/84803
3843 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3844 for refs DR analysis didn't process.
3846 2018-03-12 Richard Biener <rguenther@suse.de>
3848 PR tree-optimization/84777
3849 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3850 force-vectorize loops ignore whether we are optimizing for size.
3852 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3855 (TARGET_MD_ASM_ADJUST): Define.
3857 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3858 Kito Cheng <kito.cheng@gmail.com>
3859 Chung-Ju Wu <jasonwucj@gmail.com>
3861 * config/nds32/nds32.c (nds32_compute_stack_frame,
3862 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3863 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3864 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3865 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3866 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3867 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3868 * config/nds32/nds32.md (prologue, epilogue): Use macro
3869 NDS32_V3PUSH_AVAILABLE_P to do checking.
3871 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3873 PR debug/58150
3874 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3875 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3876 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3877 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3878 addition of most attributes on !orig_type_die or the attribute not
3879 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3881 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3882 Chung-Ju Wu <jasonwucj@gmail.com>
3884 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3885 __NDS32_VH__ macro.
3886 * config/nds32/nds32.opt (mvh): New option.
3888 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3889 Chung-Ju Wu <jasonwucj@gmail.com>
3891 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3892 function.
3893 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3894 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3895 definition.
3897 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3898 Chung-Ju Wu <jasonwucj@gmail.com>
3900 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3901 function.
3902 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3903 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3905 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3906 Kito Cheng <kito.cheng@gmail.com>
3907 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3910 UNSPEC_FFMISM and UNSPEC_FLMISM.
3911 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3912 for ffb, ffmism and flmism.
3913 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3914 (unspec_ffmism): Ditto.
3915 (unspec_flmism): Ditto.
3916 (nds32_expand_builtin_impl): Check if string extension is available.
3917 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3918 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3920 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3922 Reverting patch:
3923 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3925 PR target/83712
3926 * lra-assigns.c (assign_by_spills): Return a flag of reload
3927 assignment failure. Do not process the reload assignment
3928 failures. Do not spill other reload pseudos if they has the same
3929 reg class.
3930 (lra_assign): Add a return arg. Set up from the result of
3931 assign_by_spills call.
3932 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3933 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3934 usage_insns if it is not NULL.
3935 (spill_hard_reg_in_range): New function.
3936 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3937 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3938 function prototypes.
3939 (lra_assign): Change prototype.
3940 * lra.c (lra): Add code to deal with fails by splitting hard reg
3941 live ranges.
3943 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3945 PR target/84807
3946 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3948 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3950 PR debug/84620
3951 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3952 (dw_val_node): Add val_symbolic_view.
3953 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3954 (symview_upper_bound): New.
3955 (new_line_info_table): Initialize symviews_since_reset.
3956 (dwarf2out_source_line): Count symviews_since_reset and set
3957 symview_upper_bound.
3958 (dw_val_equal_p): Handle symview.
3959 (add_AT_symview): New.
3960 (print_dw_val): Handle symview.
3961 (attr_checksum, attr_checksum_ordered): Likewise.
3962 (same_dw_val_p, size_of_die): Likewise.
3963 (value_format, output_die): Likewise.
3964 (add_high_low_attributes): Use add_AT_symview for entry_view.
3965 (dwarf2out_finish): Reset symview_upper_bound, clear
3966 zero_view_p.
3968 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3970 PR target/83969
3971 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3972 Add strict argument and use it.
3973 (rs6000_split_multireg_move): Update for new strict argument.
3974 (mem_operand_gpr): Disallow all non-offsettable addresses.
3975 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3977 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3979 PR target/84772
3980 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3981 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3982 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3984 PR c++/84767
3985 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3986 decl, use remap_type if we want to use the type.
3988 2018-03-09 Martin Sebor <msebor@redhat.com>
3990 PR tree-optimization/84526
3991 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3992 Remove dead code.
3993 (builtin_access::generic_overlap): Be prepared to handle non-array
3994 base objects.
3996 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3998 PR rtl-optimization/84682
3999 * lra-constraints.c (process_address_1): Check is_address flag
4000 for address constraints.
4001 (process_alt_operands): Likewise.
4002 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4003 preprocess_constraints.
4004 * recog.h (preprocess_constraints): Add oploc parameter.
4005 Adjust callers.
4006 * recog.c (preprocess_constraints): Test address_operand for
4007 CT_ADDRESS constraints.
4009 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4011 PR target/83712
4012 * lra-assigns.c (assign_by_spills): Return a flag of reload
4013 assignment failure. Do not process the reload assignment
4014 failures. Do not spill other reload pseudos if they has the same
4015 reg class.
4016 (lra_assign): Add a return arg. Set up from the result of
4017 assign_by_spills call.
4018 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4019 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4020 usage_insns if it is not NULL.
4021 (spill_hard_reg_in_range): New function.
4022 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4023 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4024 function prototypes.
4025 (lra_assign): Change prototype.
4026 * lra.c (lra): Add code to deal with fails by splitting hard reg
4027 live ranges.
4029 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4031 PR target/83193
4032 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4033 Accept complain bool parameter. Only emit errors if it is true.
4034 (arm_parse_cpu_option_name): Likewise.
4035 (arm_target_thumb_only): Adjust callers of the above.
4036 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4037 prototype to take a default true bool parameter.
4038 (arm_parse_arch_option_name): Likewise.
4040 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4041 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4043 PR jit/64089
4044 PR jit/84288
4045 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4046 * configure: Regenerate.
4047 * configure.ac ("linker --version-script option"): New.
4048 ("linker soname option"): New.
4050 2018-03-09 Richard Biener <rguenther@suse.de>
4052 PR tree-optimization/84775
4053 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4054 immediate uses of predicate stmts and mark them modified.
4056 Revert
4057 PR tree-optimization/84178
4058 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4059 to caller.
4060 (version_loop_for_if_conversion): Delay update_ssa call.
4061 (tree_if_conversion): Delay update_ssa until after predicate
4062 insertion.
4064 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4066 PR target/84763
4067 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4068 when the function accesses prior frames.
4070 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4072 PR debug/84456
4073 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4074 gen_llsym, otherwise call maybe_gen_llsym.
4076 PR inline-asm/84742
4077 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4078 has ',' character inside of it.
4080 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4082 PR target/84748
4083 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4084 as clobbering CC_REGNUM.
4086 2018-03-08 Richard Biener <rguenther@suse.de>
4088 PR middle-end/84552
4089 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4090 (follow_copies_to_constant): Do not follow SSA names registered
4091 for update.
4093 2018-03-08 Richard Biener <rguenther@suse.de>
4095 PR tree-optimization/84178
4096 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4097 to caller.
4098 (version_loop_for_if_conversion): Delay update_ssa call.
4099 (tree_if_conversion): Delay update_ssa until after predicate
4100 insertion.
4102 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4104 PR tree-optimization/84178
4105 * tree-if-conv.c (release_bb_predicate): Remove the
4106 the assertion that the stmts have NULL use_ops.
4107 Discard the statements, asserting that they haven't
4108 yet been added to a BB.
4110 2018-03-08 Richard Biener <rguenther@suse.de>
4112 PR tree-optimization/84746
4113 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4114 (phi_translate): Pass in destination ANTIC_OUT set.
4115 (phi_translate_1): Likewise. For a simplified result lookup
4116 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4117 (phi_translate_set): Adjust.
4118 (do_pre_regular_insertion): Likewise.
4119 (do_pre_partial_partial_insertion): Likewise.
4121 2018-03-08 Martin Liska <mliska@suse.cz>
4123 PR gcov-profile/84735
4124 * doc/gcov.texi: Document usage of profile files.
4125 * gcov-io.h: Document changes in the format.
4127 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4129 PR debug/84404
4130 PR debug/84408
4131 * dwarf2out.c (struct dw_line_info_table): Update comments for
4132 view == -1.
4133 (FORCE_RESET_NEXT_VIEW): New.
4134 (FORCE_RESETTING_VIEW_P): New.
4135 (RESETTING_VIEW_P): Check for -1 too.
4136 (ZERO_VIEW_P): Likewise.
4137 (new_line_info_table): Force-reset next view.
4138 (dwarf2out_begin_function): Likewise.
4139 (dwarf2out_source_line): Simplify zero_view_p initialization.
4140 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4141 view directly. Omit view when omitting .loc at line 0.
4143 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4145 PR tree-optimization/84740
4146 * tree-switch-conversion.c (process_switch): Call build_constructors
4147 only if info.phi_count is non-zero.
4149 PR tree-optimization/84739
4150 * tree-tailcall.c (find_tail_calls): Check call arguments against
4151 DECL_ARGUMENTS (current_function_decl) rather than
4152 DECL_ARGUMENTS (func) when checking for tail recursion.
4154 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4156 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4157 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4158 Volker Reichelt's entry and add entries for people that perform
4159 GCC fuzzy testing and report numerous bugs.
4161 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4163 PR target/82411
4164 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4165 readonly data in sdata, if that is disabled.
4166 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4167 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4168 -mreadonly-in-sdata option.
4170 2018-03-07 Martin Sebor <msebor@redhat.com>
4172 PR tree-optimization/84468
4173 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4174 basic block when looking for nul assignment.
4176 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4178 PR target/84277
4179 * except.h (output_function_exception_table): Adjust prototype.
4180 * except.c (output_function_exception_table): Remove FNNAME parameter
4181 and add SECTION parameter. Ouput one part of the table at a time.
4182 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4183 the first part of the exception table and emit unwind directives.
4184 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4185 (i386_pe_seh_cold_init): Likewise.
4186 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4187 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4188 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4189 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4190 * config/i386/winnt.c: Include except.h.
4191 (struct seh_frame_state): Add reg_offset, after_prologue and
4192 in_cold_section fields.
4193 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4194 (i386_pe_seh_cold_init): New function.
4195 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4196 to seh->in_cold_section.
4197 (seh_emit_push): Record the offset of the push.
4198 (seh_emit_save): Record the offet of the save.
4199 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4200 Test seh->after_prologue to disregard the epilogue.
4201 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4202 (i386_pe_end_cold_function): New function.
4204 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4206 PR fortran/84565
4207 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4208 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4210 PR c++/84704
4211 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4212 on tmp_var.
4213 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4214 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4216 PR middle-end/84723
4217 * multiple_target.c: Include tree-inline.h and intl.h.
4218 (expand_target_clones): Diagnose and fail if node->definition and
4219 !tree_versionable_function_p (node->decl).
4221 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4223 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4224 sprint_ul.
4225 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4226 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4227 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4229 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4231 PR target/84710
4232 * combine.c (try_combine): Use reg_or_subregno instead of handling
4233 just paradoxical SUBREGs and REGs.
4235 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4237 * config/arc/arc.c (arc_finalize_pic): Remove function.
4238 (arc_must_save_register): We use single base PIC register, remove
4239 checks to save/restore the PIC register.
4240 (arc_expand_prologue): Likewise.
4241 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4242 Remove.
4243 (arc_verify_short): Likewise.
4244 (arc_attr_type): Likewise.
4245 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4246 (walk_stores): Likewise.
4247 (arc_address_cost): Make it static.
4248 (arc_verify_short): Likewise.
4249 (branch_dest): Likewise.
4250 (arc_attr_type): Likewise.
4251 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4252 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4253 (arc_final_prescan_insn): Remove inserting the nops due to
4254 hardware hazards. It is done in reorg step.
4255 (insn_length_variant_t): Remove.
4256 (insn_length_parameters_t): Likewise.
4257 (arc_insn_length_parameters): Likewise.
4258 (arc_get_insn_variants): Likewise.
4259 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4261 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4263 PR inline-asm/84683
4264 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4265 assertion failure.
4267 PR tree-optimization/84687
4268 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4269 on new_node->decl.
4270 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4272 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4274 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4275 Rename to ppc_speculation_barrier.
4276 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4277 __builtin_ppc_speculation_barrier.
4279 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4281 PR target/84700
4282 * combine.c (combine_simplify_rtx): Don't try to simplify if
4283 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4284 are equal to x.
4286 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4288 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4289 to 32 bytes when compiling for POWER9.
4291 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4293 PR target/84564
4294 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4295 regparm >= 3 with no arg reg available also for calls with
4296 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4298 PR target/84524
4299 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4300 orig,vex.
4301 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4303 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4305 PR target/84264
4306 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4308 2018-03-05 Richard Biener <rguenther@suse.de>
4310 PR tree-optimization/84486
4311 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4312 When inserting a __builtin_assume_aligned call set the LHS
4313 SSA name alignment info accordingly.
4315 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4317 PR tree-optimization/84114
4318 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4319 Avoid reassociation of FLOAT_MODE addition.
4321 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4323 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4324 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4325 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4326 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4327 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4328 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4329 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4330 and -mwbnoinvd.
4331 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4332 __builtin_ia32_wbinvd): New builtins.
4333 (SPECIAL_ARGS2): New.
4334 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4335 (SPECIAL_ARGS2): New.
4336 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4337 (ix86_valid_target_attribute_inner_p): Ditto.
4338 (ix86_init_mmx_sse_builtins): Add special_args2.
4339 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4340 TARGET_WBNOINVD_P): New.
4341 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4342 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4343 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4344 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4345 * config/i386/pconfigintrin.h: New file.
4346 * config/i386/wbnoinvdintrin.h: Ditto.
4347 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4348 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4350 2018-03-05 Richard Biener <rguenther@suse.de>
4352 PR tree-optimization/84670
4353 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4354 member.
4355 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4356 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4357 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4358 doesn't grow if all successors (recursively) were visited at least
4359 once.
4361 2018-03-05 Richard Biener <rguenther@suse.de>
4363 PR tree-optimization/84650
4364 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4365 if executed in the loop pipeline.
4367 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4369 * doc/configfiles.texi (Configuration Files): Move info about
4370 conditionalizing $target-protos.h to...
4371 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4372 differs from $target-protos.h.
4374 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4375 Chung-Ju Wu <jasonwucj@gmail.com>
4377 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4378 * config/nds32/nds32-multiple.md (setmemsi): Define.
4379 * config/nds32/nds32-memory-manipulation.c
4380 (nds32_gen_dup_4_byte_to_word_value): New.
4381 (emit_setmem_word_loop): New.
4382 (emit_setmem_byte_loop): New.
4383 (nds32_expand_setmem_loop): New.
4384 (nds32_expand_setmem_loop_v3m): New.
4385 (nds32_expand_setmem_unroll): New.
4386 (nds32_expand_setmem): New.
4388 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4389 Chung-Ju Wu <jasonwucj@gmail.com>
4391 * config/nds32/nds32-memory-manipulation.c
4392 (nds32_emit_load_store): New.
4393 (nds32_emit_post_inc_load_store): New.
4394 (nds32_emit_mem_move): New.
4395 (nds32_emit_mem_move_block): New.
4396 (nds32_expand_movmemsi_loop_unknown_size): New.
4397 (nds32_expand_movmemsi_loop_known_size): New.
4398 (nds32_expand_movmemsi_loop): New.
4399 (nds32_expand_movmemsi_unroll): New.
4400 (nds32_expand_movmemqi): Rename ...
4401 (nds32_expand_movmemsi): ... to this.
4402 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4403 (movmemsi): ... to this.
4404 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4405 (nds32_expand_movmemsi): ... to this.
4407 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4408 Monk Chiang <sh.chiang04@gmail.com>
4409 Chung-Ju Wu <jasonwucj@gmail.com>
4411 * config/nds32/nds32-protos.h
4412 (nds32_expand_load_multiple): New arguments.
4413 (nds32_expand_store_multiple): Ditto.
4414 (nds32_valid_multiple_load_store): Rename ...
4415 (nds32_valid_multiple_load_store_p): ... to this.
4416 * config/nds32/nds32-memory-manipulation.c
4417 (nds32_expand_load_multiple): Refine implementation.
4418 (nds32_expand_store_multiple): Ditto.
4419 * config/nds32/nds32-multiple.md
4420 (load_multiple): Update nds32_expand_load_multiple interface.
4421 (store_multiple): Update nds32_expand_store_multiple interface.
4422 * config/nds32/nds32-predicates.c
4423 (nds32_valid_multiple_load_store): Rename ...
4424 (nds32_valid_multiple_load_store_p): ... to this and refine
4425 implementation.
4426 * config/nds32/predicates.md
4427 (nds32_load_multiple_and_update_address_operation): New predicate.
4428 (nds32_store_multiple_and_update_address_operation): New predicate.
4430 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4431 Chung-Ju Wu <jasonwucj@gmail.com>
4433 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4434 (combo): New attribute.
4435 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4437 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4439 * config/nds32/nds32.opt: Change -mcmodel= default value.
4441 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4442 Monk Chiang <sh.chiang04@gmail.com>
4443 Chung-Ju Wu <jasonwucj@gmail.com>
4445 * config/nds32/constants.md (unspec_element): New enum.
4446 * config/nds32/constraints.md (Umw): New constraint.
4447 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4448 * config/nds32/nds32-intrinsic.md: Likewise.
4449 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4450 (nds32_valid_smw_lwm_base_p): New.
4451 (nds32_output_smw_single_word): New.
4452 (nds32_output_lmw_single_word): New.
4453 (nds32_expand_unaligned_load): New.
4454 (nds32_expand_unaligned_store): New.
4455 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4456 (nds32_output_smw_single_word): Declare.
4457 (nds32_output_lmw_single_word): Declare.
4458 (nds32_expand_unaligned_load): Declare.
4459 (nds32_expand_unaligned_store): Declare.
4460 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4461 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4462 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4463 NDS32_BUILTIN_UASTORE_DW.
4464 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4465 predicate.
4467 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4468 Kito Cheng <kito.cheng@gmail.com>
4469 Chung-Ju Wu <jasonwucj@gmail.com>
4471 * config/nds32/nds32-intrinsic.c
4472 (nds32_expand_builtin_null_ftype_reg): Delete.
4473 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4474 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4475 (nds32_read_argument): New.
4476 (nds32_legitimize_target): Ditto.
4477 (nds32_legitimize_argument): Ditto.
4478 (nds32_check_constant_argument): Ditto.
4479 (nds32_expand_unop_builtin): Ditto.
4480 (nds32_expand_unopimm_builtin): Ditto.
4481 (nds32_expand_binop_builtin): Ditto.
4482 (nds32_builtin_decl_impl): Ditto.
4483 (builtin_description): Ditto.
4484 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4485 (nds32_init_builtins_impl): Ditto.
4486 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4487 (nds32_builtin_decl): New.
4488 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4489 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4491 2018-03-02 Jeff Law <law@redhat.com>
4493 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4494 (redundant_insn, fill_simple_delay_slots): Likewise.
4495 (fill_slots_from_thread): Likewise.
4496 * resource.c (mark_referenced_resources): Likewise.
4497 (mark_set_resources, find_dead_or_set_registers): Likewise.
4499 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4501 * substring-locations.h (format_warning_va): Formatting fix for
4502 ATTRIBUTE_GCC_DIAG.
4503 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4504 argument.
4505 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4506 * substring-locations.c: Include intl.h.
4507 (format_warning_va): Turned into small wrapper around
4508 format_warning_n_va, renamed to ...
4509 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4510 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4511 use ngettext.
4512 (format_warning_at_substring_n): New function.
4513 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4514 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4515 format_warning_at_substring with just a shorter name instead of
4516 const function pointer.
4517 (fmtwarn_n): New function.
4518 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4519 appropriate, get rid of all the fmtstr temporaries, move conditionals
4520 with G_() wrapped string literals directly into fmtwarn arguments,
4521 cast dir.len to (int), formatting fixes.
4523 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4525 * doc/invoke.texi: Remove "Cilk Plus" references.
4527 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4528 Richard Biener <rguenther@suse.de>
4530 PR ipa/84628
4531 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4532 for error or warning attributes if CALL_FROM_THUNK_P is set.
4533 Formatting fixes.
4535 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4537 PR target/56540
4538 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4539 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4541 PR target/56540
4542 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4543 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4545 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4546 instead of -1U in last predictors element's probability member.
4548 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4550 PR ipa/83983
4551 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4552 arguments if they are comparable.
4554 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4556 PR tree-optimization/84634
4557 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4558 masks and masked_loop_p with a single loop_masks, making sure it's
4559 null for bb vectorization.
4561 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4563 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4564 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4565 loop->force_vectorize to check whether there is no alias.
4567 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4569 PR target/84614
4570 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4571 prototypes.
4572 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4573 comments.
4574 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4575 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4576 instead of a loop around prev_real_insn.
4577 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4578 prev_real_insn.
4580 PR inline-asm/84625
4581 * config/i386/i386.c (ix86_print_operand): Use conditional
4582 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4583 zero vector.
4585 2018-03-02 Richard Biener <rguenther@suse.de>
4587 PR tree-optimization/84427
4588 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4589 (bitmap_set_subtract_values): Rewrite to handle multiple
4590 exprs per value.
4591 (clean): Likewise.
4592 (prune_clobbered_mems): Likewise.
4593 (phi_translate): Take edge instead of pred/phiblock.
4594 (phi_translate_1): Likewise.
4595 (phi_translate_set): Likewise. Insert all translated
4596 exprs for a value into the set, keeping possibly multiple
4597 expressions per value.
4598 (compute_antic_aux): Adjust for phi_translate changes.
4599 When intersecting union the expressions and prune those
4600 not in the final value set, keeping possibly multiple
4601 expressions per value. Do not use value-insertion
4602 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4603 all expressions. Add verification that the value-sets
4604 only shrink during iteration.
4605 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4606 (do_pre_regular_insertion): Likewise.
4607 (do_pre_partial_partial_insertion): Likewise.
4609 2018-03-02 Richard Biener <rguenther@suse.de>
4611 PR target/82005
4612 * config/darwin.c (saved_debug_info_level): New static global.
4613 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4614 (darwin_asm_lto_end): Restore debug info generation settings.
4616 2018-03-01 Martin Liska <mliska@suse.cz>
4618 PR sanitizer/82484
4619 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4620 volatile arguments.
4622 2018-03-01 Richard Biener <rguenther@suse.de>
4624 PR debug/84645
4625 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4626 type annotation with LTO when debug was disabled at compile-time.
4628 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4630 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4631 XINT with INTVAL.
4632 (mips_final_postscan_insn): Likewise.
4634 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4636 PR rtl-optimization/84528
4637 * alias.c (init_alias_target): Add commentary.
4638 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4639 a unique base value if the frame pointer is not eliminated
4640 to the stack pointer.
4642 2018-03-01 Tom de Vries <tom@codesourcery.com>
4644 PR rtl-optimization/83327
4645 * lra-int.h (hard_regs_spilled_into): Declare.
4646 * lra.c (hard_regs_spilled_into): Define.
4647 (init_reg_info): Init hard_regs_spilled_into.
4648 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4649 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4650 (process_bb_lives): Handle hard_regs_spilled_into.
4651 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4652 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4654 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4656 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4657 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4658 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4659 * config/rs6000/aix72.h: New file.
4661 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4663 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4664 instead of warning_at with conditional singular and plural messages
4665 where possible.
4667 PR target/52991
4668 * stor-layout.c (update_alignment_for_field): For
4669 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4670 && !DECL_PACKED (field), do the alignment update, just use
4671 only desired_align instead of MAX (type_align, desired_align)
4672 as the alignment.
4673 (place_field): Don't do known_align < desired_align handling
4674 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4675 is non-NULL, instead do it after rli->prev_field handling and
4676 only if not within a bitfield word. For DECL_PACKED (field)
4677 use type_align of BITS_PER_UNIT.
4679 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4681 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4682 superfluous parentheses and trailing spaces.
4684 2018-02-28 Richard Biener <rguenther@suse.de>
4686 PR tree-optimization/84584
4687 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4688 SCoPs with fake exit edge.
4690 2018-02-28 Martin Liska <mliska@suse.cz>
4692 PR testsuite/84597
4693 * timevar.c (timer::print): Fix format to properly print 100%
4694 values.
4696 2018-02-28 Richard Biener <rguenther@suse.de>
4698 PR middle-end/84607
4699 * genmatch.c (capture_info::walk_match): Do not mark
4700 captured expressions without operands as expr_p given
4701 they act more like predicates and should be subject to
4702 "lost tail" side-effect preserving.
4704 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4706 PR rtl-optimization/81611
4707 * auto-inc-dec.c (attempt_change): Move dead note from
4708 mem_insn if it's the next use of regno
4709 (find_address): Take address use of reg holding
4710 non-incremented value. Add parm to limit search to the named
4711 reg only.
4712 (merge_in_block): Attempt to use a mem insn that is the next
4713 use of the original regno.
4715 2018-02-27 Martin Sebor <msebor@redhat.com>
4717 PR c++/83871
4718 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4719 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4721 2018-02-27 Martin Sebor <msebor@redhat.com>
4723 PR translation/84207
4724 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4725 n argument to unsigned HOST_WIDE_INT.
4726 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4727 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4728 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4729 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4731 2018-02-27 Richard Biener <rguenther@suse.de>
4733 PR tree-optimization/84512
4734 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4735 Do not use the estimate returned from record_stmt_cost for
4736 the scalar iteration cost but sum properly using add_stmt_cost.
4738 2018-02-27 Richard Biener <rguenther@suse.de>
4740 PR tree-optimization/84466
4741 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4742 Adjust last change to less strictly validate use operands.
4744 2018-02-27 Martin Liska <mliska@suse.cz>
4746 PR gcov-profile/84548
4747 * gcov.c (process_file): Allow partial overlap and consider it
4748 also as group functions.
4749 (output_lines): Properly calculate range of lines for a group.
4751 2018-02-27 Martin Liska <mliska@suse.cz>
4753 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4754 'ggc' suffixes. Change first column width.
4755 (timer::print): Fix formatting of the column.
4757 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4759 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4760 preserve inline entry blocks for the sake of debug inline
4761 entry point markers alone.
4762 (remove_unused_locals): Suggest in comments a better place to
4763 force the preservation of inline entry blocks that are
4764 otherwise unused, but do not preserve them.
4766 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4768 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4770 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4772 PR target/84039
4773 * config/i386/constraints.md (Bs): Replace
4774 ix86_indirect_branch_register with
4775 TARGET_INDIRECT_BRANCH_REGISTER.
4776 (Bw): Likewise.
4777 * config/i386/i386.md (indirect_jump): Likewise.
4778 (tablejump): Likewise.
4779 (*sibcall_memory): Likewise.
4780 (*sibcall_value_memory): Likewise.
4781 Peepholes of indirect call and jump via memory: Likewise.
4782 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4783 (*sibcall_value_GOT_32): Likewise.
4784 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4785 (GOT_memory_operand): Likewise.
4786 (call_insn_operand): Likewise.
4787 (sibcall_insn_operand): Likewise.
4788 (GOT32_symbol_operand): Likewise.
4789 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4791 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4793 PR rtl-optimization/83496
4794 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4795 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4796 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4797 redundant insn, if any.
4798 (relax_delay_slots): Likewise.
4799 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4801 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4803 PR tree-optimization/83965
4804 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4805 that grouped statements are part of a reduction chain. Return
4806 true if the statement is not marked as a reduction itself but
4807 is part of a group.
4808 (vect_recog_dot_prod_pattern): Don't check whether the statement
4809 is part of a group here.
4810 (vect_recog_sad_pattern): Likewise.
4811 (vect_recog_widen_sum_pattern): Likewise.
4813 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4815 PR debug/84545
4816 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4818 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4820 PR target/84530
4821 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4822 the bool argument.
4823 (ix86_output_indirect_function_return): New prototype.
4824 (ix86_split_simple_return_pop_internal): Likewise.
4825 * config/i386/i386.c (indirect_return_via_cx): New.
4826 (indirect_return_via_cx_bnd): Likewise.
4827 (indirect_thunk_name): Handle return va CX_REG.
4828 (output_indirect_thunk_function): Create alias for
4829 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4830 (ix86_output_indirect_jmp): Remove the bool argument.
4831 (ix86_output_indirect_function_return): New function.
4832 (ix86_split_simple_return_pop_internal): Likewise.
4833 * config/i386/i386.md (*indirect_jump): Don't pass false
4834 to ix86_output_indirect_jmp.
4835 (*tablejump_1): Likewise.
4836 (simple_return_pop_internal): Change it to define_insn_and_split.
4837 Call ix86_split_simple_return_pop_internal to split it for
4838 -mfunction-return=.
4839 (simple_return_indirect_internal): Call
4840 ix86_output_indirect_function_return instead of
4841 ix86_output_indirect_jmp.
4843 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4845 PR bootstrap/84405
4846 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4847 memset and value initialization afterwards.
4849 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4851 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4853 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4855 PR target/84521
4856 * common/config/aarch64/aarch64-common.c
4857 (aarch_option_optimization_table[]): Switch
4858 off fomit-frame-pointer
4860 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4861 Chung-Ju Wu <jasonwucj@gmail.com>
4863 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4864 volatile memory.
4865 (store_multiple): Ditto.
4867 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4869 * config.gcc: Add --with-cpu support for nds32 target.
4870 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4871 * config/nds32/nds32.opt: Add -mcpu= option.
4873 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4875 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4876 isel=yes): Warn for these deprecated options.
4878 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4880 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4881 ISA_2_5_MASKS_EMBEDDED.
4883 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4885 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4886 p->max as pointers rather than using iterative_hash_expr.
4888 2018-02-23 Carl Love <cel@us.ibm.com>
4890 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4891 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4892 BU_P8V_OVERLOAD_2.
4893 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4894 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4895 P8V_BUILTIN_VEC_VUNSIGNED2.
4897 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4899 PR target/81572
4900 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4901 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4902 LRA_UNKNOWN_ALT.
4903 * lra-constraints.c (curr_insn_transform): Set up
4904 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4905 LRA_UNKNOWN_ALT.
4906 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4907 * lra-eliminations.c (spill_pseudos): Ditto.
4908 (process_insn_for_elimination): Ditto.
4909 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4910 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4911 LRA_NON_CLOBBERED_ALT.
4913 2018-02-22 Martin Sebor <msebor@redhat.com>
4915 PR tree-optimization/84480
4916 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4917 to maybe_diag_stxncpy_trunc. Call it.
4918 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4919 from gimple_fold_builtin_strcpy. Print inlining stack.
4920 (handle_builtin_stxncpy): Print inlining stack.
4921 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4923 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4925 PR target/84176
4926 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4927 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4928 and -fcheck-pointer-bounds are used together.
4929 (indirect_thunk_prefix): New enum.
4930 (indirect_thunk_need_prefix): New function.
4931 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4932 "_nt" instead of "_bnd" for NOTRACK prefix.
4933 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4934 (output_indirect_thunk_function): Likewise.
4935 (): Likewise.
4936 (ix86_code_end): Update output_indirect_thunk_function calls.
4937 (ix86_output_indirect_branch_via_reg): Replace
4938 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4939 (ix86_output_indirect_branch_via_push): Likewise.
4940 (ix86_output_function_return): Likewise.
4941 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4942 incompatible with -fcf-protection=branch and
4943 -fcheck-pointer-bounds.
4945 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4947 PR target/83335
4948 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4949 Change gcc_assert call to output_operand_lossage.
4951 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4953 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4955 2018-02-22 DJ Delorie <dj@redhat.com>
4956 Sebastian Perta <sebastian.perta@renesas.com>
4957 Oleg Endo <olegendo@gcc.gnu.org>
4959 * config/rx/rx.c (rx_rtx_costs): New function.
4960 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4962 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4964 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4966 2018-02-22 Martin Liska <mliska@suse.cz>
4968 PR driver/83193
4969 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4970 Add "native" as a possible value.
4972 2018-02-22 Martin Liska <mliska@suse.cz>
4974 PR driver/83193
4975 * config/i386/i386.c (ix86_option_override_internal):
4976 Add "native" as a possible value for -march and -mtune.
4978 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4980 PR target/84502
4981 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4982 to all type variants.
4984 PR tree-optimization/84503
4985 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4986 width as info->bitpos + info->bitsize - start.
4987 (merged_store_group::merge_overlapping): Simplify width computation.
4988 (check_no_overlap): New function.
4989 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4990 start + width and last_order of the group, fail if check_no_overlap
4991 fails.
4992 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4993 to group if check_no_overlap fails.
4995 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4997 * config/rs6000/altivec.md: Delete contraint arguments to
4998 define_expand, define_split, and define_peephole2, and in
4999 define_insn_and_split if always unused.
5000 * config/rs6000/darwin.md: Ditto.
5001 * config/rs6000/dfp.md: Ditto.
5002 * config/rs6000/rs6000.md: Ditto.
5003 * config/rs6000/sync.md: Ditto.
5004 * config/rs6000/vector.md: Ditto.
5005 * config/rs6000/vsx.md: Ditto.
5007 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5009 * config/rs6000/altivec.md: Write output control strings as braced
5010 blocks instead of double-quoted strings.
5011 * config/rs6000/darwin.md: Ditto.
5012 * config/rs6000/rs6000.md: Ditto.
5013 * config/rs6000/vector.md: Ditto.
5014 * config/rs6000/vsx.md: Ditto.
5016 2018-02-21 Jason Merrill <jason@redhat.com>
5018 PR c++/84314 - ICE with templates and fastcall attribute.
5019 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5021 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5023 * ipa-cp.c (determine_versionability): Fix comment typos.
5025 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5027 PR c/84229
5028 * ipa-cp.c (determine_versionability): Do not version functions caling
5029 va_arg_pack.
5031 2018-02-21 Martin Liska <mliska@suse.cz>
5033 PR driver/83193
5034 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5035 Add "native" as a possible value.
5036 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5037 the macro when native cpu detection is available.
5039 2018-02-21 Martin Liska <mliska@suse.cz>
5041 PR driver/83193
5042 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5043 Add "native" as a possible value.
5044 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5045 when native cpu detection is available.
5047 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5048 Martin Sebor <msebor@redhat.com>
5050 PR tree-optimization/84478
5051 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5052 false.
5053 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5054 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5055 support which is conservatively correct, for 2 only stay conservative
5056 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5057 argument to the 2 argument get_range_strlen, adjust 6 arg
5058 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5059 false.
5060 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5061 (gimple_fold_builtin_strlen): Pass true as last argument to
5062 get_range_strlen.
5064 2018-02-20 Martin Sebor <msebor@redhat.com>
5066 PR middle-end/84095
5067 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5068 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5069 (builtin_memref::builtin_memref): Factor out parts into
5070 set_base_and_offset and call it.
5072 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5074 PR middle-end/84406
5075 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5076 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5077 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5078 search at the associated MODE_INT.
5080 2018-02-20 Jeff Law <law@redhat.com>
5082 PR middle-end/82123
5083 PR tree-optimization/81592
5084 PR middle-end/79257
5085 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5086 for range data rather than using global data.
5087 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5088 range data rather than using global data.
5089 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5090 pass it to children as needed.
5091 (struct directive::fmtresult): Similarly.
5092 (struct directive::set_width): Similarly.
5093 (struct directive::set_precision): Similarly.
5094 (format_integer, format_directive, parse_directive): Similarly.
5095 (format_none): Accept unnamed vr_values parameter.
5096 (format_percent, format_floating, format_character): Similarly.
5097 (format_string, format_plain): Similarly.
5098 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5099 the EVRP range analyzer for range data rather than using global data.
5100 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5101 gimple-ssa-evrp-analyze.h
5102 (class sprintf_dom_walker): Add after_dom_children member function.
5103 Add evrp_range_analyzer member.
5104 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5105 range analyzer as needed.
5106 (sprintf_dom_walker::after_dom_children): New member function.
5107 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5108 if not optimizing.
5109 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5110 (evrp_range_analyzer::pop_to_marker): Likewise.
5112 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5114 PR tree-optimization/84419
5115 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5116 with the required type if its current type is compatible but
5117 different.
5119 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5121 PR middle-end/82004
5122 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5123 after vectorization.
5125 2018-02-20 Martin Liska <mliska@suse.cz>
5127 PR driver/83193
5128 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5129 possible values if we don't have a hint.
5131 2018-02-20 Martin Liska <mliska@suse.cz>
5133 PR c/84310
5134 PR target/79747
5135 * final.c (shorten_branches): Build align_tab array with one
5136 more element.
5137 * opts.c (finish_options): Add alignment option limit check.
5138 (MAX_CODE_ALIGN): Likewise.
5139 (MAX_CODE_ALIGN_VALUE): Likewise.
5140 * doc/invoke.texi: Document maximum allowed option value for
5141 all -falign-* options.
5143 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5145 PR target/84146
5146 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5147 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5148 * var-tracking.c (emit_note_insn_var_location): Remove all references
5149 to NOTE_INSN_CALL_ARG_LOCATION.
5150 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5151 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5152 Use copy_rtx_if_shared.
5153 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5154 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5155 (dwarf2out_var_location): Remove handling of
5156 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5157 on call_insn.
5158 * final.c (final_scan_insn): Remove all references to
5159 NOTE_INSN_CALL_ARG_LOCATION.
5160 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5161 before dumping final insns.
5162 * except.c (emit_note_eh_region_end): Remove all references to
5163 NOTE_INSN_CALL_ARG_LOCATION.
5164 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5165 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5166 * config/arc/arc.c (hwloop_optimize): Likewise.
5167 * config/arm/arm.c (create_fix_barrier): Likewise.
5168 * config/s390/s390.c (s390_chunkify_start): Likewise.
5169 * config/sh/sh.c (find_barrier): Likewise.
5170 * config/i386/i386.c (rest_of_insert_endbranch,
5171 ix86_seh_fixup_eh_fallthru): Likewise.
5172 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5173 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5174 * config/frv/frv.c (frv_function_prologue): Likewise.
5175 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5176 reg note.
5177 (note_outside_basic_block_p): Remove all references to
5178 NOTE_INSN_CALL_ARG_LOCATION.
5179 * gengtype.c (adjust_field_rtx_def): Likewise.
5180 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5181 Likewise.
5182 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5183 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5185 PR c++/84444
5186 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5187 is ADDR_EXPR.
5189 PR tree-optimization/84452
5190 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5191 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5192 is NULL.
5194 2018-02-19 Martin Liska <mliska@suse.cz>
5196 PR sanitizer/82183
5197 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5199 2018-02-19 Martin Liska <mliska@suse.cz>
5200 Richard Sandiford <richard.sandiford@linaro.org>
5202 PR tree-optimization/82491
5203 * gimple-fold.c (get_base_constructor): Make earlier bail out
5204 to prevent ubsan.
5206 2018-02-19 Carl Love <cel@us.ibm.com>
5208 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5209 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5210 BU_P8V_OVERLOAD_1.
5211 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5212 P8V_BUILTIN_VEC_NEG.
5214 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5216 * config/rl78/rl78.md (movdf): New define expand.
5218 2018-02-19 Martin Liska <mliska@suse.cz>
5220 PR other/80589
5221 * doc/invoke.texi: Fix typo.
5222 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5224 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5226 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5227 handle rs6000_single_float and rs6000_double_float specially for
5228 e500 family CPUs.
5230 2018-02-16 Jeff Law <law@redhat.com>
5232 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5233 (pop_regs): Use it.
5235 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5237 PR ipa/84425
5238 * ipa-inline.c (inline_small_functions): Fix a typo.
5240 2018-02-16 Nathan Sidwell <nathan@acm.org>
5242 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5244 2018-02-16 Carl Love <cel@us.ibm.com>
5246 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5247 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5248 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5249 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5250 expansion to P8V_BUILTIN_VEC_FLOAT2.
5252 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5254 PR rtl-optimization/70023
5255 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5256 src_regno into account.
5258 2018-02-16 Carl Love <cel@us.ibm.com>
5260 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5261 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5262 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5263 * config/rs6000/rs6000.c: Remove case statements for
5264 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5265 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5266 and P9V_BUILTIN_VEC_VINSERT4B.
5267 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5268 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5269 * config/rs6000/vsx.md:
5270 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5271 vec_insert4b.
5273 2018-02-16 Carl Love <cel@us.ibm.com>
5275 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5276 vec_insert4b.
5277 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5278 definitions.
5279 * config/rs6000/rs6000-c.c: Add the definitions for
5280 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5281 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5282 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5283 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5284 definition for insert4b and define insn *insert3b_internal.
5285 * doc/extend.texi: Add documentation for vec_extract4b.
5287 2018-02-16 Nathan Sidwell <nathan@acm.org>
5289 * doc/extend.texi (Backwards Compatibility): Mention friend
5290 injection. Note for-scope is deprecated.
5291 * doc/invoke.texi (-ffriend-injection): Deprecate.
5293 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5295 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5296 that moved to I2, also allow destinations that are a paradoxical
5297 subreg (instead of a normal reg).
5299 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5301 PR target/83831
5302 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5303 to QImode.
5305 2018-02-16 Richard Biener <rguenther@suse.de>
5307 PR tree-optimization/84037
5308 PR tree-optimization/84016
5309 PR target/82862
5310 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5311 Adjust vec_construct for the fact we need additional higher latency
5312 128bit inserts for AVX256 and AVX512 vector builds.
5313 (ix86_add_stmt_cost): Scale vector construction cost for
5314 elementwise loads.
5316 2018-02-16 Richard Biener <rguenther@suse.de>
5318 PR tree-optimization/84417
5319 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5320 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5321 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5323 2018-02-16 Martin Liska <mliska@suse.cz>
5325 PR sanitizer/84307
5326 * internal-fn.def (ASAN_CHECK): Set proper flags.
5327 (ASAN_MARK): Likewise.
5329 2018-02-16 Julia Koval <julia.koval@intel.com>
5331 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5332 from PTA_CANNONLAKE.
5334 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5336 PR target/84272
5337 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5338 Use ++iter rather than iter++ for std::list iterators.
5339 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5340 defer deleting them until all nodes in the forest are processed. Do
5341 free even leaf nodes. Change to_process into auto_vec.
5343 PR bootstrap/84405
5344 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5345 * vec.h (vec_default_construct): Use memset instead of placement new
5346 if BROKEN_VALUE_INITIALIZATION is defined.
5347 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5348 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5349 is defined.
5351 PR rtl-optimization/83723
5352 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5353 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5354 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5355 recursive calls.
5356 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5357 callers.
5358 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5360 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5362 PR rtl-optimization/81443
5363 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5364 from inner REGs to paradoxical SUBREGs.
5366 2018-02-16 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84399
5369 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5370 For operands we can analyze at their definition make sure we can
5371 analyze them at each use as well.
5373 2018-02-16 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84190
5376 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5377 volatile accesses if the decl isn't volatile.
5379 2018-02-15 Jason Merrill <jason@redhat.com>
5381 PR c++/84314 - ICE with templates and fastcall attribute.
5382 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5383 TYPE_CANONICAL on an existing type.
5385 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5387 PR tree-optimization/84383
5388 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5389 dstoff nor call operand_equal_p if dstbase is NULL.
5391 PR tree-optimization/84334
5392 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5393 also a CONSTANT_CLASS_P, punt.
5395 2018-02-14 Jim Wilson <jimw@sifive.com>
5397 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5398 first SMALL_OPERAND check. New local min_second_step. Move assert
5399 to where locals are set. Add TARGET_RVC support.
5400 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5402 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5404 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5406 2018-02-14 Martin Sebor <msebor@redhat.com>
5408 PR tree-optimization/83698
5409 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5410 arrays constrain the offset range to their bounds.
5411 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5412 (builtin_access::overlap): Avoid setting the size of overlap if it's
5413 already been set.
5414 (maybe_diag_overlap): Also consider arrays when deciding what values
5415 of offsets to include in diagnostics.
5417 2018-02-14 Martin Sebor <msebor@redhat.com>
5419 PR c/84108
5420 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5421 that correspond to the kind of a declaration.
5423 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5425 PR target/83984
5426 * config/pa/pa.md: Load address of PIC label using the linkage table
5427 if the label is nonlocal.
5429 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5431 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5432 warning message if user requests -maltivec=be.
5433 * doc/invoke.texi: Document deprecation of -maltivec=be.
5435 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5437 PR target/84220
5438 * config/rs6000/rs6000-c.c: Update definitions for
5439 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5440 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5442 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5444 PR target/84239
5445 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5446 add _get_ssp intrinsics. Remove argument from
5447 __builtin_ia32_rdssp[d|q].
5448 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5449 * config/i386/i386-builtin.def: Remove argument from
5450 __builtin_ia32_rdssp[d|q].
5451 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5452 ix86_expand_special_args_builtin for _rdssp[d|q].
5453 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5454 Clear register before usage.
5455 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5456 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5458 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5460 PR tree-optimization/84357
5461 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5462 operand 1 of an ARRAY_REF too.
5464 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5466 PR target/83831
5467 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5468 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5469 declarations.
5470 (set_of_reg): New struct.
5471 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5472 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5473 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5474 functions.
5475 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5476 Split into bitclr, bitset, bitinvert patterns if appropriate.
5477 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5478 use rx_fuse_in_memory_bitop.
5479 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5480 to named insn, correct maximum insn length.
5482 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5484 PR target/79242
5485 * machmode.def: Define a complex mode for PARTIAL_INT.
5486 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5487 MODE_PARTIAL_INT.
5488 * doc/rtl.texi: Document CSPImode.
5489 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5490 handling.
5491 (msp430_hard_regno_nregs_with_padding): Likewise.
5493 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5495 PR target/84279
5496 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5498 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5500 PR rtl-optimization/84169
5501 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5502 we generated a parallel as new i3 and we split that to new i2 and i3
5503 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5504 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5505 those to i2, not i1. Partially rewrite this scan code.
5507 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5509 PR c/82210
5510 * stor-layout.c (place_field): For variable length fields, adjust
5511 offset_align afterwards not just based on the field's alignment,
5512 but also on the size.
5514 PR middle-end/84309
5515 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5516 of exps and logs in the use_exp2 case.
5518 2018-02-13 Jeff Law <law@redhat.com>
5520 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5521 entry for "vector".
5523 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5524 ARGS as unused.
5526 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5528 PR debug/84342
5529 PR debug/84319
5530 * common.opt (gas-loc-support, gas-locview-support): New.
5531 (ginline-points, ginternal-reset-location-views): New.
5532 * doc/invoke.texi: Document them. Use @itemx where intended.
5533 (gvariable-location-views): Adjust.
5534 * target.def (reset_location_view): New.
5535 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5536 (TARGET_RESET_LOCATION_VIEW): New.
5537 * doc/tm.texi: Rebuilt.
5538 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5539 (dwarf2out_default_as_locview_support): New.
5540 (output_asm_line_debug_info): Use option variables.
5541 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5542 (output_loc_list): Likewise.
5543 (add_high_low_attributes): Check option variables.
5544 Don't output entry view attribute in strict mode.
5545 (gen_inlined_subroutine_die): Check option variables.
5546 (dwarf2out_inline_entry): Likewise.
5547 (init_sections_and_labels): Likewise.
5548 (dwarf2out_early_finish): Likewise.
5549 (maybe_reset_location_view): New, from...
5550 (dwarf2out_var_location): ... here. Call it.
5551 * debug.h (dwarf2out_default_as_loc_support): Declare.
5552 (dwarf2out_default_as_locview_support): Declare.
5553 * hooks.c (hook_int_rtx_insn_0): New.
5554 * hooks.h (hook_int_rtx_insn_0): Declare.
5555 * toplev.c (process_options): Take -gas-loc-support and
5556 -gas-locview-support from dwarf2out. Enable
5557 -gvariable-location-views by default only with locview
5558 assembler support. Enable -ginternal-reset-location-views by
5559 default only if the target defines the corresponding hook.
5560 Enable -ginline-points by default if location views are
5561 enabled; force it disabled if statement frontiers are
5562 disabled.
5563 * tree-inline.c (expand_call_inline): Check option variables.
5564 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5566 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5568 PR tree-optimization/84321
5569 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5570 handling. Also check whether the anti-range contains any values
5571 that satisfy the mask; switch to a VR_RANGE if not.
5573 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5575 PR sanitizer/84340
5576 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5578 2018-02-13 Martin Jambor <mjambor@suse.cz>
5580 PR c++/83990
5581 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5582 of call statements, also set location of a load to a temporary.
5584 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5586 * config/rl78/rl78.c (add_vector_labels): New function.
5587 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5588 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5589 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5590 which checks that no arguments are passed.
5591 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5592 * doc/extend.texi: Documentation for the new attribute.
5594 2018-02-13 Andreas Schwab <schwab@suse.de>
5596 * config/riscv/linux.h (CPP_SPEC): Define.
5598 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5600 PR target/84335
5601 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5602 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5603 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5604 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5605 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5606 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5607 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5608 temporarily for AES and PCLMUL builtins.
5610 PR tree-optimization/84339
5611 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5612 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5613 Formatting fixes.
5615 PR middle-end/84309
5616 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5617 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5618 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5619 inline function.
5620 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5621 inline function.
5622 * omp-simd-clone.h: New file.
5623 * omp-simd-clone.c: Include omp-simd-clone.h.
5624 (expand_simd_clones): No longer static.
5625 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5626 cgraph.h and omp-simd-clone.h.
5627 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5628 (vect_recog_widen_shift_pattern): Formatting fix.
5629 (vect_pattern_recog_1): Don't check optab for calls.
5631 PR target/84336
5632 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5633 operands[2] into a REG before using gen_lowpart on it.
5635 2018-02-12 Jeff Law <law@redhat.com>
5637 PR target/83760
5638 * config/sh/sh.c (find_barrier): Consider a sibling call
5639 a barrier as well.
5641 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5642 successfully back substituting a reg.
5644 2018-02-12 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/84037
5647 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5648 parameter, move visited init to caller.
5649 (vect_slp_analyze_operations): Separate cost from validity
5650 check, initialize visited once for all instances.
5651 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5652 for all instances.
5653 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5654 out an assert.
5655 (vect_model_promotion_demotion_cost): Likewise.
5656 (vectorizable_bswap): Guard cost modeling with !slp_node
5657 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5658 SLP stmts.
5659 (vectorizable_call): Likewise.
5660 (vectorizable_conversion): Likewise.
5661 (vectorizable_assignment): Likewise.
5662 (vectorizable_shift): Likewise.
5663 (vectorizable_operation): Likewise.
5664 (vectorizable_store): Likewise.
5665 (vectorizable_load): Likewise.
5666 (vectorizable_condition): Likewise.
5667 (vectorizable_comparison): Likewise.
5669 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5671 PR sanitizer/84307
5672 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5673 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5674 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5676 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5678 PR middle-end/83665
5679 * params.def (inline-min-speedup): Increase from 8 to 15.
5680 (max-inline-insns-auto): Decrease from 40 to 30.
5681 * ipa-split.c (consider_split): Add some buffer for function to
5682 be considered inlining candidate.
5683 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5684 default values.
5686 2018-02-12 Richard Biener <rguenther@suse.de>
5688 PR tree-optimization/84037
5689 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5690 matched stmts if we cannot swap the non-matched ones.
5692 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5694 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5695 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5696 _mm_maskz_scalef_round_ss): New intrinsics.
5697 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5698 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5699 __builtin_ia32_scalefss_round): Remove.
5700 (__builtin_ia32_scalefsd_mask_round,
5701 __builtin_ia32_scalefss_mask_round): New intrinsics.
5702 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5703 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5704 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5705 "<round_constraint>")): Changed to ...
5706 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5707 "<round_scalar_constraint>")): ... this.
5708 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5709 %0, %1, %2<round_op3>}"): Changed to ...
5710 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5711 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5712 %2<round_scalar_mask_op3>}"): ... this.
5713 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5715 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5717 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5718 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5719 (_mm_maskz_sqrt_round_ss): New intrinsics.
5720 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5721 (__builtin_ia32_sqrtsd_mask_round)
5722 (__builtin_ia32_sqrtss_mask_round): New builtins.
5723 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5724 (__builtin_ia32_sqrtss_round): Remove.
5725 (__builtin_ia32_sqrtsd_mask_round)
5726 (__builtin_ia32_sqrtss_mask_round): New builtins.
5727 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5728 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5729 ((match_operand:VF_128 1 "vector_operand"
5730 "xBm,<round_constraint>")): Changed to ...
5731 ((match_operand:VF_128 1 "vector_operand"
5732 "xBm,<round_scalar_constraint>")): ... this.
5733 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5734 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5735 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5736 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5737 %<iptr>1<round_scalar_mask_op3>}): ... this.
5738 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5739 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5741 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5743 PR target/84266
5744 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5745 Cast vec_cmpeq result to correct type.
5746 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5747 Cast vec_cmpgt result to correct type.
5749 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5751 * final.c (final_scan_insn_1): Renamed from...
5752 (final_scan_insn): ... this. New wrapper, to recover
5753 seen from the outermost call in recursive ones.
5754 * config/sparc/sparc.c (output_return): Drop seen from call.
5755 (output_sibcall): Likewise.
5756 * config/visium/visium.c (output_branch): Likewise.
5758 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5760 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5761 function label.
5763 2018-02-10 Alan Modra <amodra@gmail.com>
5765 PR target/84300
5766 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5767 Specify LR as an input.
5769 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5771 PR sanitizer/83987
5772 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5773 remove_member_access_dummy_vars): New functions.
5774 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5775 lower_omp_1, execute_lower_omp): Use them.
5777 PR rtl-optimization/84308
5778 * shrink-wrap.c (spread_components): Release todo vector.
5780 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5782 PR rtl-optimization/57193
5783 * ira-color.c (struct allocno_color_data): Add member
5784 conflict_allocno_hard_prefs.
5785 (update_conflict_allocno_hard_prefs): New.
5786 (bucket_allocno_compare_func): Add a preference based on
5787 conflict_allocno_hard_prefs.
5788 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5789 (color_allocnos): Remove a dead code. Initiate
5790 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5792 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5794 PR target/84226
5795 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5796 constraint from =wa to wa. Avoid a subreg on the output operand,
5797 instead use a pseudo and subreg it in a move.
5798 (p9_xxbrd_<mode>): Changed to ...
5799 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5800 (p9_xxbrd_v2df): New expander.
5801 (p9_xxbrw_<mode>): Changed to ...
5802 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5803 (p9_xxbrw_v4sf): New expander.
5805 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5807 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5809 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5811 PR target/83926
5812 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5813 multiply in 32-bit mode.
5814 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5815 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5816 mode.
5818 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5820 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5821 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5822 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5823 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5825 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5827 PR lto/84213
5828 * dwarf2out.c (is_trivial_indirect_ref): New function.
5829 (dwarf2out_late_global_decl): Do not generate a location
5830 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5831 and that are not defined in the current unit.
5833 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5835 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5836 instead of a libcall for UNORDERED.
5838 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5840 PR target/82641
5841 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5842 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5844 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5846 PR target/PR84295
5847 * config/s390/s390.c (s390_set_current_function): Invoke
5848 s390_indirect_branch_settings also if fndecl didn't change.
5850 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5852 * config/rs6000/rs6000.md (blockage): Set length to zero.
5854 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5856 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5858 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5860 PR sanitizer/84285
5861 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5862 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5863 -static-lib*san.
5865 PR debug/84252
5866 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5867 PARALLEL incoming that failed vt_get_decl_and_offset check.
5869 PR middle-end/84237
5870 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5871 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5872 TREE_READONLY bit.
5873 (get_variable_section): For decls in named .bss* sections pass true as
5874 second argument to bss_initializer_p.
5876 2018-02-09 Marek Polacek <polacek@redhat.com>
5877 Jakub Jelinek <jakub@redhat.com>
5879 PR c++/83659
5880 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5881 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5882 Sync some changes from cxx_fold_indirect_ref.
5884 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5886 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5887 markers.
5888 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5889 (BLOCK_INLINE_ENTRY_LABEL): New.
5890 (dwarf2out_var_location): Disregard inline entry markers.
5891 (inline_entry_data): New struct.
5892 (inline_entry_data_hasher): New hashtable type.
5893 (inline_entry_data_hasher::hash): New.
5894 (inline_entry_data_hasher::equal): New.
5895 (inline_entry_data_table): New variable.
5896 (add_high_low_attributes): Add DW_AT_entry_pc and
5897 DW_AT_GNU_entry_view attributes if a pending entry is found
5898 in inline_entry_data_table. Add old entry_pc attribute only
5899 if debug nonbinding markers are disabled.
5900 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5901 markers are enabled.
5902 (block_within_block_p, dwarf2out_inline_entry): New.
5903 (dwarf2out_finish): Check that no entries remained in
5904 inline_entry_data_table.
5905 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5906 (final_scan_insn, notice_source_line): Likewise.
5907 (rest_of_clean_state): Skip inline entry markers.
5908 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5909 markers.
5910 * gimple.c (gimple_build_debug_inline_entry): New.
5911 * gimple.h (enum gimple_debug_subcode): Add
5912 GIMPLE_DEBUG_INLINE_ENTRY.
5913 (gimple_build_debug_inline_entry): Declare.
5914 (gimple_debug_inline_entry_p): New.
5915 (gimple_debug_nonbind_marker_p): Adjust.
5916 * insn-notes.def (INLINE_ENTRY): New.
5917 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5918 inline entry marker notes.
5919 (print_insn): Likewise.
5920 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5921 (INSN_DEBUG_MARKER_KIND): Likewise.
5922 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5923 * tree-inline.c (expand_call_inline): Build and insert
5924 debug_inline_entry stmt.
5925 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5926 inline entry blocks early, if nonbind markers are enabled.
5927 (dump_scope_block): Dump fragment info.
5928 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5929 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5930 (gimple_build_debug_inline_entry): New.
5931 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5932 Enable/disable inline entry points too.
5933 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5934 (DEBUG_INSN): Describe inline entry markers.
5936 * common.opt (gvariable-location-views): New.
5937 (gvariable-location-views=incompat5): New.
5938 * config.in: Rebuilt.
5939 * configure: Rebuilt.
5940 * configure.ac: Test assembler for view support.
5941 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5942 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5943 * dwarf2out.c (var_loc_view): New typedef.
5944 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5945 (dwarf2out_locviews_in_attribute): New.
5946 (dwarf2out_locviews_in_loclist): New.
5947 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5948 (enum dw_line_info_opcode): Add LI_adv_address.
5949 (struct dw_line_info_table): Add view.
5950 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5951 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5952 (zero_view_p): New variable.
5953 (ZERO_VIEW_P): New macro.
5954 (output_asm_line_debug_info): New.
5955 (struct var_loc_node): Add view.
5956 (add_AT_view_list, AT_loc_list): New.
5957 (add_var_loc_to_decl): Add view param. Test it against last.
5958 (new_loc_list): Add view params. Record them.
5959 (AT_loc_list_ptr): Handle loc and view lists.
5960 (view_list_to_loc_list_val_node): New.
5961 (print_dw_val): Handle dw_val_class_view_list.
5962 (size_of_die): Likewise.
5963 (value_format): Likewise.
5964 (loc_list_has_views): New.
5965 (gen_llsym): Set vl_symbol too.
5966 (maybe_gen_llsym, skip_loc_list_entry): New.
5967 (dwarf2out_maybe_output_loclist_view_pair): New.
5968 (output_loc_list): Output view list or entries too.
5969 (output_view_list_offset): New.
5970 (output_die): Handle dw_val_class_view_list.
5971 (output_dwarf_version): New.
5972 (output_compilation_unit_header): Use it.
5973 (output_skeleton_debug_sections): Likewise.
5974 (output_rnglists, output_line_info): Likewise.
5975 (output_pubnames, output_aranges): Update version comments.
5976 (output_one_line_info_table): Output view numbers in asm comments.
5977 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5978 Call maybe_gen_llsym.
5979 (loc_list_from_tree_1): Adjust.
5980 (add_AT_location_description): Create view list attribute if
5981 needed, check it's absent otherwise.
5982 (convert_cfa_to_fb_loc_list): Adjust.
5983 (maybe_emit_file): Call output_asm_line_debug_info for test.
5984 (dwarf2out_var_location): Reset views as needed. Precompute
5985 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5986 attribute. Set view.
5987 (new_line_info_table): Reset next view.
5988 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5989 (dwarf2out_source_line): Likewise. Output view resets and labels to
5990 the assembler, or select appropriate line info opcodes.
5991 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5992 (optimize_string_length): Catch it. Adjust.
5993 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5994 dw_val_class_view_list, and remove it if no longer needed.
5995 (hash_loc_list): Hash view numbers.
5996 (loc_list_hasher::equal): Compare them.
5997 (optimize_location_lists): Check whether a view list symbol is
5998 needed, and whether the locview attribute is present, and
5999 whether they match. Remove the locview attribute if no longer
6000 needed.
6001 (index_location_lists): Call skip_loc_list_entry for test.
6002 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6003 Use output_dwarf_version.
6004 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6005 (struct dw_val_node): Add val_view_list.
6006 * final.c (SEEN_NEXT_VIEW): New.
6007 (set_next_view_needed): New.
6008 (clear_next_view_needed): New.
6009 (maybe_output_next_view): New.
6010 (final_start_function): Rename to...
6011 (final_start_function_1): ... this. Take pointer to FIRST,
6012 add SEEN parameter. Emit param bindings in the initial view.
6013 (final_start_function): Reintroduce SEEN-less interface.
6014 (final): Rename to...
6015 (final_1): ... this. Take SEEN parameter. Output final pending
6016 next view at the end.
6017 (final): Reintroduce seen-less interface.
6018 (final_scan_insn): Output pending next view before switching
6019 sections or ending a block. Mark the next view as needed when
6020 outputting variable locations. Notify debug backend of section
6021 changes, and of location view changes.
6022 (rest_of_handle_final): Adjust.
6023 * toplev.c (process_options): Autodetect value for debug variable
6024 location views option. Warn on incompat5 without -gdwarf-5.
6025 * doc/invoke.texi (gvariable-location-views): New.
6026 (gvariable-location-views=incompat5): New.
6027 (gno-variable-location-views): New.
6029 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6031 PR tree-optimization/84136
6032 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6033 that the result of find_edge is non-NULL.
6035 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6037 PR target/83008
6038 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6039 storing integer register in SImode. Fix cost of 256 and 512
6040 byte aligned SSE register store.
6042 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6044 * config/i386/i386.c (ix86_multiplication_cost): Fix
6045 multiplication cost for TARGET_AVX512DQ.
6047 2018-02-08 Marek Polacek <polacek@redhat.com>
6049 PR tree-optimization/84238
6050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6051 get_range_strlen.
6053 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6055 PR tree-optimization/84265
6056 * tree-vect-stmts.c (vectorizable_store): Don't treat
6057 VMAT_CONTIGUOUS accesses as grouped.
6058 (vectorizable_load): Likewise.
6060 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6062 PR tree-optimization/81635
6063 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6064 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6065 (test_round_for_mask): New functions.
6066 (wide_int_cc_tests): Call test_round_for_mask.
6067 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6068 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6069 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6070 range returned by get_range_info.
6072 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6074 PR ipa/81360
6075 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6076 * symtab.c: Include builtins.h
6077 (symtab_node::output_to_lto_symbol_table_p): Move here
6078 from lto-streamer-out.c:output_symbol_p.
6079 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6080 (output_symbol_p): Move all logic to symtab.c
6081 (produce_symtab): Update.
6083 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6085 * config/s390/s390-opts.h (enum indirect_branch): Define.
6086 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6087 (s390_indirect_branch_via_thunk)
6088 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6089 (enum s390_indirect_branch_type): Define.
6090 * config/s390/s390.c (struct s390_frame_layout, struct
6091 machine_function): Remove.
6092 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6093 (indirect_branch_table_label_no, indirect_branch_table_name):
6094 Define variables.
6095 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6096 (enum s390_indirect_branch_option): Define.
6097 (s390_return_addr_from_memory): New function.
6098 (s390_handle_string_attribute): New function.
6099 (s390_attribute_table): Add new attribute handler.
6100 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6101 (s390_indirect_branch_via_thunk): New function.
6102 (s390_indirect_branch_via_inline_thunk): New function.
6103 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6104 sibling call optimization for non z10 compiles.
6105 (s390_emit_call): Force indirect branch target to be a single
6106 register. Add r1 clobber for non-z10 compiles.
6107 (s390_emit_epilogue): Emit return jump via return_use expander.
6108 (s390_reorg): Handle JUMP_INSNs as execute targets.
6109 (s390_option_override_internal): Perform validity checks for the
6110 new command line options.
6111 (s390_indirect_branch_attrvalue): New function.
6112 (s390_indirect_branch_settings): New function.
6113 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6114 (s390_output_indirect_thunk_function): New function.
6115 (s390_code_end): Implement target hook.
6116 (s390_case_values_threshold): Implement target hook.
6117 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6118 macros.
6119 * config/s390/s390.h (struct s390_frame_layout)
6120 (struct machine_function): Move here from s390.c.
6121 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6122 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6123 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6124 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6125 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6126 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6127 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6128 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6129 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6130 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6131 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6132 (mnemonic attribute): Add values which aren't recognized
6133 automatically.
6134 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6135 pattern for branch conversion. Fix mnemonic attribute.
6136 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6137 indirect branch via thunk if requested.
6138 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6139 ("*indirect_jump"): Disable for branch conversion using out of
6140 line thunks.
6141 ("indirect_jump_via_thunk<mode>_z10")
6142 ("indirect_jump_via_thunk<mode>")
6143 ("indirect_jump_via_inlinethunk<mode>_z10")
6144 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6145 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6146 ("casesi_jump_via_inlinethunk<mode>_z10")
6147 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6148 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6149 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6150 ("*indirect2_jump"): Disable for branch conversion.
6151 ("casesi_jump"): Turn into expander and expand patterns for branch
6152 conversion.
6153 ("return_use"): New expander.
6154 ("*return"): Emit return via thunk and rename it to ...
6155 ("*return<mode>"): ... this one.
6156 * config/s390/s390.opt: Add new options and and enum for the
6157 option values.
6159 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6161 * lra-constraints.c (match_reload): Unconditionally use
6162 gen_lowpart_SUBREG, rather than selecting between that
6163 and equivalent gen_rtx_SUBREG code.
6165 2018-02-08 Richard Biener <rguenther@suse.de>
6167 PR tree-optimization/84233
6168 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6169 changed flag instead of boguously re-using phi_inserted.
6171 2018-02-08 Martin Jambor <mjambor@suse.cz>
6173 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6174 static local variables.
6176 2018-02-08 Richard Biener <rguenther@suse.de>
6178 PR tree-optimization/84278
6179 * tree-vect-stmts.c (vectorizable_store): When looking for
6180 smaller vector types to perform grouped strided loads/stores
6181 make sure the mode is supported by the target.
6182 (vectorizable_load): Likewise.
6184 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6186 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6187 Increase LDP/STP opportunities by adding adjacent callee-saves.
6189 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6191 PR rtl-optimization/84068
6192 PR rtl-optimization/83459
6193 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6195 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6197 PR tree-optimization/84224
6198 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6199 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6200 non-zero arguments.
6202 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6204 PR target/84113
6205 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6206 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6207 count, remove one USE.
6209 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6211 * doc/install.texi (Configuration): Document the
6212 --with-long-double-format={ibm,ieee} PowerPC configuration
6213 options.
6215 PR target/84154
6216 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6217 Convert from define_expand to be define_insn_and_split. Rework
6218 float/double/_Float128 conversions to QI/HI/SImode to work with
6219 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6220 conversions to QI/HImode types did a store and then a load to
6221 truncate the value. For conversions to VSX registers, don't split
6222 the insn, instead emit the code directly. Use the code iterator
6223 any_fix to combine signed and unsigned conversions.
6224 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6225 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6226 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6227 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6228 (fix_<mode>di2_hw): Likewise.
6229 (fixuns_<mode>di2_hw): Likewise.
6230 (fix_<mode>si2_hw): Likewise.
6231 (fixuns_<mode>si2_hw): Likewise.
6232 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6233 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6234 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6235 fix<uns>_trunc<SFDF:mode>si2_p8.
6236 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6237 used.
6238 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6239 (fix<uns>_<mode>_mem): Likewise.
6240 (fctiw<u>z_<mode>_mem): Likewise.
6241 (fix<uns>_<mode>_mem): Likewise.
6242 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6243 the register allocator from doing a direct move to the GPRs to do
6244 a store, and instead use the ISA 3.0 store byte/half-word from
6245 vector register instruction. For IEEE 128-bit floating point,
6246 also optimize stores of 32-bit ints.
6247 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6249 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6251 * genextract.c (push_pathstr_operand): New function to support
6252 [a-zA-Z].
6253 (walk_rtx): Call push_pathstr_operand.
6254 (print_path): Support [a-zA-Z].
6256 2018-02-07 Richard Biener <rguenther@suse.de>
6258 PR tree-optimization/84037
6259 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6260 (cse_and_gimplify_to_preheader): Declare.
6261 (vect_get_place_in_interleaving_chain): Likewise.
6262 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6263 ivexpr_map.
6264 (_loop_vec_info::~_loop_vec_info): Delete it.
6265 (cse_and_gimplify_to_preheader): New function.
6266 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6267 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6268 (vectorizable_load): Likewise. For grouped stores always base
6269 the IV on the first element.
6270 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6271 condition before gimplifying.
6273 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6275 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6276 *DIV_EXPR and *MOD_EXPR.
6278 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6280 PR target/84248
6281 * config/i386/i386.c (ix86_option_override_internal): Mask out
6282 the CF_SET bit when checking -fcf-protection.
6284 2018-02-07 Tom de Vries <tom@codesourcery.com>
6286 PR libgomp/84217
6287 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6288 enough.
6290 2018-02-07 Richard Biener <rguenther@suse.de>
6292 PR tree-optimization/84204
6293 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6294 this place.
6296 PR tree-optimization/84205
6297 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6298 special-case isl_ast_op_zdiv_r.
6300 PR tree-optimization/84223
6301 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6302 Only add conditions from within the region.
6303 (gather_bbs::after_dom_children): Adjust.
6305 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6307 PR target/84209
6308 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6309 * config/avr/avr.md: Only post-reload split REG-REG moves if
6310 either register is GENERAL_REG_P.
6312 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6314 PR tree-optimization/84235
6315 * tree-ssa-scopedtables.c
6316 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6317 if the subtraction is performed in floating point type where NaNs are
6318 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6319 build 1. Formatting fix.
6321 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6323 PR target/84146
6324 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6325 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6326 and skip it regardless of bb boundaries. Use CALL_P macro,
6327 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6328 unnecessarily, formatting fix.
6330 2018-02-06 Michael Collison <michael.collison@arm.com>
6332 * config/arm/thumb2.md:
6333 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6334 (*thumb_mov_notscc): Ditto.
6336 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6338 PR target/84154
6339 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6340 unsigned_fix, not "s".
6342 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6344 * configure.ac (gcc_fn_eh_frame_ro): New function.
6345 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6346 correct .eh_frame permissions.
6347 * configure: Regenerate.
6349 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6351 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6352 irrelevant options.
6354 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6356 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6357 Display warning message for -mno-speculate-indirect-jumps.
6359 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6361 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6362 Undocumented.
6363 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6365 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6367 PR tree-optimization/84225
6368 * tree-eh.c (find_trapping_overflow): Only call
6369 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6371 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6373 PR target/84145
6374 * config/i386/i386.c: Reimplement the check of possible options
6375 -mibt/-mshstk conbination. Change error messages.
6376 * doc/invoke.texi: Fix a typo: remove extra '='.
6378 2018-02-06 Marek Polacek <polacek@redhat.com>
6380 PR tree-optimization/84228
6381 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6383 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6385 PR target/82641
6386 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6387 emitted arch directives.
6388 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6389 __ARM_FEATURE_COPROC before changing architectures.
6391 2018-02-06 Richard Biener <rguenther@suse.de>
6393 * config/i386/i386.c (print_reg): Fix typo.
6394 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6396 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6398 * configure: Regenerate.
6400 2018-02-05 Martin Sebor <msebor@redhat.com>
6402 PR tree-optimization/83369
6403 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6404 inlining context.
6406 2018-02-05 Martin Liska <mliska@suse.cz>
6408 * doc/invoke.texi: Cherry-pick upstream r323995.
6410 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6412 * ira.c (ira_init_register_move_cost): Adjust comment.
6414 2018-02-05 Martin Liska <mliska@suse.cz>
6416 PR gcov-profile/84137
6417 * doc/gcov.texi: Fix typo in documentation.
6419 2018-02-05 Martin Liska <mliska@suse.cz>
6421 PR gcov-profile/83879
6422 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6423 using dlopen functionality.
6425 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6427 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6428 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6429 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6430 _mm_maskz_range_round_ss): New intrinsics.
6431 (__builtin_ia32_rangesd128_round)
6432 (__builtin_ia32_rangess128_round): Remove.
6433 (__builtin_ia32_rangesd128_mask_round,
6434 __builtin_ia32_rangess128_mask_round): New builtins.
6435 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6436 __builtin_ia32_rangess128_round): Remove.
6437 (__builtin_ia32_rangesd128_mask_round,
6438 __builtin_ia32_rangess128_mask_round): New builtins.
6439 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6440 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6441 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6442 "<round_saeonly_constraint>")): Changed to ...
6443 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6444 "<round_saeonly_scalar_constraint>")): ... this.
6445 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6446 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6447 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6448 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6449 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6451 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6453 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6454 options.
6455 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6456 Remove all values except native, 8540 and 8548.
6458 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6460 * config/i386/i386.c (ix86_output_function_return): Pass
6461 INVALID_REGNUM, instead of -1, as invalid register number to
6462 indirect_thunk_name and output_indirect_thunk.
6464 2018-02-02 Julia Koval <julia.koval@intel.com>
6466 * config.gcc: Add -march=icelake.
6467 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6468 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6469 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6470 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6471 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6472 (processor_target_table): Add icelake.
6473 (ix86_option_override_internal): Handle new PTAs.
6474 (get_builtin_code_for_version): Handle icelake.
6475 (M_INTEL_COREI7_ICELAKE): New.
6476 (fold_builtin_cpu): Handle icelake.
6477 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6478 * doc/invoke.texi: Add -march=icelake.
6480 2018-02-02 Julia Koval <julia.koval@intel.com>
6482 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6483 to wide_int_bitmask.
6484 * wide-int-bitmask.h: New.
6486 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6488 PR target/84066
6489 * config/i386/i386.md: Replace Pmode with word_mode in
6490 builtin_setjmp_setup and builtin_longjmp to support x32.
6492 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6494 PR target/56010
6495 PR target/83743
6496 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6497 #include "opts.h".
6498 (rs6000_supported_cpu_names): New static variable.
6499 (linux_cpu_translation_table): Likewise.
6500 (elf_platform) <cpu>: Define new static variable and use it.
6501 Translate kernel AT_PLATFORM name to canonical name if needed.
6502 Error if platform name is unknown.
6504 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6506 PR target/84089
6507 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6509 2018-02-01 Jeff Law <law@redhat.com>
6511 PR target/84128
6512 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6513 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6514 the scratch if RELEASE_VIA_POP is false.
6515 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6516 If we have to save a temporary register, decrement SIZE appropriately.
6517 Pass new arguments to release_scratch_register_on_entry.
6518 (ix86_adjust_stack_and_probe): Likewise.
6519 (ix86_emit_probe_stack_range): Pass new arguments to
6520 release_scratch_register_on_entry.
6522 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6524 PR rtl-optimization/84157
6525 * combine.c (change_zero_ext): Use REG_P predicate in
6526 front of HARD_REGISTER_P predicate.
6528 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6530 * config/avr/avr.c (avr_option_override): Move disabling of
6531 -fdelete-null-pointer-checks to...
6532 * common/config/avr/avr-common.c (avr_option_optimization_table):
6533 ...here.
6535 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6537 PR tree-optimization/81635
6538 * tree-data-ref.c (split_constant_offset_1): For types that
6539 wrap on overflow, try to use range info to prove that wrapping
6540 cannot occur.
6542 2018-02-01 Renlin Li <renlin.li@arm.com>
6544 PR target/83370
6545 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6546 TAILCALL_ADDR_REGS.
6547 (aarch64_register_move_cost): Likewise.
6548 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6549 TAILCALL_ADDR_REGS.
6550 (REG_CLASS_NAMES): Likewise.
6551 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6552 TAILCALL_ADDR_REGS. Remove IP registers.
6553 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6555 2018-02-01 Richard Biener <rguenther@suse.de>
6557 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6558 for specifying RPO order and allow NULL for that.
6559 * domwalk.c (dom_walker::dom_walker): Likewise.
6560 (dom_walker::walk): Handle NULL RPO order.
6561 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6562 in RPO order.
6563 (rewrite_update_dom_walker): Likewise.
6564 (mark_def_dom_walker): Likewise.
6566 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6568 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6569 (aarch64_maybe_expand_sve_subreg_move): Declare.
6570 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6571 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6572 predicate.
6573 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6574 that are semantically a reverse operation.
6575 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6576 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6577 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6578 functions.
6579 (aarch64_can_change_mode_class): For big-endian, forbid changes
6580 between two SVE modes if they have different element sizes.
6582 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6584 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6585 the TImode handling for big-endian targets.
6587 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6589 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6590 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6591 not just bytes.
6592 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6593 Remove BSWAP handing for big-endian targets and use the form of
6594 LD1RQ appropariate for the mode.
6596 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6598 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6599 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6600 duplicated element.
6602 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6604 PR tearget/83845
6605 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6606 check for operands that need to go through aarch64_sve_reload_be.
6608 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6610 PR tree-optimization/81661
6611 PR tree-optimization/84117
6612 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6613 * tree-eh.c: Include gimplify.h.
6614 (find_trapping_overflow, replace_trapping_overflow,
6615 rewrite_to_non_trapping_overflow): New functions.
6616 * tree-vect-loop.c: Include tree-eh.h.
6617 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6618 * tree-data-ref.c: Include tree-eh.h.
6619 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6621 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6623 PR rtl-optimization/84123
6624 * combine.c (change_zero_ext): Check if hard register satisfies
6625 can_change_dest_mode before calling gen_lowpart_SUBREG.
6627 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6629 PR target/82444
6630 * ira.c (ira_init_register_move_cost): Remove assert.
6632 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6634 PR rtl-optimization/84071
6635 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6636 * doc/tm.texi: Regenerate.
6638 2018-01-31 Richard Biener <rguenther@suse.de>
6640 PR tree-optimization/84132
6641 * tree-data-ref.c (analyze_miv_subscript): Properly
6642 check whether evolution_function_is_affine_multivariate_p
6643 before calling gcd_of_steps_may_divide_p.
6645 2018-01-31 Julia Koval <julia.koval@intel.com>
6647 PR target/83618
6648 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6649 * config/i386/i386.md (rdpid_rex64) New.
6650 (rdpid): Make 32bit only.
6652 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6654 PR lto/84105
6655 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6656 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6658 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6660 Revert
6661 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6663 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6665 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6667 PR rtl-optimization/84071
6668 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6669 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6671 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6673 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6674 (arc_attribute_table): Add 'aux' attribute.
6675 (arc_in_small_data_p): Consider aux like variables.
6676 (arc_is_aux_reg_p): New function.
6677 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6678 (arc_get_aux_arg): New function.
6679 (prepare_move_operands): Handle aux-register access.
6680 (arc_handle_aux_attribute): New function.
6681 * doc/extend.texi (ARC Variable attributes): Add subsection.
6683 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6685 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6686 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6687 (arc_attribute_table): Add 'uncached' attribute.
6688 (arc_print_operand): Print '.di' flag for uncached memory
6689 accesses.
6690 (arc_in_small_data_p): Do not consider for small data the uncached
6691 types.
6692 (arc_is_uncached_mem_p): New function.
6693 * config/arc/predicates.md (compact_store_memory_operand): Check
6694 for uncached memory accesses.
6695 (nonvol_nonimm_operand): Likewise.
6696 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6698 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6700 PR c/84100
6701 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6702 falign-loops=): Add Optimization flag.
6704 2018-01-30 Jeff Law <law@redhat.com>
6706 PR target/84064
6707 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6708 INT_REGISTERS_SAVED. Check it prior to calling
6709 get_scratch_register_on_entry.
6710 (ix86_adjust_stack_and_probe): Similarly.
6711 (ix86_emit_probe_stack_range): Similarly.
6712 (ix86_expand_prologue): Corresponding changes.
6714 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6716 PR target/40411
6717 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6718 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6720 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6722 PR target/84112
6723 * lra-constraints.c (curr_insn_transform): Process AND in the
6724 address.
6726 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6728 PR rtl-optimization/83986
6729 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6730 dependence against last_pending_memory_flush in addition to
6731 pending_jump_insns.
6733 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6735 PR tree-optimization/81611
6736 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6737 copies.
6739 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6741 PR target/83758
6742 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6743 a reg rtx.
6745 2018-01-30 Richard Biener <rguenther@suse.de>
6746 Jakub Jelinek <jakub@redhat.com>
6748 PR tree-optimization/84111
6749 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6750 inner loops added during recursion, as they don't have up-to-date
6751 SSA form.
6753 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6755 PR ipa/81360
6756 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6757 (can_inline_edge_by_limits_p): ... here.
6758 (can_early_inline_edge_p, check_callers,
6759 update_caller_keys, update_callee_keys, recursive_inlining,
6760 add_new_edges_to_heap, speculation_useful_p,
6761 inline_small_functions,
6762 inline_small_functions, flatten_function,
6763 inline_to_all_callers_1): Update.
6765 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6767 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6768 zeros correctly.
6770 2018-01-30 Richard Biener <rguenther@suse.de>
6772 PR tree-optimization/83008
6773 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6774 invariant and constant vector uses in stmts when they need
6775 more than one stmt.
6777 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6779 PR bootstrap/84017
6780 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6781 * configure: Regenerate.
6783 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6785 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6786 pattern.
6787 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6788 Use gen_rtx_REG rather than gen_lowpart.
6790 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6792 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6793 rather than 0 when creating partial subregs.
6795 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6797 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6798 of usage.
6800 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6802 PR target/81550
6803 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6804 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6805 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6806 flags. This restores the settings used before the 2017-07-24.
6807 Turning off pre increment/decrement/modify allows IVOPTS to
6808 optimize DF/SF loops where the index is an int.
6810 2018-01-29 Richard Biener <rguenther@suse.de>
6811 Kelvin Nilsen <kelvin@gcc.gnu.org>
6813 PR bootstrap/80867
6814 * tree-vect-stmts.c (vectorizable_call): Don't call
6815 targetm.vectorize_builtin_md_vectorized_function if callee is
6816 NULL.
6818 2018-01-22 Carl Love <cel@us.ibm.com>
6820 * doc/extend.tex: Fix typo in second arg in
6821 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6823 2018-01-29 Richard Biener <rguenther@suse.de>
6825 PR tree-optimization/84086
6826 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6827 (flush_ssaname_freelist): When SSA names were released reset
6828 the SCEV hash table.
6830 2018-01-29 Richard Biener <rguenther@suse.de>
6832 PR tree-optimization/84057
6833 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6834 removed paths when removing edges.
6836 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6838 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6839 -mfunction-return=@var{choice}.
6841 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6843 PR diagnostic/84034
6844 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6845 Handle CR like TAB.
6846 (layout::print_source_line): Likewise.
6847 (test_get_line_width_without_trailing_whitespace): Add test cases.
6849 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6851 PR middle-end/84040
6852 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6853 debug insns.
6855 2018-01-26 Jim Wilson <jimw@sifive.com>
6857 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6859 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6860 specified.
6862 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6864 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6865 and CMP + SUB-immediate -> SUBS.
6867 2018-01-26 Martin Sebor <msebor@redhat.com>
6869 PR tree-optimization/83896
6870 * tree-ssa-strlen.c (get_string_len): Rename...
6871 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6872 Avoid assuming length is constant.
6873 (handle_char_store): Use HOST_WIDE_INT for string length.
6875 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6877 PR target/81763
6878 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6879 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6881 2018-01-26 Richard Biener <rguenther@suse.de>
6883 PR rtl-optimization/84003
6884 * dse.c (record_store): Only record redundant stores when
6885 the earlier store aliases at least all accesses the later one does.
6887 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6889 PR rtl-optimization/83985
6890 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6891 REG_CFA_RESTORE insns.
6892 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6893 REG_CFA_RESTORE insns here.
6895 PR c/83989
6896 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6897 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6899 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6901 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6902 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6903 (arc_init): Likewise.
6904 (arc_override_options): Likewise.
6905 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6906 value.
6907 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6908 support.
6909 * config/arc/arc.h (TARGET_DBNZ): Define.
6910 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6911 properly set the tune attribute.
6912 (dbnz): Use TARGET_DBNZ guard.
6913 * config/arc/arc.opt (mtune): Add core3 option.
6915 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6917 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6918 recognize new pic like addresses.
6919 (arc_delegitimize_address): Clean up.
6921 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6923 * config/arc/arc-arches.def: Option mrf16 valid for all
6924 architectures.
6925 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6926 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6927 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6928 * config/arc/arc-tables.opt: Regenerate.
6929 * config/arc/arc.c (arc_conditional_register_usage): Handle
6930 reduced register file case.
6931 (arc_file_start): Set must have build attributes.
6932 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6933 mrf16 option value.
6934 * config/arc/arc.opt (mrf16): Add new option.
6935 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6936 * config/arc/genmultilib.awk: Handle new mrf16 option.
6937 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6938 * config/arc/t-multilib: Regenerate.
6939 * doc/invoke.texi (ARC Options): Document mrf16 option.
6941 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6943 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6944 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6945 (arc_attribute_table): Add 'secure_call' attribute.
6946 (arc_print_operand): Print secure call operand.
6947 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6948 secure.
6949 (arc_is_secure_call_p): New function. * config/arc/arc.md
6950 (call_i): Add support for sjli instruction.
6951 (call_value_i): Likewise.
6952 * config/arc/constraints.md (Csc): New constraint.
6954 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6955 John Eric Martin <John.Martin@emmicro-us.com>
6957 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6958 * config/arc/arc.c (_arc_jli_section): New struct.
6959 (arc_jli_section): New type.
6960 (rc_jli_sections): New static variable.
6961 (arc_handle_jli_attribute): New function.
6962 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6963 (arc_file_end): New function.
6964 (TARGET_ASM_FILE_END): Define.
6965 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6966 (arc_add_jli_section): New function.
6967 (jli_call_scan): Likewise.
6968 (arc_reorg): Call jli_call_scan.
6969 (arc_output_addsi): Remove 'S' from printing asm operand.
6970 (arc_is_jli_call_p): New function.
6971 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6972 operand.
6973 (movhi_insn): Likewise.
6974 (movsi_insn): Likewise.
6975 (movsi_set_cc_insn): Likewise.
6976 (loadqi_update): Likewise.
6977 (load_zeroextendqisi_update): Likewise.
6978 (load_signextendqisi_update): Likewise.
6979 (loadhi_update): Likewise.
6980 (load_zeroextendhisi_update): Likewise.
6981 (load_signextendhisi_update): Likewise.
6982 (loadsi_update): Likewise.
6983 (loadsf_update): Likewise.
6984 (movsicc_insn): Likewise.
6985 (bset_insn): Likewise.
6986 (bxor_insn): Likewise.
6987 (bclr_insn): Likewise.
6988 (bmsk_insn): Likewise.
6989 (bicsi3_insn): Likewise.
6990 (cmpsi_cc_c_insn): Likewise.
6991 (movsi_ne): Likewise.
6992 (movsi_cond_exec): Likewise.
6993 (clrsbsi2): Likewise.
6994 (norm_f): Likewise.
6995 (normw): Likewise.
6996 (swap): Likewise.
6997 (divaw): Likewise.
6998 (flag): Likewise.
6999 (sr): Likewise.
7000 (kflag): Likewise.
7001 (ffs): Likewise.
7002 (ffs_f): Likewise.
7003 (fls): Likewise.
7004 (call_i): Remove 'S' asm letter, add jli instruction.
7005 (call_value_i): Likewise.
7006 * config/arc/arc.op (mjli-always): New option.
7007 * config/arc/constraints.md (Cji): New constraint.
7008 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7009 operand.
7010 (subsf3_fpx): Likewise.
7011 (mulsf3_fpx): Likewise.
7012 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7013 asm operand.
7014 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7015 function attrbutes.
7016 * doc/invoke.texi (ARC): Document mjli-always option.
7018 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7020 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7021 avoid addition with 0 and use incw and decw where possible.
7023 2018-01-26 Richard Biener <rguenther@suse.de>
7025 PR tree-optimization/81082
7026 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7027 association if it requires casting to unsigned.
7028 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7029 from fold_plusminus_mult_expr to catch important cases late when
7030 range info is available.
7032 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7034 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7035 * configure.ac (hidden_linkonce): New test.
7036 * configure: Regenerate.
7037 * config.in: Regenerate.
7039 2018-01-26 Julia Koval <julia.koval@intel.com>
7041 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7042 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7043 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7044 _mm_mask_bitshuffle_epi64_mask): Fix type.
7045 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7046 USI_FTYPE_V4DI_V4DI_USI): Remove.
7047 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7048 __builtin_ia32_vpshufbitqmb256_mask,
7049 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7050 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7051 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7053 2018-01-26 Alan Modra <amodra@gmail.com>
7055 PR target/84033
7056 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7057 UNSPEC_VBPERMQ. Sort other unspecs.
7059 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7061 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7063 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7065 PR middle-end/83055
7066 * predict.c (drop_profile): Do not push/pop cfun; update also
7067 node->count.
7068 (handle_missing_profiles): Fix logic looking for zero profiles.
7070 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7072 PR middle-end/83977
7073 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7074 on functions with #pragma omp declare simd or functions with simd
7075 attribute.
7076 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7077 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7078 Remove trailing \n from warning_at calls.
7080 2018-01-25 Tom de Vries <tom@codesourcery.com>
7082 PR target/84028
7083 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7084 for neutered workers.
7086 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7088 PR target/68467
7089 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7090 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7092 2018-01-24 Jeff Law <law@redhat.com>
7094 PR target/83994
7095 * i386.c (get_probe_interval): Move to earlier point.
7096 (ix86_compute_frame_layout): If -fstack-clash-protection and
7097 the frame is larger than the probe interval, then use pushes
7098 to save registers rather than reg->mem moves.
7099 (ix86_expand_prologue): Remove conditional for int_registers_saved
7100 assertion.
7102 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7104 PR target/84014
7105 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7106 min/max for never referenced object.
7108 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7110 PR middle-end/83977
7111 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7112 here.
7113 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7114 attributes from DECL_ATTRIBUTES (decl) without affecting
7115 DECL_ATTRIBUTES (current_function_decl).
7116 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7117 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7119 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7121 PR tree-optimization/83979
7122 * fold-const.c (fold_comparison): Use constant_boolean_node
7123 instead of boolean_{true,false}_node.
7125 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7127 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7128 with zero counts.
7130 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7132 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7133 Simplify the clause that sets the length attribute.
7134 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7135 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7136 clause that sets the length attribute.
7137 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7139 2018-01-24 Tom de Vries <tom@codesourcery.com>
7141 PR target/83589
7142 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7143 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7144 Add strict parameter.
7145 (prevent_branch_around_nothing): Insert dummy insn between branch to
7146 label and label with no ptx insn inbetween.
7147 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7149 2018-01-24 Tom de Vries <tom@codesourcery.com>
7151 PR target/81352
7152 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7153 for neutered threads in warp.
7154 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7156 2018-01-24 Richard Biener <rguenther@suse.de>
7158 PR tree-optimization/83176
7159 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7160 operands.
7162 2018-01-24 Richard Biener <rguenther@suse.de>
7164 PR tree-optimization/82819
7165 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7166 code generating pluses that are no-ops in the target precision.
7168 2018-01-24 Richard Biener <rguenther@suse.de>
7170 PR middle-end/84000
7171 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7173 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7175 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7176 to merge probabilities.
7177 * predict.c (probably_never_executed): Also mark as cold functions
7178 with global 0 profile and guessed local profile.
7179 * profile-count.c (profile_probability::combine_with_count): New
7180 member function.
7181 * profile-count.h (profile_probability::operator*,
7182 profile_probability::operator*=, profile_probability::operator/,
7183 profile_probability::operator/=): Reduce precision to adjusted
7184 and set value to guessed on contradictory divisions.
7185 (profile_probability::combine_with_freq): Remove.
7186 (profile_probability::combine_wiht_count): Declare.
7187 (profile_count::force_nonzero):: Set to adjusted.
7188 (profile_count::probability_in):: Set quality to adjusted.
7189 * tree-ssa-tail-merge.c (replace_block_by): Use
7190 combine_with_count.
7192 2018-01-23 Andrew Waterman <andrew@sifive.com>
7193 Jim Wilson <jimw@sifive.com>
7195 * config/riscv/riscv.c (riscv_stack_boundary): New.
7196 (riscv_option_override): Set riscv_stack_boundary. Handle
7197 riscv_preferred_stack_boundary_arg.
7198 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7199 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7200 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7201 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7202 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7203 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7205 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7207 PR target/83905
7208 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7209 of struct ix86_frame.
7210 (ix86_expand_epilogue): Likewise. Add a local variable for
7211 the reg_save_offset field in struct ix86_frame.
7213 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7215 PR tree-optimization/82604
7216 * tree-loop-distribution.c (enum partition_kind): New enum item
7217 PKIND_PARTIAL_MEMSET.
7218 (partition_builtin_p): Support above new enum item.
7219 (generate_code_for_partition): Ditto.
7220 (compute_access_range): Differentiate cases that equality can be
7221 proven at all loops, the innermost loops or no loops.
7222 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7223 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7224 (finalize_partitions, distribute_loop): Don't fuse partition of
7225 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7226 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7227 parloop is enabled.
7229 2018-01-23 Martin Liska <mliska@suse.cz>
7231 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7232 order to ignore the predictor.
7233 (PRED_POLYMORPHIC_CALL): Likewise.
7234 (PRED_RECURSIVE_CALL): Likewise.
7236 2018-01-23 Martin Liska <mliska@suse.cz>
7238 * tree-profile.c (tree_profiling): Print function header to
7239 aware reader which function we are working on.
7240 * value-prof.c (gimple_find_values_to_profile): Do not print
7241 not interesting value histograms.
7243 2018-01-23 Martin Liska <mliska@suse.cz>
7245 * profile-count.h (enum profile_quality): Add
7246 profile_uninitialized as the first value. Do not number values
7247 as they are zero based.
7248 (profile_count::verify): Update sanity check.
7249 (profile_probability::verify): Likewise.
7251 2018-01-23 Nathan Sidwell <nathan@acm.org>
7253 * doc/invoke.texi (ffor-scope): Deprecate.
7255 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7257 PR tree-optimization/83510
7258 * domwalk.c (set_all_edges_as_executable): New function.
7259 (dom_walker::dom_walker): Convert bool param
7260 "skip_unreachable_blocks" to enum reachability. Move setup of
7261 edge flags to set_all_edges_as_executable and only do it when
7262 reachability is REACHABLE_BLOCKS.
7263 * domwalk.h (enum dom_walker::reachability): New enum.
7264 (dom_walker::dom_walker): Convert bool param
7265 "skip_unreachable_blocks" to enum reachability.
7266 (set_all_edges_as_executable): New decl.
7267 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7268 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7269 "reachability".
7270 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7271 but converting true to REACHABLE_BLOCKS.
7272 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7273 * tree-vrp.c
7274 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7275 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7276 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7277 REACHABLE_BLOCKS.
7278 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7279 if check_all_array_refs will be called.
7281 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7283 * tree.c (selftest::test_location_wrappers): Add more test
7284 coverage.
7286 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7288 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7289 (selftest::test_bit_in_range): Likewise.
7291 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7293 PR testsuite/83888
7294 * doc/sourcebuild.texi (vect_float): Say that the selector
7295 only describes the situation when -funsafe-math-optimizations is on.
7296 (vect_float_strict): Document.
7298 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7300 PR tree-optimization/83965
7301 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7302 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7303 instead of checking only for a reduction.
7304 (vect_recog_widen_sum_pattern): Likewise.
7306 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7308 * predict.c (probably_never_executed): Only use precise profile info.
7309 (compute_function_frequency): Skip after inlining hack since we now
7310 have quality checking.
7312 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7314 * profile-count.h (profile_probability::very_unlikely,
7315 profile_probability::unlikely, profile_probability::even): Set
7316 precision to guessed.
7318 2018-01-23 Richard Biener <rguenther@suse.de>
7320 PR tree-optimization/83963
7321 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7322 Properly terminate dominator walk when crossing the exit edge not
7323 when visiting its source block.
7325 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7327 PR c++/83918
7328 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7329 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7331 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7333 PR tree-optimization/83957
7334 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7335 semicolon after for body surrounded by braces.
7337 PR tree-optimization/83081
7338 * profile-count.h (profile_probability::split): New method.
7339 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7340 Use profile_probability::split.
7341 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7342 when splitting a single conditional jump into 2.
7344 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7346 PR tree-optimization/69452
7347 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7348 decl.
7350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7352 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7353 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7354 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7356 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7358 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7359 * config/rl78/rl78.md (movdi): New define_expand.
7360 * config/rl78/rl78.c (rl78_split_movdi): New function.
7362 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7364 PR target/83862
7365 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7366 no longer used.
7367 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7368 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7369 128-bit to produce an UNSPEC move to get the double word with the
7370 signbit and then a shift directly to do signbit.
7371 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7372 implementation with a new version that just does either a direct
7373 move or a regular move. Move memory interface to separate insns.
7374 Move insns so they are next to the expander.
7375 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7376 with signbit move. Split big and little endian case.
7377 (signbit<mode>2_dm_mem_le): Likewise.
7378 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7379 (signbit<mode>2_dm2): Likewise.
7381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7383 * config/rl78/rl78.md (anddi3): New define_expand.
7385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7387 * config/rl78/rl78.md (umindi3): New define_expand.
7389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7391 * config/rl78/rl78.md (smindi3): New define_expand.
7393 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7395 * config/rl78/rl78.md (smaxdi3): New define_expand.
7397 2018-01-22 Carl Love <cel@us.ibm.com>
7399 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7400 LVX_V1TI): Add macro expansion.
7401 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7402 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7403 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7404 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7405 Change check to determine if the instruction is a byte reversing
7406 entry. Fix typo in comment.
7407 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7408 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7409 Add def_builtin calls for new builtins.
7410 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7411 Add define_insn expansion.
7413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7415 * config/rl78/rl78.md (umaxdi3): New define_expand.
7417 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7419 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7420 for non-QImode registers.
7422 2018-01-22 Richard Biener <rguenther@suse.de>
7424 PR tree-optimization/83963
7425 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7426 including the loop exit block.
7427 (scop_detection::merge_sese): Likewise.
7428 (scop_detection::add_scop): Do it here instead.
7430 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7432 * doc/sourcebuild.texi (arm_softfloat): Document.
7434 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7436 PR gcc/77734
7437 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7438 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7439 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7441 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7442 David Edelsohn <dje.gcc@gmail.com>
7444 PR target/83946
7445 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7446 Change "crset eq" to "crset 2".
7447 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7448 (*call_indirect_aix<mode>_nospec): Likewise.
7449 (*call_value_indirect_aix<mode>_nospec): Likewise.
7450 (*call_indirect_elfv2<mode>_nospec): Likewise.
7451 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7452 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7453 change assembly output from . to $.
7454 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7455 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7456 (*tablejump<mode>_internal1_nospec): Likewise.
7458 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7460 PR target/80870
7461 * config/sh/sh_optimize_sett_clrt.cc:
7462 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7464 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7466 PR tree-optimization/83940
7467 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7468 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7469 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7470 use it to pass back the definition type.
7471 (vect_check_store_rhs): Likewise rhs_dt_out.
7472 (vect_build_gather_load_calls): Add a mask_dt argument and use
7473 it instead of a call to vect_is_simple_use.
7474 (vectorizable_store): Update calls to vect_check_load_store_mask
7475 and vect_check_store_rhs. Use the dt returned by the latter instead
7476 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7477 instead of calls to vect_is_simple_use. Pass the scalar rather
7478 than the vector operand to vect_is_simple_use when handling
7479 second and subsequent copies of an rhs value.
7480 (vectorizable_load): Update calls to vect_check_load_store_mask
7481 and vect_build_gather_load_calls. Use the cached mask_dt and
7482 gs_info.offset_dt instead of calls to vect_is_simple_use.
7484 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7486 PR middle-end/83945
7487 * tree-emutls.c: Include gimplify.h.
7488 (lower_emutls_2): New function.
7489 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7490 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7491 it before further processing.
7493 PR target/83930
7494 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7495 UINTVAL (trueop1) instead of INTVAL (op1).
7497 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7499 PR debug/81570
7500 PR debug/83728
7501 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7502 INCOMING_FRAME_SP_OFFSET if not defined.
7503 (scan_trace): Add ENTRY argument. If true and
7504 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7505 emit a note to adjust the CFA offset.
7506 (create_cfi_notes): Adjust scan_trace callers.
7507 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7508 INCOMING_FRAME_SP_OFFSET in the CIE.
7509 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7510 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7511 Likewise.
7512 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7513 * doc/tm.texi: Regenerated.
7515 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7517 PR rtl-optimization/83147
7518 * lra-constraints.c (remove_inheritance_pseudos): Use
7519 lra_substitute_pseudo_within_insn.
7521 2018-01-19 Tom de Vries <tom@codesourcery.com>
7522 Cesar Philippidis <cesar@codesourcery.com>
7524 PR target/83920
7525 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7527 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7529 PR target/83790
7530 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7531 spaces for function labels.
7533 2018-01-19 Martin Liska <mliska@suse.cz>
7535 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7536 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7537 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7538 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7539 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7540 (PRED_CONST_RETURN): Change from 69 to 65.
7541 (PRED_NULL_RETURN): Change from 91 to 71.
7542 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7543 (PRED_LOOP_GUARD): Change from 66 to 73.
7545 2018-01-19 Martin Liska <mliska@suse.cz>
7547 * predict.c (predict_insn_def): Add new assert.
7548 (struct branch_predictor): Change type to signed integer.
7549 (test_prediction_value_range): Amend test to cover
7550 PROB_UNINITIALIZED.
7551 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7552 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7553 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7554 (PRED_LOOP_IV_COMPARE): Likewise.
7555 * predict.h (PROB_UNINITIALIZED): Define new constant.
7557 2018-01-19 Martin Liska <mliska@suse.cz>
7559 * predict.c (dump_prediction): Add new format for
7560 analyze_brprob.py script which is enabled with -details
7561 suboption.
7562 * profile-count.h (precise_p): New function.
7564 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7566 PR tree-optimization/83922
7567 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7568 there are no statements that need masking.
7569 (vect_active_double_reduction_p): New function.
7570 (vect_analyze_loop_operations): Use it when handling phis that
7571 are not in the loop header.
7573 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7575 PR tree-optimization/83914
7576 * tree-vect-loop.c (vectorizable_induction): Don't convert
7577 init_expr or apply the peeling adjustment for inductions
7578 that are nested within the vectorized loop.
7580 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7582 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7583 instead of NEG.
7585 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7587 PR sanitizer/81715
7588 PR testsuite/83882
7589 * function.h (gimplify_parameters): Add gimple_seq * argument.
7590 * function.c: Include gimple.h and options.h.
7591 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7592 for the added local temporaries if needed.
7593 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7594 if there are any parameter cleanups, wrap whole body into a
7595 try/finally with the cleanups.
7597 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7599 PR target/82964
7600 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7601 Use GET_MODE_CLASS for scalar floating point.
7603 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7605 PR ipa/82256
7606 patch by PaX Team
7607 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7608 Fix call of call_cgraph_insertion_hooks.
7610 2018-01-18 Martin Sebor <msebor@redhat.com>
7612 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7614 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7616 PR ipa/83619
7617 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7618 frequencies.
7620 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7622 PR other/70268
7623 * common.opt: (-ffile-prefix-map): New option.
7624 * opts.c (common_handle_option): Defer it.
7625 * opts-global.c (handle_common_deferred_options): Handle it.
7626 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7627 * file-prefix-map.h: New file.
7628 (remap_debug_filename, add_debug_prefix_map): ...here.
7629 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7630 * final.c (debug_prefix_map, add_debug_prefix_map
7631 remap_debug_filename): Move to...
7632 * file-prefix-map.c: New file.
7633 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7634 generalize, get rid of alloca(), use strrchr() instead of strchr().
7635 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7636 Implement in terms of add_prefix_map().
7637 (remap_macro_filename, remap_debug_filename): Implement in term of
7638 remap_filename().
7639 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7640 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7641 * dbxout.c: Include file-prefix-map.h.
7642 * varasm.c: Likewise.
7643 * vmsdbgout.c: Likewise.
7644 * xcoffout.c: Likewise.
7645 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7646 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7647 * doc/invoke.texi (-ffile-prefix-map): Document.
7648 (-fdebug-prefix-map): Update description.
7650 2018-01-18 Martin Liska <mliska@suse.cz>
7652 * config/i386/i386.c (indirect_thunk_name): Document that also
7653 lfence is emitted.
7654 (output_indirect_thunk): Document why both instructions
7655 (pause and lfence) are generated.
7657 2018-01-18 Richard Biener <rguenther@suse.de>
7659 PR tree-optimization/83887
7660 * graphite-scop-detection.c
7661 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7662 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7663 (scop_detection::merge_sese): Re-implement with a flood-fill
7664 algorithm that properly finds a SESE region if it exists.
7666 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7668 PR c/61240
7669 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7670 pointer_diff optimizations use view_convert instead of convert.
7672 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7674 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7675 Generate different code for -mno-speculate-indirect-jumps.
7676 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7677 (*call_indirect_aix<mode>): Disable for
7678 -mno-speculate-indirect-jumps.
7679 (*call_indirect_aix<mode>_nospec): New define_insn.
7680 (*call_value_indirect_aix<mode>): Disable for
7681 -mno-speculate-indirect-jumps.
7682 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7683 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7684 -mno-speculate-indirect-jumps.
7685 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7687 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7689 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7690 long double type, set the flags for noting the default long double
7691 type, even if we don't pass or return a long double type.
7693 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7695 PR ipa/83051
7696 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7697 failure.
7699 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7701 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7702 support for merge[hl].
7703 (fold_mergehl_helper): New helper function.
7704 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7705 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7706 (altivec_vmrglw_direct): Add xxmrglw insn.
7708 2018-01-17 Andrew Waterman <andrew@sifive.com>
7710 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7711 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7713 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7715 PR lto/83121
7716 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7717 call the lto_location_cache before reading the
7718 DECL_SOURCE_LOCATION of the types.
7720 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7721 Richard Sandiford <richard.sandiford@linaro.org>
7723 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7724 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7725 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7726 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7727 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7728 Add declaration.
7729 * config/aarch64/constraints.md (aarch64_movti_operand):
7730 Limit immediates.
7731 * config/aarch64/predicates.md (Uti): Add new constraint.
7733 2018-01-17 Carl Love <cel@us.ibm.com>
7735 * config/rs6000/vsx.md (define_expand xl_len_r,
7736 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7737 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7738 lxvll.
7739 (define_expand, define_insn): Move the shift left from the
7740 define_insn to the define_expand for lxvl and stxvl instructions.
7741 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7742 and XL_LEN_R definitions to PURE.
7744 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7746 * config/i386/i386.c (indirect_thunk_name): Declare regno
7747 as unsigned int. Compare regno with INVALID_REGNUM.
7748 (output_indirect_thunk): Ditto.
7749 (output_indirect_thunk_function): Ditto.
7750 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7751 in the call to output_indirect_thunk_function.
7753 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7755 PR middle-end/83884
7756 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7757 rather than the size of inner_type to determine the stack slot size
7758 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7760 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7762 PR target/83546
7763 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7764 to PTA_SILVERMONT.
7766 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7768 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7769 endian Linux systems to optionally enable multilibs for selecting
7770 the long double type if the user configured an explicit type.
7771 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7772 have no long double multilibs if not defined.
7773 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7774 warn if the user used -mabi={ieee,ibm}longdouble and we built
7775 multilibs for long double.
7776 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7777 appropriate multilib option.
7778 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7779 multilib options.
7780 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7781 for building long double multilibs.
7782 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7784 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7786 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7787 copies.
7789 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7790 64 bits.
7791 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7792 128 bits.
7794 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7795 variables.
7797 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7798 return value.
7800 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7802 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7803 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7805 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7807 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7808 different rtl trees depending on TARGET_64BIT.
7809 (rs6000_gen_lvx): Likewise.
7811 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7813 * config/visium/visium.md (nop): Tweak comment.
7814 (hazard_nop): Likewise.
7816 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7818 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7819 -mspeculate-indirect-jumps.
7820 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7821 for -mno-speculate-indirect-jumps.
7822 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7823 (*call_value_indirect_elfv2<mode>): Disable for
7824 -mno-speculate-indirect-jumps.
7825 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7826 (indirect_jump): Emit different RTL for
7827 -mno-speculate-indirect-jumps.
7828 (*indirect_jump<mode>): Disable for
7829 -mno-speculate-indirect-jumps.
7830 (*indirect_jump<mode>_nospec): New define_insn.
7831 (tablejump): Emit different RTL for
7832 -mno-speculate-indirect-jumps.
7833 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7834 (tablejumpsi_nospec): New define_expand.
7835 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7836 (tablejumpdi_nospec): New define_expand.
7837 (*tablejump<mode>_internal1): Disable for
7838 -mno-speculate-indirect-jumps.
7839 (*tablejump<mode>_internal1_nospec): New define_insn.
7840 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7841 option.
7843 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7845 * caller-save.c (insert_save): Drop unnecessary parameter. All
7846 callers updated.
7848 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7849 Richard Biener <rguenth@suse.de>
7851 PR libgomp/83590
7852 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7853 return early, inline manually is_gimple_sizepos. Make sure if we
7854 call gimplify_expr we don't end up with a gimple constant.
7855 * tree.c (variably_modified_type_p): Don't return true for
7856 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7857 * gimplify.h (is_gimple_sizepos): Remove.
7859 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7861 PR tree-optimization/83857
7862 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7863 vectorizable_live_operation for pure SLP statements.
7864 (vectorizable_live_operation): Handle PHIs.
7866 2018-01-16 Richard Biener <rguenther@suse.de>
7868 PR tree-optimization/83867
7869 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7870 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7872 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7874 PR c/83844
7875 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7876 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7877 If off is not INTEGER_CST, issue a may not be aligned warning
7878 rather than isn't aligned. Use isn%'t rather than isn't.
7879 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7880 into MULT_EXPR.
7881 <case MULT_EXPR>: Improve the case when bottom and one of the
7882 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7883 operand, in that case check if the other operand is multiple of
7884 bottom divided by the INTEGER_CST operand.
7886 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7888 PR target/83858
7889 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7890 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7891 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7892 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7893 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7894 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7895 (pa_function_arg_size): New function.
7897 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7899 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7900 in a separate statement.
7902 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7904 PR tree-optimization/83847
7905 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7906 group gathers and scatters.
7908 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7910 PR rtl-optimization/86620
7911 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7913 PR rtl-optimization/83213
7914 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7915 to last if both are JUMP_INSNs.
7917 PR tree-optimization/83843
7918 * gimple-ssa-store-merging.c
7919 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7920 store_immediate_info for bswap/nop orig_stores.
7922 2018-01-15 Andrew Waterman <andrew@sifive.com>
7924 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7925 !TARGET_MUL.
7926 <UDIV>: Increase cost if !TARGET_DIV.
7928 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7930 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7931 (define_attr "cr_logical_3op"): New.
7932 (cceq_ior_compare): Adjust.
7933 (cceq_ior_compare_complement): Adjust.
7934 (*cceq_rev_compare): Adjust.
7935 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7936 (is_cracked_insn): Adjust.
7937 (insn_must_be_first_in_group): Adjust.
7938 * config/rs6000/40x.md: Adjust.
7939 * config/rs6000/440.md: Adjust.
7940 * config/rs6000/476.md: Adjust.
7941 * config/rs6000/601.md: Adjust.
7942 * config/rs6000/603.md: Adjust.
7943 * config/rs6000/6xx.md: Adjust.
7944 * config/rs6000/7450.md: Adjust.
7945 * config/rs6000/7xx.md: Adjust.
7946 * config/rs6000/8540.md: Adjust.
7947 * config/rs6000/cell.md: Adjust.
7948 * config/rs6000/e300c2c3.md: Adjust.
7949 * config/rs6000/e500mc.md: Adjust.
7950 * config/rs6000/e500mc64.md: Adjust.
7951 * config/rs6000/e5500.md: Adjust.
7952 * config/rs6000/e6500.md: Adjust.
7953 * config/rs6000/mpc.md: Adjust.
7954 * config/rs6000/power4.md: Adjust.
7955 * config/rs6000/power5.md: Adjust.
7956 * config/rs6000/power6.md: Adjust.
7957 * config/rs6000/power7.md: Adjust.
7958 * config/rs6000/power8.md: Adjust.
7959 * config/rs6000/power9.md: Adjust.
7960 * config/rs6000/rs64.md: Adjust.
7961 * config/rs6000/titan.md: Adjust.
7963 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7965 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7966 ix86_indirect_branch_register logic.
7968 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7970 * config/i386/constraints.md (Bs): Update
7971 ix86_indirect_branch_register check. Don't check
7972 ix86_indirect_branch_register with GOT_memory_operand.
7973 (Bw): Likewise.
7974 * config/i386/predicates.md (GOT_memory_operand): Don't check
7975 ix86_indirect_branch_register here.
7976 (GOT32_symbol_operand): Likewise.
7978 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7980 * config/i386/predicates.md (constant_call_address_operand):
7981 Rewrite ix86_indirect_branch_register logic.
7982 (sibcall_insn_operand): Likewise.
7984 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7986 * config/i386/constraints.md (Bs): Replace
7987 ix86_indirect_branch_thunk_register with
7988 ix86_indirect_branch_register.
7989 (Bw): Likewise.
7990 * config/i386/i386.md (indirect_jump): Likewise.
7991 (tablejump): Likewise.
7992 (*sibcall_memory): Likewise.
7993 (*sibcall_value_memory): Likewise.
7994 Peepholes of indirect call and jump via memory: Likewise.
7995 * config/i386/i386.opt: Likewise.
7996 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7997 (GOT_memory_operand): Likewise.
7998 (call_insn_operand): Likewise.
7999 (sibcall_insn_operand): Likewise.
8000 (GOT32_symbol_operand): Likewise.
8002 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8004 PR middle-end/83837
8005 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8006 type rather than type addr's type points to.
8007 (expand_omp_atomic_mutex): Likewise.
8008 (expand_omp_atomic): Likewise.
8010 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8012 PR target/83839
8013 * config/i386/i386.c (output_indirect_thunk_function): Use
8014 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8015 for __x86_return_thunk.
8017 2018-01-15 Richard Biener <rguenther@suse.de>
8019 PR middle-end/83850
8020 * expmed.c (extract_bit_field_1): Fix typo.
8022 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8024 PR target/83687
8025 * config/arm/iterators.md (VF): New mode iterator.
8026 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8027 Remove integer-related logic from pattern.
8028 (neon_vabd<mode>_3): Likewise.
8030 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8032 PR middle-end/82694
8033 * common.opt (fstrict-overflow): No longer an alias.
8034 (fwrapv-pointer): New option.
8035 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8036 also for pointer types based on flag_wrapv_pointer.
8037 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8038 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8039 opts->x_flag_wrapv got set.
8040 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8041 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8042 POINTER_TYPE_OVERFLOW_UNDEFINED.
8043 * match.pd: Likewise in address comparison pattern.
8044 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8046 2018-01-15 Richard Biener <rguenther@suse.de>
8048 PR lto/83804
8049 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8050 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8051 Reset type names to their identifier if their TYPE_DECL doesn't
8052 have linkage (and thus is used for ODR and devirt).
8053 (save_debug_info_for_decl): Remove.
8054 (save_debug_info_for_type): Likewise.
8055 (add_tree_to_fld_list): Adjust.
8056 * tree-pretty-print.c (dump_generic_node): Make dumping of
8057 type names more robust.
8059 2018-01-15 Richard Biener <rguenther@suse.de>
8061 * BASE-VER: Bump to 8.0.1.
8063 2018-01-14 Martin Sebor <msebor@redhat.com>
8065 PR other/83508
8066 * builtins.c (check_access): Avoid warning when the no-warning bit
8067 is set.
8069 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8071 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8072 * ira-color (allocno_hard_regs_compare): Likewise.
8074 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8076 PR target/83013
8077 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8078 Use .pushsection/.popsection.
8080 2018-01-14 Martin Sebor <msebor@redhat.com>
8082 PR c++/81327
8083 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8085 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8087 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8088 entry from extra_headers.
8089 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8090 extra_headers, make the list bitwise identical to the i?86-*-* one.
8092 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8094 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8095 -mcmodel=large with -mindirect-branch=thunk,
8096 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8097 -mfunction-return=thunk-extern.
8098 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8099 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8100 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8102 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8104 * config/i386/i386.c (print_reg): Print the name of the full
8105 integer register without '%'.
8106 (ix86_print_operand): Handle 'V'.
8107 * doc/extend.texi: Document 'V' modifier.
8109 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8111 * config/i386/constraints.md (Bs): Disallow memory operand for
8112 -mindirect-branch-register.
8113 (Bw): Likewise.
8114 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8115 (GOT_memory_operand): Likewise.
8116 (call_insn_operand): Likewise.
8117 (sibcall_insn_operand): Likewise.
8118 (GOT32_symbol_operand): Likewise.
8119 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8120 for -mindirect-branch-register.
8121 (tablejump): Likewise.
8122 (*sibcall_memory): Likewise.
8123 (*sibcall_value_memory): Likewise.
8124 Disallow peepholes of indirect call and jump via memory for
8125 -mindirect-branch-register.
8126 (*call_pop): Replace m with Bw.
8127 (*call_value_pop): Likewise.
8128 (*sibcall_pop_memory): Replace m with Bs.
8129 * config/i386/i386.opt (mindirect-branch-register): New option.
8130 * doc/invoke.texi: Document -mindirect-branch-register option.
8132 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8134 * config/i386/i386-protos.h (ix86_output_function_return): New.
8135 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8136 set function_return_type.
8137 (indirect_thunk_name): Add ret_p to indicate thunk for function
8138 return.
8139 (output_indirect_thunk_function): Pass false to
8140 indirect_thunk_name.
8141 (ix86_output_indirect_branch_via_reg): Likewise.
8142 (ix86_output_indirect_branch_via_push): Likewise.
8143 (output_indirect_thunk_function): Create alias for function
8144 return thunk if regno < 0.
8145 (ix86_output_function_return): New function.
8146 (ix86_handle_fndecl_attribute): Handle function_return.
8147 (ix86_attribute_table): Add function_return.
8148 * config/i386/i386.h (machine_function): Add
8149 function_return_type.
8150 * config/i386/i386.md (simple_return_internal): Use
8151 ix86_output_function_return.
8152 (simple_return_internal_long): Likewise.
8153 * config/i386/i386.opt (mfunction-return=): New option.
8154 (indirect_branch): Mention -mfunction-return=.
8155 * doc/extend.texi: Document function_return function attribute.
8156 * doc/invoke.texi: Document -mfunction-return= option.
8158 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8160 * config/i386/i386-opts.h (indirect_branch): New.
8161 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8162 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8163 with local indirect jump when converting indirect call and jump.
8164 (ix86_set_indirect_branch_type): New.
8165 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8166 (indirectlabelno): New.
8167 (indirect_thunk_needed): Likewise.
8168 (indirect_thunk_bnd_needed): Likewise.
8169 (indirect_thunks_used): Likewise.
8170 (indirect_thunks_bnd_used): Likewise.
8171 (INDIRECT_LABEL): Likewise.
8172 (indirect_thunk_name): Likewise.
8173 (output_indirect_thunk): Likewise.
8174 (output_indirect_thunk_function): Likewise.
8175 (ix86_output_indirect_branch_via_reg): Likewise.
8176 (ix86_output_indirect_branch_via_push): Likewise.
8177 (ix86_output_indirect_branch): Likewise.
8178 (ix86_output_indirect_jmp): Likewise.
8179 (ix86_code_end): Call output_indirect_thunk_function if needed.
8180 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8181 needed.
8182 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8183 (ix86_attribute_table): Add indirect_branch.
8184 * config/i386/i386.h (machine_function): Add indirect_branch_type
8185 and has_local_indirect_jump.
8186 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8187 to true.
8188 (tablejump): Likewise.
8189 (*indirect_jump): Use ix86_output_indirect_jmp.
8190 (*tablejump_1): Likewise.
8191 (simple_return_indirect_internal): Likewise.
8192 * config/i386/i386.opt (mindirect-branch=): New option.
8193 (indirect_branch): New.
8194 (keep): Likewise.
8195 (thunk): Likewise.
8196 (thunk-inline): Likewise.
8197 (thunk-extern): Likewise.
8198 * doc/extend.texi: Document indirect_branch function attribute.
8199 * doc/invoke.texi: Document -mindirect-branch= option.
8201 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8203 PR ipa/83051
8204 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8206 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8208 * ipa-inline.c (want_inline_small_function_p): Return false if
8209 inlining has already failed with CIF_FINAL_ERROR.
8210 (update_caller_keys): Call want_inline_small_function_p before
8211 can_inline_edge_p.
8212 (update_callee_keys): Likewise.
8214 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8216 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8217 New function.
8218 (rs6000_quadword_masked_address_p): Likewise.
8219 (quad_aligned_load_p): Likewise.
8220 (quad_aligned_store_p): Likewise.
8221 (const_load_sequence_p): Add comment to describe the outer-most loop.
8222 (mimic_memory_attributes_and_flags): New function.
8223 (rs6000_gen_stvx): Likewise.
8224 (replace_swapped_aligned_store): Likewise.
8225 (rs6000_gen_lvx): Likewise.
8226 (replace_swapped_aligned_load): Likewise.
8227 (replace_swapped_load_constant): Capitalize argument name in
8228 comment describing this function.
8229 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8230 and stores that access quad-word aligned addresses and replace
8231 with stvx or lvx instructions when appropriate.
8232 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8233 New function prototype.
8234 (rs6000_quadword_masked_address_p): Likewise.
8235 (rs6000_gen_lvx): Likewise.
8236 (rs6000_gen_stvx): Likewise.
8237 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8238 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8239 when memory address is aligned.
8240 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8241 this split to select lvx instruction when memory address is aligned.
8242 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8243 instruction when memory address is aligned.
8244 (*vsx_le_perm_load_v16qi): Likewise.
8245 (four unnamed splitters): Modify to select the stvx instruction
8246 when memory is aligned.
8248 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8250 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8251 which appears in the queue multiple times.
8253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8254 Alan Hayward <alan.hayward@arm.com>
8255 David Sherwood <david.sherwood@arm.com>
8257 * tree-vectorizer.h (vec_lower_bound): New structure.
8258 (_loop_vec_info): Add check_nonzero and lower_bounds.
8259 (LOOP_VINFO_CHECK_NONZERO): New macro.
8260 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8261 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8262 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8263 fields. Make seg_len the distance travelled, not including the
8264 access size.
8265 (dr_direction_indicator): Declare.
8266 (dr_zero_step_indicator): Likewise.
8267 (dr_known_forward_stride_p): Likewise.
8268 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8269 tree-ssanames.h.
8270 (runtime_alias_check_p): Allow runtime alias checks with
8271 variable strides.
8272 (operator ==): Compare access_size and align.
8273 (prune_runtime_alias_test_list): Rework for new distinction between
8274 the access_size and seg_len.
8275 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8276 segment lengths.
8277 (get_segment_min_max): New function.
8278 (create_intersect_range_checks): Use it.
8279 (dr_step_indicator): New function.
8280 (dr_direction_indicator): Likewise.
8281 (dr_zero_step_indicator): Likewise.
8282 (dr_known_forward_stride_p): Likewise.
8283 * tree-loop-distribution.c (data_ref_segment_size): Return
8284 DR_STEP * (niters - 1).
8285 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8286 constructor.
8287 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8288 (vect_preserves_scalar_order_p): New function, split out from...
8289 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8290 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8291 (vect_vfa_access_size): New function.
8292 (vect_vfa_align): Likewise.
8293 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8294 (dump_lower_bound): New function.
8295 (vect_check_lower_bound): Likewise.
8296 (vect_small_gap_p): Likewise.
8297 (vectorizable_with_step_bound_p): Likewise.
8298 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8299 depencies if the vectorization factor is 1. Convert the checks
8300 for nonzero steps into checks on the bounds of DR_STEP. Try using
8301 a bunds check for variable steps if the minimum required step is
8302 relatively small. Update calls to the dr_with_seg_len
8303 constructor and to vect_compile_time_alias.
8304 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8305 function.
8306 (vect_loop_versioning): Call it.
8307 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8308 when retrying.
8309 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8312 Alan Hayward <alan.hayward@arm.com>
8313 David Sherwood <david.sherwood@arm.com>
8315 * doc/sourcebuild.texi (vect_scatter_store): Document.
8316 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8317 optabs.
8318 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8319 Document.
8320 * genopinit.c (main): Add supports_vec_scatter_store and
8321 supports_vec_scatter_store_cached to target_optabs.
8322 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8323 IFN_MASK_SCATTER_STORE.
8324 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8325 functions.
8326 * internal-fn.h (internal_store_fn_p): Declare.
8327 (internal_fn_stored_value_index): Likewise.
8328 * internal-fn.c (scatter_store_direct): New macro.
8329 (expand_scatter_store_optab_fn): New function.
8330 (direct_scatter_store_optab_supported_p): New macro.
8331 (internal_store_fn_p): New function.
8332 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8333 IFN_MASK_SCATTER_STORE.
8334 (internal_fn_mask_index): Likewise.
8335 (internal_fn_stored_value_index): New function.
8336 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8337 for scatter stores.
8338 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8339 * optabs-query.c (supports_vec_scatter_store_p): New function.
8340 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8341 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8342 true for scatter stores.
8343 (vect_gather_scatter_fn_p): Handle scatter stores too.
8344 (vect_check_gather_scatter): Consider using scatter stores if
8345 supports_vec_scatter_store_p.
8346 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8347 scatter stores too.
8348 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8349 internal_fn_stored_value_index.
8350 (check_load_store_masking): Handle scatter stores too.
8351 (vect_get_store_rhs): Make public.
8352 (vectorizable_call): Use internal_store_fn_p.
8353 (vectorizable_store): Handle scatter store internal functions.
8354 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8355 when deciding whether the end of the group has been reached.
8356 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8357 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8358 (mask_scatter_store<mode>): New insns.
8360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8361 Alan Hayward <alan.hayward@arm.com>
8362 David Sherwood <david.sherwood@arm.com>
8364 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8365 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8366 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8367 function.
8368 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8369 Use vect_truncate_gather_scatter_offset if we can't treat the
8370 operation as a normal gather load or scatter store.
8371 (get_group_load_store_type): Take the gather_scatter_info
8372 as argument. Try using a gather load or scatter store for
8373 single-element groups.
8374 (get_load_store_type): Update calls to get_group_load_store_type
8375 and vect_use_strided_gather_scatters_p.
8377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8378 Alan Hayward <alan.hayward@arm.com>
8379 David Sherwood <david.sherwood@arm.com>
8381 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8382 optional tree argument.
8383 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8384 null target hooks.
8385 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8386 but continue to use the current value as a fallback.
8387 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8388 to compare the updates.
8389 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8390 (get_load_store_type): Use it when handling a strided access.
8391 (vect_get_strided_load_store_ops): New function.
8392 (vect_get_data_ptr_increment): Likewise.
8393 (vectorizable_load): Handle strided gather loads. Always pass
8394 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8397 Alan Hayward <alan.hayward@arm.com>
8398 David Sherwood <david.sherwood@arm.com>
8400 * doc/md.texi (gather_load@var{m}): Document.
8401 (mask_gather_load@var{m}): Likewise.
8402 * genopinit.c (main): Add supports_vec_gather_load and
8403 supports_vec_gather_load_cached to target_optabs.
8404 * optabs-tree.c (init_tree_optimization_optabs): Use
8405 ggc_cleared_alloc to allocate target_optabs.
8406 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8407 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8408 functions.
8409 * internal-fn.h (internal_load_fn_p): Declare.
8410 (internal_gather_scatter_fn_p): Likewise.
8411 (internal_fn_mask_index): Likewise.
8412 (internal_gather_scatter_fn_supported_p): Likewise.
8413 * internal-fn.c (gather_load_direct): New macro.
8414 (expand_gather_load_optab_fn): New function.
8415 (direct_gather_load_optab_supported_p): New macro.
8416 (direct_internal_fn_optab): New function.
8417 (internal_load_fn_p): Likewise.
8418 (internal_gather_scatter_fn_p): Likewise.
8419 (internal_fn_mask_index): Likewise.
8420 (internal_gather_scatter_fn_supported_p): Likewise.
8421 * optabs-query.c (supports_at_least_one_mode_p): New function.
8422 (supports_vec_gather_load_p): Likewise.
8423 * optabs-query.h (supports_vec_gather_load_p): Declare.
8424 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8425 and memory_type field.
8426 (NUM_PATTERNS): Bump to 15.
8427 * tree-vect-data-refs.c: Include internal-fn.h.
8428 (vect_gather_scatter_fn_p): New function.
8429 (vect_describe_gather_scatter_call): Likewise.
8430 (vect_check_gather_scatter): Try using internal functions for
8431 gather loads. Recognize existing calls to a gather load function.
8432 (vect_analyze_data_refs): Consider using gather loads if
8433 supports_vec_gather_load_p.
8434 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8435 (vect_get_gather_scatter_offset_type): Likewise.
8436 (vect_convert_mask_for_vectype): Likewise.
8437 (vect_add_conversion_to_patterm): Likewise.
8438 (vect_try_gather_scatter_pattern): Likewise.
8439 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8440 (vect_vect_recog_func_ptrs): Add it.
8441 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8442 internal_fn_mask_index and internal_gather_scatter_fn_p.
8443 (check_load_store_masking): Take the gather_scatter_info as an
8444 argument and handle gather loads.
8445 (vect_get_gather_scatter_ops): New function.
8446 (vectorizable_call): Check internal_load_fn_p.
8447 (vectorizable_load): Likewise. Handle gather load internal
8448 functions.
8449 (vectorizable_store): Update call to check_load_store_masking.
8450 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8451 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8452 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8453 (aarch64_gather_scale_operand_d): New predicates.
8454 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8455 (mask_gather_load<mode>): New insns.
8457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8458 Alan Hayward <alan.hayward@arm.com>
8459 David Sherwood <david.sherwood@arm.com>
8461 * optabs.def (fold_left_plus_optab): New optab.
8462 * doc/md.texi (fold_left_plus_@var{m}): Document.
8463 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8464 * internal-fn.c (fold_left_direct): Define.
8465 (expand_fold_left_optab_fn): Likewise.
8466 (direct_fold_left_optab_supported_p): Likewise.
8467 * fold-const-call.c (fold_const_fold_left): New function.
8468 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8469 * tree-parloops.c (valid_reduction_p): New function.
8470 (gather_scalar_reductions): Use it.
8471 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8472 (vect_finish_replace_stmt): Declare.
8473 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8474 (needs_fold_left_reduction_p): New function, split out from...
8475 (vect_is_simple_reduction): ...here. Accept reductions that
8476 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8477 (vect_force_simple_reduction): Also store the reduction type in
8478 the assignment's STMT_VINFO_REDUC_TYPE.
8479 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8480 (merge_with_identity): New function.
8481 (vect_expand_fold_left): Likewise.
8482 (vectorize_fold_left_reduction): Likewise.
8483 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8484 scalar phi in place for it. Check for target support and reject
8485 cases that would reassociate the operation. Defer the transform
8486 phase to vectorize_fold_left_reduction.
8487 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8488 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8489 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8493 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8494 call to ifc_temp_var.
8496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8497 Alan Hayward <alan.hayward@arm.com>
8498 David Sherwood <david.sherwood@arm.com>
8500 * target.def (legitimize_address_displacement): Take the original
8501 offset as a poly_int.
8502 * targhooks.h (default_legitimize_address_displacement): Update
8503 accordingly.
8504 * targhooks.c (default_legitimize_address_displacement): Likewise.
8505 * doc/tm.texi: Regenerate.
8506 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8507 as an argument, moving assert of ad->disp == ad->disp_term to...
8508 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8509 Try calling targetm.legitimize_address_displacement before expanding
8510 the address rather than afterwards, and adjust for the new interface.
8511 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8512 Match the new hook interface. Handle SVE addresses.
8513 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8514 new hook interface.
8516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8518 * Makefile.in (OBJS): Add early-remat.o.
8519 * target.def (select_early_remat_modes): New hook.
8520 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8521 * doc/tm.texi: Regenerate.
8522 * targhooks.h (default_select_early_remat_modes): Declare.
8523 * targhooks.c (default_select_early_remat_modes): New function.
8524 * timevar.def (TV_EARLY_REMAT): New timevar.
8525 * passes.def (pass_early_remat): New pass.
8526 * tree-pass.h (make_pass_early_remat): Declare.
8527 * early-remat.c: New file.
8528 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8529 function.
8530 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8533 Alan Hayward <alan.hayward@arm.com>
8534 David Sherwood <david.sherwood@arm.com>
8536 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8537 vfm1 with a bound_epilog parameter.
8538 (vect_do_peeling): Update calls accordingly, and move the prologue
8539 call earlier in the function. Treat the base bound_epilog as 0 for
8540 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8541 this base when peeling for gaps.
8542 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8543 with fully-masked loops.
8544 (vect_estimate_min_profitable_iters): Handle the single peeled
8545 iteration in that case.
8547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8548 Alan Hayward <alan.hayward@arm.com>
8549 David Sherwood <david.sherwood@arm.com>
8551 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8552 single-element interleaving even if the size is not a power of 2.
8553 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8554 accesses for single-element interleaving if the group size is
8555 not a power of 2.
8557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8558 Alan Hayward <alan.hayward@arm.com>
8559 David Sherwood <david.sherwood@arm.com>
8561 * doc/md.texi (fold_extract_last_@var{m}): Document.
8562 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8563 * optabs.def (fold_extract_last_optab): New optab.
8564 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8565 * internal-fn.c (fold_extract_direct): New macro.
8566 (expand_fold_extract_optab_fn): Likewise.
8567 (direct_fold_extract_optab_supported_p): Likewise.
8568 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8569 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8570 EXTRACT_LAST_REDUCTION.
8571 (get_initial_def_for_reduction): Do not create an initial vector
8572 for EXTRACT_LAST_REDUCTION reductions.
8573 (vectorizable_reduction): Leave the scalar phi in place for
8574 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8575 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8576 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8577 transform phase to vectorizable_condition.
8578 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8579 split out from...
8580 (vect_finish_stmt_generation): ...here.
8581 (vect_finish_replace_stmt): New function.
8582 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8583 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8584 pattern.
8585 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8588 Alan Hayward <alan.hayward@arm.com>
8589 David Sherwood <david.sherwood@arm.com>
8591 * doc/md.texi (extract_last_@var{m}): Document.
8592 * optabs.def (extract_last_optab): New optab.
8593 * internal-fn.def (EXTRACT_LAST): New internal function.
8594 * internal-fn.c (cond_unary_direct): New macro.
8595 (expand_cond_unary_optab_fn): Likewise.
8596 (direct_cond_unary_optab_supported_p): Likewise.
8597 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8598 loops using EXTRACT_LAST.
8599 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8600 (extract_last_<mode>): ...this optab.
8601 (vec_extract<mode><Vel>): Update accordingly.
8603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8604 Alan Hayward <alan.hayward@arm.com>
8605 David Sherwood <david.sherwood@arm.com>
8607 * target.def (empty_mask_is_expensive): New hook.
8608 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8609 * doc/tm.texi: Regenerate.
8610 * targhooks.h (default_empty_mask_is_expensive): Declare.
8611 * targhooks.c (default_empty_mask_is_expensive): New function.
8612 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8613 if the target says that empty masks are expensive.
8614 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8615 New function.
8616 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8619 Alan Hayward <alan.hayward@arm.com>
8620 David Sherwood <david.sherwood@arm.com>
8622 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8623 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8624 (vect_use_loop_mask_for_alignment_p): New function.
8625 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8626 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8627 niters_skip argument. Make sure that the first niters_skip elements
8628 of the first iteration are inactive.
8629 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8630 Update call to vect_set_loop_masks_directly.
8631 (get_misalign_in_elems): New function, split out from...
8632 (vect_gen_prolog_loop_niters): ...here.
8633 (vect_update_init_of_dr): Take a code argument that specifies whether
8634 the adjustment should be added or subtracted.
8635 (vect_update_init_of_drs): Likewise.
8636 (vect_prepare_for_masked_peels): New function.
8637 (vect_do_peeling): Skip prologue peeling if we're using a mask
8638 instead. Update call to vect_update_inits_of_drs.
8639 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8640 mask_skip_niters.
8641 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8642 alignment. Do not include the number of peeled iterations in
8643 the minimum threshold in that case.
8644 (vectorizable_induction): Adjust the start value down by
8645 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8646 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8647 Take the number of skipped iterations into account when calculating
8648 the loop bounds.
8649 * tree-vect-stmts.c (vect_gen_while_not): New function.
8651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8652 Alan Hayward <alan.hayward@arm.com>
8653 David Sherwood <david.sherwood@arm.com>
8655 * doc/sourcebuild.texi (vect_fully_masked): Document.
8656 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8657 default value to 0.
8658 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8659 split out from...
8660 (vect_analyze_loop_2): ...here. Don't check the vectorization
8661 factor against the number of loop iterations if the loop is
8662 fully-masked.
8664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8665 Alan Hayward <alan.hayward@arm.com>
8666 David Sherwood <david.sherwood@arm.com>
8668 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8669 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8670 (dump_groups): Update accordingly.
8671 (iv_use::mem_type): New member variable.
8672 (address_p): New function.
8673 (record_use): Add a mem_type argument and initialize the new
8674 mem_type field.
8675 (record_group_use): Add a mem_type argument. Use address_p.
8676 Remove obsolete null checks of base_object. Update call to record_use.
8677 (find_interesting_uses_op): Update call to record_group_use.
8678 (find_interesting_uses_cond): Likewise.
8679 (find_interesting_uses_address): Likewise.
8680 (get_mem_type_for_internal_fn): New function.
8681 (find_address_like_use): Likewise.
8682 (find_interesting_uses_stmt): Try find_address_like_use before
8683 calling find_interesting_uses_op.
8684 (addr_offset_valid_p): Use the iv mem_type field as the type
8685 of the addressed memory.
8686 (add_autoinc_candidates): Likewise.
8687 (get_address_cost): Likewise.
8688 (split_small_address_groups_p): Use address_p.
8689 (split_address_groups): Likewise.
8690 (add_iv_candidate_for_use): Likewise.
8691 (autoinc_possible_for_pair): Likewise.
8692 (rewrite_groups): Likewise.
8693 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8694 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8695 (get_alias_ptr_type_for_ptr_address): New function.
8696 (rewrite_use_address): Rewrite address uses in calls that were
8697 identified by find_address_like_use.
8699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8700 Alan Hayward <alan.hayward@arm.com>
8701 David Sherwood <david.sherwood@arm.com>
8703 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8704 TARGET_MEM_REFs.
8705 * gimple-expr.h (is_gimple_addressable: Likewise.
8706 * gimple-expr.c (is_gimple_address): Likewise.
8707 * internal-fn.c (expand_call_mem_ref): New function.
8708 (expand_mask_load_optab_fn): Use it.
8709 (expand_mask_store_optab_fn): Likewise.
8711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8712 Alan Hayward <alan.hayward@arm.com>
8713 David Sherwood <david.sherwood@arm.com>
8715 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8716 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8717 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8718 (cond_umax@var{mode}): Document.
8719 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8720 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8721 (cond_umin_optab, cond_umax_optab): New optabs.
8722 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8723 (COND_IOR, COND_XOR): New internal functions.
8724 * internal-fn.h (get_conditional_internal_fn): Declare.
8725 * internal-fn.c (cond_binary_direct): New macro.
8726 (expand_cond_binary_optab_fn): Likewise.
8727 (direct_cond_binary_optab_supported_p): Likewise.
8728 (get_conditional_internal_fn): New function.
8729 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8730 Cope with reduction statements that are vectorized as calls rather
8731 than assignments.
8732 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8733 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8734 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8735 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8736 (UNSPEC_COND_EOR): New unspecs.
8737 (optab): Add mappings for them.
8738 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8739 (sve_int_op, sve_fp_op): New int attributes.
8741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8742 Alan Hayward <alan.hayward@arm.com>
8743 David Sherwood <david.sherwood@arm.com>
8745 * optabs.def (while_ult_optab): New optab.
8746 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8747 * internal-fn.def (WHILE_ULT): New internal function.
8748 * internal-fn.h (direct_internal_fn_supported_p): New override
8749 that takes two types as argument.
8750 * internal-fn.c (while_direct): New macro.
8751 (expand_while_optab_fn): New function.
8752 (convert_optab_supported_p): Likewise.
8753 (direct_while_optab_supported_p): New macro.
8754 * wide-int.h (wi::udiv_ceil): New function.
8755 * tree-vectorizer.h (rgroup_masks): New structure.
8756 (vec_loop_masks): New typedef.
8757 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8758 and fully_masked_p.
8759 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8760 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8761 (vect_max_vf): New function.
8762 (slpeel_make_loop_iterate_ntimes): Delete.
8763 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8764 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8765 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8766 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8767 internal-fn.h, stor-layout.h and optabs-query.h.
8768 (vect_set_loop_mask): New function.
8769 (add_preheader_seq): Likewise.
8770 (add_header_seq): Likewise.
8771 (interleave_supported_p): Likewise.
8772 (vect_maybe_permute_loop_masks): Likewise.
8773 (vect_set_loop_masks_directly): Likewise.
8774 (vect_set_loop_condition_masked): Likewise.
8775 (vect_set_loop_condition_unmasked): New function, split out from
8776 slpeel_make_loop_iterate_ntimes.
8777 (slpeel_make_loop_iterate_ntimes): Rename to..
8778 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8779 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8780 (vect_do_peeling): Update call accordingly.
8781 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8782 loops.
8783 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8784 mask_compare_type, can_fully_mask_p and fully_masked_p.
8785 (release_vec_loop_masks): New function.
8786 (_loop_vec_info): Use it to free the loop masks.
8787 (can_produce_all_loop_masks_p): New function.
8788 (vect_get_max_nscalars_per_iter): Likewise.
8789 (vect_verify_full_masking): Likewise.
8790 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8791 retries, and free the mask rgroups before retrying. Check loop-wide
8792 reasons for disallowing fully-masked loops. Make the final decision
8793 about whether use a fully-masked loop or not.
8794 (vect_estimate_min_profitable_iters): Do not assume that peeling
8795 for the number of iterations will be needed for fully-masked loops.
8796 (vectorizable_reduction): Disable fully-masked loops.
8797 (vectorizable_live_operation): Likewise.
8798 (vect_halve_mask_nunits): New function.
8799 (vect_double_mask_nunits): Likewise.
8800 (vect_record_loop_mask): Likewise.
8801 (vect_get_loop_mask): Likewise.
8802 (vect_transform_loop): Handle the case in which the final loop
8803 iteration might handle a partial vector. Call vect_set_loop_condition
8804 instead of slpeel_make_loop_iterate_ntimes.
8805 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8806 (check_load_store_masking): New function.
8807 (prepare_load_store_mask): Likewise.
8808 (vectorizable_store): Handle fully-masked loops.
8809 (vectorizable_load): Likewise.
8810 (supportable_widening_operation): Use vect_halve_mask_nunits for
8811 booleans.
8812 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8813 (vect_gen_while): New function.
8814 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8815 (aarch64_uqdec<mode>): New insn.
8817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8818 Alan Hayward <alan.hayward@arm.com>
8819 David Sherwood <david.sherwood@arm.com>
8821 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8822 (reduc_xor_scal_optab): New optabs.
8823 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8824 (reduc_xor_scal_@var{m}): Document.
8825 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8826 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8827 internal functions.
8828 * fold-const-call.c (fold_const_call): Handle them.
8829 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8830 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8831 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8832 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8833 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8834 (UNSPEC_XORV): New unspecs.
8835 (optab): Add entries for them.
8836 (BITWISEV): New int iterator.
8837 (bit_reduc_op): New int attributes.
8839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8840 Alan Hayward <alan.hayward@arm.com>
8841 David Sherwood <david.sherwood@arm.com>
8843 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8844 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8845 * optabs.def (vec_shl_insert_optab): New optab.
8846 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8847 (duplicate_and_interleave): Likewise.
8848 * tree-vect-loop.c: Include internal-fn.h.
8849 (neutral_op_for_slp_reduction): New function, split out from
8850 get_initial_defs_for_reduction.
8851 (get_initial_def_for_reduction): Handle option 2 for variable-length
8852 vectors by loading the neutral value into a vector and then shifting
8853 the initial value into element 0.
8854 (get_initial_defs_for_reduction): Replace the code argument with
8855 the neutral value calculated by neutral_op_for_slp_reduction.
8856 Use gimple_build_vector for constant-length vectors.
8857 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8858 but the first group_size elements have a neutral value.
8859 Use duplicate_and_interleave otherwise.
8860 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8861 Update call to get_initial_defs_for_reduction. Handle SLP
8862 reductions for variable-length vectors by creating one vector
8863 result for each scalar result, with the elements associated
8864 with other scalar results stubbed out with the neutral value.
8865 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8866 Require IFN_VEC_SHL_INSERT for double reductions on
8867 variable-length vectors, or SLP reductions that have
8868 a neutral value. Require can_duplicate_and_interleave_p
8869 support for variable-length unchained SLP reductions if there
8870 is no neutral value, such as for MIN/MAX reductions. Also require
8871 the number of vector elements to be a multiple of the number of
8872 SLP statements when doing variable-length unchained SLP reductions.
8873 Update call to vect_create_epilog_for_reduction.
8874 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8875 and remove initial values.
8876 (duplicate_and_interleave): Make public.
8877 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8878 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8881 Alan Hayward <alan.hayward@arm.com>
8882 David Sherwood <david.sherwood@arm.com>
8884 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8885 (can_duplicate_and_interleave_p): New function.
8886 (vect_get_and_check_slp_defs): Take the vector of statements
8887 rather than just the current one. Remove excess parentheses.
8888 Restriction rejectinon of vect_constant_def and vect_external_def
8889 for variable-length vectors to boolean types, or types for which
8890 can_duplicate_and_interleave_p is false.
8891 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8892 (duplicate_and_interleave): New function.
8893 (vect_get_constant_vectors): Use gimple_build_vector for
8894 constant-length vectors and suitable variable-length constant
8895 vectors. Use duplicate_and_interleave for other variable-length
8896 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8903 min_profitable_iters doesn't go negative.
8905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8906 Alan Hayward <alan.hayward@arm.com>
8907 David Sherwood <david.sherwood@arm.com>
8909 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8910 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8911 * optabs.def (vec_mask_load_lanes_optab): New optab.
8912 (vec_mask_store_lanes_optab): Likewise.
8913 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8914 (MASK_STORE_LANES): Likewise.
8915 * internal-fn.c (mask_load_lanes_direct): New macro.
8916 (mask_store_lanes_direct): Likewise.
8917 (expand_mask_load_optab_fn): Handle masked operations.
8918 (expand_mask_load_lanes_optab_fn): New macro.
8919 (expand_mask_store_optab_fn): Handle masked operations.
8920 (expand_mask_store_lanes_optab_fn): New macro.
8921 (direct_mask_load_lanes_optab_supported_p): Likewise.
8922 (direct_mask_store_lanes_optab_supported_p): Likewise.
8923 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8924 parameter.
8925 (vect_load_lanes_supported): Likewise.
8926 * tree-vect-data-refs.c (strip_conversion): New function.
8927 (can_group_stmts_p): Likewise.
8928 (vect_analyze_data_ref_accesses): Use it instead of checking
8929 for a pair of assignments.
8930 (vect_store_lanes_supported): Take a masked_p parameter.
8931 (vect_load_lanes_supported): Likewise.
8932 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8933 vect_store_lanes_supported and vect_load_lanes_supported.
8934 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8935 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8936 parameter. Don't allow gaps for masked accesses.
8937 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8938 and vect_load_lanes_supported.
8939 (get_load_store_type): Take a masked_p parameter and update
8940 call to get_group_load_store_type.
8941 (vectorizable_store): Update call to get_load_store_type.
8942 Handle IFN_MASK_STORE_LANES.
8943 (vectorizable_load): Update call to get_load_store_type.
8944 Handle IFN_MASK_LOAD_LANES.
8946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8947 Alan Hayward <alan.hayward@arm.com>
8948 David Sherwood <david.sherwood@arm.com>
8950 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8951 modes for SVE.
8952 * config/aarch64/aarch64-protos.h
8953 (aarch64_sve_struct_memory_operand_p): Declare.
8954 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8955 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8956 (VPRED, vpred): Handle SVE structure modes.
8957 * config/aarch64/constraints.md (Utx): New constraint.
8958 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8959 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8960 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8961 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8962 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8963 structure modes. Split into pieces after RA.
8964 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8965 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8966 New patterns.
8967 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8968 SVE structure modes.
8969 (aarch64_classify_address): Likewise.
8970 (sizetochar): Move earlier in file.
8971 (aarch64_print_operand): Handle SVE register lists.
8972 (aarch64_array_mode): New function.
8973 (aarch64_sve_struct_memory_operand_p): Likewise.
8974 (TARGET_ARRAY_MODE): Redefine.
8976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8977 Alan Hayward <alan.hayward@arm.com>
8978 David Sherwood <david.sherwood@arm.com>
8980 * target.def (array_mode): New target hook.
8981 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8982 * doc/tm.texi: Regenerate.
8983 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8984 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8985 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8986 targetm.array_mode.
8987 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8988 type sizes.
8990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8991 Alan Hayward <alan.hayward@arm.com>
8992 David Sherwood <david.sherwood@arm.com>
8994 * fold-const.c (fold_binary_loc): Check the argument types
8995 rather than the result type when testing for a vector operation.
8997 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8999 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9000 * doc/tm.texi: Regenerate.
9002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9003 Alan Hayward <alan.hayward@arm.com>
9004 David Sherwood <david.sherwood@arm.com>
9006 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9007 (sve): Document new AArch64 extension.
9008 * doc/md.texi (w): Extend the description of the AArch64
9009 constraint to include SVE vectors.
9010 (Upl, Upa): Document new AArch64 predicate constraints.
9011 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9012 enum.
9013 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9014 (msve-vector-bits=): New option.
9015 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9016 SVE when these are disabled.
9017 (sve): New extension.
9018 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9019 modes. Adjust their number of units based on aarch64_sve_vg.
9020 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9021 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9022 aarch64_addr_query_type.
9023 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9024 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9025 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9026 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9027 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9028 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9029 (aarch64_simd_imm_zero_p): Delete.
9030 (aarch64_check_zero_based_sve_index_immediate): Declare.
9031 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9032 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9033 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9034 (aarch64_sve_float_mul_immediate_p): Likewise.
9035 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9036 rather than an rtx.
9037 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9038 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9039 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9040 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9041 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9042 (aarch64_regmode_natural_size): Likewise.
9043 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9044 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9045 left one place.
9046 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9047 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9048 for VG and the SVE predicate registers.
9049 (V_ALIASES): Add a "z"-prefixed alias.
9050 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9051 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9052 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9053 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9054 (REG_CLASS_NAMES): Add entries for them.
9055 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9056 and the predicate registers.
9057 (aarch64_sve_vg): Declare.
9058 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9059 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9060 (REGMODE_NATURAL_SIZE): Define.
9061 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9062 SVE macros.
9063 * config/aarch64/aarch64.c: Include cfgrtl.h.
9064 (simd_immediate_info): Add a constructor for series vectors,
9065 and an associated step field.
9066 (aarch64_sve_vg): New variable.
9067 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9068 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9069 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9070 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9071 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9072 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9073 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9074 (aarch64_get_mask_mode): New functions.
9075 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9076 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9077 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9078 predicate modes and predicate registers. Explicitly restrict
9079 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9080 to store a vector mode if it is recognized by
9081 aarch64_classify_vector_mode.
9082 (aarch64_regmode_natural_size): New function.
9083 (aarch64_hard_regno_caller_save_mode): Return the original mode
9084 for predicates.
9085 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9086 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9087 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9088 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9089 functions.
9090 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9091 does not overlap dest if the function is frame-related. Handle
9092 SVE constants.
9093 (aarch64_split_add_offset): New function.
9094 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9095 them aarch64_add_offset.
9096 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9097 and update call to aarch64_sub_sp.
9098 (aarch64_add_cfa_expression): New function.
9099 (aarch64_expand_prologue): Pass extra temporary registers to the
9100 functions above. Handle the case in which we need to emit new
9101 DW_CFA_expressions for registers that were originally saved
9102 relative to the stack pointer, but now have to be expressed
9103 relative to the frame pointer.
9104 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9105 functions above.
9106 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9107 IP0 and IP1 values for SVE frames.
9108 (aarch64_expand_vec_series): New function.
9109 (aarch64_expand_sve_widened_duplicate): Likewise.
9110 (aarch64_expand_sve_const_vector): Likewise.
9111 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9112 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9113 into the register, rather than emitting a SET directly.
9114 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9115 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9116 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9117 (offset_9bit_signed_scaled_p): New functions.
9118 (aarch64_replicate_bitmask_imm): New function.
9119 (aarch64_bitmask_imm): Use it.
9120 (aarch64_cannot_force_const_mem): Reject expressions involving
9121 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9122 (aarch64_classify_index): Handle SVE indices, by requiring
9123 a plain register index with a scale that matches the element size.
9124 (aarch64_classify_address): Handle SVE addresses. Assert that
9125 the mode of the address is VOIDmode or an integer mode.
9126 Update call to aarch64_classify_symbol.
9127 (aarch64_classify_symbolic_expression): Update call to
9128 aarch64_classify_symbol.
9129 (aarch64_const_vec_all_in_range_p): New function.
9130 (aarch64_print_vector_float_operand): Likewise.
9131 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9132 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9133 and the FP immediates 1.0 and 0.5.
9134 (aarch64_print_address_internal): Handle SVE addresses.
9135 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9136 (aarch64_regno_regclass): Handle predicate registers.
9137 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9138 data modes.
9139 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9140 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9141 (aarch64_convert_sve_vector_bits): New function.
9142 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9143 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9144 rather than an rtx.
9145 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9146 Handle SVE vector and predicate modes. Accept VL-based constants
9147 that need only one temporary register, and VL offsets that require
9148 no temporary registers.
9149 (aarch64_conditional_register_usage): Mark the predicate registers
9150 as fixed if SVE isn't available.
9151 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9152 Return true for SVE vector and predicate modes.
9153 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9154 rather than an unsigned int. Handle SVE modes.
9155 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9156 SVE modes.
9157 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9158 if SVE is enabled.
9159 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9160 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9161 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9162 (aarch64_sve_float_mul_immediate_p): New functions.
9163 (aarch64_sve_valid_immediate): New function.
9164 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9165 Explicitly reject structure modes. Check for INDEX constants.
9166 Handle PTRUE and PFALSE constants.
9167 (aarch64_check_zero_based_sve_index_immediate): New function.
9168 (aarch64_simd_imm_zero_p): Delete.
9169 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9170 vector modes. Accept constants in the range of CNT[BHWD].
9171 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9172 ask for an Advanced SIMD mode.
9173 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9174 (aarch64_simd_vector_alignment): Handle SVE predicates.
9175 (aarch64_vectorize_preferred_vector_alignment): New function.
9176 (aarch64_simd_vector_alignment_reachable): Use it instead of
9177 the vector size.
9178 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9179 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9180 functions.
9181 (MAX_VECT_LEN): Delete.
9182 (expand_vec_perm_d): Add a vec_flags field.
9183 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9184 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9185 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9186 for SVE modes.
9187 (aarch64_evpc_rev): Rename to...
9188 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9189 (aarch64_evpc_rev_global): New function.
9190 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9191 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9192 MAX_VECT_LEN.
9193 (aarch64_evpc_sve_tbl): New function.
9194 (aarch64_expand_vec_perm_const_1): Update after rename of
9195 aarch64_evpc_rev. Handle SVE permutes too, trying
9196 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9197 than aarch64_evpc_tbl.
9198 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9199 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9200 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9201 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9202 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9203 (aarch64_expand_sve_vcond): New functions.
9204 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9205 of aarch64_vector_mode_p.
9206 (aarch64_dwarf_poly_indeterminate_value): New function.
9207 (aarch64_compute_pressure_classes): Likewise.
9208 (aarch64_can_change_mode_class): Likewise.
9209 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9210 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9211 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9212 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9213 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9214 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9215 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9216 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9217 constraints.
9218 (Dn, Dl, Dr): Accept const as well as const_vector.
9219 (Dz): Likewise. Compare against CONST0_RTX.
9220 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9221 of "vector" where appropriate.
9222 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9223 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9224 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9225 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9226 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9227 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9228 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9229 (v_int_equiv): Extend to SVE modes.
9230 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9231 mode attributes.
9232 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9233 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9234 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9235 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9236 (SVE_COND_FP_CMP): New int iterators.
9237 (perm_hilo): Handle the new unpack unspecs.
9238 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9239 attributes.
9240 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9241 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9242 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9243 (aarch64_equality_operator, aarch64_constant_vector_operand)
9244 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9245 (aarch64_sve_nonimmediate_operand): Likewise.
9246 (aarch64_sve_general_operand): Likewise.
9247 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9248 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9249 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9250 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9251 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9252 (aarch64_sve_float_arith_immediate): Likewise.
9253 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9254 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9255 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9256 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9257 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9258 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9259 (aarch64_sve_float_arith_operand): Likewise.
9260 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9261 (aarch64_sve_float_mul_operand): Likewise.
9262 (aarch64_sve_vec_perm_operand): Likewise.
9263 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9264 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9265 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9266 as well as const_vector.
9267 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9268 in file. Use CONST0_RTX and CONSTM1_RTX.
9269 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9270 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9271 Use aarch64_simd_imm_zero.
9272 * config/aarch64/aarch64-sve.md: New file.
9273 * config/aarch64/aarch64.md: Include it.
9274 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9275 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9276 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9277 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9278 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9279 (sve): New attribute.
9280 (enabled): Disable instructions with the sve attribute unless
9281 TARGET_SVE.
9282 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9283 aarch64_expand_mov_immediate.
9284 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9285 CNT[BHSD] immediates.
9286 (movti): Split CONST_POLY_INT moves into two halves.
9287 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9288 Split additions that need a temporary here if the destination
9289 is the stack pointer.
9290 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9291 (*add<mode>3_poly_1): New instruction.
9292 (set_clobber_cc): New expander.
9294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9296 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9297 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9298 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9299 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9300 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9301 Change innermode from fixed_mode_size to machine_mode.
9302 (simplify_subreg): Update call accordingly. Handle a constant-sized
9303 subreg of a variable-length CONST_VECTOR.
9305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9306 Alan Hayward <alan.hayward@arm.com>
9307 David Sherwood <david.sherwood@arm.com>
9309 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9310 (add_offset_to_base): New function, split out from...
9311 (create_mem_ref): ...here. When handling a scale other than 1,
9312 check first whether the address is valid without the offset.
9313 Add it into the base if so, leaving the index and scale as-is.
9315 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9317 PR c++/83778
9318 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9319 fold_for_warn before checking if arg2 is INTEGER_CST.
9321 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9323 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9324 (store_multiple_operation): Delete.
9325 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9326 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9327 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9328 guarded by TARGET_STRING.
9329 (rs6000_output_load_multiple): Delete.
9330 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9331 OPTION_MASK_STRING / TARGET_STRING handling.
9332 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9333 (const rs6000_opt_masks) <"string">: Change mask to 0.
9334 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9335 (MASK_STRING): Delete.
9336 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9337 parts. Simplify.
9338 (load_multiple): Delete.
9339 (*ldmsi8): Delete.
9340 (*ldmsi7): Delete.
9341 (*ldmsi6): Delete.
9342 (*ldmsi5): Delete.
9343 (*ldmsi4): Delete.
9344 (*ldmsi3): Delete.
9345 (store_multiple): Delete.
9346 (*stmsi8): Delete.
9347 (*stmsi7): Delete.
9348 (*stmsi6): Delete.
9349 (*stmsi5): Delete.
9350 (*stmsi4): Delete.
9351 (*stmsi3): Delete.
9352 (movmemsi_8reg): Delete.
9353 (corresponding unnamed define_insn): Delete.
9354 (movmemsi_6reg): Delete.
9355 (corresponding unnamed define_insn): Delete.
9356 (movmemsi_4reg): Delete.
9357 (corresponding unnamed define_insn): Delete.
9358 (movmemsi_2reg): Delete.
9359 (corresponding unnamed define_insn): Delete.
9360 (movmemsi_1reg): Delete.
9361 (corresponding unnamed define_insn): Delete.
9362 * config/rs6000/rs6000.opt (mno-string): New.
9363 (mstring): Replace by deprecation warning stub.
9364 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9366 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9368 * regrename.c (regrename_do_replace): If replacing the same
9369 reg multiple times, try to reuse last created gen_raw_REG.
9371 PR debug/81155
9372 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9373 main to workaround a bug in GDB.
9375 2018-01-12 Tom de Vries <tom@codesourcery.com>
9377 PR target/83737
9378 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9380 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9382 PR rtl-optimization/80481
9383 * ira-color.c (get_cap_member): New function.
9384 (allocnos_conflict_by_live_ranges_p): Use it.
9385 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9386 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9388 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9390 PR target/83628
9391 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9392 (*saddl_se_1): Ditto.
9393 (*ssubsi_1): Ditto.
9394 (*ssubl_se_1): Ditto.
9396 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9398 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9399 rather than wi::to_widest for DR_INITs.
9400 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9401 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9402 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9403 INTEGER_CSTs.
9404 (vect_analyze_group_access_1): Note that here.
9406 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9408 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9409 polynomial type sizes.
9411 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9413 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9414 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9415 (gimple_add_tmp_var): Likewise.
9417 2018-01-12 Martin Liska <mliska@suse.cz>
9419 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9420 (gimple_alloc_sizes): Likewise.
9421 (dump_gimple_statistics): Use PRIu64 in printf format.
9422 * gimple.h: Change uint64_t to int.
9424 2018-01-12 Martin Liska <mliska@suse.cz>
9426 * tree-core.h: Use uint64_t instead of int.
9427 * tree.c (tree_node_counts): Likewise.
9428 (tree_node_sizes): Likewise.
9429 (dump_tree_statistics): Use PRIu64 in printf format.
9431 2018-01-12 Martin Liska <mliska@suse.cz>
9433 * Makefile.in: As qsort_chk is implemented in vec.c, add
9434 vec.o to linkage of gencfn-macros.
9435 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9436 passing the info to record_node_allocation_statistics.
9437 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9438 and pass the info.
9439 * ggc-common.c (struct ggc_usage): Add operator== and use
9440 it in operator< and compare function.
9441 * mem-stats.h (struct mem_usage): Likewise.
9442 * vec.c (struct vec_usage): Remove operator< and compare
9443 function. Can be simply inherited.
9445 2018-01-12 Martin Jambor <mjambor@suse.cz>
9447 PR target/81616
9448 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9449 * tree-ssa-math-opts.c: Include domwalk.h.
9450 (convert_mult_to_fma_1): New function.
9451 (fma_transformation_info): New type.
9452 (fma_deferring_state): Likewise.
9453 (cancel_fma_deferring): New function.
9454 (result_of_phi): Likewise.
9455 (last_fma_candidate_feeds_initial_phi): Likewise.
9456 (convert_mult_to_fma): Added deferring logic, split actual
9457 transformation to convert_mult_to_fma_1.
9458 (math_opts_dom_walker): New type.
9459 (math_opts_dom_walker::after_dom_children): New method, body moved
9460 here from pass_optimize_widening_mul::execute, added deferring logic
9461 bits.
9462 (pass_optimize_widening_mul::execute): Moved most of code to
9463 math_opts_dom_walker::after_dom_children.
9464 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9465 * config/i386/i386.c (ix86_option_override_internal): Added
9466 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9468 2018-01-12 Richard Biener <rguenther@suse.de>
9470 PR debug/83157
9471 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9472 inline instance vars.
9474 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9476 PR target/81819
9477 * config/rx/rx.c (rx_is_restricted_memory_address):
9478 Handle SUBREG case.
9480 2018-01-12 Richard Biener <rguenther@suse.de>
9482 PR tree-optimization/80846
9483 * target.def (split_reduction): New target hook.
9484 * targhooks.c (default_split_reduction): New function.
9485 * targhooks.h (default_split_reduction): Declare.
9486 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9487 target requests first reduce vectors by combining low and high
9488 parts.
9489 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9490 (get_vectype_for_scalar_type_and_size): Export.
9491 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9492 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9493 * doc/tm.texi: Regenerate.
9494 * config/i386/i386.c (ix86_split_reduction): Implement
9495 TARGET_VECTORIZE_SPLIT_REDUCTION.
9497 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9499 PR target/83368
9500 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9501 in PIC mode except for TARGET_VXWORKS_RTP.
9502 * config/sparc/sparc.c: Include cfgrtl.h.
9503 (TARGET_INIT_PIC_REG): Define.
9504 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9505 (sparc_pic_register_p): New predicate.
9506 (sparc_legitimate_address_p): Use it.
9507 (sparc_legitimize_pic_address): Likewise.
9508 (sparc_delegitimize_address): Likewise.
9509 (sparc_mode_dependent_address_p): Likewise.
9510 (gen_load_pcrel_sym): Remove 4th parameter.
9511 (load_got_register): Adjust call to above. Remove obsolete stuff.
9512 (sparc_expand_prologue): Do not call load_got_register here.
9513 (sparc_flat_expand_prologue): Likewise.
9514 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9515 (sparc_use_pseudo_pic_reg): New function.
9516 (sparc_init_pic_reg): Likewise.
9517 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9518 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9520 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9522 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9523 Add item for branch_cost.
9525 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9527 PR rtl-optimization/83565
9528 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9529 not extend the result to a larger mode for rotate operations.
9530 (num_sign_bit_copies1): Likewise.
9532 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9534 PR target/40411
9535 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9536 -symbolic.
9537 Use values-Xc.o for -pedantic.
9538 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9540 2018-01-12 Martin Liska <mliska@suse.cz>
9542 PR ipa/83054
9543 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9544 New function.
9545 (possible_polymorphic_call_targets): Use it.
9546 (ipa_devirt): Likewise.
9548 2018-01-12 Martin Liska <mliska@suse.cz>
9550 * profile-count.h (enum profile_quality): Use 0 as invalid
9551 enum value of profile_quality.
9553 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9555 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9556 -mext-string options.
9558 2018-01-12 Richard Biener <rguenther@suse.de>
9560 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9561 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9562 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9563 Likewise.
9564 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9566 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9568 * configure.ac (--with-long-double-format): Add support for the
9569 configuration option to change the default long double format on
9570 PowerPC systems.
9571 * config.gcc (powerpc*-linux*-*): Likewise.
9572 * configure: Regenerate.
9573 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9574 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9575 used without modification.
9577 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9579 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9580 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9581 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9582 MISC_BUILTIN_SPEC_BARRIER.
9583 (rs6000_init_builtins): Likewise.
9584 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9585 enum value.
9586 (speculation_barrier): New define_insn.
9587 * doc/extend.texi: Document __builtin_speculation_barrier.
9589 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9591 PR target/83203
9592 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9593 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9594 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9595 iterators.
9596 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9597 integral modes instead of "ss" and "sd".
9598 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9599 vectors with 32-bit and 64-bit elements.
9600 (vecdupssescalarmodesuffix): New mode attribute.
9601 (vec_dup<mode>): Use it.
9603 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9605 PR target/83330
9606 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9607 frame if argument is passed on stack.
9609 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9611 PR target/82682
9612 * ree.c (combine_reaching_defs): Optimize also
9613 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9614 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9616 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9618 PR middle-end/83189
9619 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9621 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9623 PR middle-end/83718
9624 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9625 after they are computed.
9627 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9629 PR tree-optimization/83695
9630 * gimple-loop-linterchange.cc
9631 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9632 reset cached scev information after interchange.
9633 (pass_linterchange::execute): Remove call to scev_reset_htab.
9635 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9637 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9638 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9639 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9640 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9641 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9642 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9643 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9644 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9645 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9646 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9647 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9648 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9649 (V_lane_reg): Likewise.
9650 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9651 New define_expand.
9652 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9653 (vfmal_lane_low<mode>_intrinsic,
9654 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9655 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9656 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9657 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9658 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9659 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9661 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9663 * config/arm/arm-cpus.in (fp16fml): New feature.
9664 (ALL_SIMD): Add fp16fml.
9665 (armv8.2-a): Add fp16fml as an option.
9666 (armv8.3-a): Likewise.
9667 (armv8.4-a): Add fp16fml as part of fp16.
9668 * config/arm/arm.h (TARGET_FP16FML): Define.
9669 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9670 when appropriate.
9671 * config/arm/arm-modes.def (V2HF): Define.
9672 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9673 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9674 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9675 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9676 vfmsl_low, vfmsl_high): New set of builtins.
9677 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9678 (vfml_op): New code attribute.
9679 (VFMLHALVES): New int iterator.
9680 (VFML, VFMLSEL): New mode attributes.
9681 (V_reg): Define mapping for V2HF.
9682 (V_hi, V_lo): New mode attributes.
9683 (VF_constraint): Likewise.
9684 (vfml_half, vfml_half_selector): New int attributes.
9685 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9686 define_expand.
9687 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9688 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9689 New define_insn.
9690 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9691 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9692 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9693 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9694 documentation.
9695 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9696 Document new effective target and option set.
9698 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9700 * config/arm/arm-cpus.in (armv8_4): New feature.
9701 (ARMv8_4a): New fgroup.
9702 (armv8.4-a): New arch.
9703 * config/arm/arm-tables.opt: Regenerate.
9704 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9705 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9706 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9707 Add matching rules for -march=armv8.4-a and extensions.
9708 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9710 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9712 PR target/81821
9713 * config/rx/rx.md (BW): New mode attribute.
9714 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9716 2018-01-11 Richard Biener <rguenther@suse.de>
9718 PR tree-optimization/83435
9719 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9720 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9721 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9723 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9724 Alan Hayward <alan.hayward@arm.com>
9725 David Sherwood <david.sherwood@arm.com>
9727 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9728 field.
9729 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9730 (aarch64_print_address_internal): Use it to check for a zero offset.
9732 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9733 Alan Hayward <alan.hayward@arm.com>
9734 David Sherwood <david.sherwood@arm.com>
9736 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9737 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9738 Return a poly_int64 rather than a HOST_WIDE_INT.
9739 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9740 rather than a HOST_WIDE_INT.
9741 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9742 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9743 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9744 final_offset from HOST_WIDE_INT to poly_int64.
9745 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9746 to_constant when getting the number of units in an Advanced SIMD
9747 mode.
9748 (aarch64_builtin_vectorized_function): Check for a constant number
9749 of units.
9750 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9751 GET_MODE_SIZE.
9752 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9753 attribute instead of GET_MODE_NUNITS.
9754 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9755 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9756 GET_MODE_SIZE for fixed-size registers.
9757 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9758 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9759 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9760 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9761 (aarch64_print_operand, aarch64_print_address_internal)
9762 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9763 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9764 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9765 Handle polynomial GET_MODE_SIZE.
9766 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9767 wider than SImode without modification.
9768 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9769 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9770 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9771 passing and returning SVE modes.
9772 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9773 rather than GEN_INT.
9774 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9775 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9776 (aarch64_allocate_and_probe_stack_space): Likewise.
9777 (aarch64_layout_frame): Cope with polynomial offsets.
9778 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9779 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9780 polynomial offsets.
9781 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9782 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9783 poly_int64 rather than a HOST_WIDE_INT.
9784 (aarch64_get_separate_components, aarch64_process_components)
9785 (aarch64_expand_prologue, aarch64_expand_epilogue)
9786 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9787 (aarch64_anchor_offset): New function, split out from...
9788 (aarch64_legitimize_address): ...here.
9789 (aarch64_builtin_vectorization_cost): Handle polynomial
9790 TYPE_VECTOR_SUBPARTS.
9791 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9792 GET_MODE_NUNITS.
9793 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9794 number of elements from the PARALLEL rather than the mode.
9795 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9796 rather than GET_MODE_BITSIZE.
9797 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9798 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9799 (aarch64_expand_vec_perm_const_1): Handle polynomial
9800 d->perm.length () and d->perm elements.
9801 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9802 Apply to_constant to d->perm elements.
9803 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9804 polynomial CONST_VECTOR_NUNITS.
9805 (aarch64_move_pointer): Take amount as a poly_int64 rather
9806 than an int.
9807 (aarch64_progress_pointer): Avoid temporary variable.
9808 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9809 the mode attribute instead of GET_MODE.
9811 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9812 Alan Hayward <alan.hayward@arm.com>
9813 David Sherwood <david.sherwood@arm.com>
9815 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9816 x exists before using it.
9817 (aarch64_add_constant_internal): Rename to...
9818 (aarch64_add_offset_1): ...this. Replace regnum with separate
9819 src and dest rtxes. Handle the case in which they're different,
9820 including when the offset is zero. Replace scratchreg with an rtx.
9821 Use 2 additions if there is no spare register into which we can
9822 move a 16-bit constant.
9823 (aarch64_add_constant): Delete.
9824 (aarch64_add_offset): Replace reg with separate src and dest
9825 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9826 Use aarch64_add_offset_1.
9827 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9828 an rtx rather than an int. Take the delta as a poly_int64
9829 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9830 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9831 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9832 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9833 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9834 and aarch64_add_sp.
9835 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9836 aarch64_add_constant.
9838 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9840 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9841 Use scalar_float_mode.
9843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9845 * config/aarch64/aarch64-simd.md
9846 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9847 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9848 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9849 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9850 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9851 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9852 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9853 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9854 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9855 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9857 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9859 PR target/83514
9860 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9861 targ_options->x_arm_arch_string is non NULL.
9863 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9865 * config/aarch64/aarch64.h
9866 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9868 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9870 PR target/82096
9871 * expmed.c (emit_store_flag_force): Swap if const op0
9872 and change VOIDmode to mode of op0.
9874 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9876 PR rtl-optimization/83761
9877 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9878 than bytes to mode_for_size.
9880 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9882 PR middle-end/83189
9883 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9884 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9885 profile.
9887 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9889 PR middle-end/83575
9890 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9891 when in layout mode.
9892 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9893 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9894 partition fixup.
9896 2018-01-10 Michael Collison <michael.collison@arm.com>
9898 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9899 * config/aarch64/aarch64-option-extension.def: Add
9900 AARCH64_OPT_EXTENSION of 'fp16fml'.
9901 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9902 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9903 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9904 * config/aarch64/constraints.md (Ui7): New constraint.
9905 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9906 (VFMLA_SEL_W): Ditto.
9907 (f16quad): Ditto.
9908 (f16mac1): Ditto.
9909 (VFMLA16_LOW): New int iterator.
9910 (VFMLA16_HIGH): Ditto.
9911 (UNSPEC_FMLAL): New unspec.
9912 (UNSPEC_FMLSL): Ditto.
9913 (UNSPEC_FMLAL2): Ditto.
9914 (UNSPEC_FMLSL2): Ditto.
9915 (f16mac): New code attribute.
9916 * config/aarch64/aarch64-simd-builtins.def
9917 (aarch64_fmlal_lowv2sf): Ditto.
9918 (aarch64_fmlsl_lowv2sf): Ditto.
9919 (aarch64_fmlalq_lowv4sf): Ditto.
9920 (aarch64_fmlslq_lowv4sf): Ditto.
9921 (aarch64_fmlal_highv2sf): Ditto.
9922 (aarch64_fmlsl_highv2sf): Ditto.
9923 (aarch64_fmlalq_highv4sf): Ditto.
9924 (aarch64_fmlslq_highv4sf): Ditto.
9925 (aarch64_fmlal_lane_lowv2sf): Ditto.
9926 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9927 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9928 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9929 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9930 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9931 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9932 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9933 (aarch64_fmlal_lane_highv2sf): Ditto.
9934 (aarch64_fmlsl_lane_highv2sf): Ditto.
9935 (aarch64_fmlal_laneq_highv2sf): Ditto.
9936 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9937 (aarch64_fmlalq_lane_highv4sf): Ditto.
9938 (aarch64_fmlsl_lane_highv4sf): Ditto.
9939 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9940 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9941 * config/aarch64/aarch64-simd.md:
9942 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9943 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9944 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9945 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9946 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9947 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9948 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9949 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9950 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9951 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9952 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9953 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9954 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9955 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9956 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9957 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9958 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9959 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9960 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9961 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9962 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9963 (vfmlsl_low_u32): Ditto.
9964 (vfmlalq_low_u32): Ditto.
9965 (vfmlslq_low_u32): Ditto.
9966 (vfmlal_high_u32): Ditto.
9967 (vfmlsl_high_u32): Ditto.
9968 (vfmlalq_high_u32): Ditto.
9969 (vfmlslq_high_u32): Ditto.
9970 (vfmlal_lane_low_u32): Ditto.
9971 (vfmlsl_lane_low_u32): Ditto.
9972 (vfmlal_laneq_low_u32): Ditto.
9973 (vfmlsl_laneq_low_u32): Ditto.
9974 (vfmlalq_lane_low_u32): Ditto.
9975 (vfmlslq_lane_low_u32): Ditto.
9976 (vfmlalq_laneq_low_u32): Ditto.
9977 (vfmlslq_laneq_low_u32): Ditto.
9978 (vfmlal_lane_high_u32): Ditto.
9979 (vfmlsl_lane_high_u32): Ditto.
9980 (vfmlal_laneq_high_u32): Ditto.
9981 (vfmlsl_laneq_high_u32): Ditto.
9982 (vfmlalq_lane_high_u32): Ditto.
9983 (vfmlslq_lane_high_u32): Ditto.
9984 (vfmlalq_laneq_high_u32): Ditto.
9985 (vfmlslq_laneq_high_u32): Ditto.
9986 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9987 (AARCH64_FL_FOR_ARCH8_4): New.
9988 (AARCH64_ISA_F16FML): New ISA flag.
9989 (TARGET_F16FML): New feature flag for fp16fml.
9990 (doc/invoke.texi): Document new fp16fml option.
9992 2018-01-10 Michael Collison <michael.collison@arm.com>
9994 * config/aarch64/aarch64-builtins.c:
9995 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9996 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9997 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9998 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9999 (AARCH64_ISA_SHA3): New ISA flag.
10000 (TARGET_SHA3): New feature flag for sha3.
10001 * config/aarch64/iterators.md (sha512_op): New int attribute.
10002 (CRYPTO_SHA512): New int iterator.
10003 (UNSPEC_SHA512H): New unspec.
10004 (UNSPEC_SHA512H2): Ditto.
10005 (UNSPEC_SHA512SU0): Ditto.
10006 (UNSPEC_SHA512SU1): Ditto.
10007 * config/aarch64/aarch64-simd-builtins.def
10008 (aarch64_crypto_sha512hqv2di): New builtin.
10009 (aarch64_crypto_sha512h2qv2di): Ditto.
10010 (aarch64_crypto_sha512su0qv2di): Ditto.
10011 (aarch64_crypto_sha512su1qv2di): Ditto.
10012 (aarch64_eor3qv8hi): Ditto.
10013 (aarch64_rax1qv2di): Ditto.
10014 (aarch64_xarqv2di): Ditto.
10015 (aarch64_bcaxqv8hi): Ditto.
10016 * config/aarch64/aarch64-simd.md:
10017 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10018 (aarch64_crypto_sha512su0qv2di): Ditto.
10019 (aarch64_crypto_sha512su1qv2di): Ditto.
10020 (aarch64_eor3qv8hi): Ditto.
10021 (aarch64_rax1qv2di): Ditto.
10022 (aarch64_xarqv2di): Ditto.
10023 (aarch64_bcaxqv8hi): Ditto.
10024 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10025 (vsha512h2q_u64): Ditto.
10026 (vsha512su0q_u64): Ditto.
10027 (vsha512su1q_u64): Ditto.
10028 (veor3q_u16): Ditto.
10029 (vrax1q_u64): Ditto.
10030 (vxarq_u64): Ditto.
10031 (vbcaxq_u16): Ditto.
10032 * config/arm/types.md (crypto_sha512): New type attribute.
10033 (crypto_sha3): Ditto.
10034 (doc/invoke.texi): Document new sha3 option.
10036 2018-01-10 Michael Collison <michael.collison@arm.com>
10038 * config/aarch64/aarch64-builtins.c:
10039 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10040 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10041 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10042 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10043 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10044 (AARCH64_ISA_SM4): New ISA flag.
10045 (TARGET_SM4): New feature flag for sm4.
10046 * config/aarch64/aarch64-simd-builtins.def
10047 (aarch64_sm3ss1qv4si): Ditto.
10048 (aarch64_sm3tt1aq4si): Ditto.
10049 (aarch64_sm3tt1bq4si): Ditto.
10050 (aarch64_sm3tt2aq4si): Ditto.
10051 (aarch64_sm3tt2bq4si): Ditto.
10052 (aarch64_sm3partw1qv4si): Ditto.
10053 (aarch64_sm3partw2qv4si): Ditto.
10054 (aarch64_sm4eqv4si): Ditto.
10055 (aarch64_sm4ekeyqv4si): Ditto.
10056 * config/aarch64/aarch64-simd.md:
10057 (aarch64_sm3ss1qv4si): Ditto.
10058 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10059 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10060 (aarch64_sm4eqv4si): Ditto.
10061 (aarch64_sm4ekeyqv4si): Ditto.
10062 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10063 (sm3part_op): Ditto.
10064 (CRYPTO_SM3TT): Ditto.
10065 (CRYPTO_SM3PART): Ditto.
10066 (UNSPEC_SM3SS1): New unspec.
10067 (UNSPEC_SM3TT1A): Ditto.
10068 (UNSPEC_SM3TT1B): Ditto.
10069 (UNSPEC_SM3TT2A): Ditto.
10070 (UNSPEC_SM3TT2B): Ditto.
10071 (UNSPEC_SM3PARTW1): Ditto.
10072 (UNSPEC_SM3PARTW2): Ditto.
10073 (UNSPEC_SM4E): Ditto.
10074 (UNSPEC_SM4EKEY): Ditto.
10075 * config/aarch64/constraints.md (Ui2): New constraint.
10076 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10077 * config/arm/types.md (crypto_sm3): New type attribute.
10078 (crypto_sm4): Ditto.
10079 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10080 (vsm3tt1aq_u32): Ditto.
10081 (vsm3tt1bq_u32): Ditto.
10082 (vsm3tt2aq_u32): Ditto.
10083 (vsm3tt2bq_u32): Ditto.
10084 (vsm3partw1q_u32): Ditto.
10085 (vsm3partw2q_u32): Ditto.
10086 (vsm4eq_u32): Ditto.
10087 (vsm4ekeyq_u32): Ditto.
10088 (doc/invoke.texi): Document new sm4 option.
10090 2018-01-10 Michael Collison <michael.collison@arm.com>
10092 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10093 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10094 (AARCH64_FL_FOR_ARCH8_4): New.
10095 (AARCH64_FL_V8_4): New flag.
10096 (doc/invoke.texi): Document new armv8.4-a option.
10098 2018-01-10 Michael Collison <michael.collison@arm.com>
10100 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10101 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10102 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10103 * config/aarch64/aarch64-option-extension.def: Add
10104 AARCH64_OPT_EXTENSION of 'sha2'.
10105 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10106 (crypto): Disable sha2 and aes if crypto disabled.
10107 (crypto): Enable aes and sha2 if enabled.
10108 (simd): Disable sha2 and aes if simd disabled.
10109 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10110 New flags.
10111 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10112 (TARGET_SHA2): New feature flag for sha2.
10113 (TARGET_AES): New feature flag for aes.
10114 * config/aarch64/aarch64-simd.md:
10115 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10116 conditional on TARGET_AES.
10117 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10118 (aarch64_crypto_sha1hsi): Make pattern conditional
10119 on TARGET_SHA2.
10120 (aarch64_crypto_sha1hv4si): Ditto.
10121 (aarch64_be_crypto_sha1hv4si): Ditto.
10122 (aarch64_crypto_sha1su1v4si): Ditto.
10123 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10124 (aarch64_crypto_sha1su0v4si): Ditto.
10125 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10126 (aarch64_crypto_sha256su0v4si): Ditto.
10127 (aarch64_crypto_sha256su1v4si): Ditto.
10128 (doc/invoke.texi): Document new aes and sha2 options.
10130 2018-01-10 Martin Sebor <msebor@redhat.com>
10132 PR tree-optimization/83781
10133 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10134 as string arrays.
10136 2018-01-11 Martin Sebor <msebor@gmail.com>
10137 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10139 PR tree-optimization/83501
10140 PR tree-optimization/81703
10142 * tree-ssa-strlen.c (get_string_cst): Rename...
10143 (get_string_len): ...to this. Handle global constants.
10144 (handle_char_store): Adjust.
10146 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10147 Jim Wilson <jimw@sifive.com>
10149 * config/riscv/riscv-protos.h (riscv_output_return): New.
10150 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10151 (riscv_attribute_table, riscv_output_return),
10152 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10153 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10154 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10155 function.
10156 (riscv_expand_prologue): Add early return for naked function.
10157 (riscv_expand_epilogue): Likewise.
10158 (riscv_function_ok_for_sibcall): Return false for naked function.
10159 (riscv_set_current_function): New.
10160 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10161 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10162 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10163 * doc/extend.texi (RISC-V Function Attributes): New.
10165 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10167 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10168 check for 128-bit long double before checking TCmode.
10169 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10170 128-bit long doubles before checking TFmode or TCmode.
10171 (FLOAT128_IBM_P): Likewise.
10173 2018-01-10 Martin Sebor <msebor@redhat.com>
10175 PR tree-optimization/83671
10176 * builtins.c (c_strlen): Unconditionally return zero for the empty
10177 string.
10178 Use -Warray-bounds for warnings.
10179 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10180 for non-constant array indices with COMPONENT_REF, arrays of
10181 arrays, and pointers to arrays.
10182 (gimple_fold_builtin_strlen): Determine and set length range for
10183 non-constant character arrays.
10185 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10187 PR middle-end/81897
10188 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10189 empty blocks.
10191 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10193 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10195 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10197 PR target/83399
10198 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10199 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10200 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10201 indexed_or_indirect_operand predicate.
10202 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10203 (*vsx_le_perm_load_v8hi): Likewise.
10204 (*vsx_le_perm_load_v16qi): Likewise.
10205 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10206 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10207 (*vsx_le_perm_store_v8hi): Likewise.
10208 (*vsx_le_perm_store_v16qi): Likewise.
10209 (eight unnamed splitters): Likewise.
10211 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10213 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10214 * config/rs6000/emmintrin.h: Likewise.
10215 * config/rs6000/mmintrin.h: Likewise.
10216 * config/rs6000/xmmintrin.h: Likewise.
10218 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10220 PR c++/43486
10221 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10222 "public_flag".
10223 * tree.c (tree_nop_conversion): Return true for location wrapper
10224 nodes.
10225 (maybe_wrap_with_location): New function.
10226 (selftest::check_strip_nops): New function.
10227 (selftest::test_location_wrappers): New function.
10228 (selftest::tree_c_tests): Call it.
10229 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10230 (maybe_wrap_with_location): New decl.
10231 (EXPR_LOCATION_WRAPPER_P): New macro.
10232 (location_wrapper_p): New inline function.
10233 (tree_strip_any_location_wrapper): New inline function.
10235 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10237 PR target/83735
10238 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10239 stack_realign_offset for the largest alignment of stack slot
10240 actually used.
10241 (ix86_find_max_used_stack_alignment): New function.
10242 (ix86_finalize_stack_frame_flags): Use it. Set
10243 max_used_stack_alignment if we don't realign stack.
10244 * config/i386/i386.h (machine_function): Add
10245 max_used_stack_alignment.
10247 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10249 * config/arm/arm.opt (-mbranch-cost): New option.
10250 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10251 account.
10253 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10255 PR target/83629
10256 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10257 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10259 2018-01-10 Richard Biener <rguenther@suse.de>
10261 PR debug/83765
10262 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10263 early out so it also covers the case where we have a non-NULL
10264 origin.
10266 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10268 PR tree-optimization/83753
10269 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10270 for non-strided grouped accesses if the number of elements is 1.
10272 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10274 PR target/81616
10275 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10276 * i386.h (TARGET_USE_GATHER): Define.
10277 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10279 2018-01-10 Martin Liska <mliska@suse.cz>
10281 PR bootstrap/82831
10282 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10283 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10284 partitioning.
10285 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10286 CLEANUP_NO_PARTITIONING is not set.
10288 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10290 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10291 for vectors, as a partial revert of r254296.
10292 * rtl.h (const_vec_p): Delete.
10293 (const_vec_duplicate_p): Don't test for vector CONSTs.
10294 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10295 * expmed.c (make_tree): Likewise.
10297 Revert:
10298 * common.md (E, F): Use CONSTANT_P instead of checking for
10299 CONST_VECTOR.
10300 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10301 checking for CONST_VECTOR.
10303 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10305 PR middle-end/83575
10306 * predict.c (force_edge_cold): Handle in more sane way edges
10307 with no prediction.
10309 2018-01-09 Carl Love <cel@us.ibm.com>
10311 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10312 V4SI, V4SF types.
10313 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10314 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10315 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10316 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10317 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10318 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10319 * config/rs6000/rs6000-protos.h: Add extern defition for
10320 rs6000_generate_float2_double_code.
10321 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10322 function.
10323 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10324 (float2_v2df): Add define_expand.
10326 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10328 PR target/83628
10329 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10330 op_mode in the force_to_mode call.
10332 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10334 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10335 instead of checking each element individually.
10336 (aarch64_evpc_uzp): Likewise.
10337 (aarch64_evpc_zip): Likewise.
10338 (aarch64_evpc_ext): Likewise.
10339 (aarch64_evpc_rev): Likewise.
10340 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10341 instead of checking each element individually. Return true without
10342 generating rtl if
10343 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10344 whether all selected elements come from the same input, instead of
10345 checking each element individually. Remove calls to gen_rtx_REG,
10346 start_sequence and end_sequence and instead assert that no rtl is
10347 generated.
10349 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10351 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10352 order of HIGH and CONST checks.
10354 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10356 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10357 if the destination isn't an SSA_NAME.
10359 2018-01-09 Richard Biener <rguenther@suse.de>
10361 PR tree-optimization/83668
10362 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10363 move prologue...
10364 (canonicalize_loop_form): ... here, renamed from ...
10365 (canonicalize_loop_closed_ssa_form): ... this and amended to
10366 swap successor edges for loop exit blocks to make us use
10367 the RPO order we need for initial schedule generation.
10369 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10371 PR tree-optimization/64811
10372 * match.pd: When optimizing comparisons with Inf, avoid
10373 introducing or losing exceptions from comparisons with NaN.
10375 2018-01-09 Martin Liska <mliska@suse.cz>
10377 PR sanitizer/82517
10378 * asan.c (shadow_mem_size): Add gcc_assert.
10380 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10382 Don't save registers in main().
10384 PR target/83738
10385 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10386 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10387 * config/avr/avr.c (avr_set_current_function): Don't error if
10388 naked, OS_task or OS_main are specified at the same time.
10389 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10390 OS_main.
10391 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10392 attribute.
10393 * common/config/avr/avr-common.c (avr_option_optimization_table):
10394 Switch on -mmain-is-OS_task for optimizing compilations.
10396 2018-01-09 Richard Biener <rguenther@suse.de>
10398 PR tree-optimization/83572
10399 * graphite.c: Include cfganal.h.
10400 (graphite_transform_loops): Connect infinite loops to exit
10401 and remove fake edges at the end.
10403 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10405 * ipa-inline.c (edge_badness): Revert accidental checkin.
10407 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10409 PR ipa/80763
10410 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10411 symbols; not inline clones.
10413 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10415 PR target/83507
10416 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10417 hard registers. Formatting fixes.
10419 PR preprocessor/83722
10420 * gcc.c (try_generate_repro): Pass
10421 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10422 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10423 do_report_bug.
10425 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10426 Kito Cheng <kito.cheng@gmail.com>
10428 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10429 (riscv_leaf_function_p): Delete.
10430 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10432 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10434 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10435 function.
10436 (do_ifelse): New function.
10437 (do_isel): New function.
10438 (do_sub3): New function.
10439 (do_add3): New function.
10440 (do_load_mask_compare): New function.
10441 (do_overlap_load_compare): New function.
10442 (expand_compare_loop): New function.
10443 (expand_block_compare): Call expand_compare_loop() when appropriate.
10444 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10445 option description.
10446 (-mblock-compare-inline-loop-limit): New option.
10448 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10450 PR target/83677
10451 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10452 Reverse order of second and third operands in first alternative.
10453 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10454 of first and second elements in UNSPEC_VPERMR vector.
10455 (altivec_expand_vec_perm_le): Likewise.
10457 2018-01-08 Jeff Law <law@redhat.com>
10459 PR rtl-optimizatin/81308
10460 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10461 (process_switch): If group_case_labels makes a change, then set
10462 cfg_altered.
10463 (pass_convert_switch::execute): If a switch is converted, then
10464 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10466 PR rtl-optimization/81308
10467 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10468 splitting insns.
10470 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10472 PR target/83663 - Revert r255946
10473 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10474 generation for cases where splatting a value is not useful.
10475 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10476 across a vec_duplicate and a paradoxical subreg forming a vector
10477 mode to a vec_concat.
10479 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10481 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10482 -march=armv8.3-a variants.
10483 * config/arm/t-multilib: Likewise.
10484 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10486 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10488 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10489 to generate rtl.
10490 (cceq_ior_compare_complement): Give it a name so I can use it, and
10491 change boolean_or_operator predicate to boolean_operator so it can
10492 be used to generate a crand.
10493 (eqne): New code iterator.
10494 (bd/bd_neg): New code_attrs.
10495 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10496 a single define_insn.
10497 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10498 decrement (bdnzt/bdnzf/bdzt/bdzf).
10499 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10500 with the new names of the branch decrement patterns, and added the
10501 names of the branch decrement conditional patterns.
10503 2018-01-08 Richard Biener <rguenther@suse.de>
10505 PR tree-optimization/83563
10506 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10507 cache.
10509 2018-01-08 Richard Biener <rguenther@suse.de>
10511 PR middle-end/83713
10512 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10514 2018-01-08 Richard Biener <rguenther@suse.de>
10516 PR tree-optimization/83685
10517 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10518 references to abnormals.
10520 2018-01-08 Richard Biener <rguenther@suse.de>
10522 PR lto/83719
10523 * dwarf2out.c (output_indirect_strings): Handle empty
10524 skeleton_debug_str_hash.
10525 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10527 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10529 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10530 (emit_store_direct): Likewise.
10531 (arc_trampoline_adjust_address): Likewise.
10532 (arc_asm_trampoline_template): New function.
10533 (arc_initialize_trampoline): Use asm_trampoline_template.
10534 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10535 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10536 * config/arc/arc.md (flush_icache): Delete pattern.
10538 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10540 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10541 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10542 munaligned-access.
10544 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10546 PR target/83681
10547 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10548 by not USED_FOR_TARGET.
10549 (make_pass_resolve_sw_modes): Likewise.
10551 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10553 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10554 USED_FOR_TARGET.
10556 2018-01-08 Richard Biener <rguenther@suse.de>
10558 PR middle-end/83580
10559 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10561 2018-01-08 Richard Biener <rguenther@suse.de>
10563 PR middle-end/83517
10564 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10566 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10568 PR middle-end/81897
10569 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10570 basic blocks with a small number of successors.
10571 (convert_control_dep_chain_into_preds): Improve handling of
10572 forwarder blocks.
10573 (dump_predicates): Split apart into...
10574 (dump_pred_chain): ...here...
10575 (dump_pred_info): ...and here.
10576 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10577 (can_chain_union_be_invalidated_p): Improve check for invalidation
10578 of paths.
10579 (uninit_uses_cannot_happen): Avoid unnecessary if
10580 convert_control_dep_chain_into_preds yielded nothing.
10582 2018-01-06 Martin Sebor <msebor@redhat.com>
10584 PR tree-optimization/83640
10585 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10586 subtracting negative offset from size.
10587 (builtin_access::overlap): Adjust offset bounds of the access to fall
10588 within the size of the object if possible.
10590 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10592 PR rtl-optimization/83699
10593 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10594 extract_bit_field_as_subreg to cases in which the extracted
10595 value is also a vector.
10597 * lra-constraints.c (process_alt_operands): Test for the equivalence
10598 substitutions when detecting a possible reload cycle.
10600 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10602 PR debug/83480
10603 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10604 by default if flag_selective_schedling{,2}. Formatting fixes.
10606 PR rtl-optimization/83682
10607 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10608 if it has non-VECTOR_MODE element mode.
10609 (vec_duplicate_p): Likewise.
10611 PR middle-end/83694
10612 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10613 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10615 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10617 PR target/83604
10618 * config/i386/i386-builtin.def
10619 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10620 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10621 Require also OPTION_MASK_ISA_AVX512F in addition to
10622 OPTION_MASK_ISA_GFNI.
10623 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10624 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10625 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10626 to OPTION_MASK_ISA_GFNI.
10627 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10628 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10629 OPTION_MASK_ISA_AVX512BW.
10630 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10631 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10632 addition to OPTION_MASK_ISA_GFNI.
10633 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10634 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10635 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10636 to OPTION_MASK_ISA_GFNI.
10637 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10638 a requirement for all ISAs rather than any of them with a few
10639 exceptions.
10640 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10641 processing.
10642 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10643 bitmasks to be enabled with 3 exceptions, instead of requiring any
10644 enabled ISA with lots of exceptions.
10645 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10646 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10647 Change avx512bw in isa attribute to avx512f.
10648 * config/i386/sgxintrin.h: Add license boilerplate.
10649 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10650 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10651 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10652 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10653 defined.
10654 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10655 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10656 temporarily sse2 rather than sse if not enabled already.
10658 PR target/83604
10659 * config/i386/sse.md (VI248_VLBW): Rename to ...
10660 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10661 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10662 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10663 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10664 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10665 mode iterator instead of VI248_VLBW.
10667 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10669 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10670 (record_modified): Skip clobbers; add debug output.
10671 (param_change_prob): Use sreal frequencies.
10673 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10675 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10676 punt for user-aligned variables.
10678 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10680 * tree-chrec.c (chrec_contains_symbols): Return true for
10681 POLY_INT_CST.
10683 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10685 PR target/82439
10686 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10687 of (x|y) == x for BICS pattern.
10689 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10691 PR tree-optimization/83605
10692 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10693 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10694 can throw.
10696 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10698 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10699 * config/epiphany/rtems.h: New file.
10701 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10702 Uros Bizjak <ubizjak@gmail.com>
10704 PR target/83554
10705 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10706 QIreg_operand instead of register_operand predicate.
10707 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10708 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10709 comments instead of -fmitigate[-_]rop.
10711 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10713 PR bootstrap/81926
10714 * cgraphunit.c (symbol_table::compile): Switch to text_section
10715 before calling assembly_start debug hook.
10716 * run-rtl-passes.c (run_rtl_passes): Likewise.
10717 Include output.h.
10719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10721 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10722 range_int_cst_p rather than !symbolic_range_p before calling
10723 extract_range_from_multiplicative_op_1.
10725 2018-01-04 Jeff Law <law@redhat.com>
10727 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10728 redundant test in assertion.
10730 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10732 * doc/rtl.texi: Document machine_mode wrapper classes.
10734 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10736 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10737 using tree_to_uhwi.
10739 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10741 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10742 the VEC_PERM_EXPR fold to fail.
10744 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10746 PR debug/83585
10747 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10748 to switched_sections.
10750 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10752 PR target/83680
10753 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10754 test for d.testing.
10756 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10758 PR target/83387
10759 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10760 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10762 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10764 PR debug/83666
10765 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10766 is BLKmode and bitpos not zero or mode change is needed.
10768 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10770 PR target/83675
10771 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10772 TARGET_VIS2.
10774 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10776 PR target/83628
10777 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10778 instead of MULT rtx. Update all corresponding splitters.
10779 (*saddl_se): Ditto.
10780 (*ssub<modesuffix>): Ditto.
10781 (*ssubl_se): Ditto.
10782 (*cmp_sadd_di): Update split patterns.
10783 (*cmp_sadd_si): Ditto.
10784 (*cmp_sadd_sidi): Ditto.
10785 (*cmp_ssub_di): Ditto.
10786 (*cmp_ssub_si): Ditto.
10787 (*cmp_ssub_sidi): Ditto.
10788 * config/alpha/predicates.md (const23_operand): New predicate.
10789 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10790 Look for ASHIFT, not MULT inner operand.
10791 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10793 2018-01-04 Martin Liska <mliska@suse.cz>
10795 PR gcov-profile/83669
10796 * gcov.c (output_intermediate_file): Add version to intermediate
10797 gcov file.
10798 * doc/gcov.texi: Document new field 'version' in intermediate
10799 file format. Fix location of '-k' option of gcov command.
10801 2018-01-04 Martin Liska <mliska@suse.cz>
10803 PR ipa/82352
10804 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10806 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10808 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10810 2018-01-03 Martin Sebor <msebor@redhat.com>
10812 PR tree-optimization/83655
10813 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10814 checking calls with invalid arguments.
10816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10818 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10819 (vectorizable_mask_load_store): Delete.
10820 (vectorizable_call): Return false for masked loads and stores.
10821 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10822 instead of gimple_assign_rhs1.
10823 (vectorizable_load): Handle IFN_MASK_LOAD.
10824 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10828 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10829 split out from..,
10830 (vectorizable_mask_load_store): ...here.
10831 (vectorizable_load): ...and here.
10833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10835 * tree-vect-stmts.c (vect_build_all_ones_mask)
10836 (vect_build_zero_merge_argument): New functions, split out from...
10837 (vectorizable_load): ...here.
10839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10841 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10842 split out from...
10843 (vectorizable_mask_load_store): ...here.
10844 (vectorizable_store): ...and here.
10846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10848 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10849 split out from...
10850 (vectorizable_mask_load_store): ...here.
10852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10854 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10855 (vect_model_store_cost): Take a vec_load_store_type instead of a
10856 vect_def_type.
10857 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10858 (vect_model_store_cost): Take a vec_load_store_type instead of a
10859 vect_def_type.
10860 (vectorizable_mask_load_store): Update accordingly.
10861 (vectorizable_store): Likewise.
10862 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10866 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10867 IFN_MASK_LOAD calls here rather than...
10868 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10871 Alan Hayward <alan.hayward@arm.com>
10872 David Sherwood <david.sherwood@arm.com>
10874 * expmed.c (extract_bit_field_1): For vector extracts,
10875 fall back to extract_bit_field_as_subreg if vec_extract
10876 isn't available.
10878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10879 Alan Hayward <alan.hayward@arm.com>
10880 David Sherwood <david.sherwood@arm.com>
10882 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10883 they are variable or constant sized.
10884 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10885 slots for constant-sized data.
10887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10888 Alan Hayward <alan.hayward@arm.com>
10889 David Sherwood <david.sherwood@arm.com>
10891 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10892 handling COND_EXPRs with boolean comparisons, try to find a better
10893 basis for the mask type than the boolean itself.
10895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10897 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10898 is calculated and how it can be overridden.
10899 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10900 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10901 if defined.
10902 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10903 if nonzero.
10905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10906 Alan Hayward <alan.hayward@arm.com>
10907 David Sherwood <david.sherwood@arm.com>
10909 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10910 Remove the mode argument.
10911 (aarch64_simd_valid_immediate): Remove the mode and inverse
10912 arguments.
10913 * config/aarch64/iterators.md (bitsize): New iterator.
10914 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10915 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10916 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10917 aarch64_simd_valid_immediate.
10918 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10919 (aarch64_reg_or_bic_imm): Likewise.
10920 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10921 with an insn_type enum and msl with a modifier_type enum.
10922 Replace element_width with a scalar_mode. Change the shift
10923 to unsigned int. Add constructors for scalar_float_mode and
10924 scalar_int_mode elements.
10925 (aarch64_vect_float_const_representable_p): Delete.
10926 (aarch64_can_const_movi_rtx_p)
10927 (aarch64_simd_scalar_immediate_valid_for_move)
10928 (aarch64_simd_make_constant): Update call to
10929 aarch64_simd_valid_immediate.
10930 (aarch64_advsimd_valid_immediate_hs): New function.
10931 (aarch64_advsimd_valid_immediate): Likewise.
10932 (aarch64_simd_valid_immediate): Remove mode and inverse
10933 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10934 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10935 and aarch64_float_const_representable_p on the result.
10936 (aarch64_output_simd_mov_immediate): Remove mode argument.
10937 Update call to aarch64_simd_valid_immediate and use of
10938 simd_immediate_info.
10939 (aarch64_output_scalar_simd_mov_immediate): Update call
10940 accordingly.
10942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10947 (mode_nunits): Likewise CONST_MODE_NUNITS.
10948 * machmode.def (ADJUST_NUNITS): Document.
10949 * genmodes.c (mode_data::need_nunits_adj): New field.
10950 (blank_mode): Update accordingly.
10951 (adj_nunits): New variable.
10952 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10953 parameter.
10954 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10955 listed in adj_nunits.
10956 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10957 listed in adj_nunits. Don't emit case statements for such modes.
10958 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10959 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10960 nothing if adj_nunits is nonnull.
10961 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10962 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10963 (emit_mode_fbit): Update use of print_maybe_const_decl.
10964 (emit_move_size): Likewise. Treat the array as non-const
10965 if adj_nunits.
10966 (emit_mode_adjustments): Handle adj_nunits.
10968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10970 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10971 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10972 (VECTOR_MODES): Use it.
10973 (make_vector_modes): Take the prefix as an argument.
10975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10976 Alan Hayward <alan.hayward@arm.com>
10977 David Sherwood <david.sherwood@arm.com>
10979 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10980 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10981 for MODE_VECTOR_BOOL.
10982 * machmode.def (VECTOR_BOOL_MODE): Document.
10983 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10984 (make_vector_bool_mode): New function.
10985 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10986 MODE_VECTOR_BOOL.
10987 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10988 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10989 Likewise.
10990 * stor-layout.c (int_mode_for_mode): Likewise.
10991 * tree.c (build_vector_type_for_mode): Likewise.
10992 * varasm.c (output_constant_pool_2): Likewise.
10993 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10994 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10995 for MODE_VECTOR_BOOL.
10996 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10997 of mode class checks.
10998 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10999 instead of a list of mode class checks.
11000 (expand_vector_scalar_condition): Likewise.
11001 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11004 Alan Hayward <alan.hayward@arm.com>
11005 David Sherwood <david.sherwood@arm.com>
11007 * machmode.h (mode_size): Change from unsigned short to
11008 poly_uint16_pod.
11009 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11010 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11011 or if measurement_type is not polynomial.
11012 (fixed_size_mode::includes_p): Check for constant-sized modes.
11013 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11014 return a poly_uint16 rather than an unsigned short.
11015 (emit_mode_size): Change the type of mode_size from unsigned short
11016 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11017 (emit_mode_adjustments): Cope with polynomial vector sizes.
11018 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11019 for GET_MODE_SIZE.
11020 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11021 for GET_MODE_SIZE.
11022 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11023 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11024 * caller-save.c (setup_save_areas): Likewise.
11025 (replace_reg_with_saved_mem): Likewise.
11026 * calls.c (emit_library_call_value_1): Likewise.
11027 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11028 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11029 (gen_lowpart_for_combine): Likewise.
11030 * convert.c (convert_to_integer_1): Likewise.
11031 * cse.c (equiv_constant, cse_insn): Likewise.
11032 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11033 (cselib_subst_to_values): Likewise.
11034 * dce.c (word_dce_process_block): Likewise.
11035 * df-problems.c (df_word_lr_mark_ref): Likewise.
11036 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11037 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11038 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11039 (rtl_for_decl_location): Likewise.
11040 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11041 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11042 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11043 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11044 (expand_expr_real_1): Likewise.
11045 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11046 (pad_below): Likewise.
11047 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11048 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11049 * ira.c (get_subreg_tracking_sizes): Likewise.
11050 * ira-build.c (ira_create_allocno_objects): Likewise.
11051 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11052 (ira_sort_regnos_for_alter_reg): Likewise.
11053 * ira-costs.c (record_operand_costs): Likewise.
11054 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11055 (resolve_simple_move): Likewise.
11056 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11057 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11058 (lra_constraints): Likewise.
11059 (CONST_POOL_OK_P): Reject variable-sized modes.
11060 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11061 (add_pseudo_to_slot, lra_spill): Likewise.
11062 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11063 * optabs-query.c (get_best_extraction_insn): Likewise.
11064 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11065 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11066 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11067 * recog.c (offsettable_address_addr_space_p): Likewise.
11068 * regcprop.c (maybe_mode_change): Likewise.
11069 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11070 * regrename.c (build_def_use): Likewise.
11071 * regstat.c (dump_reg_info): Likewise.
11072 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11073 (find_reloads, find_reloads_subreg_address): Likewise.
11074 * reload1.c (eliminate_regs_1): Likewise.
11075 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11076 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11077 (simplify_binary_operation_1, simplify_subreg): Likewise.
11078 * targhooks.c (default_function_arg_padding): Likewise.
11079 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11080 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11081 (verify_gimple_assign_ternary): Likewise.
11082 * tree-inline.c (estimate_move_cost): Likewise.
11083 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11084 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11085 (get_address_cost_ainc): Likewise.
11086 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11087 (vect_supportable_dr_alignment): Likewise.
11088 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11089 (vectorizable_reduction): Likewise.
11090 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11091 (vectorizable_operation, vectorizable_load): Likewise.
11092 * tree.c (build_same_sized_truth_vector_type): Likewise.
11093 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11094 * var-tracking.c (emit_note_insn_var_location): Likewise.
11095 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11096 (ADDR_VEC_ALIGN): Likewise.
11098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11103 unsigned short.
11104 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11105 or if measurement_type is polynomial.
11106 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11107 * combine.c (make_extraction): Likewise.
11108 * dse.c (find_shift_sequence): Likewise.
11109 * dwarf2out.c (mem_loc_descriptor): Likewise.
11110 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11111 (extract_bit_field, extract_low_bits): Likewise.
11112 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11113 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11114 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11115 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11116 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11117 * reload.c (find_reloads): Likewise.
11118 * reload1.c (alter_reg): Likewise.
11119 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11120 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11121 * tree-if-conv.c (predicate_mem_writes): Likewise.
11122 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11123 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11124 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11125 * valtrack.c (dead_debug_insert_temp): Likewise.
11126 * varasm.c (mergeable_constant_section): Likewise.
11127 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
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 * expr.c (expand_assignment): Cope with polynomial mode sizes
11134 when assigning to a CONCAT.
11136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11137 Alan Hayward <alan.hayward@arm.com>
11138 David Sherwood <david.sherwood@arm.com>
11140 * machmode.h (mode_precision): Change from unsigned short to
11141 poly_uint16_pod.
11142 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11143 short.
11144 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11145 or if measurement_type is not polynomial.
11146 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11147 in which the mode is already known to be a scalar_int_mode.
11148 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11149 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11150 initializer.
11151 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11152 for GET_MODE_PRECISION.
11153 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11154 for GET_MODE_PRECISION.
11155 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11156 as polynomial.
11157 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11158 (expand_field_assignment, make_extraction): Likewise.
11159 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11160 (get_last_value): Likewise.
11161 * convert.c (convert_to_integer_1): Likewise.
11162 * cse.c (cse_insn): Likewise.
11163 * expr.c (expand_expr_real_1): Likewise.
11164 * lra-constraints.c (simplify_operand_subreg): Likewise.
11165 * optabs-query.c (can_atomic_load_p): Likewise.
11166 * optabs.c (expand_atomic_load): Likewise.
11167 (expand_atomic_store): Likewise.
11168 * ree.c (combine_reaching_defs): Likewise.
11169 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11170 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11171 * tree.h (type_has_mode_precision_p): Likewise.
11172 * ubsan.c (instrument_si_overflow): Likewise.
11174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11175 Alan Hayward <alan.hayward@arm.com>
11176 David Sherwood <david.sherwood@arm.com>
11178 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11179 polynomial numbers of units.
11180 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11181 (valid_vector_subparts_p): New function.
11182 (build_vector_type): Remove temporary shim and take the number
11183 of units as a poly_uint64 rather than an int.
11184 (build_opaque_vector_type): Take the number of units as a
11185 poly_uint64 rather than an int.
11186 * tree.c (build_vector_from_ctor): Handle polynomial
11187 TYPE_VECTOR_SUBPARTS.
11188 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11189 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11190 (build_vector_from_val): If the number of units is variable,
11191 use build_vec_duplicate_cst for constant operands and
11192 VEC_DUPLICATE_EXPR otherwise.
11193 (make_vector_type): Remove temporary is_constant ().
11194 (build_vector_type, build_opaque_vector_type): Take the number of
11195 units as a poly_uint64 rather than an int.
11196 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11197 VECTOR_CST_NELTS.
11198 * cfgexpand.c (expand_debug_expr): Likewise.
11199 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11200 (store_constructor, expand_expr_real_1): Likewise.
11201 (const_scalar_mask_from_tree): Likewise.
11202 * fold-const-call.c (fold_const_reduction): Likewise.
11203 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11204 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11205 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11206 (fold_relational_const): Likewise.
11207 (native_interpret_vector): Likewise. Change the size from an
11208 int to an unsigned int.
11209 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11210 TYPE_VECTOR_SUBPARTS.
11211 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11212 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11213 duplicating a non-constant operand into a variable-length vector.
11214 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11215 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11216 * ipa-icf.c (sem_variable::equals): Likewise.
11217 * match.pd: Likewise.
11218 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11219 * print-tree.c (print_node): Likewise.
11220 * stor-layout.c (layout_type): Likewise.
11221 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11222 * tree-cfg.c (verify_gimple_comparison): Likewise.
11223 (verify_gimple_assign_binary): Likewise.
11224 (verify_gimple_assign_ternary): Likewise.
11225 (verify_gimple_assign_single): Likewise.
11226 * tree-pretty-print.c (dump_generic_node): Likewise.
11227 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11228 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11229 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11230 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11231 (vect_shift_permute_load_chain): Likewise.
11232 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11233 (expand_vector_condition, optimize_vector_constructor): Likewise.
11234 (lower_vec_perm, get_compute_type): Likewise.
11235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11236 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11237 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11238 (vect_recog_mask_conversion_pattern): Likewise.
11239 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11240 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11241 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11242 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11243 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11244 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11245 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11246 (supportable_widening_operation): Likewise.
11247 (supportable_narrowing_operation): Likewise.
11248 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11249 Likewise.
11250 * varasm.c (output_constant): Likewise.
11252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11253 Alan Hayward <alan.hayward@arm.com>
11254 David Sherwood <david.sherwood@arm.com>
11256 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11257 so that both the length == 3 and length != 3 cases set up their
11258 own permute vectors. Add comments explaining why we know the
11259 number of elements is constant.
11260 (vect_permute_load_chain): Likewise.
11262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11263 Alan Hayward <alan.hayward@arm.com>
11264 David Sherwood <david.sherwood@arm.com>
11266 * machmode.h (mode_nunits): Change from unsigned char to
11267 poly_uint16_pod.
11268 (ONLY_FIXED_SIZE_MODES): New macro.
11269 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11270 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11271 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11272 New typedefs.
11273 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11274 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11275 or if measurement_type is not polynomial.
11276 * genmodes.c (ZERO_COEFFS): New macro.
11277 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11278 poly_uint16.
11279 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11280 Use ZERO_COEFFS when emitting initializers.
11281 * data-streamer.h (bp_pack_poly_value): New function.
11282 (bp_unpack_poly_value): Likewise.
11283 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11284 for GET_MODE_NUNITS.
11285 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11286 for GET_MODE_NUNITS.
11287 * tree.c (make_vector_type): Remove temporary shim and make
11288 the real function take the number of units as a poly_uint64
11289 rather than an int.
11290 (build_vector_type_for_mode): Handle polynomial nunits.
11291 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11292 * emit-rtl.c (const_vec_series_p_1): Likewise.
11293 (gen_rtx_CONST_VECTOR): Likewise.
11294 * fold-const.c (test_vec_duplicate_folding): Likewise.
11295 * genrecog.c (validate_pattern): Likewise.
11296 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11297 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11298 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11299 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11300 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11301 * rtlanal.c (subreg_get_info): Likewise.
11302 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11303 (vect_grouped_load_supported): Likewise.
11304 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11305 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11306 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11307 (simplify_const_unary_operation, simplify_binary_operation_1)
11308 (simplify_const_binary_operation, simplify_ternary_operation)
11309 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11310 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11311 instead of CONST_VECTOR_NUNITS.
11312 * varasm.c (output_constant_pool_2): Likewise.
11313 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11314 explicit-encoded elements in the XVEC for variable-length vectors.
11316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11318 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11321 Alan Hayward <alan.hayward@arm.com>
11322 David Sherwood <david.sherwood@arm.com>
11324 * coretypes.h (fixed_size_mode): Declare.
11325 (fixed_size_mode_pod): New typedef.
11326 * builtins.h (target_builtins::x_apply_args_mode)
11327 (target_builtins::x_apply_result_mode): Change type to
11328 fixed_size_mode_pod.
11329 * builtins.c (apply_args_size, apply_result_size, result_vector)
11330 (expand_builtin_apply_args_1, expand_builtin_apply)
11331 (expand_builtin_return): Update accordingly.
11333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11335 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11336 * cselib.c (cselib_hash_rtx): Likewise.
11337 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11338 CONST_VECTOR encoding.
11340 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11341 Jeff Law <law@redhat.com>
11343 PR target/83641
11344 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11345 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11346 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11347 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11349 PR target/83641
11350 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11351 explicitly probe *sp in a noreturn function if there were any callee
11352 register saves or frame pointer is needed.
11354 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11356 PR debug/83621
11357 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11358 BLKmode for ternary, binary or unary expressions.
11360 PR debug/83645
11361 * var-tracking.c (delete_vta_debug_insn): New inline function.
11362 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11363 insns from get_insns () to NULL instead of each bb separately.
11364 Use delete_vta_debug_insn. No longer static.
11365 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11366 delete_vta_debug_insns callers.
11367 * rtl.h (delete_vta_debug_insns): Declare.
11368 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11369 instead of variable_tracking_main.
11371 2018-01-03 Martin Sebor <msebor@redhat.com>
11373 PR tree-optimization/83603
11374 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11375 arguments past the endof the argument list in functions declared
11376 without a prototype.
11377 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11378 Avoid checking when arguments are null.
11380 2018-01-03 Martin Sebor <msebor@redhat.com>
11382 PR c/83559
11383 * doc/extend.texi (attribute const): Fix a typo.
11384 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11385 issuing -Wsuggest-attribute for void functions.
11387 2018-01-03 Martin Sebor <msebor@redhat.com>
11389 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11390 offset_int::from instead of wide_int::to_shwi.
11391 (maybe_diag_overlap): Remove assertion.
11392 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11393 * gimple-ssa-sprintf.c (format_directive): Same.
11394 (parse_directive): Same.
11395 (sprintf_dom_walker::compute_format_length): Same.
11396 (try_substitute_return_value): Same.
11398 2018-01-03 Jeff Law <law@redhat.com>
11400 PR middle-end/83654
11401 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11402 non-constant residual for zero at runtime and avoid probing in
11403 that case. Reorganize code for trailing problem to mirror handling
11404 of the residual.
11406 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11408 PR tree-optimization/83501
11409 * tree-ssa-strlen.c (get_string_cst): New.
11410 (handle_char_store): Call get_string_cst.
11412 2018-01-03 Martin Liska <mliska@suse.cz>
11414 PR tree-optimization/83593
11415 * tree-ssa-strlen.c: Include tree-cfg.h.
11416 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11417 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11418 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11419 to false.
11420 (strlen_dom_walker::before_dom_children): Call
11421 gimple_purge_dead_eh_edges. Dump tranformation with details
11422 dump flags.
11423 (strlen_dom_walker::before_dom_children): Update call by adding
11424 new argument cleanup_eh.
11425 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11427 2018-01-03 Martin Liska <mliska@suse.cz>
11429 PR ipa/83549
11430 * cif-code.def (VARIADIC_THUNK): New enum value.
11431 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11432 thunks.
11434 2018-01-03 Jan Beulich <jbeulich@suse.com>
11436 * sse.md (mov<mode>_internal): Tighten condition for when to use
11437 vmovdqu<ssescalarsize> for TI and OI modes.
11439 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11441 Update copyright years.
11443 2018-01-03 Martin Liska <mliska@suse.cz>
11445 PR ipa/83594
11446 * ipa-visibility.c (function_and_variable_visibility): Skip
11447 functions with noipa attribure.
11449 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11451 * gcc.c (process_command): Update copyright notice dates.
11452 * gcov-dump.c (print_version): Ditto.
11453 * gcov.c (print_version): Ditto.
11454 * gcov-tool.c (print_version): Ditto.
11455 * gengtype.c (create_file): Ditto.
11456 * doc/cpp.texi: Bump @copying's copyright year.
11457 * doc/cppinternals.texi: Ditto.
11458 * doc/gcc.texi: Ditto.
11459 * doc/gccint.texi: Ditto.
11460 * doc/gcov.texi: Ditto.
11461 * doc/install.texi: Ditto.
11462 * doc/invoke.texi: Ditto.
11464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11466 * vector-builder.h (vector_builder::m_full_nelts): Change from
11467 unsigned int to poly_uint64.
11468 (vector_builder::full_nelts): Update prototype accordingly.
11469 (vector_builder::new_vector): Likewise.
11470 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11471 (vector_builder::operator ==): Likewise.
11472 (vector_builder::finalize): Likewise.
11473 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11474 Take the number of elements as a poly_uint64 rather than an
11475 unsigned int.
11476 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11477 from unsigned int to poly_uint64.
11478 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11479 (vec_perm_indices::new_vector): Likewise.
11480 (vec_perm_indices::length): Likewise.
11481 (vec_perm_indices::nelts_per_input): Likewise.
11482 (vec_perm_indices::input_nelts): Likewise.
11483 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11484 number of elements per input as a poly_uint64 rather than an
11485 unsigned int. Use the original encoding for variable-length
11486 vectors, rather than clamping each individual element.
11487 For the second and subsequent elements in each pattern,
11488 clamp the step and base before clamping their sum.
11489 (vec_perm_indices::series_p): Handle polynomial element counts.
11490 (vec_perm_indices::all_in_range_p): Likewise.
11491 (vec_perm_indices_to_tree): Likewise.
11492 (vec_perm_indices_to_rtx): Likewise.
11493 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11494 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11495 (tree_vector_builder::new_binary_operation): Handle polynomial
11496 element counts. Return false if we need to know the number
11497 of elements at compile time.
11498 * fold-const.c (fold_vec_perm): Punt if the number of elements
11499 isn't known at compile time.
11501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11503 * vec-perm-indices.h (vec_perm_builder): Change element type
11504 from HOST_WIDE_INT to poly_int64.
11505 (vec_perm_indices::element_type): Update accordingly.
11506 (vec_perm_indices::clamp): Handle polynomial element_types.
11507 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11508 (vec_perm_indices::all_in_range_p): Likewise.
11509 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11510 than shwi trees.
11511 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11512 polynomial vec_perm_indices element types.
11513 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11514 * fold-const.c (fold_vec_perm): Likewise.
11515 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11516 * tree-vect-generic.c (lower_vec_perm): Likewise.
11517 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11518 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11519 element type to HOST_WIDE_INT.
11521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11522 Alan Hayward <alan.hayward@arm.com>
11523 David Sherwood <david.sherwood@arm.com>
11525 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11526 rather than an int. Use plus_constant.
11527 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11528 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 (emit_call_1, expand_call): Change struct_value_size from
11535 a HOST_WIDE_INT to a poly_int64.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * calls.c (load_register_parameters): Cope with polynomial
11542 mode sizes. Require a constant size for BLKmode parameters
11543 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11544 forces a parameter to be padded at the lsb end in order to
11545 fill a complete number of words, require the parameter size
11546 to be ordered wrt UNITS_PER_WORD.
11548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11549 Alan Hayward <alan.hayward@arm.com>
11550 David Sherwood <david.sherwood@arm.com>
11552 * reload1.c (spill_stack_slot_width): Change element type
11553 from unsigned int to poly_uint64_pod.
11554 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
11561 (reload_inner_reg_of_subreg, push_reload): Use it.
11563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11564 Alan Hayward <alan.hayward@arm.com>
11565 David Sherwood <david.sherwood@arm.com>
11567 * lra-constraints.c (process_alt_operands): Reject matched
11568 operands whose sizes aren't ordered.
11569 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11576 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11583 Use plus_constant instead of gen_rtx_PLUS.
11585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11586 Alan Hayward <alan.hayward@arm.com>
11587 David Sherwood <david.sherwood@arm.com>
11589 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11590 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11591 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11592 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11593 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11594 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11595 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11596 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11597 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11598 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11599 a poly_int64.
11600 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11601 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11602 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11603 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11604 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11605 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11606 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11607 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11608 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11609 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11610 function.
11611 * expr.c (emit_move_resolve_push): Treat the input and result
11612 of PUSH_ROUNDING as a poly_int64.
11613 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11614 (emit_push_insn): Likewise.
11615 * lra-eliminations.c (mark_not_eliminable): Likewise.
11616 * recog.c (push_operand): Likewise.
11617 * reload1.c (elimination_effects): Likewise.
11618 * rtlanal.c (nonzero_bits1): Likewise.
11619 * calls.c (store_one_arg): Likewise. Require the padding to be
11620 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11627 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11634 rather than an int.
11636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11637 Alan Hayward <alan.hayward@arm.com>
11638 David Sherwood <david.sherwood@arm.com>
11640 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11641 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11642 via stack temporaries. Treat the mode size as polynomial too.
11644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11645 Alan Hayward <alan.hayward@arm.com>
11646 David Sherwood <david.sherwood@arm.com>
11648 * expr.c (expand_expr_real_2): When handling conversions involving
11649 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11650 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11651 as a poly_uint64 too.
11653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11654 Alan Hayward <alan.hayward@arm.com>
11655 David Sherwood <david.sherwood@arm.com>
11657 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11660 Alan Hayward <alan.hayward@arm.com>
11661 David Sherwood <david.sherwood@arm.com>
11663 * combine.c (can_change_dest_mode): Handle polynomial
11664 REGMODE_NATURAL_SIZE.
11665 * expmed.c (store_bit_field_1): Likewise.
11666 * expr.c (store_constructor): Likewise.
11667 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11668 and polynomial REGMODE_NATURAL_SIZE.
11669 (gen_lowpart_common): Likewise.
11670 * reginfo.c (record_subregs_of_mode): Likewise.
11671 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11678 numbers of elements.
11680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11681 Alan Hayward <alan.hayward@arm.com>
11682 David Sherwood <david.sherwood@arm.com>
11684 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11691 in a POINTER_PLUS_EXPR.
11693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11694 Alan Hayward <alan.hayward@arm.com>
11695 David Sherwood <david.sherwood@arm.com>
11697 * omp-simd-clone.c (simd_clone_subparts): New function.
11698 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11699 (ipa_simd_modify_function_body): Likewise.
11701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11702 Alan Hayward <alan.hayward@arm.com>
11703 David Sherwood <david.sherwood@arm.com>
11705 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11706 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11707 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11708 (expand_vector_condition, vector_element): Likewise.
11709 (subparts_gt): New function.
11710 (get_compute_type): Use subparts_gt.
11711 (count_type_subparts): Delete.
11712 (expand_vector_operations_1): Use subparts_gt instead of
11713 count_type_subparts.
11715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11716 Alan Hayward <alan.hayward@arm.com>
11717 David Sherwood <david.sherwood@arm.com>
11719 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11720 (vect_compile_time_alias): ...this new function. Do the calculation
11721 on poly_ints rather than trees.
11722 (vect_prune_runtime_alias_test_list): Update call accordingly.
11724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11725 Alan Hayward <alan.hayward@arm.com>
11726 David Sherwood <david.sherwood@arm.com>
11728 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11729 numbers of units.
11730 (vect_schedule_slp_instance): Likewise.
11732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11733 Alan Hayward <alan.hayward@arm.com>
11734 David Sherwood <david.sherwood@arm.com>
11736 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11737 constant and extern definitions for variable-length vectors.
11738 (vect_get_constant_vectors): Note that the number of units
11739 is known to be constant.
11741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11742 Alan Hayward <alan.hayward@arm.com>
11743 David Sherwood <david.sherwood@arm.com>
11745 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11746 of units as polynomial. Choose between WIDE and NARROW based
11747 on multiple_p.
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 (simd_clone_subparts): New function.
11754 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11761 vectors as polynomial. Use build_index_vector for
11762 IFN_GOMP_SIMD_LANE.
11764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11765 Alan Hayward <alan.hayward@arm.com>
11766 David Sherwood <david.sherwood@arm.com>
11768 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11769 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11770 for variable-length vectors.
11771 (vectorizable_mask_load_store): Treat the number of units as
11772 polynomial, asserting that it is constant if the condition has
11773 already been enforced.
11774 (vectorizable_store, vectorizable_load): Likewise.
11776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11777 Alan Hayward <alan.hayward@arm.com>
11778 David Sherwood <david.sherwood@arm.com>
11780 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11781 of units as polynomial. Punt if we can't tell at compile time
11782 which vector contains the final result.
11784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11785 Alan Hayward <alan.hayward@arm.com>
11786 David Sherwood <david.sherwood@arm.com>
11788 * tree-vect-loop.c (vectorizable_induction): Treat the number
11789 of units as polynomial. Punt on SLP inductions. Use an integer
11790 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11791 cast of such a series for variable-length floating-point
11792 reductions.
11794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11795 Alan Hayward <alan.hayward@arm.com>
11796 David Sherwood <david.sherwood@arm.com>
11798 * tree.h (build_index_vector): Declare.
11799 * tree.c (build_index_vector): New function.
11800 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11801 of units as polynomial, forcibly converting it to a constant if
11802 vectorizable_reduction has already enforced the condition.
11803 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11804 to create a {1,2,3,...} vector.
11805 (vectorizable_reduction): Treat the number of units as polynomial.
11806 Choose vectype_in based on the largest scalar element size rather
11807 than the smallest number of units. Enforce the restrictions
11808 relied on above.
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 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11815 number of units as polynomial.
11817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11818 Alan Hayward <alan.hayward@arm.com>
11819 David Sherwood <david.sherwood@arm.com>
11821 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11822 * target.def (autovectorize_vector_sizes): Return the vector sizes
11823 by pointer, using vector_sizes rather than a bitmask.
11824 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11825 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11826 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11827 Likewise.
11828 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11829 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11830 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11831 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11832 * omp-general.c (omp_max_vf): Likewise.
11833 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11834 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11835 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11836 * tree-vect-slp.c (vect_slp_bb): Likewise.
11837 * doc/tm.texi: Regenerate.
11838 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11839 to a poly_uint64.
11840 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11841 the vector size as a poly_uint64 rather than an unsigned int.
11842 (current_vector_size): Change from an unsigned int to a poly_uint64.
11843 (get_vectype_for_scalar_type): Update accordingly.
11844 * tree.h (build_truth_vector_type): Take the size and number of
11845 units as a poly_uint64 rather than an unsigned int.
11846 (build_vector_type): Add a temporary overload that takes
11847 the number of units as a poly_uint64 rather than an unsigned int.
11848 * tree.c (make_vector_type): Likewise.
11849 (build_truth_vector_type): Take the number of units as a poly_uint64
11850 rather than an unsigned int.
11852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11853 Alan Hayward <alan.hayward@arm.com>
11854 David Sherwood <david.sherwood@arm.com>
11856 * target.def (get_mask_mode): Take the number of units and length
11857 as poly_uint64s rather than unsigned ints.
11858 * targhooks.h (default_get_mask_mode): Update accordingly.
11859 * targhooks.c (default_get_mask_mode): Likewise.
11860 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11861 * doc/tm.texi: Regenerate.
11863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11864 Alan Hayward <alan.hayward@arm.com>
11865 David Sherwood <david.sherwood@arm.com>
11867 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11868 * omp-general.c (omp_max_vf): Likewise.
11869 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11870 (expand_omp_simd): Handle polynomial safelen.
11871 * omp-low.c (omplow_simd_context): Add a default constructor.
11872 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11873 (lower_rec_simd_input_clauses): Update accordingly.
11874 (lower_rec_input_clauses): Likewise.
11876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11877 Alan Hayward <alan.hayward@arm.com>
11878 David Sherwood <david.sherwood@arm.com>
11880 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11881 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11882 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11883 (vect_analyze_slp_cost): Likewise.
11884 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11885 (vect_model_load_cost): Likewise.
11887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11888 Alan Hayward <alan.hayward@arm.com>
11889 David Sherwood <david.sherwood@arm.com>
11891 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11892 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11893 from an unsigned int * to a poly_uint64_pod *.
11894 (calculate_unrolling_factor): New function.
11895 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11898 Alan Hayward <alan.hayward@arm.com>
11899 David Sherwood <david.sherwood@arm.com>
11901 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11902 from an unsigned int to a poly_uint64.
11903 (_loop_vec_info::slp_unrolling_factor): Likewise.
11904 (_loop_vec_info::vectorization_factor): Change from an int
11905 to a poly_uint64.
11906 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11907 (vect_get_num_vectors): New function.
11908 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11909 (vect_get_num_copies): Use vect_get_num_vectors.
11910 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11911 to an unsigned int *.
11912 (vect_analyze_data_refs): Change min_vf from an int * to a
11913 poly_uint64 *.
11914 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11915 than an unsigned HOST_WIDE_INT.
11916 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11917 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11918 to an unsigned int *.
11919 (vect_analyze_data_ref_dependences): Likewise.
11920 (vect_compute_data_ref_alignment): Handle polynomial vf.
11921 (vect_enhance_data_refs_alignment): Likewise.
11922 (vect_prune_runtime_alias_test_list): Likewise.
11923 (vect_shift_permute_load_chain): Likewise.
11924 (vect_supportable_dr_alignment): Likewise.
11925 (dependence_distance_ge_vf): Take the vectorization factor as a
11926 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11927 (vect_analyze_data_refs): Change min_vf from an int * to a
11928 poly_uint64 *.
11929 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11930 vfm1 as a poly_uint64 rather than an int. Make the same change
11931 for the returned bound_scalar.
11932 (vect_gen_vector_loop_niters): Handle polynomial vf.
11933 (vect_do_peeling): Likewise. Update call to
11934 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11935 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11936 be constant.
11937 * tree-vect-loop.c (vect_determine_vectorization_factor)
11938 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11939 (vect_get_known_peeling_cost): Likewise.
11940 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11941 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11942 (vect_transform_loop): Likewise. Use the lowest possible VF when
11943 updating the upper bounds of the loop.
11944 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11945 rather than an int.
11946 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11947 polynomial unroll factors.
11948 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11949 (vect_make_slp_decision): Likewise.
11950 (vect_supported_load_permutation_p): Likewise, and polynomial
11951 vf too.
11952 (vect_analyze_slp_cost): Handle polynomial vf.
11953 (vect_slp_analyze_node_operations): Likewise.
11954 (vect_slp_analyze_bb_1): Likewise.
11955 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11956 than an unsigned HOST_WIDE_INT.
11957 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11958 (vectorizable_load): Handle polynomial vf.
11959 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11960 a poly_uint64.
11961 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11964 Alan Hayward <alan.hayward@arm.com>
11965 David Sherwood <david.sherwood@arm.com>
11967 * match.pd: Handle bit operations involving three constants
11968 and try to fold one pair.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11972 * tree-vect-loop-manip.c: Include gimple-fold.h.
11973 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11974 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11975 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11976 Add a path that uses a step of VF instead of 1, but disable it
11977 for now.
11978 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11979 and niters_no_overflow parameters. Update calls to
11980 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11981 Create a new SSA name if the latter choses to use a ste other
11982 than zero, and return it via niters_vector_mult_vf_var.
11983 * tree-vect-loop.c (vect_transform_loop): Update calls to
11984 vect_do_peeling, vect_gen_vector_loop_niters and
11985 slpeel_make_loop_iterate_ntimes.
11986 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11987 (vect_gen_vector_loop_niters): Update declarations after above changes.
11989 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11991 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11992 128-bit round to integer instructions.
11993 (ceil<mode>2): Likewise.
11994 (btrunc<mode>2): Likewise.
11995 (round<mode>2): Likewise.
11997 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11999 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12000 unaligned VSX load/store on P8/P9.
12001 (expand_block_clear): Allow the use of unaligned VSX
12002 load/store on P8/P9.
12004 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12006 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12007 New function.
12008 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12009 swap associated with both a load and a store.
12011 2018-01-02 Andrew Waterman <andrew@sifive.com>
12013 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12014 * config/riscv/riscv.md (clear_cache): Use it.
12016 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12018 * web.c: Remove out-of-date comment.
12020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12022 * expr.c (fixup_args_size_notes): Check that any existing
12023 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12024 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12025 (emit_single_push_insn): ...here.
12027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12029 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12030 (const_vector_encoded_nelts): New function.
12031 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12032 (const_vector_int_elt, const_vector_elt): Declare.
12033 * emit-rtl.c (const_vector_int_elt_1): New function.
12034 (const_vector_elt): Likewise.
12035 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12036 of CONST_VECTOR_ELT.
12038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12040 * expr.c: Include rtx-vector-builder.h.
12041 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12042 directly on the tree encoding.
12043 (const_vector_from_tree): Likewise.
12044 * optabs.c: Include rtx-vector-builder.h.
12045 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12046 sequence of "u" values.
12047 * vec-perm-indices.c: Include rtx-vector-builder.h.
12048 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12049 directly on the vec_perm_indices encoding.
12051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12053 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12054 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12055 * rtx-vector-builder.h: New file.
12056 * rtx-vector-builder.c: Likewise.
12057 * rtl.h (rtx_def::u2): Add a const_vector field.
12058 (CONST_VECTOR_NPATTERNS): New macro.
12059 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12060 (CONST_VECTOR_DUPLICATE_P): Likewise.
12061 (CONST_VECTOR_STEPPED_P): Likewise.
12062 (CONST_VECTOR_ENCODED_ELT): Likewise.
12063 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12064 (unwrap_const_vec_duplicate): Likewise.
12065 (const_vec_series_p): Check for a non-duplicated vector encoding.
12066 Say that the function only returns true for integer vectors.
12067 * emit-rtl.c: Include rtx-vector-builder.h.
12068 (gen_const_vec_duplicate_1): Delete.
12069 (gen_const_vector): Call gen_const_vec_duplicate instead of
12070 gen_const_vec_duplicate_1.
12071 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12072 (gen_const_vec_duplicate): Use rtx_vector_builder.
12073 (gen_const_vec_series): Likewise.
12074 (gen_rtx_CONST_VECTOR): Likewise.
12075 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12076 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12077 Build a new vector rather than modifying a CONST_VECTOR in-place.
12078 (handle_special_swappables): Update call accordingly.
12079 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12080 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12081 Build a new vector rather than modifying a CONST_VECTOR in-place.
12082 (handle_special_swappables): Update call accordingly.
12084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12086 * simplify-rtx.c (simplify_const_binary_operation): Use
12087 CONST_VECTOR_ELT instead of XVECEXP.
12089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12091 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12092 the selector elements to be different from the data elements
12093 if the selector is a VECTOR_CST.
12094 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12095 ssizetype for the selector.
12097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12099 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12100 before testing each element individually.
12101 * tree-vect-generic.c (lower_vec_perm): Likewise.
12103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12105 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12106 * selftest-run-tests.c (selftest::run_tests): Call it.
12107 * vector-builder.h (vector_builder::operator ==): New function.
12108 (vector_builder::operator !=): Likewise.
12109 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12110 (vec_perm_indices::all_from_input_p): New function.
12111 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12112 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12113 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12114 instead of reading the VECTOR_CST directly. Detect whether both
12115 vector inputs are the same before constructing the vec_perm_indices,
12116 and update the number of inputs argument accordingly. Use the
12117 utility functions added above. Only construct sel2 if we need to.
12119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12121 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12122 the broadcast of the low byte.
12123 (expand_mult_highpart): Use an explicit encoding for the permutes.
12124 * optabs-query.c (can_mult_highpart_p): Likewise.
12125 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12126 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12127 (vectorizable_bswap): Likewise.
12128 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12129 explicit encoding for the power-of-2 permutes.
12130 (vect_permute_store_chain): Likewise.
12131 (vect_grouped_load_supported): Likewise.
12132 (vect_permute_load_chain): Likewise.
12134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12136 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12137 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12138 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12139 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12140 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12141 (vect_gen_perm_mask_any): Likewise.
12143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12145 * int-vector-builder.h: New file.
12146 * vec-perm-indices.h: Include int-vector-builder.h.
12147 (vec_perm_indices): Redefine as an int_vector_builder.
12148 (auto_vec_perm_indices): Delete.
12149 (vec_perm_builder): Redefine as a stand-alone class.
12150 (vec_perm_indices::vec_perm_indices): New function.
12151 (vec_perm_indices::clamp): Likewise.
12152 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12153 (vec_perm_indices::new_vector): New function.
12154 (vec_perm_indices::new_expanded_vector): Update for new
12155 vec_perm_indices class.
12156 (vec_perm_indices::rotate_inputs): New function.
12157 (vec_perm_indices::all_in_range_p): Operate directly on the
12158 encoded form, without computing elided elements.
12159 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12160 encoding. Update for new vec_perm_indices class.
12161 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12162 the given vec_perm_builder.
12163 (expand_vec_perm_var): Update vec_perm_builder constructor.
12164 (expand_mult_highpart): Use vec_perm_builder instead of
12165 auto_vec_perm_indices.
12166 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12167 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12168 or double series encoding as appropriate.
12169 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12170 vec_perm_indices instead of auto_vec_perm_indices.
12171 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12172 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12173 (vect_permute_store_chain): Likewise.
12174 (vect_grouped_load_supported): Likewise.
12175 (vect_permute_load_chain): Likewise.
12176 (vect_shift_permute_load_chain): Likewise.
12177 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12178 (vect_transform_slp_perm_load): Likewise.
12179 (vect_schedule_slp_instance): Likewise.
12180 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12181 (vectorizable_mask_load_store): Likewise.
12182 (vectorizable_bswap): Likewise.
12183 (vectorizable_store): Likewise.
12184 (vectorizable_load): Likewise.
12185 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12186 vec_perm_indices instead of auto_vec_perm_indices. Use
12187 tree_to_vec_perm_builder to read the vector from a tree.
12188 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12189 vec_perm_builder instead of a vec_perm_indices.
12190 (have_whole_vector_shift): Use vec_perm_builder and
12191 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12192 truncation to calc_vec_perm_mask_for_shift.
12193 (vect_create_epilog_for_reduction): Likewise.
12194 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12195 from auto_vec_perm_indices to vec_perm_indices.
12196 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12197 instead of changing individual elements.
12198 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12199 the vector in d.perm.
12200 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12201 from auto_vec_perm_indices to vec_perm_indices.
12202 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12203 instead of changing individual elements.
12204 (arm_vectorize_vec_perm_const): Use new_vector to install
12205 the vector in d.perm.
12206 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12207 Update vec_perm_builder constructor.
12208 (rs6000_expand_interleave): Likewise.
12209 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12210 (rs6000_expand_interleave): Likewise.
12212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12214 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12215 to qimode could truncate the indices.
12216 * optabs.c (expand_vec_perm_var): Likewise.
12218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12220 * Makefile.in (OBJS): Add vec-perm-indices.o.
12221 * vec-perm-indices.h: New file.
12222 * vec-perm-indices.c: Likewise.
12223 * target.h (vec_perm_indices): Replace with a forward class
12224 declaration.
12225 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12226 * optabs.h: Include vec-perm-indices.h.
12227 (expand_vec_perm): Delete.
12228 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12229 (expand_vec_perm_const): Declare.
12230 * target.def (vec_perm_const_ok): Replace with...
12231 (vec_perm_const): ...this new hook.
12232 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12233 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12234 * doc/tm.texi: Regenerate.
12235 * optabs.def (vec_perm_const): Delete.
12236 * doc/md.texi (vec_perm_const): Likewise.
12237 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12238 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12239 expand_vec_perm for constant permutation vectors. Assert that
12240 the mode of variable permutation vectors is the integer equivalent
12241 of the mode that is being permuted.
12242 * optabs-query.h (selector_fits_mode_p): Declare.
12243 * optabs-query.c: Include vec-perm-indices.h.
12244 (selector_fits_mode_p): New function.
12245 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12246 is defined, instead of checking whether the vec_perm_const_optab
12247 exists. Use targetm.vectorize.vec_perm_const instead of
12248 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12249 fit in the vector mode before using a variable permute.
12250 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12251 vec_perm_indices instead of an rtx.
12252 (expand_vec_perm): Replace with...
12253 (expand_vec_perm_const): ...this new function. Take the selector
12254 as a vec_perm_indices rather than an rtx. Also take the mode of
12255 the selector. Update call to shift_amt_for_vec_perm_mask.
12256 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12257 Use vec_perm_indices::new_expanded_vector to expand the original
12258 selector into bytes. Check whether the indices fit in the vector
12259 mode before using a variable permute.
12260 (expand_vec_perm_var): Make global.
12261 (expand_mult_highpart): Use expand_vec_perm_const.
12262 * fold-const.c: Includes vec-perm-indices.h.
12263 * tree-ssa-forwprop.c: Likewise.
12264 * tree-vect-data-refs.c: Likewise.
12265 * tree-vect-generic.c: Likewise.
12266 * tree-vect-loop.c: Likewise.
12267 * tree-vect-slp.c: Likewise.
12268 * tree-vect-stmts.c: Likewise.
12269 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12270 Delete.
12271 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12272 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12273 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12274 (aarch64_vectorize_vec_perm_const): ...this new function.
12275 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12276 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12277 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12278 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12279 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12280 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12281 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12282 into...
12283 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12284 check for NEON modes.
12285 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12286 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12287 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12288 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12289 into...
12290 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12291 the old VEC_PERM_CONST conditions.
12292 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12293 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12294 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12295 (ia64_vectorize_vec_perm_const_ok): Merge into...
12296 (ia64_vectorize_vec_perm_const): ...this new function.
12297 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12298 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12299 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12300 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12301 * config/mips/mips.c (mips_expand_vec_perm_const)
12302 (mips_vectorize_vec_perm_const_ok): Merge into...
12303 (mips_vectorize_vec_perm_const): ...this new function.
12304 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12305 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12306 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12307 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12308 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12309 (rs6000_expand_vec_perm_const): Delete.
12310 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12311 Delete.
12312 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12313 (altivec_expand_vec_perm_const_le): Take each operand individually.
12314 Operate on constant selectors rather than rtxes.
12315 (altivec_expand_vec_perm_const): Likewise. Update call to
12316 altivec_expand_vec_perm_const_le.
12317 (rs6000_expand_vec_perm_const): Delete.
12318 (rs6000_vectorize_vec_perm_const_ok): Delete.
12319 (rs6000_vectorize_vec_perm_const): New function.
12320 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12321 an element count and rtx array.
12322 (rs6000_expand_extract_even): Update call accordingly.
12323 (rs6000_expand_interleave): Likewise.
12324 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12325 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12326 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12327 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12328 (rs6000_expand_vec_perm_const): Delete.
12329 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12330 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12331 (altivec_expand_vec_perm_const_le): Take each operand individually.
12332 Operate on constant selectors rather than rtxes.
12333 (altivec_expand_vec_perm_const): Likewise. Update call to
12334 altivec_expand_vec_perm_const_le.
12335 (rs6000_expand_vec_perm_const): Delete.
12336 (rs6000_vectorize_vec_perm_const_ok): Delete.
12337 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12338 reference to the SPE evmerge intructions.
12339 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12340 an element count and rtx array.
12341 (rs6000_expand_extract_even): Update call accordingly.
12342 (rs6000_expand_interleave): Likewise.
12343 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12344 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12345 new function.
12346 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12350 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12351 vector mode and that that mode matches the mode of the data
12352 being permuted.
12353 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12354 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12355 directly using expand_vec_perm_1 when forcing selectors into
12356 registers.
12357 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12361 * optabs-query.h (can_vec_perm_p): Delete.
12362 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12363 * optabs-query.c (can_vec_perm_p): Split into...
12364 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12365 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12366 particular selector is valid.
12367 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12368 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12369 (vect_grouped_load_supported): Likewise.
12370 (vect_shift_permute_load_chain): Likewise.
12371 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12372 (vect_transform_slp_perm_load): Likewise.
12373 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12374 (vectorizable_bswap): Likewise.
12375 (vect_gen_perm_mask_checked): Likewise.
12376 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12377 implementations of variable permutation vectors into account
12378 when deciding which selector to use.
12379 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12380 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12381 with a false third argument.
12382 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12383 to test whether the constant selector is valid and can_vec_perm_var_p
12384 to test whether a variable selector is valid.
12386 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12388 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12389 * optabs-query.c (can_vec_perm_p): Likewise.
12390 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12391 instead of vec_perm_indices.
12392 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12393 (vect_gen_perm_mask_checked): Likewise,
12394 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12395 (vect_gen_perm_mask_checked): Likewise,
12397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12399 * optabs-query.h (qimode_for_vec_perm): Declare.
12400 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12401 (qimode_for_vec_perm): ...this new function.
12402 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12404 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12406 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12407 does not have a conditional at the top.
12409 2018-01-02 Richard Biener <rguenther@suse.de>
12411 * ipa-inline.c (big_speedup_p): Fix expression.
12413 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12415 PR target/81616
12416 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12417 for generic 4->6.
12419 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12421 PR target/81616
12422 Generic tuning.
12423 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12424 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12425 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12426 cond_taken_branch_cost 3->4.
12428 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12430 PR tree-optimization/83581
12431 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12432 TODO_cleanup_cfg if any changes have been made.
12434 PR middle-end/83608
12435 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12436 convert_modes if target mode has the right side, but different mode
12437 class.
12439 PR middle-end/83609
12440 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12441 last argument when extracting from CONCAT. If either from_real or
12442 from_imag is NULL, use expansion through memory. If result is not
12443 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12444 the parts directly to inner mode, if even that fails, use expansion
12445 through memory.
12447 PR middle-end/83623
12448 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12449 check for bswap in mode rather than HImode and use that in expand_unop
12450 too.
12451 \f
12452 Copyright (C) 2018 Free Software Foundation, Inc.
12454 Copying and distribution of this file, with or without modification,
12455 are permitted in any medium without royalty provided the copyright
12456 notice and this notice are preserved.